UnderstandingQualitativeResearch PDF
UnderstandingQualitativeResearch PDF
iii
Edited by
Contents
viii
Introduction
Robin Usher
23
24
39
52
67
81
83
96
107
119
10
132
134
vii
Jackie Powell
12
150
13
167
14
190
15
203
16
214
Notes on Contributors
226
Index
229
The purpose of the Social Research and Educational Studies series is to provide authoritative
guides to key issues in educational research. The series includes overviews of fields,
guidance on good practice and discussions of the practical implications of social and
educational research. In particular, the series deals with a variety of approaches to
conducting social and educational research. Contributors to this series review recent
work, raise critical concerns that are particular to the field of education and reflect on the
implications of research for educational policy and practice.
Each volume in the series draws on material that will be relevant for an international
audience. The contributors to this series all have wide experience of teaching, conducting
and using educational research. The volumes are written so that they will appeal to a wide
audience of students, teachers and researchers. Altogether the volumes in the Social
Research and Educational Studies series provide a comprehensive guide for anyone concerned
with contemporary educational research.
The series will include individually authored books and edited volumes on a range of
themes in education including qualitative research, survey research, the interpretation of
data, self-evaluation, research and social policy, analyzing data, action research and the
politics and ethics of research.
The last five years have witnessed considerable discussion and debate about the shape
and scope of research training for postgraduate students in general, and social science
postgraduates in particular. The contributors to this volume provide a wide ranging
discussion of issues and themes that are of central concern for the education and training of
those engaged in social and educational research. The papers offer a starting point for new
researchers that will guide their work and promote discussion and debate in graduate
seminars. It is to be hoped that all together these papers will stimulate a range of new
thinking and new work among those engaged in social and educational research at
graduate level and beyond.
Robert Burgess
University of Warwick
1
Introduction
Robin Usher
This collection comprises the thoughts of individual social researchers who approach their
work from a variety of perspectives. The diverse approaches represented here reflect the
diversity of socio-economic problems which are likely to be found amongst any group of
social scientists. If this collection has any one single theme that stands out above all others,
it is that in the realm of social and economic research there is no single correct practice
and no superordinate methodology. This theme is exemplified in the papers. Their
diversity and variety both in terms of research methods and disciplinary focus, far from
being a sign of weakness, is rather a sign of difference at work. Far from being a matter of
regret, it is in our view a matter of celebration and a mark of the sophistication and
complexity of the process of social research. Our hope is therefore, that these papers will
stimulate processes of exploration and interrogation amongst social researchers.
In order to express this diversity, contributors were left free to develop their chosen
topic in their own way, subject only to the condition first, that their contribution should
be rooted in their own practice as researchers but that they should try to be as reflexive as
possible about this, and second, that they endeavour to highlight the epistemological
commitments embedded in their research practice and the methodological choices that
follow from these commitments. The hope is, that as a consequence, readers will be better
able to understand the connection between epistemology, methodology and research
practice, and in doing so enrich their own research. However, it is important to
emphasize that this is no DIY manual of research methods since the message that emerges
very clearly, taking the contributions collectively, is that there can be no such manual.
Methods are not part of some universal algorithm of how to do research but are a
function of concrete research practice, rooted in research traditions and paradigms.
What can be achieved, and this is clearly revealed in the contributions, is that research
methods need to be, and indeed can be, subjected to critical scrutiny. This critical
scrutiny is, however, somewhat different from what would be normally understood by
this term, for what is most in need of scrutiny is not the outcomes of research (which is
the way methods are usually assessed) but also the epistemological commitments of any
research. Even if researchers feel confident methodologically and see no need to review
their methods, it is increasingly the case that social research now demands not only an
evaluation of outcomes, but a reflexive analysis of the research process and of the place of
the researcher within that.
2 ROBIN USHER
INTRODUCTION 3
4 ROBIN USHER
involve the application of scientific method; third, that the kinds of things that can be
known are those that are directly observable and quantifiable.
This seems to imply that scientific method is an abstract set of logical rules,
independent of the world and its social practices (as it were, made in heaven), and
universal in their applicability, i.e. all knowledge claims can be differentiated in the same
way. Critics of positivism argue however, that not only is there no one single
epistemology, no one single test or set of rules of what is to count as knowledge, but that
epistemology should not be understood as defining a set of universal logical rules. Instead,
whatever rules there are should be seen as a cultural artefact, historically-located and
value-laden. Rationality is neither universal, culturally neutral nor invariant in its form.
Scientific method, as we have seen for instance, has itself evolved historically with the
growth of the natural sciences and of Western philosophy. Furthermore, there is no single
scientific method, rather it is more aptly understood as ways of working specific to
particular research paradigms and to particular disciplinary pursuits.
Different epistemologies provide different versions of how things can be known.
Epistemologies in this sense are linked to disciplines with different disciplines having
different ways of knowing the world. They are also linked to ontologies, different
versions of what kinds of things exist in the world. Any research paradigm or tradition has
its own epistemology, its own way of validating its knowledge claims (as the contribution
by McKenzie on economic method and Hamlin on political economy show very clearly).
Disciplines are located within a paradigm (or in the social sciences more than one
paradigm as many of the contributions show). Paradigms delineate what questions can be
asked, what can be researched, what is an appropriate methodology, what constitutes
data, and what kind of tests enable beliefs to be counted as knowledge. Any research
method or procedure is therefore inextricably embedded in commitments to particular
versions of the world and to particular ways of knowing it, and the researcher, by using
these methods and procedures rather than others, reproduces and strengthens these
commitments. Epistemology is not just a technical philosophical procedure but a
commitment to a particular way of understanding the world and acting within it through
research. It follows from this that there is no means of carrying out research which is
neutral and self-validating, any method in the final analysis being dependent on its location
in disciplinary paradigms and research traditions and on an epistemological-ontological
rationale and position.
Of course, this is not to say that the situation in the disciplines of the social sciences is
one of easy-going plurality and the ready toleration of difference. On the contrary,
epistemologies in the social sciences are sites of struggle and this is particularly the case in
newly emerging disciplines. The contributions by Payne on Nursing and Avison on
Information Systems, both newly emerging disciplines, make this very clear. By
highlighting the question of where disciplines are to be located within the spectrum of the
social sciences, issues to do with boundaries and exclusions are raised in an acute form, both
within disciplines and between disciplines.
Some epistemologies have more credibility and dominance that others and this is not so
much a matter of their natural goodness in describing the world but because they are
powerful. The most powerful is still a positivist/empiricist epistemology that holds up the
INTRODUCTION 5
methods and procedures of the natural sciences as the model for all research and which
implicitly understands itself as a universal epistemology. The consequences of the
dominance of positivist/empiricist epistemology are by now well-known but are still
worth mentioning in this context. First, in the social sciences and in social research a preeminent place has been accorded to the production of generalizable knowledge and the
discovery of law-like regularities. Second, there has been a privileging of the language,
methods and quantification which supposedly characterize research in the natural sciences
(Scotts chapter on qualitative research in education clearly demonstrates the taken-for
grantedness and therefore power of quantitative approaches). Third, feminist research has
shown that the influence of positivist/empiricist epistemology in the social sciences has
been to focus mainly on questions about social life that appear problematic from within
the social experiences that are characteristic for men (Harding, 1987; p. 16). Hence the
tests to which knowledge claims are subjected are always gendered, male definitions
prevail of the things in the world that can be known and, as a consequence, the experience
of women becomes invisible. In more general terms, Hand in his chapter shows clearly
the significance for the research process of the initial question asked.
Of course, the dominance of positivist/empiricist epistemology has not gone
unchallenged and the contributors show that there are now a number of counterepistemologies in the disciplines of the social sciences. For example, hermeneutic/
interpretive epistemology argues that the model to be followed is not an idealized and
universal logic of scientific discovery and justification because this is an inappropriate
model for the social sciences. In social research, it is argued, the test of knowledge should
not be generalization and prediction but interpretive power, meaning and illumination.
The focus should be on human action and interaction which by its nature is meaningful and
hence has to be interpreted. An idealized logic drawn from the natural sciences and a
limiting of what can be known to the empirically given cannot hope to elucidate these
meanings and thus cannot hope to portray the rich diversity of the social world.
Yet concluding from this that an unbridgeable chasm exists between the natural and
social sciences may not be the only possibility. Kuhns view of natural science as it is
actually practised, rather than how it understands itself and is itself understood,
emphasizes the importance of normative consensus and commitment within research
communities as the means by which a paradigm is maintained and realized in practice. In
effect, this means that the natural and social sciences are actually not that different
although in a reverse way to that which is normally assumed. The social sciences need not
strive to be more like the natural sciences, since the latter might actually be more like the
former than we might at first think. Kuhn (1970) helps us to see that the way research in
the natural sciences is practised does not follow a positivist/empiricist epistemologyon
the contrary, there is a significant hermeneutic/interpretive dimension. The difference,
for example, between the natural sciences and the social sciences is not that the former
are more objective whilst the latter are subjective. Since any research practice has a
hermeneutic dimension, the natural sciences are just as subjective in this sense as the
social sciences. Here, it is important to note however that subjective does not refer to
individual subjectivity but to the subjective as socially defined.
6 ROBIN USHER
Paradigms are frameworks that function as maps or guides for research communities,
whether in the natural or the social sciences. They provide ways of looking at and ways of
working in the world. As we have noted earlier, they define the objects, direction and
methods of research. To this extent, they provide a social subjective. Knowledge is not
the product of an individual consciousness nor is the latter the means by which knowledge
is validated. Rather than research or finding out about the world being a matter of
epistemological individualism, we should see it as the outcome of active and historically
evolving communities. Furthermore, if research is a social practice carried out by research
communities this means that what constitutes objectivity and the objective will be
defined by the community and the paradigm which shapes its workin other words, it is
the community rather than the world or a set of universal rules which decides what is
objective.
Research in the positivist/empiricist mode proceeds from hypothesis formation to data
collection and to the verification or disconfirmation of hypotheses in the light of the data.
More interpretive approaches such as grounded research and action research (discussed by
Avison, Bartlett and Payne, Powell, Payne, Scott and Nandhakumar) start with data and
generate hypotheses or theory from the data in a backwards and forwards, or dialectical
movement, between theory and data. Here the emphasis is more on searching for order
rather than assuming it is already there as a positivist approach does. The feminist
approach (as discussed by Pat Usher, Orme and Payne) is concerned with critiquing all
research traditions by exposing their roots in androcentrism and patriarchal control and
exclusion. This however does not lead to a rejection of existing research methods but rather
to a more openended and critical approach where a diversity of methods can be employed
within an inter-disciplinary environment
Postmodern research (as discussed by R.Usher) on the other hand, whilst recognizing
that there is no postmodern method(s) as such, questions both the assumption of, and the
search for, a pre-existing order by highlighting the privileging and exclusion characteristic
of all research paradigms and traditions. This emphasizes the need for reflexivity, a
recognition of the way in which research constructs a world to be researched.
Furthermore, in contrast to the positivists narrow empiricism, the postmodern empirical
is much wider, and one with greater potential for the social scientist. Here empirical
testing can relate to practice in close and fruitful ways. When the test of a theory is not its
correspondence with the empirical but its correspondence with the practical, the
empirical includes, rather than excludes, the practical. The merit of a theory then is found
in its practical implication and efficacy in solving problems of the discipline. By
questioning the hierarchically structuring and rigid binary oppositions, for example
between theory and practice, the universal and the local, the abstract and the specific, the
rigorous and the relevant, postmodern researchers seek therefore to replace order,
homogeneity and determinacy as the prime goal of research with diversity, difference and
indeterminacy. By foregrounding the power inherent in all disciplinary-based research
they draw our attention to the very ambivalence of disciplinethat it is at one and the
same time, both knowledge and social control.
INTRODUCTION 7
References
HARDING, S. (1987) Introductionis there a feminist method?, in HARDING S. (Ed) Feminism
and Methodology, Buckingham, Open University Press.
KUHN, T. (1970) The Structure of Scientific Revolutions, Chicago, University of Chicago Press.
2
The Age of Reason or the Age of Innocence?
George McKenzie
Introduction
Research is a response to a challenge. It is undertaken in an attempt to solve a problem. We
undertake research everyday whether in our personal lives, as students or as part of out
professional career. We may need to find the best route to travel from Southampton to
Birmingham or the cheapest food market. Natural scientists may undertake research with
the view of developing more efficient means for generating electric power or methods for
increasing farm productivity. As social scientists we are concerned with identifying the
causes of unemployment, poverty and drug abuse and for designing policies which seek to
alleviate these social problems and improve the quality of life. Whatever the nature of the
problem, the challenge and response process is similar. There is a problem. We may have
a vision as to how to deal with that problem and this might take the form of some
conjectures or prior theories as to the best way to proceed. That is, we may have some
idea as to the cheapest shop or about the chemical makeup of farm soil or the causes of
unemployment. Information is gathered, prices are compared, experiments or interviews
are undertaken. Conclusions are drawn and decisions taken.
But this sounds too easy. Wouldnt it be nice if we really could follow the procedure of
theorizing and data collection to solve the problems that we face. If this were the case
then all that would be involved is time and effort which, of course, we are all prepared to
expend. Unfortunately, the process of research is embedded with a terrible anxiety: how
do we know that our conclusions are the correct ones? I may not have time to compare all
prices in all shops. That would be a very expensive activity, so I may simply guess on the
basis of limited information. A more efficient means of generating electricity may be
designed but it may have unforeseen effects on the environment. Or a policy designed to
alleviate unemployment may lead to higher taxes for those employed who in turn as the
majority seek redress through the ballot box. How do I know when I am correct? What
criteria can I use to decide?
The research process is further complicated by the fact that in the face of uncertainty
others may have reached different conclusions. One member of the family may prefer one
shop over another because their father always shopped there. They do not want to look
elsewhere. One group of researchers may argue that coal-based electricity generation is to
be preferred to nuclear-based because it preserves the traditions of the mining
GEORGE MCKENZIE 9
communities. Some policy makers may argue that unemployment is not as important as
inflation. Not only may we be unsure as to the results of our research activity, we have to
justify it in the face of competing claims which not only arise from different interpretations
of available information but different prejudices on the part of the researcher. What criteria
can be used to choose between the competing claims?
The implication is that the process of research is embedded in a churning vortex of
constructive and destructive tensions. In other words, the research process is itself
problematic. Constructive energy is generated through the creation of new modes of
doing things or new ideas which enable us to live and work in more satisfying ways. But in
the wake is left destruction. Traditional modes of production which are part of a way of
life for many are destroyed. Researchers who may have devoted a lifetime to developing
and refining a particular idea now find that this idea is no longer valued. As we seek to
improve our understanding of the world in which we live, we are inevitably drawn to
conclusions that challenge or destroy the current orthodoxy. Thus research openly seeks
to be destructive.
But what about the new propositions? If they are widely accepted and turn out to be
reasonably correct then it is possible that many people, possibly everyone, can be made
better off. But of course the propositions may not be correct. Disastrous errors could be
made as a result. And even if the propositions were true, they could be misused by
unscrupulous individuals. Nuclear power offers many a cheap source of electric power
but it also has huge destructive potential that can be used to threaten national or ethnic
control at the expense of others. New sources of cattle feed that were both cheaper and more
nutritious are today found to be the cause of disease in both cattle and humans. While we
may like to think that our research could lead to blissful order, it can and does cause
chaos. Because this possibility is not fully appreciated, there exists a tension between
those who seek to impose upon the process of research rules which seek to preserve
order, and those who argue that research is itself inherently chaotic and can never be
ordered.
This tension is not new. It is not a product of the Industrial Revolution of the
eighteenth century or of the Electronic Revolution of the late twentieth century. It has
been recognized and discussed since the times of the ancient Greeks at least. Despite the
fact that the world of 2000 AD is remarkably different from the world of 400 BC, the debate
about what we know and how we know it has continued unabated, basically focusing on
the same questions: is it possible to establish rules which define objectively our knowledge
and hence define the process whereby we search for that knowledge? Or is it impossible
to separate the search for knowledge and, indeed, the very concept of knowledge from our
sensory experience. What is the relationship between what we see and understand and
that which is reality? In other words, how do we go about creating knowledge about the
world in which we live?
Inevitably, those who sought to develop rules found themselves continually pressed to
explain exceptions to the existing rules. The uncertain nature of the world and the
uncertain nature of the research process itself were addressed through the use of closure
statements which enabled gaps in reasoning and interpretation to be bridged. Conclusions
could then be drawn by reference to metaphysical concepts such as the soul, as with the
ancient Greeks, with God, as in medieval times, or with the rational mind as during the
Enlightenment of the seventeenth and eighteenth centuries and subsequently in modern
times. By their nature, the closure rules involved leaps of faith and consequently
disbelievers would be marginalized. Clubs in the form of the Academy, the Church and
modern disciplines, such as economics and sociology, would seek to provide fora in which
the believers could refine their ideas and methods. Perhaps not unsurprisingly they would
also seek to monopolize the process of knowledge creation.
Throughout recorded history the sceptics have argued that recognition of the
impossibility of creating universal truths is an important part of any research process
designed to enhance understanding, whether by an individual or by society as a whole.
Our world today is markedly different from that of the ancient Greeks, yet the same
issues are being addressed today as they were 2300 years ago. Thus, although the
development of this chapter will appear to be linear and historical, it really concerns the
ebb and flow in the continuing debate as to what human understanding is about.
I begin with the attempts by ancient Greek thinkers to define the nature of existence
(ontology). This is followed with a discussion of the approach taken by Descartes who
shifted the emphasis away from studying the meaning of existence to the process of
knowledge formation (epistemology). His work is credited with laying the foundations for
the Age of Enlightenment during the eighteenth century when it came to be believed that
rational human thought could provide the basis for all human endeavour. From this era,
the modern natural and social sciences, as we know them, evolved. More recently,
however, postmodern and feminist scholars have argued that there is an inconsistency
between what we claim that we are doing in our research and what we actually do. These
inconsistencies are discussed in the later sections of the chapter.
The Academy
At the heart of Greek thinking was the relationship between reality and perception. This
can be illustrated by the following simple but classic textbook example. Consider a stick
which is partly submerged in water. Viewed for the first time, we would be inclined to
conclude that the stick was bent. In reality, of course, the stick is straight. The proposition
that the stick is bent can be verified or falsified by removing it from the water. This is, of
course, a very simple problem. Suppose instead that we are interested in the process
whereby the water causes wood to rot. To understand that, we would not only need to
see how all the atoms and molecules move and are linked to each other but to explain why
they are linked and why they move. We would need to get inside the wood. Similarly, no
such experiment is available to those researchers trying to identify the causes of war, poverty
or financial collapse. To fully comprehend such problems would require investigators to
put themselves in the shoes of all persons, politicians, the poor, the rich, investors, et al.,
in order to understand their circumstances and motivations in responding to those
circumstances. But of course this is impossible. Others may give us answers to our
questions, but these answers may not reflect what they are actually thinking. They may
have reasons for trying to mislead us either by lying or selectively choosing the
information provided. Or their subconscious mind may be influencing their answer in
GEORGE MCKENZIE 11
some way unknown even to themselves. Whereas, we could remove the distorting effects
of the water in our experiment so as to verify that the stick is straight, it is impossible to
remove the gap between what we may say to others and what we actually think and do.
No matter how hard we try to understand and appreciate another persons actions, it is
impossible to crawl into their mind. This impossibility defines in large measure the limits
of our research enterprise and creates a tension between any claims to objective research
and what is actually being achieved.
Such issues were addressed by Platos Academy, devoted to understanding the
reasoning mind. It was based on the ideas of Plato (circa 427347 BC) and involved his
pupils, including Aristotle, and their successors until around 240 BC. The focus of the
Academy was the essence of existence. In Theaetetus and The Republic, Plato sought to
advance the view that knowledge must be unchanging. According to his thinking,
experience derived from the senses could not count as knowledge since objects viewed in
this way were subject to change. Knowledge could only be achieved through the use of
reason. This would enable generalizations based on mathematics and logic to be drawn
about the world around us. The method involves two steps. First, it is necessary to search
for unchanging objects and when they are identified to describe them. Second, it is
necessary to illustrate how they could be known through the use of reason. This would be
achieved by identifying necessary and sufficient conditions that would define or delimit an
object. This search for knowledge takes the form of trying to identify counter-examples to
a proposed definition. When a definition is immune to counter-examples, it becomes
knowledge for it must then satisfy the criteria of being unchanging. These steps provide
the basis of modern research methodologies that began to evolve rapidly in the
seventeenth century and in the eighteenth century during the Enlightenment. First, the
process involves the establishment of universal propositions or knowledge. Second, the
process involves the assumption that those searching for knowledge were rational,
although it was argued in ancient times that not everyone could achieve this lofty
endeavour. In other words, these intellectual activities could only be undertaken by the
Academy. Only certain kinds of men could achieve the status of a learner and the power
associated with it. Closure would be achieved by restricting discussion as to what is or is
not knowledge to an elite. The Academy was thus a forerunner to the modern disciplines
such as physics or sociology.
The objects that satisfied Platos criteria of knowledge are severely limitedthey refer
to form: triangularity or whiteness are properties of objects that may change. The
definition of a triangle or of the colour white does not change and therefore they count as
knowledge. Platos search for definitions represents his search for knowledge, which itself
required definition. Although it may be possible to follow the method of providing
counter-examples to eliminate particular propositions for their failure to satisfy the
criterion of being unchanging, it is not the same as falsification, which Plato argued was
impossible. In Theaetetus, Plato defines knowledge as justified true belief. It is impossible
to falsify beliefs. You may think that you know something, but this is only thinking that
you know, without actually knowing. Justification for a belief must be provided.
Unfortunately, Plato was unable to provide us with an idea as to what the process of
justification involves.
Aristotle (384322 BC) pushed the ideas of his teacher further and in so doing
attempted to identify more clearly the limits of our ability to know and understand.
Basically Aristotle argued that research was a circular process. Observations of
phenomena would be made and from these explanatory principles would be derived
through induction. Then on the basis of these principles further characteristics of the
phenomena would be deduced. Aristotle stated four criteria that such premises or
assumptions should satisfy. First, they should be true. Second, premises should be
indemonstrable. Third, they should be better known than the conclusion. Fourth, they
should be better known than the conclusion.
In explaining his approach, Aristotle emphasized that it was highly subjective and hence
the search for knowledge was not straightforward. First, observations would be made on
the basis of sensory experience, sight, sound, taste, smell, touch. As Plato had
emphasized, sensory perception was not necessarily objective. Two individuals might
perceive the same phenomena in different ways. Inductive reasoning could then take two
forms. One could enumerate or catalogue observations and then seek to draw
generalizations from the results, or one might seek to draw generalizations on the basis of
insight. Some unexplained intuitive process enabled the investigator to identify
generalizations. Aristotle was quick to point out the limitations of the inductive-deductive
process. First, he recognized that a particular phenomena could be deduced from several
sets of premises. Thus he excluded as valid arguments those statements which have true
conclusions but false premises. Second, Aristotle recognized that scientific enquiry must
accept that some premises are not capable of verification, otherwise there would be no
end to any investigation. Thus uncertainty is a fundamental property of the research process.
This recognition of uncertainty then leads Aristotle to conclude that knowledge resides in
the mind and that the mind is part of the soul. The concept of the soul whose existence
can neither be proved or disproved serves as a mythological construct to fill a gap in
understanding what exists and what is knowable. This is analogous to Platos concept of
true belief.
Warning flags have been hoisted deliberately. If knowledge claims are restricted to
propositions based on immutable concepts such as white or a triangle, then it would
seem that there is little that we can do to enhance our knowledge about contemporary social
and economic problems. For example, let us take a brief diversion and consider the
problem of unemployment. One measure of unemployment is the difference between the
amount of time that individuals desire to work and the amount of time that employers
actually offer at current wages. However, this definition is based on human desire which
is changeable and differs across individuals. Not everyone will desire to work the same
amount of time. Another definition is based upon the difference between the amount of
work that is socially just and that which is actually offered. But since everyone has different
notions of justice this definition also fails Platos criterion. Yet another definition involves
the difference between the amount of work which would be offered by a free competitive
market where labour demand and supply were equal and that actually offered. However,
since we can never observe such a free market equilibrium this definition fails as well. In
other words, it is impossible for contemporary social scientists to agree on a definition of
unemployment.
GEORGE MCKENZIE 13
Over subsequent years, many sceptical voices emerged, perhaps inspired by the
statement by Socrates (469399 BC) that the only thing that he knew was that he knew
nothing. A vacuum was being created because of the failure on the part of thinkers to
develop concepts of knowledge creation which satisfied their criteria based upon a
dichotomy between the object of investigation and the perceptions of the investigator.
Theological precepts began to be used to fill the gap. St. Augustine (354430 AD)
claimed that human knowledge would be impossible if God did not enlighten the human
mind and thereby allow it to comprehend ideas. In pursuing his vision he sought to
demolish the ideas of the New Academy, intellectual descendants of The Academy. These
ideas were characterized by propositions advanced by Cicero (10643 BC), namely that
since we can never know what is true, we must act according to what seems probable.
Augustine thereby characterizes Cicero as someone for whom knowledge is uncertain. In
attempting to deal with this position which was unacceptable to him, Augustine sought to
revive Platonic criteria: knowledge must be timeless, immutable and accessible to the
mind. However, the human mind was itself immaterial to the process except in the sense
that it was enlightened by God. Indeed, Augustine proposed that the human mind was
part of God. This doctrine was to apply to all knowledge creation. In a sense what
Augustine sought to do was very much what both physical and social scientists do today
when they reach an impasse. A conjectural story or myth is created to fill some gap. In
this case, the impasse was the understanding of the process of knowledge creation and the
role of the individual which through Augustines epistemology could be brought to the
forefront as part of God. Although such stories could never be proved or disproved,
provided that people believed them on faith and their structure was not completely
inconsistent with the worldly environment, then such myths could provide a basis for
stable societies. Inconsistencies were characterized as mysteries, themselves the product of
Gods enlightening influence. The role of reconciling the inability to define knowledge in
some universal had shifted from the Academy to God.
The Age of Enlightenment
Beginning in the twelfth century, Aristotles works on scientific enquiry began to be
translated from Greek and Arabic into Latin thereby making them accessible to a wider
European audience and for the following four centuries scholars interpreted and debated
the meaning of Aristotles thoughts and those of the Academy. By the seventeenth and
eighteenth centuries the fruits of scientific enquiry were being more widely enjoyed. The
world was no longer conceived as being flat. Trade with the New World and Asia was
leading to increased prosperity. Demand for new and better consumption goods increased
and this led to pressures on productive processes and this challenge led to the Industrial
Revolution. Many believed that the scientific principles which were transforming both
industry and agriculture could be used to transform society and the way in which it was
organized. The Enlightenment began.
Some historians have sought to date the Enlightenment as the period between the
English Revolution of 1688 and the French Revolution of 1789, roughly the eighteenth
century. However, two of the most prominent and enduring influences on the
Enlightenment were thinkers from the early seventeenth century, Ren Descartes (1596
1650) and Francis Bacon (15611626). Both were influenced by the writing of Aristotle.
Both sought to overcome the limitations on the search for knowledge which Aristotle had
identified. In their thinking we can identify the seeds of the positivist and interpretivist
methodologies which dominate social and economic research today.
Descartes was a mathematician whom many credit with the invention of analytical
geometry (e.g. the so-called Cartesian co-ordinates). Like Plato, he believed that the
essence of science was the discovery of relationships that could be expressed in
mathematical form. In his mind, mathematics provided the structure for the creation of
certain knowledge. Descartes argued that deductive reasoning based upon mathematics
should be the basis of all human enquiry. Such propositions are not produced by sense
experience but independently by reason, he argued. Descartes expressed this approach in
the form of rules which he believed would expand the realm of human understanding.
Basically these rules stated that research should only be undertaken when it was known
that a certain outcome was feasible. As with Plato, Aristotle and Augustine, the emphasis
is on the search for truth and certainty. Any research that produces knowledge which is
uncertain is to be avoided.
However, rather than elaborate extensively on Cartesian logic, I am going to turn to a
much earlier debate between St. Anselm of Canturbury (circa 10331109) and the monk
Gaunilo. Few contemporary philosophers, I suspect, will attach much importance to this.
However, to my mind, it pre-emptively identifies the limitations of rational thought that
underpinned the Age of Enlightenment and Modern Times. Descartes was a believer in
the proof of the existence of God advanced by St. Anselm of Canterbury in his Proslogion.
This is a masterpiece of deductive reasoning. Assumptions are made and a result is
derived, entirely in the spirit of contemporary economic and sociological analysis. St.
Anselms proof is about a page long and is well worth studying. Its argumentation is based
on the axiom that God is the thing than which a greater cannot be conceived. If a thought
exists in the mind, then it exists in reality. Since there is no greater thought than God,
then God must exist. Anselm argued that his proof was so transparent that it had to be
accepted by any fool who denied Gods existence. In retort the monk Gaunilo argued in
his A Reply on Behalf of the Fool that Anselms reasoning could be used to prove anything.
In essence, propositions based upon deductive reasoning are conditional upon
assumptions. Although the logical process of reasoning involved may be invariate, as in
mathematics, the assumptions may not. Of course if the underlying assumptions are
shared by many people, then they may form the basis for a religion based upon those
shared beliefs. However, such thoughts need not exist for everybody. Deduction could not
enable a particular idea to be advanced as universal. This was Gaunilos conclusion.
If Descartes deductive approach can be characterized as bottom-up in the sense that
one works from assumptions to conclusions, then Bacons inductive approach is very
much top down. With Descartes and Bacon, we have the destruction of Aristotles
inductive-deductive circle. Like Descartes, Bacon believed that the rational human mind
was capable of knowledge creation. However, this would not take the form of rational,
deductive thought, but of rationally conceived empirical investigations. Organized
experiments would be undertaken with a systematic recording of results. Attempts at
GEORGE MCKENZIE 15
world of certainty, optimism and a rationally constructed social order. The edifice of
Descartes and Bacon was being extended. Adam Smith (1723 1790) and Karl Marx
(18181883), as examples, both applied deductive argumentation in their attempts to
understand the evolution of society. So powerful are some of their propositions that they
continue to dominate economic and social policy formation implicitly if not explicitly some
200 years later. For example, Adam Smith utilized deductive argument to advance the
case for the abolition of monopolistic organizations and for free international trade. His
theorem concerning the benefits of free trade, though refined over the years by
economists, forms the basis for the European Community, the North American Free Trade
Area, US and European anti-monopoly legislation and many policies being implemented
in the former Soviet Union. As a result many have argued that free competition should be
the universal norm governing all societies. But wherever we see claims to universality, we
know that there will exist some equally strong counter argument. In this case, free trade
argument says nothing about how the gains from trade are distributed, a concern which
continues to provide the basis for considerable political debate in most countries.
Auguste Comte (17981857) sought to lay the foundations for a new discipline which
he called social physics and ultimately labelled sociology. He advocated a positive
method involving the correlation of facts in an attempt to understand the nature of
society. He viewed this positive science as the third and final stage of human
development. The first stage involves the interaction between the theocracy and the
secular state. In the second or metaphysical stage, there is rebellion against secular and
religious authorities out of which change emerges. This is followed by a positive period
during which the scientific method forms the basis for a durable and ordered society.
However, in 1844 Comte began to reconsider his ideas when he became friends with the
journalist Clotilde de Vaux and they both developed the thesis that rational thought
needed to be subordinated to love of people, and that only motivated in this way could
scientific enquiry promote progress.
Max Weber, however, argued that the subjective and objective aspects of research
needed to be kept separate even though subjective concerns were driving the research
endeavour, viz.
What is really at issue is the intrinsically simple demand that the investigator and
teacher should keep unconditionally separate the establishment of empirical facts
(including the value-oriented conduct of the empirical individual whom he is
investigating) and his own practical evaluations, i.e. his evaluation of these facts as
satisfactory or unsatisfactory (including among these facts evaluations made by the
empirical persons who are the objects of investigation.) These two things are
logically different and to deal with them as though they were the same represents a
confusion of entirely heterogeneous problems. (1949, p. 11)
Whereas Descartes and Bacon claimed that deductive or inductive methods could be used
in all areas of human inquiry, Weber would argue that there is a difference between the
natural and social sciences involving the nature of the object of research. While care
would need to be taken in distinguishing between the subjective and objective, Weber
GEORGE MCKENZIE 17
argues that once this is done, then the scientific method can be applied. For many
researchers, however, the subjective and objective cannot be compartmentalized in this
dichotomous fashion. Knowingly or unknowingly, we build our individual prejudices into
our research.
Another attempt at trying to identify a universally applicable research methodology is
associated with Karl Popper (19021994). He argued that the purpose of research is to
falsify. Indeed he argued that rules should be established which only admit hypotheses
which are falsifiable (1965, p. 49). In a sense he is applying the logic of Descartes. He
seeks certainty and universality in falsification. For Popper, the only interesting
propositions are those that are falsifiable. Consider his classic example based on the
proposition that all swans are white. Either it is true or it is false. As soon as a black
swan is observed, the proposition is falsified. There is complete certainty. Indeed we have
a proposition that would satisfy Plato. Unfortunately, problems of interest in social and
economic research are never posed in such clearcut terms. For example, many believe
that government policies to curb inflation by strictly controlling the availability of money
is the prime cause of unemployment. This proposition can never be falsified because it
would involve a detailed modelling of all aspects of human life including the
interrelationships between the financial sector, the industrial and service sectors and
labour markets. This has proven to be impossible.
The existence of such tensions were noted by Baudelaire. In 1863 he sought to define
modernity as the transient, the fleeting, the contingent. It is the one half of art, the other
being the eternal and the immutable. Modern social research seeks to develop claims
which are eternal and immutable. In contrast, we need to be aware of the transient and
contingent nature of our understanding. It is the tension in this duality which in this
writers opinion needs to be drawn out and made explicit as we pursue our research
projects.
But where does this leave us in our search for a unified scientific approach to research?
This is the theme which the physicist and philosopher of science, Thomas Kuhn (1922),
sought to develop in his History of Scientific Revolutions (1962, 1974). He argues that
science itself is a social process in which scientists solve puzzles posed by the dominant
methodology of the particular scientific community in which they participate. Each
community has its paradigms or modes of thought which define a research tradition.
Expressed in such terms, scientific enquiry is bound to appear to be very conservative.
Thus Kuhn argues that progress in understanding can only be achieved if existing paradigms
are destroyed. The challenge of an anomaly in accepted reasoning should lead to a
response urging change,
Because it demands large-scale paradigm destruction and major shifts in the
problems and techniques of normal science, the emergence of new theories is
generally preceded by as period of pronounced professional insecurity. As one
might expect, that insecurity is generated by the persistent failure of the puzzles of
normal science to come out as they should. Failure of existing rules is the prelude
to a search for new ones. (pp. 678)
Postmodern Perspectives
Postmodernism is a rejection of modernity and its emphasis on reason, science and
rationality that was spawned after the Age of Enlightenment. It rejects modern attempts
at identifying universal rules applicable to the organization of society and, in the context
of this volume, to the undertaking of research designed to understand the problems of
that society. It thus denies any possibility of constructing rules which define knowledge in
a unique and enduring way. This does not mean that research should not be undertaken or
that it should be unsystematic. In my view, postmodern perspectives assist us in focusing
upon problems we had not thought existed. It is not a universal method because such a
method does not exist. Rather, it is one of many ways of thinking. It suggests that greater
account needs to be taken of behaviour that has not been defined as rational, that research
is an uncertain process which is only capable of enhancing understanding and not
developing universal laws, and that research necessarily requires the use of diverse
methodologies. It suggests that whenever we appear to identify regularities in human
behaviour, we should also devote effort to seeking to understand why some individuals or
groups do not conform to the pattern. If we return to the duality used by Baudelaire to
characterize modernity, we might conclude that modern social science has been obsessed
with identifying propositions which are eternal and immutable whereas postmodernism
concentrates on the the transient, the fleeting, the contingent.
It is impossible to do justice to the diversity of postmodern perspectives. Some are
discussed by Robin Usher in Chapter 3, hence I will focus on only a few ideas that are
related to the themes that I have raised in previous sections. We have seen how thinkers
from the times of the Greek Academy have queried our ability to establish universal
propositions about the world about us. But then what is it that we are doing? We have moved
very quickly over the past decades from the latter stages of the Industrial Revolution into
the early stages of the Electronic Revolution. Productive processes now require fewer
workers on the production line. However, they now require more people to process the
information necessary to design this machinery and to monitor its output. Financial capital
has become a major ingredient in the Electronic Revolution, with the result that even
more information is being created about the complex properties of diverse financial
instruments. If we cannot hope to develop universal propositions about this data, what
then can we do with it? Derrida (1930) has suggested that what we are really doing is
creating texts based upon our own interpretation. These texts are read by others who in
turn may give them another interpretation. These texts and their interpretations then form
a complex and unexplainable, uncertain dynamic within and around which we make
judgments and decisions. In the social sciences, the rapid growth in the number of data
bases would seem to offer more opportunities for understanding our world, yet all this
new information will tax our capability to tell stories about it. Hence the continuing
search for some universal rule or filter (e.g. statistical analysis) which will tell a story for
us. Literally, we are writing formula novels since the assumptions and style are
unchanging.
Gadamer also seeks to understand the nature of what we write and how we interpret what
we read. He proposes that there is an overarching discipline hermeneutics, the theory and
GEORGE MCKENZIE 19
practice of interpretation. One of his major theses is that the dichotomy between subject
and object which is the basis of the natural sciences has limited validity for the human
sciences. Fundamental to an appreciation of his reasoning is the recognition of the role of
prejudice and tradition. Gadamer argues that contemporary society has a prejudice against
prejudice. He suggests that we should recognize the impossibility of eliminating prejudice
from our work and instead should seek to make its basis explicit. To do this, he argues that
every statement must be seen as an answer to a question and it is the question which first
must be understood. In Gadamers hermeneutics the notion of an unbiased subject
investigating an object is no longer a valid dichotomy. In the human as well as the natural
sciences, our perspective is influenced by the thinking which has preceded it as well as by
our individual perspective. In the human sciences the subject and object are intertwined
through different or similar prejudices and traditions. Gadamer admits that this process of
unwinding is problematical. First, it is in principle endless. Second, there is a limit to
what we can understand about the role of the unconscious in influencing our research
efforts.
A second theme, already alluded to, of postmodernism is the existence of binary
oppositions. Derrida argues that the systems of thought upon which text construction and
interpretation are based is trapped in a world of binary oppositions. He shows that one
term in each pair is always given a privileged position over the other which is marginalized
or excluded. He therefore proposes to examine the term which is marginalized and in so
doing seeks to show that the position of the superior term over the marginalized term is
untenable. The privileged term achieves its status by what it suppresses. In this way power
is achieved. The process by which binary oppositions are unwound is called deconstruction.
In the context of social and economic research, this critique suggests that instead of, or in
addition to, seeking to identify generalizations about human behaviour, we should also
study difference. Instead of focusing on the average or representative person, efforts
should also be made to understand those who have been categorized as marginal. Where
binary oppositions appear, we should seek to identify reasons why one of the terms is
marginalized. But care needs to be taken. This is not a method but a way of thinking.
This dichotomy is particularly apparent in the binary pair involving the subject, or
investigator, and the object, the investigated. Traditionally, it has been the investigator
who has been privileged, developing theory, designing questionnaires and undertaking the
empirical investigation of the objects who are treated as separate, independent entities in
the manner suggested by Weber, as noted previously. The procedure of grounded theory
and action research, in contrast, involve the object in the process of project design. For
example, this might involve consulting with the objects as to their concerns and enabling
them to formulate questions to be included in any survey. Critics suggest that the objects
may be prejudiced and that this will render the results of the survey invalid. However,
this implies that the investigator (subject) is objective.
Similar themes are also apparent in the work of Foucault. His systematic studies of the
marginalized, those labelled as mad or deviant from what is considered to be the norm,
illustrate how madness or deviancy are relative concepts used to control the
contradictions in society in a manner so as to create a society which is governable.
Disciplines create the concepts of madness and deviancy and thereby create subjects both
GEORGE MCKENZIE 21
their problems or in the design of research strategies. Sawicki (1991) illustrates this point
in respect of health policy and research.
This has led to an alternative approach involving the recognition of difference as a basis
for action. Women have different problems from men for a variety of historical, cultural,
social and physiological reasons. It is not surprising, therefore, that many feminist
researchers have sought to integrate postmodern perspectives into their research agenda.
They seek greater awareness about the prejudices that influence education and research.
This involves recognition that social structures (including those within educational and
research institutions) whether based along gender, racial, religious or ethnic lines are
designed to perpetuate the power of the relevant establishment, that there is no
representative woman, let alone representative man, that the role of the unconscious (as
opposed to the rational, conscious self), is important and that progress needs to be made
in unwinding the subject/object dichotomy prevalent in university activities. As such,
contemporary feminist research strategy is increasingly emphasizing the validity of diverse
research methodologies, thereby enabling the use of whatever concepts are required to
understand a particular problem. The question is posed what are the best ways to
understand a problem? rather than I have a methodwhat problems can I apply it to?.
(The reader is referred to the chapters in this volume by Pat Usher and Joan Orme and to
the thoughts of feminist scholars such as Hekman (1990), Flax (1990) and Lather
(1991)).
There are no rules or strict agenda associated with the feminist project, hence it is not a
discipline in the conventional sense. Like postmodernism, it is a perspective. Feminist
research focuses upon the problems facing women and from a perspective that is
impossible for a male. In so doing it challenges the patriarchal discourse of the traditional
disciplines. Because feminist research strategy does not consist of rules that would enable
it be classified as a discipline, feminist scholars leave themselves open to criticisms from
those to seek to defend and advance the patriarchal views that emanated from The
Enlightenment. But, any such defence is bound to fail. In my opinion the open-endedness
of feminist research practice offers a solid basis for all contemporary social and economic
research. It explicitly recognizes the tensions that have been the subject of so much
discussion over the centuries. Attempts to create rules delimiting the creation of
knowledge have been shown to be incapable of providing solutions to the major
contemporary socio-economic problems of unemployment and poverty. Feminist
researchers thus recognize the futility of more than 2500 years of search for universal
propositions governing the process of knowledge creation. By adopting such an openended approach to my own research, I am able to draw upon whatever methodologies and
methods which seem appropriate to me when trying to understand a particular problem.
The focus of my research is the problem rather than the methodology.
References
ANSELM OF CANTERBURY (1077/78) Proslogion, translated by Charlesworth, M.J. (1965),
London, Oxford University Press.
ARISTOTLE (1994) Posterior Analytics, translated by Jonathan Barnes, Oxford Clarendon Press.
AUGUSTINE (1950) Against the Academicians, translated by OMeara, J. London,
BACON FRANCIS, (1878) Novum Organum, Oxford, Clarendon Press.
BAUDELAIRE CHANLES, (1992) Selected Writing on Art and Literature, London, Peugvin.
COMTE AUGUSTE, (1830/1842) Course on Positive Philosophy in ANDRESKI, S. (Ed.) The Essential
Comte, London, Croom Helm.
DERRIDA J. (1976) La Voix et le Phenome, Paris, Presses Universitaires de France.
DESCARTES, REN (1628) Regulae ad directionem ingenii, translated by COTTINGHAM, J.
STOOTHOFF R., and MURDOCH, D. in Descartes: Selected Philosophical Writings (1988),
Cambridge University Press.
FLAX, J. (1990) Thinking Fragments: Psychoanalysis, Feminism, and Postmodernism in the Contemporary West,
Berkeley, University of California Press.
FOUCAULT, M. (1972) The Discourse on Language, in The Archaeology of Knowledge, New York,
Pantheon Books.
GADAMER, H-G. (1981) Reason in the Age of Science, Cambridge, Mass., MIT Press.
GAUNILO (1078) A Reply on Behalf of the Fool, translated by M.J.Charlesworth, published with
ANSELMS Proslogion, London, Oxford University Press.
HEKMAN, S. (1990) Gender and Knowledge, Cambridge England, Polity Press.
KUHN, T. (1962, 1974) The Structure of Scientific Revolutions, second edition, Chicago: University of
Chicago Press.
LATHER, P. (1991) Getting Smart: Feminist Research and Pedagogy With/In the Postmodern, London,
Routledge.
LYOTARD, J-F. (1992) The Postmodern Explained to Children, London, Turnaround.
MARX, KARL (1974) Das Capital, London, Peugvin.
PLATO, (1883) Theaetetus, translated by Campbell, L. Oxford, Clarendon Press.
PLATO, (1976) Republic, translated by Liusay, A.D., London, J.M.Dent
POPPER, K. (1965) The Logic of Scientific Discovery, New York, Harper and Row.
SAWICKI, J. (1991) Disciplining Foucault, New York, Routledge.
SCHUMPETER, J. (1954) History of Economic Analysis, New York, Oxford University Press.
SMITH, A. (1950) The Wealth of Nations, London, J.M.Dent.
WEBER, M. (1949) The Methodology of the Social Sciences, New York, The Free Press of Glencoe.
Part 1
The Nature of Enquiry
This first group of chapters addresses many of the philosophical issues embedded in
social and economic research. In Chapter 3, Robin Usher pursues the theme that research
is really a form of story telling and that the search for deep realities underlying the
modernist perspective is misleading. He focuses upon the postmodern perspective which
questions many aspects of research which many take for granted, for example that data is
independent of its interpretations or that there exist universally accepted criteria for
discriminating amongst theories. In the second part of his chapter, Usher develops the
theme that research is a social practice which expresses itself in textual form, and that the
autobiography of both the researcher and the reader are essential to the interpretation of
the text.
In Chapter 4 Pat Usher also questions the modernist project with its emphasis on
rationality, order and certainty. In so doing she argues that it is patriarchal in nature and
suggests that various themes should form the basis of feminist research. These include,
amongst others, recognition that: a) gender is a pervasive influence that guides the
questions and criteria involved in research projects, b) that all theories are a product of
historical and sociopolitical influences, and, c) that a multiplicity of research methods
should be used within an interdisciplinary environment. Chapters 5 and 6 examine the
application of the economic method to disciplines other than economics, namely politics
and sociology. The economic method involves deductive theorizing which takes as given
that economic agents, i.e. individuals and firms, act rationally so as maximize their
individual objectives. As Alan Hamlin points out, such objectives need not be narrowly
conceived and could include a wide range of altruistic aims. He seeks to develop this mode
of thought as a basis for researching public policy. In Chapter 6, George McKenzie lays
out some of the formal structure of the economic method against the background of Gary
Beckers analysis of family behaviour. His prime objective is encourage the reader to work
through an important, modern mode of analysis as a basis for understanding it as a truth
producing practice within the discourse of a particular discipline.
3
Telling a Story about Research and Research
as Story-telling: Postmodern Approaches to
Social Research
Robin Usher
In his postmodern text America, Jean Baudrillard writes that the point is not to write
the sociology or psychology of the car, the point is to drivethat way you learn more
about this society than all academia could tell you (Baudrillard, 1988; p. 54). For
Baudrillard, participating in American driving behaviour is a better way to understand
contemporary American society than through research as conventionally understood.
Although we can understand what Baudrillard is getting at, its very likely that we find
his remarks disturbing. We are so used to thinking of research as providing a special kind
of methodologically validated knowledge about society which merely driving could not
possibly provide that we suspect Baudrillard of not being entirely serious in his remarks.
For the same reason, its not easy to accept the notion of research as story-telling. We
think of story-telling as unserious, as fictional, whereas our image of research is that it is
about truth and is therefore an altogether more serious business. Equally, its not easy to
accept that an account of research is an example of telling a story. To attempt to explicate
the nature of research through a story does not somehow seem appropriate.
However, another and different reading of Baudrillards remarks, would be that the
task of understanding or finding out about something is best approached indirectly and
obliquely. It is this message which informs my starting-point because I want to present
postmodern approaches to research, initially at least, through a reading of Umberto Ecos
novelsThe Name of the Rose (Eco, 1984) and Foucaults Pendulum (Eco, 1990). This may
seem a rather indirect and oblique way of proceeding but my argument is that a
postmodern understanding of social research is best secured in this way. Thus I will try to
exemplify through the approach itself something important about the postmodern. I will
try to show the textuality, fictionality, and narrativity or story-telling dimensions of
research. I shall present research as story-telling by first telling a story (Ecos) which
although not about research can nonetheless help us to understand research from a
postmodern perspective.
The stories that unfold in Ecos novels have often been seen as a metaphor for the
epistemological quest of the modernist project, a quest for knowledge of the deep
underlying causes of events, for unitary meaning and the total explanation of phenomena.
In Foucault s Pendulum, the three protagonists through their work as publishers of
esoterica, become intrigued by conspiracy theory writings that purport to explain history
in terms of a grand plan cleverly hidden by its authorsthe epistemological quest taken
to its ultimate (and irrational) end. On the assumption that the more unlikely the
connections made the more convincing the plot which ensues, they set out to devise a plan
of their own using random computer-generated associations. When a group committed to
a conspiracy theory of history hears of the plan it proceeds to hunt them down for their
hidden knowledge. The more they protest that there is no plan in reality, the more the
group believes that there is. They end up meeting bizarre deaths at the hands of this group
for a knowledge which does not exist.
At one level, Foucaults Pendulum is questioning the modernist separation of knowledge
from power. For its authors, the plan is nonsense, the knowledge it purveys fictional and
harmless. Yet in the event it is powerful in its effects and in ways not obvious to those
concocting it. The message then is that knowledge which purports to explain reality in
terms of its deep underlying meanings is dangerous and must be treated with caution lest
it overwhelm those who create it and those who become subject to itin Foucault s
Pendulum these are one and the same.
More generally, Foucaults Pendulum is also a cautionary tale to all those in the social
sciences whose objective, in line with their commitment to the modernist project (the
search for deep underlying realities), has been the development of explanatory grand
designs. The story of the plan can be read as an allegory of all the grand designs that
have sought to erect structures and systems, overarching theoriesdesigns which
fabricate the world to the extent that this is what the world ends up becoming. The
thinking that lies behind this is that if only the jigsaw can be constructed from the pieces
of data randomly strewn around we can have the big picture which will explain and give
a meaning to the disorder and contingency of the world. Armed with this deep
knowledge, the world can then be controlled and changed and we, through this, can
become empowered. This is the totalizing dream of the modernist project, a dream which
in the end has nothing to do with reason but everything to do with desirethe desire for
mastery and ultimately controland hence is itself totalitarian.
In The Name of the Rose a monk, William of Baskerville, is called in to solve a number of
inexplicable murders at a monastery with a library which contains the most extensive
collection of books and manuscripts in the Christian world. This quest then becomes
entangled with a quest for the book whose identity is unknown yet whose possession is the
motive for the murders.
The plot of the novel is centred on the librarya library which is itself a labyrinth with
many hidden secrets, the foremost being the mysterious book around which the action
revolves. The library is where all knowledge is to be foundif you know how to find it
and only someone like a detective, the epistemological searcher par excellence with
deeply penetrating observation and highly developed powers of reasoning can unlock the
secrets of the library-labyrinth and thereby know the truth (the identity both of the
murderer and the book).
As McHale (1992) points out the essence of the modernist project is encapsulated in
the question how can the world be truthfully known? The detective is therefore an apt
metaphor for the modernist social researcher. He (and Eco deliberately makes him a he)
seeks the truth usually in the form of a quest for a missing or hidden item of knowledge
and does not rest until this item is found and the truth discovered. The problem, then, which
26 ROBIN USHER
both the detective and the researcher grapple with is epistemological, a problem of the
accessibility and circulation of knowledge where the individual mind, aided only by its
power of reasoning, has to grapple with an elusive, hidden reality to find the deep hidden
meaning that underlies chaotic and disordered events.
Yet, William of Baskerville, the epistemological hero, the prototype detectivescientist, ultimately fails in his quest. He discovers the truth (or rather a truth because
the library contains many truths) but only by stumbling upon it rather than by a successful
chain of reasoning. As he himself finally admits, there was no mastermind, no plot, and no
pattern underlying the murders.
Thus the champion of empiricism and reason only solves the mystery through constant
misinterpretation of the evidence and even then only through irrational associational leaps
prompted by a dream and a grammatical error, both recounted by his adolescent sidekick.
This element of the irrational (or the non-rational) in Williams detective work represents
a challenge to the very adequacy of and faith in systematic reason and scientific method. In
the novel, it subverts the detective storys epistemological structure and in the parallel
world of the modernist project it subverts the epistemological quest at the very heart of
that project. What takes its place is another, alternative concerna postmodern concern
displacing the modernist concern with how the world is to be truly known with the
ontological questionwhat is this world and what is to be done in it?
It is this ontological notion of constructed worlds which mark Ecos novels as
postmodern. In Foucaults Pendulum, the three protagonists by putting together a story
construct a world which becomes murderously real. In The Name of the Rose there are a
plurality of largely incommensurable worlds created at a variety of different levels which
confront each other. Both texts embody the very post-modernist notion that different
languages, different registers of the same language, different discourses each construct the
world differently; in effect, different worlds are knowledged or languaged into being.
At the same time, however, just as these worlds are created they are destabilizedin
other words, their status as constructed worlds is laid bare and the very process and
strategy of world-making is thus foregrounded. In The Name of the Rose there is a
doubling or mirroring of a world within a world. The structure of the library for example
is a scale model of the known world, its floor plan reproducing the map of the world. It is
a labyrinth that represents a labyrinthine world. This doubling of worlds opens up an
abyss of a potentially infinite regress which destabilizes the world constructed by the text
and by so doing foregrounds the concern with world-making and the role of language in
the construction of worlds. The text becomes self-referential or self-reflexive by drawing
attention to its own world-making through the use of textual strategies such as doubling.
It is perhaps worth remembering at this point that stylistically postmodernism is often
said to have its origins in architecture. Modernist structures are homogeneous, oriented
within right-angled grids and sharply bounded, displaying the tyranny of the straight line
(Kvale, 1992; p. 37); they are meant to be triumphs of reason aptly reflecting the
epistemological project of modernity. Postmodernist structures, on the other hand, are
skewed and distorted, displaying shifts and transformations, ambiguous and
unpredictable, with no fixed boundaries. They suspend the notion of a clear final ordering
and instead emphasize ambiguity, the intermixing of orders and styles in a general irony,
pastiche and playfulness. Above all, they reflexively draw attention to themselves as built
structures. In The Name of the Rose Eco projects, through the space of the library, a
reflection of this postmodern conception of the world. As readers, we experience it as
both an interior and physical space of uncertainty, complexity and disorientation. In Ecos
texts there are spaces where worlds are both made and subverted. Like postmodern
structures, we become aware of complexity, are plunged into uncertainty and
disoriented. Yet this in itself is productive because it draws our attention to the process of
world-making and directs us towards a reflexive stance.
Jameson (1991) suggests that postmodern spaces should be read as allegories of the
contemporary ontological condition, the almost unthinkably complex, interrelated and
interactive, global system of multinational capitalism, the information superhighway and
the condition of hyper-reality where the signs and symbols of reality have become more
real than reality itself. Postmodernist space points to the ungraspable of this world and
thus as it can be seen as standing for the futility of our attempts to master it through
totalizing knowledge and the discovery of deep underlying meanings. This is the lesson
that both William of Baskerville and the protagonists of Foucaults Pendulum learn.
Some Features of a Postmodern Approach
As we have seen therefore postmodernism displaces epistemology. There is a general
postmodern scepticism about epistemologys traditional aim of distinguishing true, certain
knowledge through a general account of the nature and limits of knowledge.
Epistemology privileges scientific method as the methodological guarantee of a true and
certain knowledge. It seeks to formulate universal rules as to what can be counted as
scientific knowledge, a set of universal characteristics that qualifies a practice as scientific,
a theory or explanation as adequate. Postmodernism therefore questions epistemologys
dominant positivist/empiricist conception of scientific knowledge which traditionally sets
the standard and provides the model for all knowledge claiming to be scientific. In
particular, it challenges the leading assumptions of the positivist/empiricist research
tradition that:
28 ROBIN USHER
within the terms and discourse of the positivist/empiricist traditionin other words, the
emphasis on the subjective instead of the objective is merely a reversal which still
works within a framework of objective-subjective as polar opposites. Instead, a
postmodern approach seeks to subvert this dichotomy and suggest alternatives which
radically challenge and critique the dominant epistemological discourse in all its various
forms.
One of these alternatives is to turn to the actual historically located practices of the
various sciences, natural as well as social. This historical approach, that for example
informs the work of Kuhn (1970), displaces the essentialist and transcendental view of
science which is to be found in both the positivist/empiricist and hermeneutic/
interpretive traditions and instead argues that all sciences are social practices. This
emphasis on the actual practice of a science highlights its specificity and situatedness and
the practice-constituted criteria for judging the validity of knowledge-claims and theory
choice. I shall say more about this later.
Another postmodern alternative arises as a consequence of the displacement of
epistemology and the foregrounding of ontology. This has two important aspects. The
first, which we have touched on when discussing Ecos texts, is to do with worldmaking through language, discourses and texts. Postmodernism sees knowledgegeneration as a practice of languaging, a practice of textual production. Here, language
is not conceived as a mirror held up to the world, as simply a transparent vehicle for
conveying the meaning of an independent external reality. Since there is always an already
existing structure of significations which gives rather than reflects meaning, referents are
an effect of language. No form of knowledge can be separated from language, discourses
and texts at work within culture. The structures, conceptuality and conventions of
language, embodied in discourses and textslanguage as a meaning-constituting system
govern what can be known and what can be communicated.
One consequence of this is that knowledge, being relative to discourses, is always
partial and perspectival. Thought and experience, the positivist/empiricist bedrock, are
not independent of socio-cultural contexts and practices. As we have seen already when
considering epistemological research traditions, they are coded by language and discourse
in terms of binary oppositions, e.g. masculine-feminine, subject-object, rational-irrational
that constitute identity through powerful and oppressive hierarchies.
The second aspect of the postmodern is a decentring of the knowing subject, the
epistemological subject with a universal and essential human natureunitary, rational,
consciousness-centred and the originary point of thought and action. This essential nature
is conceived as allowing subjects to be autonomous of the world and to occupy an
Archimedean point that transcends their own subjectivity, history and socio-cultural
location. This stance of objectivity, where the subject is a pure experiencer and reasoner
transcending particularity, partiality and contingency, is the condition for the
interchangeability of knowing subjects and hence the public verifiability of scientific
knowledge (Code, 1993).
Postmodernism challenges and displaces this abstract, transcendental subject arguing
instead that subjects cannot be separated from their subjectivity, history and socio-cultural
location. In the postmodern, there are no Archimedean points, the subject is instead
30 ROBIN USHER
re-written and re-read. Hence there is neither an originary point of knowledge nor a
final interpretation. However, as I have just noted, some readings are more powerful than
others. The most powerful readings are those imposed by the violence of closure and the
metaphysics of presence the claim to a direct and unmediated knowledge of the real
that can only be provided by scientific method. Yet given that all readings are subject to
contingency and the historical moment in which they are read, and given that the object
of scientific research is always open to contest, then all claims to presence, to unmediated
knowledgeability, are always problematic.
Postmodernism challenges foundationalism, the position that knowledge is founded in
disciplines, and the consequent boundary-defining and maintenance that is characteristic
of disciplinary knowledge. There are two aspects to this. The first is that social scientists
located as they are in the modernist epistemological project, want to give reasoned,
connected and totalizing accounts. But as we have seen, the world they investigate and
seek to explain is not one that can readily be reflected in their theories and accounts. It is,
in other words, not organized naturally into disciplinary compartments. They are
therefore always as (Acker et al., 1991; p. 149) point out in the business of attempting to
systematically reconstruct social reality and to put these reconstructions into the form of
a social theory.
The social sciences conceive of themselves as representing the real whereas what they
are actually doing is writing it. Social reality does not exist as an extra-discursive
context, rather the real and the discursive are intimately interwoventhe social is
writtenthere is no extra-discursive real outside cultural [i.e. meaning] systemsthe
social world does not consist of ready-made objects that are put into representation
(Game, 1991; p. 4). If disciplinary theorizing is itself a practice of writing, then theory
cannot be tested against the real. The question then becomes by what discursive strategies
does a discipline maintain its claims to the status of knowledge. In the case of the social
sciences, this is done through a privileged representation of social reality based on the
binary opposition real-representation and the consequent repression of the fictionality or
textuality of the social sciences.
Social Research in a Postmodern Mode
Text
Earlier on, I introduced the idea that all research is a social practice. I want to argue that
all social research is located in knowledge-producing communities. Although these are not
exactly communities in the Kuhnian sense with a single and settled research paradigm and
a single disciplinary matrix (Kuhn, 1970), the production of any kind of systematic
theoretical or scientific knowledge (as against informal or craft knowledge) always takes
place and indeed requires a knowledge or research community of some sort, no matter
how flexible and loosely structured it may be.
This inseparability from a knowledge-producing community is what makes research a
social practice. One consequence of this is that it is never a matter of researchers doing
their own thing, although this individualistic conception is a powerful one. Even though
it is not always carried out collectively, research is always collective in two important
senses. First, in the sense that it is intertextual (I shall discuss this further later). Second,
that it is a delimited set of activities legitimated by a relevant community where certain
activities are judged appropriate and function as criteria for validating knowledge
outcomes, whilst others are ruled out of order and excluded. Communities define rules of
exclusion, set boundaries and impose closures. This narrows what can be done and what
will count as legitimate research, valid knowledge outcomes and truth.
The existence of a knowledge-producing community is a taken-for-granted
background, implicit yet largely unacknowledged in specific research activity. As
researchers, our awareness of the activity of research is normally in terms of a process
systems metaphor, e.g. that research is a process or system with the following stages
literature search, hypothesis generation, generation of field data, analysis of data (usually
by quantitative means), validation of hypothesis, conclusions and recommendations. Of
course, the notion that research is a system does have a certain heuristic value but it is
also very limiting because it makes research seem a mechanistic and algorithmic, a step-bystep linear, teleological and finite activity. Most significantly, it projects a model of
research as both disembeddedan essentially ahistorical, apolitical and technical activity,
a transcendental, contextless set of procedures and methodsand disembodied in the
sense of being carried out by abstracted, asocial, genderless individuals without a history
or culture.
If instead we see research as a social practice we are better able to recognize that it is not
a universal process of applying a set of general methods or of following an algorithmic
procedure. Rather methods and procedures are themselves a function of the knowledgecommunitys practice, its culture, its networks of implicit beliefs and pre-suppositions,
whose rules, boundaries and exclusions, no matter how flexible, legitimate and sanction
certain kinds of activity and exclude others.
Consider for a moment one of the effects of a practices exclusions. Lets take as an
example the fact that social and educational research cannot be presented in the form of a
literary text. This may seem a wild example but it demonstrates an important point
through its very extremeness. If research is a social practice, a practice of producing
knowledge that is socially validated, then it is the socially located activity of research that
constructs a world to be researched. Research can be seen as knowledgeing, with the
implication that whats going on is not simply a matter of representing, reflecting or
reporting a world that already exists. Research therefore does not passively reflect, it
actively constructsbut it does so in a particular way.
Now, of course, literature (or fictioning) is also a social practice where worlds are
created. We have seen examples of this earlier in considering Ecos texts. In this sense,
research is just as fictional as literature even though both are equally real. Presenting
research in the form of a literary text is just as much creating a world, albeit a very
different one, as presenting research in terms of the linear process model. Im not arguing
that social research is the same as literature but I am saying that both are world-making
social practices. More specifically, both are textual practices, examples of world-making
through particular ways of reading and writing, interpreting and understanding, the
32 ROBIN USHER
world. Social research is therefore a textual practice which, however, cannot be presented
in the form of a literary text because that is outside the boundaries of what is constituted as
acceptable by the knowledge communities of social science.
The world-making, constructive quality of research can either be deniedas it is,
for example, in the positivist/empiricist traditionor it can be acknowledged. To
acknowledge it involves laying bare the activity of knowledging in our action is our
knowing (Lather, 1991; p. xv)and foregrounding what is involved in world-making.
It is in a sense to research the research, to bend the research back upon itself, to ask by
what practices, strategies and devices is world-making achieved?. By asking this question,
the research act is made self-referential or reflexive. This is an activity which is familiar
enough in literary texts (as we have seen with Ecos texts) but the positivist/empiricist
tradition that is dominant in the knowledge-producing communities of social research denies
it through its own textual practices. Yet since all research, including positivist/empiricist
research, is a textual practice and since reflexivity is a function of textuality then any
denial of reflexivity misses out an important dimension of the activity of research (Usher,
1993).
Traditionally, reflexivity has been seen as a problem to be dealt with through
scientific method. It is in this way that claims to presence, of direct and unmediated
knowledge of the world, could supposedly be established. Yet since the activity of
knowing is not simply a passive representation of the world and since how the world is
known influences what is known, then presence is itself a function of those activities. In
other words, presence can only be achieved by presencing practiceshence a direct
and unmediated knowledge of the world is impossible. The world always comes already
interpretedin other words, as a text. As such, it has an openness, an indeterminacy
which as we have seen earlier can only be closed, made determinate, measurable, lawful,
by violence.
To foreground the textuality of research and the absence of presence leads
immediately to the question why do research if you cannot say anything about what is out
there and all research is self-reflexive? (Steier, 1991; p. 10). But this is to assume that
because the knower is constitutively inseparable from the activity of knowing then
reflexivity must always be a problem. There is, however, an alternative which is to see
reflexivity as a resource, to acknowledge and work with it, but doing so requires
foregrounding research as a textual, presencing practice rather than a neutral, technicist
process. Another way of putting this is that reflexivity can be a resource when the place of
the active researcher is recognized why do research for which you must deny
responsibility for what you [i.e. the researcher] have found? (Steier, ibid). Yet we are
hesitant to see reflexivity as a resource since it seems to make research subjective, a
subtle form of writing the self; a different, non-literary way of writing ones
autobiography?
I want to argue that a postmodern approach to social research highlights the need for
researchers to be reflexive and to subject themselves, as knowledgers, to critical selfscrutiny. Being reflexive is not a lapse into subjectivity, it involves more than laying bare
ones psychological self or revealing ones biases. Hence research is more than just writing
the self in an individualistic sense. Here I want to suggest that an autobiographical stance
which on the face of it appears to enshrine research as writing the self actually helps us to
see reflexivity differently (in a socio-cultural rather than an individualistic sense) and thus
provides critical conceptual resources for interrogating the production (writing) and
consumption (reading) of research texts.
Con-text
If research is a textual practice, a textualizing of the world through the production and
consumption of authoritative knowledge-claims in the form of texts, then these always
have a con-text, in the sense of that which is with the text. What is with the text in this
sense is the situated autobiography of the researcher/reader.
Here, what is being highlighted is the socio-cultural subjective, the contextual self or to
put it another way, the embodied and embedded self. By asking reflexive questions about
con-text we can scrutinize the knowledging effects of the self that researches (writes)
and the self that reads research, a self with an autobiography marked by the significations
of gender, sexuality, ethnicity, class, etc. These significations are socio-cultural products
that effect, through writing, the very forms, outcomes and consequences of research and
the way in which research is read. They are not simply biases which can be eliminated by
first admitting them and then placing ourselves under methodological control. These are
biases, ineliminably part of us, which can be recognized but not willed away. They are
the marks of the trajectory of our desires and emotional investments in the research act.
Pre-text
If research is a textual practice, then language as a signifying system assumes a central
place. Here language needs to be seen not simply as a neutral vehicle for conveying
representations but as an activity of languaging, the means through which the
representational work of research is carried out. Consequently, research texts have a pretext in the sense of that which is before the text; language as the repository of meaning,
discourses as particular ways of organizing meanings, the textual strategies, literary
conventions and rhetorical devices of writing. As Shotter points out:
It is by the use of such rhetorical devicesas reference to special methods of
investigation, objective evidence, special methods of proof, independent
witnesses, etc.that those with competence in such procedures can construct
their factual statements and claim authority for them as revealing a special true
reality behind appearances (Shotter, 1993; p. 25)
This is an aspect of research that we pay little attention towe focus on methods and
outcomes and dont ask how meanings are created and received. As Game (1991; p. 28)
points out if research is understood as writing, critical attention is drawn to the process
of textual production which is research, as opposed to a final writing-up of research
results. This is due to the academic form that most research reports take and which
34 ROBIN USHER
actually functions to conceal their own textuality; the pre-text repressed by the form
itself.
However, it is by asking reflexive questions about pre-text that we can better
understand that the self as researcher writes within prescribed forms (and this is why, to
answer an earlier question, research cannot be presented as a literary text) that create the
meanings of the textualized world that is researched and of the texts that make knowledge
claims about that world.
Sub-text
I referred earlier to research as a practice of presencing, of the making of authoritative
claims to unmediated knowledgeability in the form of truthful representations of the
world. What this implies is that claims to presence are implicated with the operation of
power, the ability to claim and establish presence. Thus research texts have a sub-text in
the sense of that which is beneath the text; the operation of research paradigms and
traditions and the power/knowledge discourses through which they are expressed and
have their effects (Foucault, 1980). The research act is constituted by power relations
between researcher and researched whose outcome is the production and consumption of
powerful texts, texts which frequently become part of regulatory mechanisms in the
domain of governmentality.
This is something we often find it difficult to accept since given the traditional
separation of knowledge from power, we tend to think of research, particulary our
personal research, as innocent, power-less, perhaps even useful and emancipatory. A
reflexive questioning of the sub-text enables us to better interrogate the implication of
our practice within discourses of power and how it becomes part of oppressive and
dominant discourses through a reflexive, in the sense of a taken-for-granted, acceptance
of the neutrality of research, its pragmatic usefulness or its emancipatory potentialand
how as writers and readers of research we become part of such discourses despite our best
intentions.
Inter-text
Research in the scientific mode generally takes the form of decontextualized knowledge
claims which are made through and against other textually-embodied claims. To
paraphrase Kuhn, researchers do not match up their claims against an independently
existing objective reality but against other claims. Thus research texts have an inter-text
in the sense of that which is between texts. Intertextuality refers to the inhabiting of any
particular text by the structure of the tracethe interlacings and resonances with other
texts (Wood, 1990; p. 47) and which works both at the conscious and unconscious
levels. In effect, intertextuality points to the place of history in textual productionthe
way in which history is inserted into the text and particular texts into history.
At the conscious level, intertextuality works through citations that are actually
present. Unconsciously, intertextual traces are always present both between the
researchers own different texts and and between the researchers text and other texts.
As Barthes puts it:
a text is not a line of words releasing a single theological meaning [the authors
meaning] but a multidimensional space in which a variety of writings, none of them
original, blend and clash. The text is a tissue of quotations drawn from the
innumerable centres of cultures. (Barthes, 1977; p. 146)
These traces, the tissue of quotations are citations that although not physically present
are present as absences. They speak through the text in an endless referability and are
potent in their effects even though they are not spoken about in the text.
Intertextuality means first, that texts have a referability, they can be cited
authoritatively without the presence of their author. Secondly, it means that texts are
productive in the sense that they both transform prior texts (texts can be re-read
through new texts) and make possible new texts (a text can only be written if there is a
culture of writing and discourses which reveal a world to be researched). However, this
productivity is bounded, socially and culturally constrained and subject to relations of
power (Fairclough, 1992). Researchers are never fully the authors of their texts.
A Postmodern Approach to Research: an Example
I shall now try to show what it means to do research in the postmodern mode by
presenting an example drawn from Patti Lathers account of her research Staying dumb?:
Student resistance to liberatory curriculum (Lather, 1991). Here she explores what it
means to do research differently although on the face of it the research is conventionally
empiricist. She collects data from interviews, research reports and entries from her own
reflective diary. The research was a three-year enquiry into student resistance to a
liberatory curriculum in an introductory womens study course.
Her aims were to:
1 Create a space from which the voices of those not normally heard could be heard,
2 Move outside conventional research texts, and the unacknowledged textual devices
found in scientific research,
3 Ask questions about the way she as researcher constructed her research text and
organized meaning,
4 Challenge the myth of a found world in research and its effective communication
outside the intrusion of language and an embodied researcher,
5 Explore a complex and heterogeneous reality which does not fit neatly into preestablished categories,
6 Be concerned with the politics of research by recognizing that categorizing is an act
of power which always marginalizes and,
7 Put the researcher back into the picture. The researcher is a social subject in relation
with others. The specificity of the researcher, for example Patti Lathers interest in
emancipatory pedagogy, shapes the process and product of her enquiry.
36 ROBIN USHER
What is particulary interesting for me about this research is that Patti Lather analyzed and
presented her data not simply in a conventional form but in the form of four tales:
1 A realist talea tale based on the assumption of a found world knowable through
adequate method and theory. In other words, a conventional account.
2 A critical talea tale based on the assumption that there are powerful determining
structures underlying the world which are invisible to everyday understandings. The
researchers job is to bring these out into the open and and by so doing challenge
their dominance.
3 A deconstructive talea tale that attempts to undermine powerful tales, such as the
two above, that appear to, as it were, tell themselves. This tale discloses the
constructed nature of other tales (that the story they tell is not simply a reflection of
the world as it really is) and in this way foregrounds its own constructed nature.
4 A reflexive talea tale which brings the teller of the tale (the researcher) back into
the narrative. Here the idea is to show that the author/researcher does not exist in a
transcendental realm but is embodied, desiring, and herself invested in contradictory
privileges and struggles.
A Postmodern Approach to Research: Some Questions
To conclude, I want to set out some of the implications of a postmodern approach to
social research. I present this in the form of a number of questions rather than a list of
certainties:
Is it more useful to see research as a practice rather than a process? What
follows from seeing research this way?
If we do see research as a practice, then what it would mean to see it
specifically as a textual practice, a practice of writing?
As a practice of this kind, is it not then akin to a fictional text in the sense
that it too, like the latter, is in the business of constructing a world or worlds?
If this is the case, then doesnt research always raise questions of reflexivity?
Does the existence of reflexivity suggest that all research, whatever else it may
be, is autobiographical? But what is autobiography and what is its relation to
the self who researches?
Does being reflexive also mean taking account of the effects of the
situatedness of researchin other words, asking where is it coming from?
What about the part played by language, textual strategies and conventions in
achieving objectivity and truth?
Do we need to be careful about what is underneath, the operation of power
and its effects through particular discourses?
If research is textual, then are research texts arguing against other texts rather
than reflecting the world?
If our concern is that research contributes to domination and oppression should
research be aimed explicitly at emancipation? But do we then need to be careful
that our desire for this might merely lead to another form of oppression, of
speaking for others?
Post-text
In the text that has unfolded I have tried to tell a story about social research and one of my
ways of doing this has been to recount Ecos storya story which on the face of it is not
about research. Yet as we have seen, by reading Eco in a particular way we can gain an
understanding (a different understanding?) of research which might otherwise not have
been possible. This is one way of following a deconstructive strategya strategy with
which postmodernism is commonly associated.
The lesson here (and Patti Lathers research shows this very clearly) is that such a
strategy is not an alternative paradigm for research, let alone a new method for doing
research, What it is, if it is anything at all, is an injunction to be constantly vigilant, to
take nothing for granted in doing research. By being vigilant we are reminded to always
ask not onlywhat is my research finding out?but alsowhere is it coming from?;
what is it doing?; and with what is it implicated? In this way, we become aware that
research is not a transcendental activity or merely the application of an invariant technical
process. Most of all, we become aware that research is both a constructed and a
constructing activity.
Furthermore, in telling this, a postmodern story, about research we can better
understand research as story-tellingas constructed and constructing. An advantage
of this approach is the emphasis on research as illuminative, insightful and emancipatory.
But what is also revealed, as Eco shows only too clearly, is that story-telling can also be
powerful, oppressive and dangerous. There are always two sides to the text of social
research.
References
ACKER, J., BARRY, K. and ESSEVELD, J. (1991) Objectivity and truth: problems in doing feminist
research, in FONOW, M. and COOK, J. Beyond Methodology, Bloomington, Indiana
University Press.
BARTHES, R. (1977) Image-Music-Text, New York, Hill and Wang.
BAUDRILLARD, J. (1988) America, London, Verso.
CODE, L. (1993) Taking subjectivity into account, in ALCOFF, L. and POTTER, E. Feminist
Epistemologies, London, Routledge.
ECO, U. (1984) The Name of the Rose, London, Picador.
ECO, U. (1990) Foucaults Pendulum, London, Picador.
FAIRCLOUGH, N. (1992) Discourse and Social Change, Oxford, Polity Press.
FOUCAULT, M. (1980) Power/Knowledge, Brighton, Harvester Press.
GAME, A. (1991) Undoing the Social, Milton Keynes, Oxford University Press.
JAMESON, F. (1991) Postmodernism: The Cultural Logic of Late Capitalism, Durham, NC, Duke
University Press.
KUHN, T.S. (1970) The Structure of Scientific Revolutions, Chicago, University of Chicago Press.
38 ROBIN USHER
4
Challenging the Power of Rationality
Pat Usher
Each chapter in this reader is attempting to provide insights on the relationship between
producing knowledge or meanings about the world, theoretical foundations in which
disciplines are embedded and the actual practice of doing research. Regardless of which
discipline each of us is working within, all of us are socialized into a methodologyby which
I mean a theoretical framework that derives from a research tradition. In order to justify
our choice of methodology, I would argue that we need to be able to articulate its
characteristics, analyze the categories of analysis that it takes for granted, be alert to the ways
that these categories structure our opinions and provide us with criteria to make
judgments about social experience; in short we need to be aware of how our choice of
research methodology forms a context and rationale against which we evaluate other
research approaches.
In this chapter I am looking at how one kind of critical theoryfeminismis raising
questions about the relationship between knowledge creation and the search for objective
truth. Who is it that creates meanings about the world, what criteria are used to decide
what constitutes valid truth and does gender come into any of this? In the first section of
this chapter I will outline the kind of critical analysis which radical feminist theory is
making of the assumptions around which modernist thought has been structured, in
particular the privileging of rational discourse and a particular model of knowledge
exemplified by the scientific method of the natural sciences which is regarded as the
guarantor of truth. I will then move on to elaborate the constructive approaches which
radical feminism is advocating in seeking to expand the criteria for what is considered to
be true, rational or valid.
But before I move into the main body of this papers argument, I want to identify the
assumptions I am making about the purposes and nature of research in the social sciences.
Breaking with Traditions of Knowledge-making
It is now commonplace to allege that Western thought is engaged in a crisis of
legitimation and that the social sciences are questioning the epistemological foundations of
the Enlightenment legacy that underpins modernist thought as we know it. My arguments
are challenging the defining characteristics of modernism: namely that disciplinary
knowledge can be objective, impartial, innocent in intention and effect and neutrally
discovered; that there is only one, true method by which knowledge is acquired; that
knowledge can be discovered by a rational subject that is distanced from its object and
who can separate herself/himself from emotions, personal self interests and political
values in creating innocent knowledge. I am arguing that it is not the task of the social
sciences to discover absolutes in the way that the Enlightenment defined epistemology
(i.e. the relationship between knowledge construction and truth)but rather the task is
the interpretation of social meaning and not a search for scientific truth; the search is for
understandings of the social world and not for scientific laws about human beings.
In other words I am challenging the view that has been derived from Enlightenment
thought that the purpose of research in the social sciences is to separate universal truth
from historical and cultural influences, which by implication, warp human thought. I am
advocating an approach to research which recognizes that all knowledge is produced out
of a relationship between human thought and human existence: that in producing
knowledge one cannot make a distinction between objective and subjective knowledge,
that all knowledge is shaped by a mixture of historical and cultural influences.
Drawing on the work of Foucault and Gadamer I am arguing that modern science, far
from offering the one true model of knowledge, is rooted in certain historically specific
assumptions about the way the world is; assumptions which are not universally valid but
related to specific contexts. This approach to knowledge undermines the universalistic
claims of scientific, objective knowledge, reveals the historical specificity of Enlightenment
thought and illustrates a further important assumption which I make in this paperthat
values, politics and knowledge are intrinsically connected. If, as I am arguing,
understanding is always from a particular perspective, researchers must acknowledge that
their values and prejudices are implicated in what they create as knowledge. So as well as
acknowledging the influence of socio-historical conditions and the interpretive role of
social research I am also arguing for a reflexivity that seeks to understand that power is
not exterior to knowledge or to social relations but is embedded within their very
conditions of existence. My aim is to illustrate this in the sections that follow.
Feminist Theory and the Challenge to Orthodoxies
The last three decades have witnessed an enormous breadth in the development of
feminist theory. The construction of theory has always been allied to a politics and each
phase has defined the focus differently and so the emphases within theory have been
different. In the 1960s feminism was largely concerned with removing barriers to
womens participation in all aspects of public affairs and to arguing for a greater share for
women in the rights, duties, privileges and responsibilities of men. This egalitarian
politics was accompanied by forms of theory which sought to challenge sexist practice.
Sexism was defined as the unwarranted differential treatment of women: it was
identifiable in empirical terms and could, it was argued, be removed. If men and women
were treated equally in the political, social and personal sense, there was the rational
possibility that womens views and activities could be invested with the same degree of
significance as mens: equal but different was the sound-bite of the egalitarian paradigm.
In the 1970s, feminists were attempting to integrate their approaches into mainstream
radical theories such as marxist, liberal or psychoanalytic, only to find them particularly
PAT USHER 41
v) by grounding all claims to truth and authority in the application of reason, conflicts
between truth, knowledge and power can be overcome. By privileging reason, truth
can serve power without distortion; by utilizing knowledge in the service of power,
both freedom and progress can flourish. As a result knowledge can be neutral
because it is grounded in reason and its effects are beneficial. Finally,
vi) the scientific paradigm is the framework to guarantee truth and its methods are
neutral and its contents socially beneficial.
Radical feminists have entered into a dialogue with postmodernist discourses as they have
begun to deconstruct notions of reason, knowledge and the self to reveal the gendered
nature of concepts and analytic categories which had been seemingly neutral and
universal. Scholars such as Hekman (1990) and Flax (1990) have argued that knowledge is
structured through a number of fundamental dichotomies such as rational/irrational,
subject/object, nature/culture which constitute a hierarchical view of knowledge. The
way that knowledge is constructed privileges certain concepts over others and rather than
being a peripheral issue, the study of epistemology is central to the way the social sciences
examine the relationship between human thought and social existence.
Hekman and Flax argue that the rationalism which is the source of modernist
epistemology is a specifically masculine mode of thought. They argue that the positivists
claim that only rational, objective, abstract, universalistic thought can lead to the truth is
in fact a specifically masculine claim. Further, they argue that each of the dualisms on
which modernist thought rests is the product of the fundamental dualism between male
and female. In each of the dualisms mentioned above rational/irrational, subject/object,
culture/nature the male is associated with the first element, the female with the second.
In each case the male element is privileged over the female and maintains its position by
its capacity to define itself as a universal standard against which the subjective, the
emotional, the aesthetic, the natural, the feminine must be judged. In the very dualities of
modernist thought womens significance is defined as inferior to the rational, objective,
abstract qualities of the scientific method which not only guarantees Truth but positions
masculinity and man as capable of finding the Truth. Once this kind of analysis is
accepted all the other common sense stereotypes which we constantly attempt to counter
fall into perspective. As Jane Flax comments, woman is defined as a deficient man in
discourses from Aristotle to Freud.
There are some interesting parallels between feminist analyses of modernist dualities
and the irresolvable disputes between positivist and interpretive approaches in the social
sciences. Many people working in an interpretive paradigm do not challenge the
subjective/objective dichotomy on which positivism rests because they believe that the
strength of social science research is that it can be objective in a subjective kind of way.
Despite all their efforts however, the objectivity of the natural sciences remains the gold
standard against which all other variantsqualitative/interpretive approaches are
judged. By not challenging the dichotomy, the privileged concept maintains its power of
definition. For feminists the same dilemma apply. Many feminists try to exalt the virtues
of female nature and the related concepts of community, care, relatedness, as opposed
to the male values of control, mastery, rationality and abstraction. But as the dispute in
PAT USHER 43
the social sciences reveals, these arguments are not convincing because such feminists
cannot succeed in privileging the female over the male because they have not challenged
the very dichotomy through which the female is defined as inferior in the first place.
Equal but different simply cannot succeed in a world where knowledge is constructed
through hierarchical meanings. Not that the reverse argument of valorizing femininity over
masculinity works either. What does emerge is the insight that if particular dualisms
cannot be reversed, perhaps the aim should be to dissolve them.
At this point in the paper it might be helpful to review how radical feminism has
developed as a perspective and where it stands in relation to other discourses.
As I indicated above radical feminism is critical of liberal feminist discourse because the
latter is tied to the dichotomies and grounding in absolutes which I analyzed above as the
characteristics of Enlightenment epistemology. Anything short of a rejection of the
rationality and dualisms of Enlightenment thought and the research methodologies which
derive from it, will not prove a successful strategy for women.
In developing its analysis that a commitment to modernist categories is in effect a
commitment to the production of gender specific theory that privileges masculinity rather
than gender neutral theory, radical feminists have been greatly helped by the work of
Thomas Kuhn on paradigm shifts. His work established the notion that science should be
examined within a specific cultural and historical context and that its statements were not
the universals they often were presented to be. Meaning and interpretation were part of
the scientists world. Although Kuhn may not have intended it, his work challenged the idea
of a fixed, absolute reality against which we test our notions about the natural world.
Given his approach to science, what appears according to scientists as objective reality
does in fact change with changing paradigms. Theory can be argued as being grounded in
distinct contextssocial, historical and politicaland there are not fixed, universal
criteria for determining one as more correct than the other. This hermeneutic or
interpretive dimension to science is very important for all social scientists but particularly
so for feminists. If the model of knowledge embodied in the scientific method of the
natural sciences is not the only paradigm of knowledge then all the dualities which have
disprivileged women can be exposed to critique and displacement.
The way is opened up for radical feminists to argue that a commitment to reason,
perspectiveless truth, objective and neutral forms of knowledge, separation of the subject
from the object of knowledge are all commitments to the production of male theory in
which reason surreptitiously defines itself by excluding categories associated with
femininitysubjectivity, the emotions, desire and specificity. Through such definitions,
the masculinity associated with theory control and its production has not been recognized
because no other theory or knowledge has emerged with which it could be compared.
Men have been able to assume the role of neutral thinkers and producers of ideas because
the categories of analysis they employ have repressed all traces of their sexual specificity.
By universalizing their knowledge they successfully tag feminism with the label of specific,
personal, incomplete, partial knowledge which can be easily dismissed and pushed to the
margins.
PAT USHER 45
from that of natural scientific, but he shows how a particular discourse creates a kind of
subjectwho is defined as mad, criminal, deviant, unreasoned.
His second great insight is that knowledge and power are inextricably linked in
discourse. His basic claim is that power and knowledge are fused in the human sciences. His
theory of power differs radically from the humanist/Marxist accounts in three basic ways:
power is exercised rather than possessed,
power is not primarily repressive but productive, and
power is analyzed as coming from the bottom up.
He is distancing himself from a preoccupation with power as a possession in order to
account for how subjects are constituted by power relations. He rejects a repressive model
of power in order to show how individuals actually obey repressive and coercive forms of
powerin other words how individuals police themselves.
What he is arguing is that in the human sciences disciplinary power has been exercised
over the bodies and minds of individualsdocile bodies and obedient souls. Disciplinary
power has spread through the production of certain forms of knowledge such as positivist
and interpretive human sciences and through the emergence of disciplinary techniques
such as surveillance, examination and discipline which facilitate the process of finding out
about individuals. Thus ways of creating knowledge in the social sciences are equated with
ways of exercising power over individuals. The different disciplines of medicine,
psychiatry, criminology, psychology are therefore in a powerful position to create the
divisions of healthy/ill, sane/mad, legal/delinquent, normal/deviant and are implicated
in effecting the means of normalization and social control.
He convincingly argues that the relations of power and knowledge are internal to
knowledges and that each discipline has a political investment in enhancing its view of the
world. His bottom-up analysis shows how all of us at the micro-level of society become
part of dominant networks of power relations. By being engaged in disciplinary enquiries
the power of individual researchers is enhanced within the rules of its discourse in relation
to the objects of their study: their power to define people and the nature of their
problems is exercised through their actions as researchers.
All of these upsetting and disturbing categories of thought are immensely helpful for
radical feminists. Firstly they allow a challenge to be made to philosophys orientation
around oneness, unity or identityone truth, one method, one logic. There is opened up
scope for perspective and a diversity in the models on which philosophy can develop. This
insistence on perspectives is advocated as a relational, not a relativist theory of knowledge
(Grosz, 1987).
Instead of being committed to truth and objectivity as privileged by both positivist and
interpretive perspectives, feminism can argue for the insight that all discourses create
subjects, objects, regimes of power and truth. Instead of being for or against the regimes
of any one research tradition, one asks what conditions facilitated the development of a
particular discourse and what criteria and assumptions does each discourse use to present
itself as speaking the truth and disguising its own desire to dominate. Knowledges must be
reflexive in coming to understand their own self development as knowledges. Feminism
must therefore accept its own status as context-specific, the product of socio-economic
and historical movements. It has no more claim to speak the truth than any other discourse
but must own up to its own points of view, specific aims, desires and political position
within power relations. No discourse can openly disassociate itself from a political
position. Thus the political is always personal. Feminist theory is neither subjective, nor
objective, neither absolutist or relativist: it strives to occupy the middle ground excluded
by these dichotomies: it argues that all theory is relational and connected to other practices
(Grosz, 1987). Radical feminism rejects the add women and stir approach of many
interpretive approaches to research. Gender divisions are fundamental categories around
which all social structures are organized and meanings created. This should be the
understanding which underpins the development of all human thought and men as well as
women should internalize the dominant conceptions we all have about masculinity and
femininity. The context for thinking about gender is that we see it as a social relation
rather than a set of opposite and different characteristics. In so dong we can illuminate
how our thinking becomes implicated in power/knowledge relationships both between the
sexes and within each categoryhow men treat other men, women treat one another.
The sub-text for such an approach is that we consciously admit the potential in our
enquiries for both expressing and uncovering relations of dominance. This may often
imply the writing of histories or giving expression to voices and views that have been
suppressed, unarticulated or denied by traditional discourse. Models of investigation
which assume homogeneity as their starting point fail to acknowledge the variety of
human beings and are oppressive in their phallocentric assumptionsthat women can be
represented as having the same characteristics and interests as men. For these kinds of
reasons, radical feminism argues that one cannot accept the separation of the subject and
object of knowledge. In accepting objectivity one must accept a knowing subject free of
personal, social, political and moral interesta positivist subject unimplicated in a social
context or uninfluenced by prior ideas and knowledge. Feminism as I represent it must
reconceptualize the inter-relationship between subjects and objects so that reason and
knowledge include a recognition of history, context and specificity.
Instead of organizing thought through dichotomies which separate and exclude
subject/object, reason/emotion, culture/nature, masculinity/femininitya feminist
position would argue for these terms to be regarded as continuities and for discourses to
be created that avoid dualities, essences and the temptation to privilege the feminine over
the masculine. The dualisms of masculine thought need to be displaced by a pluralism and
fluidity that transforms categories which have continually positioned and named women as
inferior. The modernist move of trying to incorporate women into the masculine
definition of rationality should be avoided in favour of deconstructing the dualism on
which it rests. Radical feminists argue that we dont need new truth but a plurality of
perspectives. Joan W.Scotts discussion of equality-versus-difference illustrates the point
well. Instead of accepting categories in fixed opposition to one another such as unity/
diversity, identity/ difference, presence/absence, universality/specificity, we argue for
their interdependence so that we reverse and then displace the binary opposition. In this
context we can refuse to place equality and difference in opposition to one another and
argue for a respect for differences as the condition of individual and collective identities,
PAT USHER 47
differences as the constant challenge to the fixing of those identities, differences as the
very meaning of equality itself (1988). Challenging the way knowledge/power is
constructed from the ground of difference seems to me a crucial insight not only for all
women but for all social scientists as well if they are to arrive at valid understandings about
the relationships between human thought and the nature of existence. Only by following
this approach can women or any other subordinated group avoid being defined as inferior
against the universalizing categories of rationality, objectivity and masculinity.
What are the Implications for Feminist Method?
In this chapter so far I have been looking at the relationship between epistemology and
methodology. I have defined epistemology as the study of how knowledge is constructed
about the world, who constructs it and what criteria do they use to create meaning and
methodology in the sense used by Harding (1987) as a theoretical and conceptual
framework with which research proceeds. In this final section I want to consider whether
the themes I have identified amount to such a thing as a feminist methodby which I
mean distinctive techniques of evidence gathering or pursuing inquiries.
I have argued that feminism is a perspective which asks all methodologies to recognize
that they too are perspectives. Arising out of this I want to argue as Reinharz (1992) does
that feminist researchers do not consider feminism to be a method; rather they see it as a
perspective that can be infiltrated into all disciplines as a means of developing innovative
methods. Feminist social research uses the methods supplied by the disciplineboth
quantitative and qualitativeand adds its own perspective in order to create diverse and
innovative approaches to analyzing human activity. At an earlier stage in the development
of feminist theory it was argued that qualitative approaches were the hallmark of feminism
because of their respect for the inclusion of womens experience which interpretive
research methodologies advocated. However for reasons I analyzed earlier, feminist
scholarship has identified the contradictions of methodologies which do not challenge
dichotomies which label womens experience as of lesser significance. Theories and
methodologies which attempt to valorize the essential nature of women and their
experience were seen to pose no challenge at all to the phallocentric nature of both
positivist and interpretive methodologies and so feminist theory now uses the full range of
methods and implicates itself in the greatest range of subject matters, challenging the false
homogeneity of much that passes for understanding of the social world.
The following themes are those characteristic of feminist research which should guide
the development of new approaches.
a) Acknowledge the pervasive influence of gender as a category of analysis and organization. Each
of us understands our own identity through the meanings we learn to attach to gender
difference: societys institutions and structures are organized around divisions that have a
gendered nature. Irrespective of the area of research in which we work, feminism would
argue that as a strategy in designing research one should be asking how meanings about
gender are implicated in the questions and criteria that guide ones thinking. These are
not the peripheral questions which can be added on after a long list of more important
questions have been asked; not to address questions of gender difference is to adhere to
patriarchal theoretical norms which collapse all issues of difference into unitary,
homogenized categories. Such approaches cannot lead to intelligible meaning about the
diversity of experience in the social world; patriarchal discourses disguise their own
specificity by portraying their frameworks and questions as universally significant.
b) Deconstruct the traditional commitments to truth, objectivity and neutrality: reverse and
then dissolve the binary oppositions which structure thought and subordinate all the
categoriesthe subjective, the privately emotional, community, practice-basedthat are
associated with the feminine and females. The aim is to expand and multiply the criteria
for what is considered to be rational and true and to be critical of norms which are
discriminatory in their assumptions.
c) Adopt an approach to knowledge creation that recognizes all theories are perspectival: that
each methodology is the product of a specific combination of historical, socio-political
influenceswhether it is positivism or post-modernismall of them are making claims
about the truth but none should be regarded as telling the truth. Feminism itself must be
treated in this light: it is not telling a better story about social relations but in making its
own claims, it expects each claim to be equally self-reflexive. This does not mean that
each perspective is equally valid but each of them has a point of view and should be
interrogated from a stance that accepts that no perspective is producing disinterested
knowledge and each of them represents particular positions within power relations.
d) Utilize a multiplicity of research methods. Feminist research reaches into all disciplines
and uses all the methods; it does not concern itself solely with the study of gender but is
concerned that research about any area of social practice is sensitive to the differential
experience of female and male. Reinharz (p. 245) agrees with a number of writers who
suggest that feminism is accumulating stages in scholarship. The current emphasis on
multiplicity of methods is because in her view feminists value inclusiveness more than
orthodoxy. The outpouring of scholarly feminist literature has contributed to what Jessie
Bernard calls the Female Renaissance or Feminist Enlightenment. A multiplicity of
methods allows women to study the broadest range of subject matters and reach a broad
set of goals. Their approaches may include interview and oral history methods, case
studies, cross-cultural research, ethnography, surveys or experiments. Giving voice to
experience, understanding variations within populations or measuring behaviours and
attitudes are only a few of the methods utilized by feminists.
e) Be open to inter-disciplinary work. The organization of knowledge into discrete
disciplines is a thorny issue that feminist researchers cannot avoid. Each of us is shaped by
disciplinary approaches but given feminisms critique of the dualities of modernist thought
which disprivilege the feminine, it follows that feminists must be alert to the ways that
disciplines function as discursive networks to ignore or deny female experience. As I said
earlier, by presenting theory as universally applicable those working within disciplines can
successfully disguise the specificity of their theory. Disciplines function as phallocentric
discursive systems and collapse the differential experience of the sexes into universal
models with homogenized categories. Anna Yeatmans discussion of sociologys resistance
to the feminist challenge illustrates the extent to which the social sciences including
sociology, economics, psychology and political science are all structured by modernist
dualisms.
PAT USHER 49
that the researcher establish an empathic understanding with the people s/he studies in a
non-exploitative manner. Relations of respect, shared information, joint involvement in
the commentary and evaluation of their own lives, openness, are all the goals of a feminist
researcher. Clearly these themes are not without ambiguity and controversy but if the task
is to represent the complexity of social reality, researchers need to form relationships that
allow for those who are the researched to express what is significant to them in their
everyday lives. It is this which provides for the validity of the research not pseudoobjectivity.
Summary
In this chapter I have tried to draw distinctions between epistemology, methodology and
methods. Epistemology is concerned with philosophical approaches to what counts as
knowledge and truth; methodology refers to the conceptual framework within which
research as a practice is located and methods are the techniques employed to gather
research materials.
I have argued that feminism is challenging the epistemological foundations of western
thought by questioning the dualities and hierarchies through which modernist thought has
constituted femininity and women as inferior.
I have analyzed the characteristics of feminism as a perspective drawing on the
postmodernist discourse theory of knowledge. In arguing that truth is constituted through
discourses, I am asking for a critical reflexivity towards the characteristics of all research
methodologies so that one can understand that all knowledge is the product of historical
and contextual circumstances.
In arguing that discourses create knowledge, I have also drawn attention to the power
hierarchies that are created within knowledge and in particular to the way that subjects
and objects are constituted through categories privileged within the discourse. It is in this
respect, that radical feminism is such a fundamental challenge to traditional intellectual
values because it rejects the Cartesian subject that creates knowledge through rational
abstraction. Feminism is arguing that it is these Englightenment assumptions that have
prevented the development of new modes of subjectivity and discourses which can resist
the power of phallocentric research traditions.
Feminism has encouraged a research practice that reveals difference because that speaks
of the the diversity of men and women and is a basis from which male domination of
theory can be effectively challenged. It advocates the use of a multiplicity of methods in
all the disciplines so that the boundaries of what constitutes research can be opened up and
new ways of knowing, new forms of analysis and new texts can be created.
References
BERNARD, S. (1987) The Female World from a Global Perspective, Bloomington, Indiana University
Press.
PAT USHER 51
FLAX, J. (1990) Thinking Fragments: Psychoanalysis, Feminism, and Postmodernism in the Contemporary West,
Berkeley, University of California Press.
FLAX, J. (1993) Disputed Subjects: Essays on Psychoanalysis, Politics and Philosophy, New York,
Routledge.
FOUCAULT, M. (1972) The discourse on language, in The Archaeology of Knowledge, New York,
Harper and Row, (pp. 21538).
GADAMER, H.G. (1975) Truth and Method, New York, Continuum.
GROSZ, E.A. (1987) Feminist Theory and the Challenge to Knowledges, Womens Studies
International Forum, 101, 5, pp. 47580.
GROSZ, E.A. (1990) Philosophy, in GUNEW, S. (Ed) Feminist Knowledge: Critique and Construct,
London, Routledge, (pp. 14774).
HARDING, S. (Ed) (1987) Introduction: Is There a Feminist Method?, Feminism and Methodology,
pp. 114.
HEKMAN, S.J. (1990) Gender and Knowledge: Elements of a Postmodern Feminism, Oxford, Polity Press.
KUHN, T.S. (1962) The Structure of Scientific Revolutions, Chicago, University of Chicago Press.
REINHARZ, S. (1992) Feminist Methods in Social Research, New York, Oxford University Press.
SCOTT, J.W. (1988) Deconstructing EqualityversusDifference: or the uses of
poststructuralist theory for feminism, Feminist Studies, 14, 1, pp. 3350.
YEATMAN, A. (1990) A Feminist Theory of Social Differentiation, NICHOLSON, L.J. (Ed)
Feminism, Postmodernism, London, Routledge, (pp. 28199).
5
Being Economical with Politics
Alan Hamlin
Introduction
Like many of those who are drawn to study economics, my major initial motivation was
an interest in economic policynot so much the macro-policy issues of inflation or the
exchange rate that seem to many to be the essence of economics, but the micro-policy
issues of taxation and welfare benefits, of regional or training policy, of policy towards the
nationalized industries (as they then were) and so on. In short, the myriad ways in which
government economic policy decisions impact on the everyday lives of individuals. As an
undergraduate, I took options in courses concerned with taxation and public finance,
industrial economics, the economics of the nationalized industries and so on. On
graduating, I decided to pursue these interests further by taking a post graduate degree at
the United Kingdom department which (at that time) specialized most strongly in the
analysis of the public sector and its policiesthe University of York. My first year at York
was spent in the relatively standard pursuit of taught courses at the Masters level, but once
I was admitted to study for a DPhil, I began to think seriously of research for the first
time.
My initial research topic was the relationship between local and central government
policies, and the possibility of policy improvements in specific areas from closer
integration between government agencies. But I wrote no thesis on this topic. Like many
research students before and since, I found that my original topic was simply a point of
departure.
The analysis of government policy that I had studied up to that date was almost entirely
normative in character. That is, it attempted to answer questions of the form What
should government do in situation X, on the assumption that the overall aim of
government policy is to maximize some particular notion of social welfare. Thus the
concern was with deriving ideal or optimal policies from a specific moral criterion. This
approach begs two obvious questions: (1) how do we analyze how governments actually
decide what to do; and (2) what particular moral criterion is relevant to the evaluation of
government policy. These questions, which we may label the political question and the
ethical question, came to dominate my research interests to the extent that I never
returned to the particular questions of the relationships between local and central
ALAN HAMLIN 53
As I have already suggested, the basic theme of public choice theory as a response to the
political question of how we analyze how governments actually decide what to do
involves the application of an essentially economic mode of analysis to questions which are
normally considered to be in the domain of politics. Even making this statement raises a
number of interesting questions: do we (and should we) identify individual social sciences
with particular domains of study (economics with the economy, politics with the polity
etc.) or with particular styles of theorizing; what is the difference, if any, between crossdisciplinary and inter-disciplinary research; what exactly is an essentially economic mode
of analysis; and what virtue might there be in applying it in the domain of politics? In this
brief essay, I hope to tackle at least some of these questions, and one or two related
topics.
In addressing these questions I have two major objectives in mindfirst the explication
of some aspects of the economic mode of analysis, particularly those associated with
individualism and rationality; and second a suggestion to the effect that this mode of
analysis is of value across the domain of social science. While my discussion is presented in
terms of the application of economic methods to politics, I hope to suggest some further
generalization. Questions of individualism and rationality are of relevance in many areas
of social research, whether they are embedded in the economic method or not, and I hope
that the discussion of these topics may be of interest to those who are neither economists
nor political scientists.
In approaching my objectives I shall be operating at a level of discussion which
fluctuates from methodology to methodat times the discussion will link back to the
grand themes of epistemology and the philosophy of social science, while at other times I
will be concerned with the more detailed and practical methods involved in economic
analysis. The reader should be aware of these different levels of debate, and I will try to make
the appropriate level explicit where confusion might arise.
I should also be explicit about what I am not attempting in this essay. In particular I
will not discuss the general question of approaches to empirical work. I will be concerned
to characterize aspects of the economic mode of analysis seen as a style of theorizing which
might contribute to our understanding of the topics under study. Some might argue that
the economic approach to social science involves not only a distinctive style of theorizing
but also a particular approach to empirical work and the vexed questions of theory
appraisal. I doubt this but, in any case, I will not be concerned with questions of empirical
method or methodology here.
The final point that I would stress in laying out my intentions is that I will not be
arguing that the economic mode of theorizing is in any way uniquely privileged relative to
other modes of thought. My aim is much more modest: all I seek to argue is that the
economic mode of theorizing is of value outside (as well as inside) the traditional domain
of economics, providing a coherent and defensible project in the broad area of social
theory. In short, that there is at least some merit in being economical with politics.
The remainder of this essay is presented in four further sections. The next section
provides a thumbnail sketch of public choice theory and so provides an overview of the
economistic approach to politics. Sections 3 and 4 then look in slightly more detail at two
significant aspects of the economistic method which are often criticized. In section 3 I
ALAN HAMLIN 55
focus on the individualistic nature of the method, while in section 4 I focus on questions
of rationality. Section 5 contains some final comments.
Public Choice Theory
To provide a caricature of public choice theoryor constitutional political economy as it
is now sometimes calledone can simply reverse the view attributed to the traditional
economist above. Rather than assuming that policy making is free of all institutional
constraint and depends only on the technical ability to derive optimal policy conclusions
from given (or assumed) moral criteria, the public choice theorist views policy as the
emergent outcome of a complex institutional procedure. Once the institutional structure
is established, individuals act rationally to achieve their own ends (whatever they may be)
within that institutional structure, and policy is whatever emerges from that process.
Policy can be evaluated relative to any moral criteria you like to specify, but policies
derive from individual and rational actions within institutional contexts.
A common analogy is with a simple parlour game. Once the rules of the game are set,
the outcome of the game (the result, score or whatever) is determined by those rules,
together with the actions of the individual players. But the actions of the players are
themselves influenced by their understanding of the rules of the game. Change the rules
and you will change both the behaviour of players and the outcome of the game. Consider
the game of chess, for example. The rules of chess define the strategies available to
players, and the players then face the task of selecting the best strategy given those rules
and whatever they can learn about their opponents strategy, where best is normally
understood as meaning most likely to win. The game is complex and strategic and
although we do not have any analytic solution to the question of the choice of the best
strategy (as we do with some simpler games such as tic-tac-toe), the game may be
understood and analyzed in terms of players attempting to find better strategies given
their understanding of the rules. Change the rules (by introducing time limits, for
example) and you will change the players behaviour and the pattern of the game.
This allows us to identify the sense in which the approach taken by public choice theory
(and economics more generally) is individualistic in nature. Complex social processes are
analyzed in terms of the actions of individuals (just as chess games are analyzed in terms of
the moves made by individual players). This is not to say that institutions are irrelevant,
indeed the whole point is to view the impact of institutions, but simply to say that
institutions are important precisely because of the influence they have on individuals (just
as the institutional rules of chess are important in influencing the strategies of players). This
basic idea of individualistic analysis will be taken up in more detail in section 3 below.
On this view of politics offering policy advice of the traditional type is largely
irrelevant. Of course, some economic advice can be expected as political players attempt
to ensure that they are reasonably well informedbut this suggests that political players
will employ specialist advisers in order to work out their best political strategies (rather as
chess players employ specialist advisers to help them construct their strategies), so that
the role of advice is very different from that envisaged when thinking about a disinterested
policy advisor to a benign government on the more traditional model. In short, each
political actor will find an economic adviser who offers appropriate advice.
Public choice theory has as its basic structural theme the idea of comparative institutional
analysis. That is the idea that the appropriate focus of attention is the institutional
structure or rules of the game. Thus, for example, the public choice theorist would argue
that the questions of institutional and constitutional design and reform are the key
questions in the political sphere, rather than the questions of policy design and reform. This
means that the primary emphasis in public choice theory is on the study of the operating
characteristics of alternative institutional structures. What are the properties of different
voting rules or committee procedures? How do pressure groups interact with members of
the legislature? What is the role of a system of parliamentary committees, or of a second
chamber of parliament? How do political parties with this or that structure contribute to
political outcomes? These are the basic questions addressed by public choice theory.
Once answers to questions of this general form can be sketched, the next step is to
evaluate alternative institutional structures. That is to compare alternative arrangements
not with some ideal but with each other. Clearly this step brings the ethical question back
in to play, since the evaluative criteria must be made explicit. I shall say relatively little
here about the ethical question beyond a brief discussion in section 3 below of the
distinction between ethical individualism and methodological individualism.
In following this general strategy of analyzing and evaluating alternative institutional
structures, the researcher must clearly say something about the way in which individuals
might be expected to operate within the relevant institutions. Without some maintained
hypothesis concerning the behaviour of individuals, it would not be possible to analyze the
properties of alternative institutional structures (just as we would not be able to analyze
chess without assuming some particular motivation on the part of the playersnormally
that each player is playing to win). In this context the typical behavioural hypothesis
maintained by public choice theorists is again drawn from the more general economic
method: The hypothesis of economic rationality. In fact, this hypothesis can be broken
into parts including the idea that individuals act from the same basic motivations
regardless of the institutional setting (I shall term this the hypothesis of singlemindedness), and the idea of a substantively egoistic motivationthe idea that individuals
act to promote their own interests. I shall discuss each of these in section 4 below.
To recap, public choice theory involves the application of economic methods to the
analysis of social and political institutions and this involves breaking with some traditions
in economics in order to focus on comparative institutional analysis, but key aspects of the
economic method are maintained. Some of the criticisms of the application of economic
ideas in the political arena focus on the alleged limitations of the economic method,
particularly in relation to individualism and rationality, so it is appropriate to view these
commitments in slightly more detail.
Individualism
The economic method is often criticized for its individualism, the argument being that the
reduction from society to individual obscures or eliminates many issues of interest and
ALAN HAMLIN 57
blinds the analyst to social forces and constraints which lie outside of the individual. A
similar criticism is often aimed at psychology by sociologists. The point here is that
economics and psychology share a commitment to the micro level of explanation, while
politics and sociology operate predominantly at the macro level.
I believe this criticism to be ill founded, partly because it conflates several different
senses of individualism and partly because it ascribes to the economic method a form of
individualism that is much more extreme than is required. In this section I intend to
simply distinguish between strands of individualism in order to suggest a defence of the
type of individualism that I believe appropriate to the economic method. This section
will, then, operate largely at the level of methodological debate.
Individualism must be distinguished into at least two major strands: methodological and
ethical. My major concern here is with methodological individualism, although I shall say
a little about ethical individualism at the end of this section. But even methodological
individualism comes in a variety of formsI shall distinguish three. At its most severe
methodological individualism holds that no non-individual social entities are real. A less
virulent strain maintains that, even if social entities are real, they are fully reducible to the
individualistic, in the sense that any sentence containing reference to a social entity can be
rewritten in purely individualistic termsso that social entities are simply a sort of
convenient shorthand notation. The mildest form of methodological individualism says
that, even if social entities are real and irreducible, still they do not compromise
individual autonomy: they do not undermine or pre-empt individual agency.
The economic method may appear to be individualistic because it depicts government
(or any other social entity: a firm, a trade union, the European Community etc.) as
secondary to the individuals who act in its name. Recall that my sketch of public choice
theory involved viewing government as the actions of individuals within the institutional
and procedural structures of government, rather than as some supra-individualistic agency.
Nevertheless, I would argue that this does not involve any commitment to
methodological individualism beyond the mildest strain, and that this mildest strain is readily
defensible.
The economic method certainly does not entail the strongest version of methodological
individualism which would require government and any other social entity to be unreal
which might be one reading of Margaret Thatchers (in)famous remark that there is no such
thing as society. The only sense to the idea that government is unreal can be captured by
an analogy. Peter Geach once introduced an entity which he named a surman. Surman
are men, and no doubt women, who share the same surname; two people named Jones
may be different and unrelated, but they are (parts of) the same surman. It is clear that
surmen are unreal, and it is also clear why. The concept of the surman plays no part in
explicating the experiential or explanatory order of the world.
Does the economic method put the government and other such social entities in the
same category as Geachs surmen? Obviously not, the approach allows us to acknowledge
that our experience of the social world and our explanations of at least some of its
patterns make considerable use of the concept of the government (and other social
entities). We can recognize that government is no artifice created for purely theoretical
purposes, but a reality that is salient in the world; indeed, we must recognize this since
the object of our study is the impact of alternative institutional structures on the world.
The reality of social entities is not denied by the economic approach (indeed, it is affirmed),
so there can be no commitment to the strongest form of methodological individualism.
But what of the reducibility of social entities and the second sort of methodological
individualism? Social entities, on the economic approach, are a certain sort of system of rules
and rightsinstitutional and procedural rules and rights granted to office holders defined
by those rules. In one sense that makes the entity reducibleit means that we can define
the property of being a government in other terms. But that sense of reducibility is not
relevant to methodological individualism, since the other terms, those involving rules and
rights, are not themselves individualistic. The relevant question of reducibility turns on
whether the property of being a social entity can be redefined strictly in individualistic
terms, that is, terms which make no reference to society or any entities other than
individuals. But there is nothing in the economic approach which requires this type of
reducibility. What is true of the economic approach is that social entities are supervenient
on individualsin the sense that if two worlds were identical in respect of all
individualistic properties, then they would also be identical in respect of their social
propertiesbut this is a much weaker claim than the claim of reducibility.
The final question is whether the economic approach requires methodological
individualism of the third and weakest type. Does it rule out the view that social entities
such as government have an intrinsic power that undermines or pre-empts individual
autonomy? Even here the answer is less than a full yes. The approach certainly rules out
the wholesale undermining of individual autonomy, for it assumes that individuals act in a
proper intentional manner when they act within social entities. But this is scarcely a
burdensome conclusion. It simply rules out the view that individuals are pawns in the
hands of non-personal entities. But there is nothing in the approach that rules out more
limited pre-emption of individual agency. Indeed, it is precisely intended that individual
agents should be seen as being influenced in their action by the institutional structures that
they recognize.
In this way, the methodological individualism of the economic method may be seen to
be both minimal and defensible. The commitment is to a form of analysis which takes
institutions, rules, rights and other such social entities as important, but as important
precisely because of their impact on individual behaviour and the outcomes generated by
that behaviour. Methodological individualism of this mild form provides a basis for the
study of the impact of social entities, and comparisons across differing institutional
structures. I would suggest that many of the critics of the economic methodparticularly
when it is applied in areas of political and social institutionsmake the mistake of taking
the legitimate criticisms of more virulent doctrines of methodological individualism as
being effective against the milder methodological individualism which characterizes the
economic method.
A further potential confusion arises between methodological and ethical individualism.
As I have indicated, the form of methodological individualism required by the economic
approach is merely that social entities are seen as supervenient on individuals and that
individual agency is not entirely eroded. Ethical individualism, by contrast, is a view
concerning the basis for ethical evaluation. Loosely speaking, ethical individualism simply
ALAN HAMLIN 59
ALAN HAMLIN 61
institutional remedies are unnecessary, while the second might be dismissed on the
grounds that there is unlikely to be any unique specification of what constitutes
immorality even if we can agree on a unique specification of morality. The assumption that
each individual is a prudent but essentially amoral maximizer of his or her own utility is
not descriptively accurate, but neither is self-interest the most pessimistic assumption.
The choice between descriptive accuracy and the moderate pessimism and moral
neutrality of assuming individuals to be amoral must be made on methodological grounds.
Some argue that, even if we had available to us a descriptively accurate model of the
moral character of individuals, we should use the more pessimistic, amoral, model when
analyzing alternative institutional frameworks with a view to institutional or constitutional
reform. The argument is essentially that institutions form a sort of insurance, that they
should be robust against variations in the character of individuals and should perform
reasonably even when individuals do not. This is the basic insight contained in the
frequently quoted remark of David Humes that: in constraining any system of
government, and fixing the several checks and controls of the constitution, every man
ought to be supposed a knave, and to have no other end, in all his actions, than private
interest.3
A second point concerns the moral neutrality of the assumption of self-interest. This
relates to the basic structure of invisible hand arguments in which desired outcomes are
generated by the institutional structurethat is the pattern and form of relationships
between agentsrather than by direct appeal to the motivations of the agents themselves.
The choice of the assumption of amoral agents then focuses attention on the search for
invisible hand mechanismsmechanisms which economize on virtue in the sense that
they do not require virtuous agents.
This is important both because of the special power of invisible hand arguments and
because of the limitations of such arguments. For example, one type of social institution
which will not appear to be effective under the assumption of amoral agents, but which
may work in practice, is the type of institution which seeks to screen in the more moral
members of a society, or socialize certain individuals to particular moral standards, in
order to place them in certain significant roles. The public service tradition so evident in
the British civil service in the late nineteenth and early twentieth century might be an
example of an attempt to socialize individuals into a particular outlook, while the method
of selecting and rewarding judges in Britain (usually involving a significant reduction in
salary) might provide an example of a screening process. Such institutions are not invisible
hand mechanisms since they depend on the existence of more moral members of
society. The aim with this type of institution is to amplify the social effects of existing
moral character, and if this is the type of institution of interest, it would be more
appropriate to base analysis on a plausibly realistic view of moral character.
In short, there is a role for both the amoral assumption and for greater descriptive
realism. Each approach focuses on a different type of institutional effectthe incentive
based invisible hand case and the selection based amplification case. Excessive reliance
on either approach may divert attention from the variety of ways in which institutions may
operate to influence outcomes for better or worse.
ALAN HAMLIN 63
Public choice theory has typically taken the line of modelling individuals as singleminded and amoral. I believe that an essentially methodological defense of this line is
possible, but I do not see these methodological arguments as necessarily compelling.
Moreover, the possibility of modelling individuals as more morally heterogeneous, so as
to view screening and socializing institutions as well as invisible hand institutions, seems to
me to be both exciting and well within the general scope of the economistic approach to
politics.4 In this sense the assumption of single-mindedness can be seen as being much
more basic to the economistic enterprise than any more detailed assumption of prudent
amorality.
Too often, critics of public choice theory (and of economics in general) focus on
motivational questions: they understand economics as defined by the assumption that all
agents act to maximize the satisfaction of their own interests, fairly narrowly conceived,
and seek to discard the whole enterprise on the basis that that motivational assumption is
empirically wrong or ethically objectionable or both. I do not dispute that de facto the
homo economicus construction plays an important role in much public choice related
literature, or that there is clear evidence that some significant practitioners do take homo
economicus to be central. The truth of the matter is, however, that most of the core results
of public choice theory are such that the assumption of wholly or even predominantly
egoistic motivation is neither necessary nor sufficient.
Consider necessity first. In public choice theory, as in the parent discipline of
economics, most of the central results are comparative static results concerning the effects
of changes in relative prices or other relevant variables or institutional structures. All that
is required for such results is the assumption that agents are concerned about their own
interests among other things. That is, homo economicus can be regarded as embodying a
partial specification of agents motivational structures, rather than a complete description.
This more modest assumption is all that is required, for example, to show that particular
institutional arrangements encourage particular (desirable) modes of behaviour. There is
no need to show that, in the absence of those institutional arrangements no-one would
undertake the desirable behaviour. We do not need to assume that all drivers would
routinely break the rules of the road in the absence of detection and penalty devices in
order to argue that detection and penalty devices may have a positive impact on
compliance. We do not need to assume that all candidates elected to life-line political
office would break campaign promises, in order to point out that the the incentive
mechanism provided by future elections may improve the rate of promise keeping. We do
not need to assume that all political agents will always exploit all genuinely discretionary
power assigned to them in order to believe that institutional devices that reduce or
monitor discretion (or select political agents with care) may reduce exploitation overall.
What this means is that those exponents of public choice theory who believe that it is
importantly constituted by the introduction of homo economicus in its strong form into
politics (and those critics who take the same view) are mistakenat least for a very wide
range of relevant propositions. Occams razor (a principle much beloved by economists)
would require us not to make a strong motivational assumption when a weaker one will
do. Even more important, however, is the sufficiency aspect. Motivational assumptions
in and of themselves are a matter of psychology, not of social analysis. The irony of the
debate about motivational assumptions in public choice theory is that those assumptions
are only an input into the real analytical game which is the analysis of institutional
structures as such. The way in which institutional arrangements operate to determine
social outcomes, and the differential operation of alternative institutions, form the central
agenda items. Indeed, the economic approach goes to some lengths to finesse
psychological questions of motivation (which is one reason why the behaviourist element
embodied in revealed preference theory has such appeal to economists). What economics,
and public choice theory, needs is the weakest possible set of motivational postulates,
because the main quarry lies elsewhere, in the discussion of how different institutions
work to produce different social outcomes from the interactions among a group of agents
with no change in their motivations. Issues of institutional performance are obscure, and
analysis often leads to results that are somewhat counter-intuitive. It is this latter
intellectual enterprise that the economic approach to politics is engaged in, and to be
distracted by psychological questions involves a sacrifice of comparative advantage.
In my view, the economic approach to politics can be readily detached from any
assumption of radical egoism as the central motivation, notwithstanding the practice (and
sometimes the explicit claims) of some public choice scholars. The central interest in
political institutions as possible remedies for principal-agent problems, the attention to
the incentive effects embodied in particular institutional arrangements, a focus on the
extent to which social outcomes emergent under particular institutional arrangements
reflect the preferences of the individuals who act within these arrangementsnone of
these issues requires an assumption of unrelieved egoism on the part of agents to be
interesting. Some analytic convenience may be bought by the strict homo economicus
construction, but the price paid in terms of unnecessary antagonism may well be too high.
Finale
Economistic models of politics offer a distinctive approach to questions of the design and
reform of political and social institutions. But the approach carries with it a certain
amount of methodological baggage. This baggage is often seen by critics of the economic
approach as being altogether too heavy, precluding useful analysis or discussion. I have
tried to suggest that in two key areasthe areas of methodological individualism and
rationalitythis baggage is considerably lighter and more defensible than many critics
claim. Public choice scholars and others interested in the economistic approach to politics
are not tied to the implausible denial of the importance of social entities, and nor are they
tied to the empirical claim that all actions are grounded in narrow self interest. It is
relatively easy to sketch a caricature of the economic approach and to dismiss that caricature
as untenable; but that is in the nature of caricatures rather than in the nature of the
economic approach.
However, it is clear that the methodological baggage of the economic approach, even
on a more subtle and sympathetic reading, is not empty. If it were there could be no
purpose to carrying it. My claim is simply that this baggage can provide some insight into
the operation of social systems, that being economical with politics is a worthwhile
ALAN HAMLIN 65
activity. But, as with any method, it is likely that social scientists will be lead astray if they
adopt a method unreflectively and uncritically.
Notes
1 A bibliographic note at the end of this essay contains references which will enable the
interested reader to pursue these and other ideas in more detail.
2 Kelman, S. Public Choice and the Public Spirit, Public Interest, 1987, pp. 8094. Quoted
from pages pp. 801.
3 Hume, D. (1985) Essays Moral, Political and Literary. Edited by E.F.Miller, Indianapolis, Liberty
Classics. Quoted from page 42.
4 A beginning is made in Brennan and Hamlin (1995b).
Bibliographical note
This note is intended to provide some points of entry to the literature relating to the
issues sketched here and to provide references to some of my own work in which these
themes are developed or discussed in more detail. The literature on social choice theory
tends to be technical in nature. Arrow (1963) initiated much of the literature; Sen (1970)
is a classic text in which the technique is largely separated from the discussion; many of
the seminal contributions are collected in Rowley (1993) which also includes a useful
introductory essay. Public Choice theory is surveyed in Mueller (1989); Buchanan and
Tullock (1962) is a locus classicus for the public choice approach to politics. Some relevant
work by Jack Wiseman is collected in Wiseman (1989). Sen (1987) provides an
introduction to the links between economics and ethics. For alternative interpretations of
individualism see Lukes (1973) and ONeill (1973). On rationalism and rationality see
Brown (1988), and Elster (1986).
As for my own work, Hamlin (1986) contains a more detailed account of rationality
and links to the analysis of the state, Hamlin and Pettit (1989) discuss individualism and
other commitments of normative political analysis, Brennan and Hamlin (1992, 1993,
1994) provide examples of the economic analysis of political institutions, Brennan and
Hamlin (1995a) offers a detailed discussion of the political philosophy of constitutional
political economy.
References
ARROW, K. (1963) Social Choice and Individual Values, (Revised Edition) New Haven, Yale
University Press.
BRENNAN, G. and HAMLIN, A.P. (1992) Bicameralism and Majoritarian Equilibrium, Public
Choice, 74, pp. 16979.
BRENNAN, G. and HAMLIN, A.P. (1993) Rationalizing Parliamentary Systems, Australian
Journal of Political Science, 28, pp. 44357.
BRENNAN, G. and HAMLIN, A.P. (1994) A Revisionist View of the Separation of Powers,
Journal of Theoretical Politics, 6, pp. 34568.
BRENNAN, G. and HAMLIN, A.P. (1995a) Constitutional Political Economy: The Political
Philosophy of Homo Economicus?, Journal of Political Philosophy, 3, pp. 280303.
BRENNAN, G. and HAMLIN, A.P. (1995b) Economizing on virtue, Constitutional Political
Economy, 6, pp. 3556.
BROWN, H. (1988) Rationality, London, Routledge.
BUCHANAN, J.M. and TULLOCK, G. (1962) The Calculus of Consent, Ann Arbor, University of
Michigan Press.
HAMLIN, A.P. (1986) Ethics, Economics and the State, Hemel Hempstead, Wheatsheaf Books.
HAMLIN, A.P. and PETTIT, P. (1989) The Normative Analysis of the State in HAMLIN, A.P.
and PETTIT, P. (Eds) The Good Polity, Oxford, Blackwell.
LUKES, S. (1973) Individualism, Oxford, Blackwell.
MUELLER, D.C. (1989) Public Choice II, Cambridge, Cambridge University Press.
ONEILL, J. (Ed) (1973) Modes of Individualism and Collectivism, London, Heinemann.
ROWLEY, C. (Ed) (1993) Social Choice Theory, (3 Volumes), Cheltenham, Edward Elgar.
SEN, A.K. (1970) Collective Choice and Social Welfare, San Francisco, Holden-Day.
SEN, A.K. (1987) On Ethics and Economics, Oxford Blackwell.
WISEMAN, J. (1989) Cost, Choice and Political Economy, Cheltenham, Edward Elgar.
6
Family Behaviour and the Economic Method
George McKenzie
A Challenge
The aim of this book and of this chapter is to enhance the readers awareness of the
different traditions that characterize social and economic research, the historical basis of
these traditions and their impact upon the actual process of research. It is our view that
such awareness can be enhanced by reading about how researchers approach their work.
Hence the implicit autobiographical nature of the several chapters in this book. However,
a deeper understanding of a particular tradition can only be attained through immersion in
the discourse of that tradition. That is, an actual involvement in the methodology and
practice of other, alternative perhaps competing approaches is required. Such
involvement not only enhances an appreciation of what others are up to but also enhances
a process of self-reflexivity which can only serve to improve our individual critical insight.
This is the nature of the challenge which this chapter seeks to set explicitlyit asks
readers to place themselves within the tradition of rational, deductive reasoning such as
Descartes advocated as the basis for human enquiry. In this chapter, as an example, we
shall explore the application of the Economic Method to an analysis of the decisions and
problems of the family. It has long been my view that it is not possible to be critical of a
particular approach to social and economic research or of the practice associated with that
tradition without submerging oneself in the discourse of that tradition. This will not be an
easy task and, in any case, we can only scratch the surface.
Readers will be asked for the moment to suppress their prejudices. This will be an
impossible task. We have all developed our own approach to research and teaching that
will have evolved from our individual personal experiences, including our educational
experience. The latter, in turn, will have been influenced by ideas generated over the
preceding centuries. Thus our prejudices, explicit or implicit, are bound to be deeply
rooted. It will not be surprising if the initial reaction of many readers to what follows is
one of deep anxiety if not anger. But there is some consolation. For the committed positivists
reading this chapter the task will be even more difficult. They will be asked to reflect
upon how the other traditions discussed in this volume would seek to understand the
specific issues discussed here. But for all readers, the theme that I am seeking to draw out
is that are many valid research methodologies and that the one discussed here is only one.
68 GEORGE MCKENZIE
Our focus will be Gary Beckers analysis of the family (1987, 1991). Becker was
awarded the 1992 Nobel Prize in Economics. This distinction was not gained through the
discovery of some new technique or research method but rather through the extension of
the discourse of deductive economic theory to a wider range of social issues such as
discrimination, marriage, divorce and fertility. His approach to these topics is outlined in
his Nobel Lecture (1993). This work illustrates the use of the central concept of economic
analysis, the doctrine of opportunity cost. Basically, this states that there is no such thing as a
free lunch. In order to achieve a certain want, something must be given up. For example,
for an individual with a limited income, purchasing a house might mean foregoing an
automobile. Becker argues that this principle is relevant for all of lifes decisions, not just
those that are explicitly economic. This can be very frustrating for non-economists
engaging in discussion with an economist. By utilizing the concept of opportunity cost, the
economist can appear to be an expert on virtually every topic!
Blau (1987) notes that Becker has laid the foundations for what has come to be called
the new home economics. From a purely historical perspective, the words home and
economics are redundant since the word economics is derived from the Greek words
meaning house and o meaning law or rule. Thus for the Ancient Greeks,
economics referred to the practice of household management. It is in this spirit that we
shall examine the application of the method of the economist to the analysis of family
structure. For those who are not economists, the method will represent a challenge to the
entire concept of interdisciplinary research. For everyone, Beckers work illustrates the
difficulty and trauma frequently experienced in the process of research. In his speech
accepting the Nobel prize, he admits to breaking publishers deadlines and then feeling so
exhausted after completion of the work that his research programme was interrupted for
two years.
Following in the tradition of Max Weber, economic analysis distinguishes between
normative or subjective statements and positive or objective statements. The latter, it is
argued, are capable of analysis by rigorous, scientific analysis. Indeed the individuals that
make up society are themselves thought to act in a scientific way, rational and capable of
undertaking extremely complex decisions. Thus in Beckers analysis, each individual seeks
to rationally maximize his or her level of satisfaction subject to constraints imposed by
income, wealth, social practice and institutions. In the traditional model of consumer
behaviour, such as studied by first-year students in economics, the individual is assumed to
choose her or his consumption of goods subject to a given level of income. However,
Becker has sought to pry economists away from narrow assumptions about self-interest
(1993, p. 385). He seeks to extend this mode of analysis to allow a wider range of
behaviour to be incorporated in individual preferences and hence to enable an
examination of the determinants of family size, marriage, divorce and the participation of
women in the labour force.
In this way of thinking, individuals maximize their satisfaction as they conceive it,
whether they be selfish, altruistic, loyal, spiteful or masochistic. (p. 386, 1993). Such
behaviour is taken as given. Rational behaviour admits to any set of preferences. No attempt
is made to enquire whether markets and actual decision making processes have an
influence upon behaviour. In the spirit of Weber, the subjective may influence objective
and observable behaviour. However, it is a one way street and no attempt is made to
examine the influence of markets, institutions or policies upon preferences themselves.
This is a theme to which I shall return in the conclusion to this chapter.
To illustrate Beckers method, I will consider the interrelationship between four topics
that figure prominently in his A Treatise on the Family: human fertility, the division of
labour between genders, marriage and divorce. A complete exposition of his framework
would require the extensive use of mathematics, something which is beyond the scope of
this paper. Consequently, I will seek to develop the main elements of his approach in a
very partial way, illuminated with the aid of diagrams.
To Have or Not to HaveA Family, That Is
Let us suppose that the preferences of male and female partners depend only upon the
quantity of goods consumed, denoted by Z, and the number of children in the family,
denoted by N. The preferences of the two as a group are represented in Figure 6.1 by a set
of indifference curves, of which the curves labelled U1, U2 and U3 are examples. Along
any curve such as U1, the couple derives equal satisfaction from the several patterns of
goods consumed and family size. That is, the couple is indifferent between the
combinations represented by the points I and II in the diagram. This reflects the
assumption that in order to obtain more goods for consumption the couple is prepared to
have a smaller family. The diagram also reflects two additional assumptions. The couple
obtains more satisfaction from consuming more goods and from a larger family. This
assumption is not necessary for the analysis which can be complicated by allowing for
satiation. However, a second assumption is crucial. The indifference curves cannot cross.
If they did, as in Figure 6.2, the couple would be considered to be irrational. That is, they
would appear to prefer situation I to II, to be indifferent between II and III and by
implication indifferent between I and III. However, this is not possible since they prefer IV
to III and are indifferent between I and IV. In other words, indifference curves that cross
are logically inconsistent. This is the positivist economists concept of irrationality. One
cannot both prefer and be indifferent between two situations. A fuller understanding of
these concepts can be obtained by reference to a good economics textbook, such as that
written by Begg, Dornbusch and Fischer (1993).
In Figure 6.3, the curve AB represents the constraints faced by the couple. Available
choice may constrained by many factors but especially household income and the cost of
rearing children. Thus at one extreme, if the couple decide to have no children, they
could consume OA. However, if children are to be reared, the couple are able to devote
less of their time to income generating work. At the other extreme, all family resources
could be devoted to rearing OB children with no consumption at all, not a feasible
outcome. Neither extreme is optimal for the
70 GEORGE MCKENZIE
Figure 6.1
Figure 6.2
couple under consideration. Instead, they rationally choose to consume OC goods and
to maintain a family of size OD. This yields the highest feasible level of satisfaction at V. A
decision to operate at VI or VII would yield a lower level of satisfaction U1. Any pattern
of family size and consumption along the indifference curve U3 is not feasible.
Figure 6.3
If the costs of rearing children were higher, then the couple would choose a smaller family
size as shown in Figures 6.4a and 6.4b. If family size and consumption were viewed by the
couple as being complementary, they would also choose to consume less (Figure 6.4a).
However, if they were substitutes (Figure 6.4b), consumption would actually increase.
This mode of analysis can be extended in a number of directions. For example, if
employment opportunities exist for children, a larger family size will generate greater
levels of satisfaction since it would also be a source of income.
At this point it is worthwhile taking a brief, methodological digression. Beckers theory
as outlined so far places great emphasis on measurability. Income is earned, goods are
consumed, children are produced. In addition, we have also spoken of individual
satisfaction as if it were a measurable quantity. In fact, this is a fiction designed to simplify
the exposition. However, this fiction is a holdover from an important debate in the discourse
of economics. In the nineteenth century, the Utilitarians explicitly spoke of individual
satisfaction as measurable in objective units of utility. However, as shown by Paul
Samuelson and John Hicks, like Becker Nobel Prize winners, this assumption was
unnecessary. All that was required was for individuals to be able to rank all feasible
opportunities. Measurement was not required for rational decision making.
The Sexual Division of Labour
The previous discussion has implicitly assumed that the male and female partners have
identical characteristics. However, this assumption is dropped when Becker turns his
attention to the analysis of labour force participation by women and decisions concerning
marriage and divorce. Although he takes care to give recognition to the role of cultural
72 GEORGE MCKENZIE
Figure 6.4a
Figure 6.4b
factors in these activities, he places a very strong emphasis on the biological differences
between men and women. Two issues are involved: 1) the economies of scale emanating
from the division of labour and 2) the preferences of women towards income generating
employment versus child rearing.
The concept of the division of labour is one the most important in economics and was
highlighted by Adam Smith in his Wealth of Nations (1776). He illustrated the idea with his
classic example of the production of a pin. He argued that it was less efficient for a
individual to undertake all the tasks involved. Rather, one individual could produce the
shaft, another would sharpen the head and yet another would create the top. By
specializing in this way, the group of workers would be able to produce more pins than if
each tried to carry out the entire job. Economies of scale would arise in the sense that if
each individual doubled their time on a particular task, their output would more than
double.
Figure 6.5
74 GEORGE MCKENZIE
Figure 6.6
Figure 6.7a
76 GEORGE MCKENZIE
In order to avoid the use of mathematical analysis, I have simplified the analysis
considerably. The method of exposition is designed to be pedagogic. In reality, of course,
the choice for men and women is not simply between work and child-rearing. There is
also the choice of leisure time. To extend the analysis in this direction would require a
three dimensional diagram. Hence a verbal presentation will have to suffice. The ultimate
Figure 6.7b
constraint on all of us is time. There are only twenty-four hours in each day. Hence, if a
woman is to pursue a career and raise a family, it is because she values these activities
more highly than her leisure time. However, if the preferences of the couple are
complementary, the term which Becker uses to describe love, then the male partner will
desire to forgo some of his leisure time and devote it to child-rearing in order that his
partner may enjoy more leisure.
Marriage and Divorce, Economics Style
Becker extends the concepts of opportunity cost to an analysis of marriage and divorce.
Here the idea of Pareto efficiency plays a crucial role. Consider the following example. A
Pareto efficient situation is said to be one in which it is impossible to reallocate production
and consumption so as to make any one individual better off without making others worse
off. Similarly, a Pareto inefficient situation is one where at least one individual can be
made better off without anyone else becoming worse off. For Becker, this way of thinking
explains the institution of marriage:
Marriage can be said to take place in a market that assigns men and women to
each other or to remain single until better opportunities come along. An optimal
assignment in an efficient market with utility-maximizing participants has the
property that persons not assigned to each other could not be made better off by
marrying each other. (1987, p. 284)
Important to this calculation will be the extent of dowries, bride prices, leisure and the
exercise of power that one or both parties may enjoy from the marriage.
The analysis of divorce follows on: a rational person will want a divorce if the
satisfaction expected from remaining married is less than that which follows a divorce,
especially if this involves a new liaison. The link between this situation and Pareto
efficiency is as follows. If men and women were always Pareto efficient, then there would
be no incentive for divorce. However, as Becker argues, new information becomes
available. Chosen partners may turn out to be less desirable than anticipated, or an
alternative partner is identified who appears to be more desirable. For Becker, this
possibility explains the basis of the marriage contract. Since women tend to specialise in
child care, for the reasons explained previously, they become economically vulnerable to
separation and divorce. Hence a divorce will usually involve legally enforceable
compensation to the woman.
Becker emphasizes that new information about a mate or a prospective mate is not the
only cause of divorce. He argues:
The rise in labour force participation of married women also lowered the gain from
remaining married because the sexual division of labour was reduced, and women
became more independent financially. (1987, p. 284)
Becker suggests that the liberalization of divorce laws has not had any permanent impact
on the rate of divorce. Rather, he suggests that the decision to remain married or not
depends upon whether family income is greater in a state of marriage than in a state of
divorce:
78 GEORGE MCKENZIE
Assume that the husband appears to gain from divorce, but the apparent loss to the
wife is greater. If divorce were unilateral, he might be tempted to seek a divorce
even when she would be greatly harmed. However, she could change his mind by
offering a bribe that would make them both better off by staying married.
This passage illustrates another important concept of economic analysis, the compensation
principle. It states that for efficiency to be improved by some action, gainers should be able
to compensate the losers. This is simply another way of restating the concept of Pareto
efficiency.
Reflections
My own research has built heavily upon many of the same foundation stones that Becker
has utilized in his research on the family. This will be apparent, for example, from
Measuring Economic Welfare. There I made most of the same assumptions made by Becker:
rationality on the part of individuals, measurability, the representative person, certainty.
The tools of mathematics that were utilized were identical. In the concluding chapter of my
book I sought to identify limits as to the nature of the understandings that I thought that I
had achieved. It seems reasonable that these same thoughts should be directed at Beckers
work as well. Before doing that, however, it is important to note a significant difference
between the two volumes. Becker is trying to extend the methodology of the economist
to a wider range of social issues. He is advocating a particular approach towards
interdisciplinary research. In my work, I was trying to undermine some of the dogma
involved in a particular discourse about how to identify changes in individual and social
well-being. This involved two objectives. The first was technical and involved pointing out
that there were logical errors involved in traditional cost/benefit measurements of gains
and losses. That is, conventional measurement techniques might show a gain in satisfaction
whereas in fact there had been a loss. Rather than simply criticize the perceived status quo,
I proposed an alternative method which was not radical but based upon the same
methodological foundations as the flawed analysis. My second aim was to provide an
antidote to those who would seek to construct measures of individual and social wellbeing that are objective. Such measures may mask changes in the distribution of income
that might be undesirable even if there were gains in efficiency.
But there are limits to all this. In order to understand these limits it is worth-while
saying a few words about the continuing debate over whether or not it is possible to
construct a social science along the same lines as the physical sciences. There are many
elements to this debate but I propose to concentrate on only one which I believe has been
relatively neglected. Physical scientists, in very many cases, can undertake experiments.
They can create situations in which certain variables are held constant and the relationship
between those variables of interest is then studied. Undertaking experiments in a vacuum
is a case in point. The molecules and atoms which form air are eliminated and hence have
no influence upon the results.
In contrast, social scientists are generally not in a position to undertake such controlled
experiments on individuals or groups of individuals. There are two reasons for this. First,
the values of Western society suggest that the undertaking of such experiments is
undesirable. It would impinge upon individual liberties. Second, once people knew that
they were part of an experiment they might change their behaviour and hence bias the
results. Indeed, it would be impossible to forecast the direction of any bias. In one
experiment, individuals might seek to reinforce the preconceived notions of the
investigators, whereas in another experiment, they might seek to undermine the
investigators.
For these reasons, economists and many other social scientists resort to intellectual
experiments. Using mathematics, geometry or verbal logic, the economist creates a
structure in which all variables are assumed to remain constant except those which are of
interest. These are the famous ceteris paribus assumptions. Very many of these experiments
are clearly identified for what they are. They are not designed to be approximations in any
way to the real world. Their aim is simply to enhance the understanding of the
investigator. The investigator then passes on his insights, as Becker has done, in the form
of articles and books. These are valid exercises in the sense that they are designed to raise
questions and to enhance awareness and understanding. They are undertaken for their
own sake.
Readers are then free to give their interpretation to the work in a variety of ways. I will
consider two possible outcomes. First, some readers may disagree with the basic
assumptions of the analysis. Modern critics of positivism have attacked the assumption of
rationality, the representative person, the nature of certainty and the emphasis on
Cartesian measurement. Alternative assumptions would need to be made. For example,
psychoanalytic theorists might suggest that marriage arises from the desire to eliminate a
lacking or rejection that followed childhood rejection by mother. Lane (1991) argues that
markets are cockpits of emotion and that this undermines the economists concept of
rationality. We have noted above, Beckers view that the marriage contract protects the
vulnerable woman from the costs of divorce. However, many women (cf. Viner (1994))
believe that marriage is institutionalized rape. Becker explicitly assumes away the
possibility of rape (1991, p. 313). These thoughts suggest that research into family
behaviour needs to extend beyond the discourse of the economic method.
If theories are simply intellectual experiments, then it does not matter whether its
assumptions are unrealistic or not. Problems arise, however, when researchers seek to
misuse theory. Even though the analysis is conceived as an intellectual experiment, this
has not prevented investigators from attempting to see how well the theories approximate
reality. In a sense, they see the theories as analogues to Newtons Law. But inevitably
none of these theories will stand the test of time because they have not been constructed
to reflect reality in its totality. They are what they are: intellectual experiments
undertaken by complex but limited human minds. The construction of theories and their
use in seeking to understand the world in which we live is, in my opinion an uncertain,
conditional, open-ended activity. The methodology of the positivist or any research
tradition provides a language whereby individuals with similar approaches to research can
communicate with each other. A very great danger, of course, is that the methodology
and its language become the end, rather than a means for understanding.
80 GEORGE MCKENZIE
References
BECKER, G. (1991) A Treatise on the Family, Cambridge Massachusetts, Harvard University Press.
BECKER, G. (1993) Nobel Lecture: The Economic Way of Looking at Behaviour, Journal of
Political Economy, 101, June, pp. 385409.
BECKER, G. (1987) The Family, The New Palgrave Dictionary of Economics, London, Macmillan, pp.
28186.
BEGG, D., DORNBUSCH, R. and FISCHER, S. (1993) Economics, London, McGraw-Hill.
BLAU, F. (1987) Gender, The New Palgrave Dictionary of Economics, London, Macmillan, 40297.
LANE, R. (1991) The Market Experience, Cambridge, Cambridge University Press.
MCKENZIE, G. (1983) Measuring Economic Welfare: New Methods, Cambridge, Cambridge University
Press.
SMITH, A. (1950) The Wealth of Nations, London, J.M.Dent.
VINER, K. (1994) Sisters with Soft Centers, The Guardian Tabloid, 15 June.
Part 2
The Nature of Disciplines
The word discipline implies order and control. Thus those individuals who form
themselves into the traditional or classical research communities such as economics,
politics, sociology, psychology or statistics are implicitly seeking the order and control that
will enable them to pursue their particular conception of systematic research. But for such
disciplines to exist there must be a consensus which forms a common basis in the search for
knowledge. The chapters in this second part address the question of what enables a
particular community to call itself a discipline?. The question is dealt with in an oblique
manner, not directly but indirectly. The first three chapters seek to identify the unifying
elements underlying new and emerging disciplines: information systems, nursing and
social work studies. However, each of the authors focuses on different attributes of their
particular discipline. Avison views Information Systems as interdisciplinary and multifaceted, drawing on computer science, sociology, semiology, economics, mathematics,
management, applied psychology, linguistics, politics, ethics, ergonomics and cultural
studies. Is it possible to define a discipline which encompasses virtually all the disciplines
comprising social sciences? Or are historical labels irrelevant? Is a discipline defined by its
community according to the functions that they seek to perform?
In seeking to define nursing as a discipline, Payne focuses not on its overlap with other,
established disciplines but on research methods. She identifies three phases in the
development of nursing research, the first based on natural science, the second involving
the use of social science paradigms and the third reflecting contemporary controversies
over the processes characterizing knowledge creation. She discusses the application of
grounded theory, ethnography, phenomenology, feminist strategies and action research to
the field of nursing. She also notes that in an attempt to reconcile the differences between
qualitative and quantitative methodologies, some researchers have suggested the use of
triangulation, the application of multiple methods with the niew that no single method is
adequate. Like Avison, she is suggesting that the process of research is open-ended.
Joan Orme takes a somewhat different tack in identifying the central theme of social
work research as practice, that is, the need to inform the practitioner. This is in contrast
to the approach of the more traditional disciplines which adopt Webers position in
arguing that the research process must be objective and independent of those being
researched. In this sense, some would argue that practitioners are a link between the
research and the researched and hence capable of introducing bias. Hence, they conclude
that practitioners should also be independent of the research process. Orme discusses
these tensions and suggests that feminist writing on epistemology and methodology
82
provides a clarification of the above dichotomy and of knowledge creation not only in
social work but in all research practice.
Philosophers of science such as Thomas Kuhn note that science disciplines are less
orderly and systematic than they claim to be. In the fourth chapter of this section, David
Hand provides concrete examples of this by focusing on the use of statistical methods in
the social sciences. This involves two steps. The first involves the specification of the
question in the language of the particular discipline. The second involves ensuring that the
statistical method used actually addresses the same question. His general conclusion is that
before we can make progress in understanding our social and economic environment, we
must focus our attention on carefully specifying the question being researched.
7
The Search for the `Discipline' of
Information Systems
David E.Avison
Background
This chapter is about some of the issues relating to establishing an academic discipline. It
discusses this in the context of information systems, but many of the concerns are found in
other subjects which are trying to establish themselves. Indeed there were similar growing
pains relating to disciplines which are now well established in universities. In the Faculty
of Social Sciences at Southampton, there are groups such as social work and accounting
which are also not well established in many universities, as well as politics and economics
which are well established. However, some of the concerns expressed here are also
relevant to established disciplines. For example, there are debates in English and
anthropology departments about the nature of their disciplines. Perhaps such debates are a
healthy part of any lively discipline and an indication that the discipline is relevant and will
remain relevant.
Some may argue that these concerns are merely academic. However, the concerns
discussed in this chapter are important in practice. The outcome of such debates will
influence what we teach and research under the umbrella of the discipline. They will also
influence how we research, that is, what research methods we might use. In some
disciplines there are well-established research methods. Here, few question the research
method used, and question only whether it was used according to the tenets of the
discipline. However, some university groups might establish a reputation through their
use of a particular approach which goes against the trend. In other disciplines, of which
information systems is one, there seems to be a very wide range of topics taught in
universities, in areas researched and in research methods used.
There are other reasons why this topic is important. Many are political and
financial. For example, on the financial side, it is essential for those working in the
discipline that the subject achieves recognition in order to gain funding from universities
and funding bodies, such as the research councils. It is also important to have good
representation in the political decision-making process. A young discipline may not be
represented in research assessment exercises. This will have a direct financial implication.
As a founding professor in information systems (IS) at the University of Southampton,
I decided to introduce the discipline of information systems to the university in my inaugural
84 DAVID E.AVISON
lecture What is IS? (Avison, 1995). As a new discipline, few people in the university had
knowledge of what the subject was about, what impact it had made and could make, what
was researched and how it was researched. Indeed, there was the widespread view that IS
was Information Technology (IT) under a different name, and much of the talk was about
the comparison of IS with IT as a subject (the latter expressed in popular books called
What is IT?).
This chapter concerns itself most with the inter-disciplinary nature of information
systems. There are many source disciplines which are relevant. However, this does not
invalidate it as a discipline and many of the social sciences are similar in this respect. Further,
well-established disciplines such as medicine are also inter-disciplinary. We look at
research methods that are relevant to information systems. Again, their plurality is
evident. We then consider one area of information systems teaching and research, that of
information systems development. By showing how many different approaches to
information systems development might be used, we return to our theme. Because of the
nature of this text, the differences in research methods used in information systems are
highlighted.
No attempt is made to suggest that those within the academic discipline should get
their act together and agree a very limited set of research methods and teaching topics.
Such an exercise would restrict the potential of this young discipline as well as being
divisive in the community. However, there is a discussion of the potential difficulties that
this richness imposes.
The Domain of Information Systems
In recent years there has been much debate about the actual state and possible future of
information systems as an academic discipline (Banville and Landry, 1989). While there
is, as yet, no consensus as to what should be included or excluded from the field, it is
generally accepted that information systems is an essentially pluralistic field, founded on
knowledge from many other well-established source disciplines. This is true of many of the
social sciences. The insights from the source disciplines have proved of relevance to
researchers, teachers and practitioners alike.
The multi-faceted nature of the field is seen if we consider a definition of the central object
of the discipline. Buckingham et al. (1987) define an information system as follows:
A system which assembles, stores, processes and delivers information relevant to an
organization (or to society) in such a way that the information is accessible and
useful to those who wish to use it, including managers, staff, clients and citizens.
An information system is a human activity (social) system which may or may not
involve computer systems. (p. 18)
Thus, information systems is not simply about information technology. Indeed, it may not
involve computers at all. The definition would seem to encompass a wide range of areas,
for example, information theory (information), semiology (delivers information),
organization theory and sociology (organization and society) and computer science and
86 DAVID E.AVISON
emerging from the data. Other studies illustrate the potential of a theory to explain an
information systems phenomenon (Orlikowski and Robey, 1991; Walsham and Han,
1993; Jones and Nandhakumar, 1993). Another non-positivist approach to information
systems research is critical epistemology (Klein and Hirschheim, 1993), which takes
account of structural contradiction within the social system.
There is also a range of research methods which are generally associated with the
positivist and non-positivist approaches to IS research. For example, much of the
information systems research within the positivist traditions are primarily surveys
investigating phenomena within a single slice of time. This method is normally associated
with a scientific discipline and quantitative data. Interpretive studies on the other hand
tend to employ longitudinal field studies, seeking to obtain in-depth understanding on
information systems phenomena. Most widely used of these techniques in information
systems research are interviews which are generally associated with qualitative data. Many
other techniques are used in conjunction with interviews in the field such as observation
(Orlikowski, 1992) and participant observation (Jones and Nandhakumar, 1993).
Interpretive studies tend to present detailed case studies from the field data to describe a
version of events from which alternative interpretations can be made. Many of these
techniques are commonly used in other disciplines such as sociology and anthropology. For
example, participant observation has its root in ethnographic research studies, where the
researcher would live in tribal villages attempting to understand cultures (Easterby-Smith
et al., 1991).
Other field methods used in IS research are action research (Avison and Wood-Harper,
1991). In this method, the researcher seeks deliberately to intervene in the situation,
often by employing specific techniques, in order to achieve a particular outcome. Action
research is most frequently adopted in organizational development (Easterby-Smith et al.,
1991).
The strength of the positivist research approach lies in its ability to provide a wide
coverage of various situations and to be fast and economical, as well as in its rigour and
replicability in the conduct of scientific research. Orlikowski and Baroudi (1991) claim
that the existing dominance of positivism in information systems research provides a
partial view and has implications on the understanding of information systems
phenomena, theory building and thus for the practice of information systems work. The
strengths and weaknesses of the non-positivist approach are fairly complementary. Thus
the main strength of interpretive approach is its ability to look at change processes over
time, to understand actors meanings, to adjust to new issues and theories as they emerge,
and to contribute to the evolution of new theories (Easterby-Smith et al., 1991).
In researching information systems development, surveys, case studies, and action
research are particularly well used, but there are advocates of all the following approaches
(Van Horn (1973), Dickson et al. (1977) Ein-Dor and Segev (1981), Galliers (1985 and
1991) and Benbasat et al. (1987)):
Conceptual study,
Mathematical modelling,
Laboratory experiment,
Field experiment,
Surveys,
Case studies,
Futures research,
Phenomenological research/hermeneutics,
Ethnography,
Grounded theory,
Longitudinal study,
Action research.
I will attempt to give the briefest of overviews of these research methods. Conceptual
study is frequently referred to as armchair research. No actual on-site experimentation is
carried out. It may precede other research. In mathematical modelling the degree of
control is absolute: all variables are known, no human subjects are required and no
context exists to affect the results. Objectivity is high. Phenomenology is concerned with
meanings that give sense and significance to our experience and attempts to find out what
things are. The context is important (as it might be in any study of social action).
Hermeneutics refutes the idea of an objective world alone, reality is flavoured by our
subjective perception. In laboratory experiments, one of the independent variables is
manipulated by the researcher. Compared to action research, more control can be
exercised by the researcher.
Action research is often confused with case study research, but whereas case study
research examines phenomena in its natural setting with the researcher as independent
outsider, in action research the researcher is participant. Grounded research might
proceed through a series of interviews or observations to hypothesis formation. No
hypothesis is assumed before the interviews in this case, thus attempting to avoid
prejudging the issuesit generates rather than tests theory. This is discussed in
Chapter 13. In participant observation (see Chapter 15), a research approach frequently
associated with ethnography, although there is active involvement by the observer, that
person does not seek to influence the situation more than would be expected from other
participants. Action research (see Chapter 14) is notable for the deliberate intervention of
the researcher. Field experiments include field studies, field tests, adaptive experiment
and group feedback analysis and demonstrate mixes of subjective and objective elements
with a moderate degree of control. Survey, or opinion research, concerns the gathering of
data from human subjects on attitudes, opinions and beliefs as well as responses which are
more objective.
I wish to make some observations about research methods in the context of information
systems. First, I have purposely mixed up the list. They are often described as being
listed in a sequence that represents some continuum, say, from quantitative to qualitative;
positivist to anti-positivist (Burrell and Morgan, 1979). Three observations are relevant
here. The first is that there might be other series of continua (for example, degree of
involvement) and also that the research methods do not sit easily in their positions on the
continuum (some aspects might be positivist and other aspects qualitativesurvey
research provides an example). Second, in any research project, several research methods
88 DAVID E.AVISON
can be used (and not only the same mix). As we shall see in Chapter 14, the research in
information systems discussed is described as action research, and indeed the cycle of theory,
action, reflection and theory modification is a cycle of action research. But this whole
research included conceptual research carried out when considering the various research
in the literature, case study research when looking at the application of the alternative
methodologies in context, and so on. Third, much research in information systems does
not fit easily into a research category. The neat categories are the realm of texts and
courses in research methods. To give one example, some research does not fit in easily to
either case study or action research: the role of the researcher may have aspects in the
project of observer and decision maker.
Information Systems Development
In order to illustrate the multi-disciplinary nature of information systems further, we look
at one aspect of information systems research, teaching and practice: that of information
systems development.
There is a wide variety of approaches to information systems development. Longworth
(1985) identifies over 300 information systems development methodologies. WoodHarper and Fitzgerald (1982) discuss two basic differences of approaches as lying either
within a systems paradigm or scientific paradigm, illustrated by soft systems method
(Checkland, 1981) and structured analysis and design (DeMarco, 1979) respectively.
Avison and Fitzgerald (1995) widen the basis for comparison and suggest that
information systems development methodologies can be compared on the basis of
philosophy, model, techniques, tools, scope, outputs, practice and product, and they
classify approaches within a number of broad themes including:
Systems,
Strategic,
Participative,
Prototyping,
Structured,
Data,
Object-oriented.
General systems theory attempts to understand the nature of systems which are large and
complex. Organizations are open systems, and the relationship between the organization
and its environment is important. Systems approaches attempt to capture this holistic
view, following Aristotles dictum that the whole is greater than the sum of the parts. By
simplifying a complex situation, we may be reductionist, and thereby distort our
understanding of the overall system.
Strategic approaches stress the pre-planning involved in developing information
systems and the need for an overall strategy. This involves top management in the analysis
of the objectives of their organization. Planning approaches counteract the possibility of
developing information systems in a piecemeal fashion.
In participative approaches, the role of all users is stressed, and the role of the
technologist may be subsumed by other stakeholders of the information system. If the
users are involved in the analysis, design and implementation of information systems
relevant to their own work, particularly if this takes the form of genuine decision-making
(as against lip-service consultation at the other extreme), these users are likely to give the
new information system their full commitment when it is implemented, and thereby
increase the likelihood of its success.
A prototype is an approximation of a type that exhibits the essential features of the final
version of that type. By implementing a prototype first, the analyst can show the users
inputs, intermediary stages, and outputs from the system. These are not diagrammatic
approximations, which tend to be looked at as abstract things, or technically-oriented
documentation, which may not be understood by the user, but the actual figures on
computer paper or on terminal or workstation screens. Data dictionaries, fourth
generation systems, CASE tools and workbenches of various kinds can all enable
prototyping. These have become more and more powerful over the last few years.
Structured methodologies are based on functional decomposition, that is, the breaking
down of a complex problem into manageable units in a disciplined way. The development
of structured methodologies in systems analysis and design stemmed from the perceived
benefits of software engineering. These approaches tend to stress techniques such as
decision trees, decision tables, data flow diagrams, data structure diagrams, and
structured English, and tools such as data dictionaries. Most of the techniques enable
complex structures to be communicated using functional decomposition as the basic
technique. Most of the documentation aids are graphic representations of the subject
matter. This is usually much easier to follow than text or computer-oriented
documentation.
Whereas structured analysis and design emphasizes processes, data analysis
concentrates on understanding and documenting data. It involves the collection, validation
and classification of the entities, attributes and relationships that exist in the area
investigated. Even if applications change, the data already collected may still be relevant to
the new or revised systems and therefore need not be collected and validated again.
Object-oriented information systems development has become the latest silver bullet.
The approach is more natural than data or process based alternatives, (the concepts of
objects and attributes, wholes and parts and classes and parts are familiar to children) and
also unifies the information systems development process.
In this brief tour around the various approaches to developing information systems, I
hope I have at least communicated the potentially diverse nature of the information
systems development process. None of these approaches is the answer. Each of the
above themes can be criticized as the basis of an information systems development
methodology. For example:
Systems approaches may not seem relevant to the practitioner who wants a quick answer
to particular problems.
Planning approaches frequently lead to priorities according to the power of managers
and not organizational need.
90 DAVID E.AVISON
It is true that many other disciplines (including medicine, French, management and
geography) do not have a simple and single disciplinary status and can further be described
as a collection of social practices. However, such disciplines can be seen by other
academics as confused and lacking in coherence and academic rigour: information systems
is often seen in the same light. In short, the discipline of information systems lacks
credibility. Indeed, Backhouse, Liebenau and Land (1991) state that Coming as it does out
of computer science, management studies, and a variety of social and technical fields, it is
hardly surprising that [information systems] does not have any theoretical clarity. I agree
with them that information systems lies in the social sciences and not engineering nor
science and the complexity is to be expected. On the other hand, others disagree. For
example, a technological emphasis is found in Gray (1992). This can be seen as confusion
by many outside the discipline (and some inside it!), and the lack of an agreed and
consistent theory is potentially disabling.
The choice of research method should depend on the area of concern. But in practice it
also depends on other factors such as what type of research is acceptable to funding
bodies, to university departments, and to assessors of various kinds. However, nonquantitative research can be rigorous and this extends to stories (see Silverman, 1993).
Likewise, the choice of information systems development methodology should depend on
its appropriateness for the particular domain, but is more likely to depend on other
factors, such as, the dominant paradigm in the organization (which may, in reality, be
rule-of-thumb and experience) or the actors familiarity with the particular methods.
The emphasis and influence of practice on the disciplineseen in the curricula and
research methodsis also a potential weakness. The reverse impact, that of academics on
practice, is less strong. Practitioners do not read (nor contribute) to information systems
journals as much as in many other applied disciplines. Academics in other disciplines
regard information systems lecturers as able to bring in students who see university as a
training ground for the few jobs available. This may be only a temporary phenomenon.
Information systems academics are also seen as able to teach fundamentals in data
processing and computing (work that they do not necessarily find stimulating), and
information systems is often not perceived as an academic discipline in its own right.
This has meant that there are few stand-alone university departments in information
systems, especially in countries outside of North America. And even in the United States
and Canada, less than half of the 1,889 IS faculty are found in standalone information
systems departments (DeGross et al., 1992). Most information systems faculty are found
in computer science, mathematics, business and management, social science, commerce
and other departments. Sometimes, groups of information systems teachers can be seen as
islands off the main influential mainland departments, and in other cases information
systems faculty are split between two separate departments. This makes communications
between information systems people difficult. Thus information systems faculty as well as
research students may find themselves isolated.
The lack of stand-alone university departments in information systems results in
departments having a particular bias, such as a technology bias of most information systems
academics in computer science departments, or the practitioner bias of those in business
schools. This means that the eclectic nature of information systems is not always reflected
92 DAVID E.AVISON
a group to interpret and act upon their environment (p. 3). But as Alexander (1990)
points out, there is extraordinary disagreement today over what culture actually is. There
is also disagreement about the relationship between culture and social structure (see, for
example, Carrithers, 1992). This last point can be seen most clearly if we look at the
differences between American cultural anthropology and British social anthropology. As a
general rule it can be said that American cultural anthropology privileges the concept of
culture (Lewis, 1985; p. 380). The focus is on enculturation, socialization, culture-andpersonality, and child-rearing practices (Caulfield, 1972; Lewis, 1985; Watson, 1984).
Historically, culture was the important concept that held American anthropology
together (Yengoyan, 1986; p. 368). By contrast, British social anthropology privileges
the concept of society (Lewis, 1985; p. 380). The primary emphasis of social
anthropologists is on social relations, and culture is treated as the vehicle or medium for
social interaction rather than an end in itself. In this view, social anthropology is more a
kind of sociology than it is a branch of general anthropology (Watson, 1984; p. 353). The
risk in using a concept such a culture, therefore, is that informa-tion systems researchers
may use a textbook definition of culture, and this usage may end up being a gross
simplification.
Observations and Conclusions
It seems unlikely that there will be a merging of minds as to what discipline or set of
disciplines is the most relevant to information systems and also as to which research
methods are most appropriate. The first division of disagreements (or, perhaps, clashes of
dogmas) occurs between those taking a technical and technological view and those taking
a social and human view; this division can again be subdivided as much as the deepest
functional decomposition diagram and can be as complex as the most complex spiders
web! In fact, such a consensus may be unhelpful as well as unrealistic because information
systems is multi-disciplinary and the different contributions enrich the discipline.
Different research methods are appropriate to the academic study of information systems.
In this chapter we have also addressed the multidisciplinary status of information
systems as a discipline. We have suggested that there are both advantages and
disadvantages in trying to use insights from other disciplines to inform the information
systems research enterprise, and that the advantages outweigh the disadvantages. The
particular combination of relevant source disciplines and research methods define the
discipline. Given this increasing recognition within information systems in recent years of
the value of a plurality of approaches and relevant source disciplines, however, we do
need to ensure that information systems researchers avoid becoming a Jack of all trades
but master of none and further that some coherence to the discipline can be established.
Acknowledgments
I am grateful to Guy Fitzgerald, Joe Nandhakumar, Michael Myers and Trevor WoodHarper for help in the research that led to some of the ideas expressed in this chapter.
94 DAVID E.AVISON
References
ALEXANDER, J.C. (1990) Analytic debates: Understanding the relative autonomy of culture, in
ALEXANDER, J.C. and SEIDMAN, S. (Eds.), Culture and Society: Contemporary Debates,
Cambridge University Press, Cambridge.
AVISON, D.E. and FITZGERALD, G. (1995) Information Systems Development: Methodologies,
Techniques and Tools, McGraw-Hill, Maidenhead.
AVISON, D.E. and MYERS, M. (1995) Information Systems and Anthropology: An
Anthropological Perspective on IT and Organizational Culture, Information Technology and
People, 8, 3.
AVISON, D.E. and WOOD-HARPER, A.T. (1990) Multiview: An Exploration in Information Systems
Development, McGraw-Hill, Maidenhead.
AVISON, D.E. and WOOD-HARPER, A.T. (1991) Information Systems Development Research:
An Exploration of Ideas in Practice, Computer Journal, 34, 2.
AVISON, D.E. (1995) What is IS?, Inaugural Lecture Series, University of Southampton.
AVISON, D.E., KENDALL, J.E. and DEGROSS, J.I. (Eds.) (1993) Human, Organizational, and
Social Dimensions of Information Systems Development, North-Holland, Amsterdam.
BACKHOUSE, J., LIEBENAU, J. and LAND, F. (1991) The Discipline of Information Systems,
Journal of Information Systems, 1, 1.
BANVILLE, C. and LANDRY, M. (1989) Can the Field of MIS be Disciplined?, Communications of
the ACM, 32, 1.
BENBASAT, I., GOLDSTEIN, D. and MEAD, M. (1987) The Case Research Strategy in Studies of
Information Systems, MIS Quarterly, February, 1987.
BOLAND, R.J. and DAY, W.F. (1989) The Experience of System Design: A Hermeneutic of
Organizational Action, Scandinavian Journal of Management, 5, 2, pp. 87104.
BUCKINGHAM, R.A., HIRSCHHEIM, R.A., LAND, F.F. and TULLY, C.J. (Eds.) (1987)
Information Systems Education: Recommendations and Implementation, CUP, Cambridge.
BURRELL, G. and MORGAN, G. (1979) Sociological Paradigms and Organizational Analysis.
Heinemann, London.
CARRITHERS, M. (1992) Why Humans Have Cultures, Oxford University Press, Oxford.
CAULFIELD, M.D. (1972) Culture and Imperialism: Proposing a New Dialectic, in HYMES, D.
(Ed.), Reinventing Anthropology, Random House Inc., New York.
CHECKLAND P.B. (1981) Systems Thinking, Systems Practice, Wiley, Chichester.
DEGROSS, J.I., DAVIS, G.B. and LITTLEFIELD, R.S. (1992) 1992 Directory of Management
Information Systems Faculty, McGraw-Hill, Minnesota.
DEMARCO, T. (1979) Structured analysis: system specifications, Prentice Hall, New York.
DICKSON, G.S, SENN, J. and CHERVANY, N.L. (1977) Research in Management Information
Systems: The Minnesota Experience, Management Science, 23, 9.
EASTERBY-SMITH, M., THORPE, R. and LOWE, A. (1991) Management Research: An Introduction,
Sage: London.
EIN-DOR, P. and SEGEV, E. (1981) A Paradigm for Management Information Systems, Prager, New
York.
FOSTER, L.W. and FLYNN, D.W. (1984) Management Information Technology: Its Effects on
Organizational Form and Function, MIS Quarterly, 8, 4, pp. 22935.
GALLIERS, R.D. (1985) In Search of a Paradigm for Information Systems Research, in Mumford
et al. (1985), Research Methods in Information Systems, North-Holland, Amsterdam.
8
Nursing Research: a Social Science?
Sheila Payne
Introduction
Research in the social sciences has a relatively short history in comparison to the natural
sciences. Their relative immaturity in theoretical and methodological terms has been
emphasized by their critics. It is proposed that nursing research represents an interesting
case example of an emergent academic discipline; one that may have similarities with other
essentially professional groups such as education, social work and para-medical therapies.
The development of nursing research over a relatively short period reflects in microcosm
the more general development of social science research. This paper aims to explore the
emergence and development of British nursing research as a separate academic entity. It will
emphasize the historial, political and social context in which nursing research is embedded.
It will describe a number of tensions in nursing research in its search to identify and
legitimize a separate focus, theoretical basis and methodology which have parallels in
other social sciences. Nursing research has drawn many of its theoretical foundations and
methodological approaches from the natural and social sciences. It is proposed that
broadly three phases can be identified; an early use of natural science methods which
aimed to establish academic credibility, the subsequent influence of social scientific
paradigms, and the current challenges presented to disciplinary integrity by health
services research. This has culminated in a claim that there is no such thing as nursing
research (Stacey, 1994). I will argue that the research agenda and methodologies adopted
by nurses are not situated in isolation or even in best practice but in what is timely,
trendy and funded.
The Context of Nursing Research
Early research which focused on nurses was conducted because of concerns about the role
and functioning of the profession within the National Health Service. These early studies
were conducted by non-nurses and addressed issues of concern to policy makers rather
than practitioners of nursing. The slow development of a nursing generated research
agenda has been attributed to medical dominance and gender inequalities (Hardey and
Mulhall, 1994). It is important to recognize that nursing education was predominantly
biomedical in content, and was largely delivered by, and examined by doctors until the
early 1980s. Thus it is unsurprising that biomedical models of health and quantitative
methodology were very influential in early research done by nurses. The Briggs Report
(1972) emphasized the importance of nursing research to the development of the
profession and was the first official document to suggest that nursing needed its own
research base for practice.
During the 1980s, nursing research developed in the context of emerging nursing
theories; a greater professionalization of nursing practice with pressure to increase the
academic content and status of nurse education; an emergence of an academic nursing
elite, and an increasing orientation to social science. It is difficult to disentangle these
elements, although I shall attempt to highlight important issues.
Critical Issues
The practitioner-academic divide in nursing, along with other practice-based disciplines
like social work, needs to generate knowledge which is useable or even useful for the
majority of practitioners, in comparison with economics, political science or sociology,
where the attainment of academic knowledge may be seen as an end in its self. According
to Rolfe,
there is a widening gap between nursing theorists and practitioners, such that the
findings and recommendations from a great deal of research is never translated into
practical nursing interventions. (1994, p. 969)
The debate about what constitutes relevant research to nursing is ongoing. It has been
argued that the current model of academic research, which aims to develop knowledge but
is largely unconcerned about its implementation in clinical practice, is elitist and does
little to contribute to improvements in nursing care. This type of model of research may
be appropriate for academic disciplines such as sociology or economics. However, the
majority of clinical nurses have little understanding of research and perhaps, even less
commitment to, or knowledge of how to implement research findings. There are similar
complaints voiced by teachers who claim that educational theories have little relevance in
actual classroom teaching.
Moreover, there is a tension within practice-based disciplines about the purpose of
research. The clients or patients can be seen as direct consumers of research findings.
Thus on the one hand there appears to be an imperative to make all research applied and
applicable to clinical contexts but such research may not be conducive to theoretical
development. A more cynical view is that research in academic institutions is concerned with
income generation and personal advancement. The location of nursing research units
within universities or institutions of higher education potentially removes researchers from
clinical contexts and daily involvement with patients or clients, which means that their
efforts are often seen to be removed from the reality of hands-on-care.
Mulhall (1995) has argued that the exposure of many nurses to research has been as
data collectors on projects led by other disciplines, usually doctors. This lowly status has
served to alienate them from the research process and made them cynical of research
98 SHEILA PAYNE
findings. It has also failed to equip them with the necessary skills to generate research
proposals and manage projects. It appears that nurses are often employed as manual
labourers on other peoples research endeavours.
The Role of Theory
Specific nursing theories have largely originated from North American nurse academics.
These theories have attempted to characterize definitions of nursing activity and human
interactions. It is debatable whether theories which relate to such a culturally specific
phenomena as health services, can be applied cross-culturally. There are similar
difficulties in other social science disciplines in researching culturally defined practices
such as parenting behaviours. The danger is that Western norms are imposed as the ideal.
This ethnocentricism has also been found in psychological research in child development.
Perhaps the most influential recent nursing theory, also proposed by North Americans,
(Benner and Wrebel, 1989) is derived from Lazaruss cognitive psychological model of
stress and coping combined with a phenomenological approach to understanding the
lived experience of health and illness. These models are largely descriptive rather than
explanatory. They are frequently used, with more or less success, to guide the recognition
of patients problems and as a framework in planning nursing interventions within the
nursing process. It is less clear that they have a useful function within nursing research,
partly because their vagueness makes it difficult to derive testable hypotheses. A review of
428 nursing studies published between 1986 and 1990 in six leading academic nursing
journals, revealed that only 7 per cent used nursing theories (Jaarsma and Dassen, 1993).
Although overall, 21 per cent of the studies were guided by theoretical frameworks, these
were predominantly derived from the social sciences. It seems safe to conclude that the
use of nursing theory to guide research appears minimal.
Professionalization of Nursing
Throughout the last decade, there has been pressure from within nursing to increase its
professional status. From a sociological viewpoint, nursing is often regarded as a semiprofession, in that it is an occupation which is predominantly female, with a non-graduate
qualifying examination, and much of the work may be defined as manual. Mackay (1989)
suggests that there are three patterns which characterize the way nurses perceive their
occupation. The majority regard it as a job in which their commitment extends no further
than functioning in a safe and appropriate manner whilst at work. A small minority regard
nursing as a vocation, in which personal characteristics such as caring, patience, duty,
devotion and unquestioning obedience are seen as important. For this group nursing is
seen as a lifestyle. Thus nurses are born rather than educated into their role. It is derived
from models of nursing provided by religious institutions and military nursing services.
Although few nurses currently espouse this view of the occupation, the vocational model
is still regarded as an ideal by the general public. A third minority group within nursing
has a professional model of the occupation. Interestingly, this group, which includes nurse
educationalists and managers, have been successful in reforming the status of basic nurse
their status in American hospitals (Glaser and Strauss, 1965). In the 1950s and 1960s, it was
not common medical practice to inform patients of their prognosis, or to speak to them
directly about dying, thus a questionnaire survey would have been entirely inappropriate.
Since it was first described, grounded theory as a method has evolved over time (Strauss
and Corbin, 1990). The method aims to generate rather than test theory. Transcripts of
tape recorded interviews and conversations, field notes and other written material are
used as raw data. The samples are selected on the basis of some defining characteristic
rather than being representative of a particular population. Typically, sample sizes are
smaller than those needed for statistical studies but it is often impossible to specify the
number of participants required at the outset because analysis of the data, and the
saturation of categories determines the completion of data collection. The analysis
commences shortly after the researcher starts collecting data. It may be helpful to
summarize briefly the five main stages of analysis.
1 Development of categories to describe the data: Close examination of the transcripts
are used to identify categories which describe the data. These categories closely fit
the data and are not imposed by previous theories.
2 Saturation of categories: The initial selection of categories is dependent upon the first
interviews analyzed. However, as additional examples of the phenomenon are
discovered in subsequent transcripts, the categories become saturated. The discovery
of new categories in later analyses may necessitate re-analysis of previously processed
transcripts.
3 Definition of categories: As sufficient instances of a particular category are developed,
it is possible to define it. The definition is used to place limits and boundaries around
a category, enabling it to be compared and contrasted with other categories. After
definition, categories are integrated to form a grounded theory. To achieve
integration, it is necessary to formulate a storyline around the central phenomenon
of the study. This involves relating the central phenomenon or core category to
other categories. The categories are arranged and re-arranged until they seem to fit
the story, and provide an analytical version.
4 Linkage with existing theory: After being immersed in the data for a number of
months/years, one begins to recognize differences and emerging patterns. At this
stage, the emerging theory may be linked with existing theoretical constructs.
5 Testing of emerging theory: Finally the constant comparative method is used to test
the emerging theory. This involves an active search of the data to find examples
which confirm or refute the new theory. Validating the theory against the data
completes its grounding. It may also be necessary at this stage to re-interview
participants to present the emerging theory for their confirmation or modification,
thus enabling them to influence the final version. It may also be necessary to recruit
new participants, especially if the phenomenon under investigation is a process
where it may be anticipated that peoples understandings change over time, to check
out specific parts of the theory. A deliberate attempt may be made to recruit people
who represent extremes on some continuum to challenge the theory.
Typically, grounded theories are mid-range theories. Further development may require
more quantitative testing of hypotheses derived from the new theory. The method has
been criticized because it is based on an assumption of the existence of categories and that
data can be forced into them. It also relies heavily on the unseen and unverifiable cognitive
processes of the researcher. Even though researchers are recommended to write analytic
memos (Turner, 1981) during the process of analysis, it is often difficult for readers to
understand how theoretical linkages are made. The use of a research diary as well as the
writing of memos may provide an audit trail. Thus it may be possible to monitor the
development of understanding and insight. From my own experience, I regard the writing
of grounded theory account as an iterative process which contributes to the refinement of
an emerging theory. These accounts should provide rich descriptions of the phenomena
which are illustrated with verbatim quotes. Another source of criticism, is how these
quotes are selected. It is desirable that material is used which both supports and refutes
the emerging theory. It is naive to assume that one can tell it as it is, as was claimed by
Melia (1982) in her account of the learning experiences of student nurses. This is
problematic in at least two ways; firstly in what participants choose to tell a researcher
may differ from what they tell to others, and secondly, how a researcher selects the final
account to present in their written report will be influenced by explicit and implicit
biases.
The grounded theory method has become very popular in nursing research, partly
because it enables processes to be explored systematically and it does not require large
sample sizes. However, less worthy explanations of its popularity include nursing
students dislike of statistics and the perception that it is easy to do a grounded theory
analysis on just a few interviews. This has given rise to a large number of poor quality studies,
with inadequate sample sizes to ensure adequate saturation of categories and a failure to
develop conceptually an analytical version of the theory. At its worst, grounded theory
analysis appears to be used as a type of poorly controlled content analysis. It may be
helpful to mention a few examples of grounded theory studies. Hawker (1983) used this
approach to investigate the interactions between nurses and patients relatives on a
number of hospital wards. She used participant observation and interviews to identify how
nurses controlled their interactions with relatives. Her observational data indicated the
range of strategies they used to avoid contact; for example, the legitimate gait involved
walking quickly with minimal eye contact to give the impression that one was in a great
hurry and thus could not be interrupted by relations. Another study examined how nurses
dealt with the patients body (Lawler, 1991). Lawler used observations (some of them
covert observations) and interviews to understand how nurses coped with potentially
embarrassing situations whilst providing intimate care. In my own research, I have used this
method to understand the process of undergoing palliative chemotherapy for breast
cancer (Payne, 1990) to explore elderly peoples experience of residential care (Smith and
Payne, 1993), and to identify palliative care patients and staffs perceptions of a good
death (Payne et al., 1996). These studies have enabled insider views of various processes
to be articulated.
My own research strategies have been shaped by the tensions I feel at being located in
an academic Psychology Department and my nursing imperative to do clinically relevant
considerably increased work load for the researcher. The technique in its self can not
resolve dilemmas in the relative weight to be given to different data sources, especially if
they are contradictory.
Where to from Here?
Whatever the theoretical positions or methodologies employed, British nursing research
is not in a healthy state. In the most recent Universities Funding Council research
assessment exercise (in 1992), nursing was rated as the lowest of all the seventy-two
subjects reviewed (Tierney, 1994). Although there was considerable variation between
departments, none were awarded the maximum 5 rating and only two reached the 4
standard, with seventeen of the twenty-nine departments receiving the lowest rating
possible. Within the specialist area of cancer nursing research, Payne (1993) has
highlighted some of the important constraints on nursing research; namely, inadequate
funding of both postgraduate level research training and postdoctoral research
programmes, which results in few people with adequate experience to offer research
supervision or initiate and conduct new research projects from a nursing perspective. A
survey conducted in Britain in 1992, revealed only ten funded cancer nursing projects, of
which eight were being undertaken by post-graduate students registered for PhDs
(Corner, 1993). Overall, the majority (85 per cent) of nursing research in Britain is
funded by the government and is policy related (Levine et al., 1993). In the past, the
Department of Health has been the main supporter of nursing research from earmarked
money, especially in the funding of research training studentships, but a recent report
(Report of the Taskforce on the Strategy for Research in Nursing, Midwifery and Health
Visiting, 1993) indicates that they no longer perceive the need for a separate funding
structure for nursing. In the future research funding will be made available via Regional
Health Authority Research and Development Committees. They aim to sponsor health
services research which is multidisciplinary and specifically addresses issues of relevance to
the health and social services. However, the problems of this approach from a nursing
perspective are that medical dominance on committees or in the way issues are
conceptualized may lead to a disease-orientated model. In addition, a scientific medical
epistemology regards the Randomized Controlled Trial as the model of all good research.
Thus studies with qualitative designs are unlikely to be funded as they fail to satisfy criteria
such as generalizability. In collaborative research endeavours, there is a danger that nurses
merely participate as data collectors. These changes in funding appear to be signalling the
demise of nursing research as a separate entity. Research funded through this route is
unlikely to be discipline specific or to address specific nursing concerns. Questions about
how the research agenda is determined will be vitally important if nurses are to retain any
influence. Moreover, it is unlikely that research funded from these sources will contribute
to theoretical developments in nursing.
Is the future for nursing research bleak? Given its poor academic performance in the last
University Assessment Exercise, does it matter? Should nursing research input be merely
seen as a contributing discipline to a larger endeavour, that of health services research?
References
BENNER, P. and WREBEL, J. (1989) The Primary of Caring, Addison-Wesley, New York.
BRIGGS, A. (1972) Report of the Committee on Nursing, HMSO, London.
CORNER, J. (1991) In search of more complete answers to research questions. Quantitative versus
qualitative research methods: is there a way forward? Journal of Advanced Nursing, 16, pp. 718
27.
CORNER, J. and WILSON-BARNETT, J. (1992) The newly registered nurse and the cancer
patient: an educational evaluation, International Journal of Nursing Studies, 29, 2, pp. 17790.
CORNER, J. (1993) Building a framework for nursing research in cancer care, European Journal of
Cancer Care, 2, pp. 112116.
COWMAN, S. (1993) Triangulation: a means of reconciliation in nursing research, Journal of
Advanced Nursing, 18, pp. 78892.
DENZIN, N. (1990) Strategies of multiple triangulation, in DENZIN, N. (Ed) The Research Act in
Sociology, Aldine, Chicago.
FIELD, P.A. and MORSE, J.M. (1985) Nursing Research, Chapman and Hall, London.
GLASER, B.G. and STRAUSS, A.L. (1965) Awareness of Dying, Aldine, Chicago.
GLASER, B.G. and STRAUSS, A.L. (1967) The Discovery of Grounded Theory: Strategies for Qualitative
Research, Aldine, New York.
HARDEY, M. and MULHALL, A. (1994) Nursing Research: Theory and Practice, Chapman and Hall,
London.
HAWKER, R. (1983) Interaction between nurses and relatives, unpublished PhD thesis, University of
Exeter.
JAARSMA, T. and DASSEN, T. (1993) The relationship of nursing theory and research: the state of
the art, Journal of Advanced Nursing, 18, pp. 78387.
LAWLER, J. (1991) Behind the Screens: nursing, Somology and the Problem of the Body, Churchill
Livingstone, London.
LEVINE, E., LEATT, P. and POULTON, K. (1993) Nursing research in the UK, in LEVINE, E.,
LEATT, P. and POULTON, K. (Eds) Nursing Practice in the UK and North America, Chapman
and Hall, London.
LINCOLN, Y.S. and GUBA, E.G. (1985) Naturalistic Inquiry, Sage, Beverley Hills.
MACKAY, L. (1989) Nursing a Problem, Open University Press, Milton Keynes.
MELIA, K.M. (1982) Tell it as it is Qualitative methodology and nursing research:
understanding the student nurses world, Journal of Advanced Nursing, 7, pp. 32735.
MILLAR, M.A. (1993) The place of research and development in nurse education, Journal of
Advanced Nursing, 18, pp. 10391042.
MULHALL, A. (1995) Nursing research: what difference does it make?, Journal of Advanced
Nursing, 21, pp. 57683.
OAKLEY, A. (1992) Social Support and Motherhood, Blackwell, Oxford.
PAYNE, S.A. (1989) Quality of Life in Women with Advanced Cancer, unpublished PhD thesis,
University of Exeter.
PAYNE, S.A. (1990) Coping with palliative chemotherapy, Journal of Advanced Nursing, 15, pp.
65258.
PAYNE, S. (1993) Constraints for nursing in developing a framework for cancer care research,
European Journal of Cancer Care, 2, pp. 117120.
PAYNE, S.A., HILLIER, R., LANGLEY-EVANS, A. and ROBERTS, T. (1996) Impact of
witnessing death on hospice patients, Social Science and Medicine, 43, 12, pp. 178594.
POLIT, D. and HUNGLER, B. (1983) Nursing Research Principles and Methods, (2nd. Edition)
Lippincott, Philadelphia.
PORTER, S. (1993) Nursing research conventions: objectivity or obfuscation? Journal of Advanced
Nursing, 18, pp. 13743.
REPORT OF THE TASKFORCE ON THE STRATEGY FOR RESEARCH IN NURSING,
MIDWIFERY AND HEALTH VISITING, 1993, Department of Health, Leeds.
ROLFE, G. (1994) Towards a new model of nursing research, Journal of Advanced Nursing, 19, pp.
96975.
SMITH, L. and PAYNE, S. (1993) A qualitative study of elderly peoples perceptions of residential
care, Elders, 2, 2, pp. 2338.
STACEY, T. (1994) Achieving effective multi-disciplinary health services research and
development, Conference on Nursing and Therapy Professions Contribution to Health
Services Research and Development, Department of Health, London, 12th May.
STRAUSS, A.L. and CORBIN, J. (1990) Basics of Qualitative Research: Grounded Theory Procedures and
Techniques, Sage, Newbury Park, CA.
TAYLOR, J.S. (1993) Resolving epistemological pluralism: a personal account of the research
process, Journal of Advanced Nursing, 18, pp. 10731076.
TIERNEY, A.J. (1994) An analysis of nursings performance in the 1992 assessment of research in
British universities, Journal of Advanced Nursing, 19, pp. 593602.
TURNER, B.A. (1981) Some practical aspects of qualitative data analysis: One way of organizing
some of the cognitive processes associated with the generation of grounded theory, Quality
and Quantity, 15, pp. 22547.
9
The Case for Research into Practice
Joan Orme
Introduction
This paper is based on the proposition that a research base is necessary for the
development and improvement of practice in social work, but that concerns about
epistemology and methodology have influenced responses to research by practitioners. It
argues that social work can benefit from the contribution that feminist writing has made to
the research debate, and that the issues for social work might be pertinent to other areas of
study, or disciplines. Therefore, it has to be read as complementary to a number of others
in the reader, particularly the papers by David Avison, Jackie Powell and Pat Usher.
That the need for research into practice is particularly relevant to social work is postulated
by Powell when she argues that evaluating social work practice leads to more fundamental
questions about the nature of social work. As she rightly points out this only becomes
significant for those who regard social work as more than a set of skills (or competencies,
in current parlance) and perceive the need for a knowledge base for social work practice
which is informed by research and reflection. Hence the practitioner moves away from
routinization and technical expertise (Schon, 1987). Moreover, Avisons account of the
search for a discipline in Information Systems echoes some of the challenges facing those
theorizing and researching in social work.
In his other paper, Avison argues the case for action research, suggesting that the
identification of a discrete methodological approach helps to establish the identity of a
discipline. Action research does have its limitations. Mies (1983), points out that it may
fail to integrate science and knowledge, and asserts that it is necessary to go beyond a
particular method. The limitations of particular methodologies leads her to focus on the
process of research, the engagement in the activity, and an exploration of how this
contributes to the construction of the knowledge. Hence she argues for the integration of
praxis and research claiming that the demand to link praxis and research consistently
follows an historical, dialectical and materialist theory of knowledge. According to this
concept the truth of a theory is not dependent upon certain methodological principles
and rules, but on its potential to orient the processes of praxis towards progressive
emancipation and humanization (1983, p. 124). Such an approach has particular
relevance to social work where a degree of resistance to the research process has arisen out
of the value base of practice.
argument that the object of research is not something static and homogeneous but an
historical, dynamic and contradictory entity (Mies, 1983; p. 125).
The tension between theory, as represented in a research base, and practice is not a
recent phenomenon in social work. The conflict which surrounded the introduction of the
first social work course in the London School of Economics (for psychiatric social
workers) in 1929 was seen to be between instilling of disciplined habits of thought and
the business of training, or imparting particular skills (Brewer and Lait, 1980; p. 37).
On the one hand, the academy became the site where purists fear the distortion of
theory by pragmatists who seek to tailor it to meet the needs of practice (Brewer and
Lait, 1980; p. 38). On the other, there was a concern that practitioners and their
practice, or more particularly their clients, might be used for research purposes but that
this might not necessarily enhance practice (Everitt et al., 1992). Hence the debate was
about who takes preeminence, the theorist or the practitioner. In exploring the notion of
praxis it can be argued that neither need to; the dialectic of doing and knowing can create
a partnership between practitioner and researcher.
While the development of theory and praxis relating to the research process did not
originate in feminist thinking, it has been feminist discussions of epistemology and
methodology which have most fruitfully questioned the process of knowledge
construction (Mies, 1983; Stanley and Wise, 1983; Stanley, 1990). The challenge to the
established methodologies presented by feminism has been documented by Pat Usher, as
are the alternative ways of knowledge formation, or at least the processes of seeking after
knowledge. Just as Powell documents her commitment to evaluation as a methodological
framework, so I hold firmly to the view that feminist approaches offer the means to move
forward for a social work research tradition.
It is not suggested that social work has accepted the arguments of feminism, but that
the concerns that are raised for social workers can be met by feminist approaches to
research. There are many reasons why social work needs to attend to the position of
women in society, not least because women constitute the majority of workers (Nottage,
1991) and clients (Hanmer and Statham, 1988). Moreover, the resonance between social
work practice and feminism provides compelling reasons for a synthesis. Both are
committed to social and personal change and both seek an understanding of private and
public worlds. More significantly, both stress the dialectic relationship between theory
and practice, but in doing so accept feelings as facts and value self knowledge (Dominelli
& McCleod, 1989). It is because of this that the feminist exploration of praxis can allay
some of the resistance of social workers to research.
Resistance
The nature of this resistance has been identified as coming from a number of sources,
often contradictory. Hanvey (1990) has suggested that it was the immaturity of social
work which has to come of age as a profession. The implication being that in doing so it
will encompass a research base. This fails to recognize the need for social work to establish
its research base before it can reach maturity. The concept of the reflective practitioner,
who can reflect on the practice (or action), and reflect on the reflection in action (Schon,
1987) is predicated on the need for research. What is contentious is the nature of the
research practice which will facilitate this reflection.
Others suggest that the modernist managerialism of social work accountability (Parton,
1994) involves the collection of statistics and data which do not relate to the quality of
practice and alienates the practitioner, making him or her resistant to the activity. This
process is challenged in Stanleys account of the role of researchers in social services
departments. Her case study method (1990), indicating the invisibility of much of the
data, highlights the manner in which official accounts, based on monitoring and statistical
information, tell little about the work that is being done. This, she suggests, is because
researchers are constrained to connect human relationships to numerical research. In
undertaking this exercise they often fail to discover all the information, and as such
produce findings of limited value. That the case study was of her own family which meant
that she had a great deal of private/personal information further illustrated the limitations
of the methodology. Such an approach challenges the orthodox understanding of
objectivity. However, it amply demonstrates the feminist claim that the personal is the
political, that one persons experience is as valid a piece of knowledge as the sum total of
the experience of many. For social work this claim is echoed in Ungersons work on the
documented experiences of individuals who are involved in caring roles and relationships
of the most intimate nature. Her title Policy is Personal (1987) reflects not only that each
individuals experience is valid, but also that global policy decisions impact on these
individuals in different ways.
The inefficiency of positivistic methods is therefore a shared concern of social work and
feminism. A further resistance to research from social work practice is that the
reductionism of research practices associated with positivism is alienating. Social sciences
research is into the sensitive areas of peoples lives. It makes public what is often kept
private. This is much more pertinent in social work where the power to pathologize
individuals because of perceived patterns of behaviour or problems can have significant
effects on the outcomes of intervention, which in turn have implications for justice and
liberty. To avoid such inappropriate uses of power and damaging interventionary
processes it is necessary to acknowledge both the individual experience and the social
circumstances which contribute to that experience. As Goldberg argues putting people
into categories and quantifying phenomena, which in the last analysis are subjective
experiences, cuts right across the social workers belief in the uniqueness of individual
experiences and the need to individualize problems in order to help people in their
difficulties (1972, p. 139).
Research Practice
Such concerns can of course be met by any challenge to positivist styles: for Everitt et al.,
interpretive epistemologies reject the possibility of objective social facts. Rather, social
phenomena are given meaning by those who define and make them explicit. Thus
different people will have different subjective understandings of social phenomena (1992,
p. 7). Such a standpoint is consistent with the definition of one of the core values of social
work, individualization. The classic explanation, that clients are not regarded as fulfilling
certain types and paradigms but as presenting a particular problem which needs to be
considered against its own particular background (Plant, 1970; p. 9) would seem to
negate a research base for social work practice, if such a base is created from an amalgam
of information about individuals, their problems and methods of intervention. While
interpretive epistemologies might reflect the concept of individualization, it is feminist
research which has drawn attention to the particular problems of women and the
particular background of patriarchy which have provided some of the significant
developments in practice in social work. These developments pertain not only to
intervention in the lives of women, but have also facilitated analysis of the delivery of
social work services to black and ethnic minority people. A more recent development has
been the acknowledgment that feminist praxis, that is a specific dialectic between theory
and practice, precipitates significant changes in service for men. Hence Carlens (1989)
work which originally focused on the treatment of women in the criminal justice system
now proposes a feminist jurisprudence, capable of bringing about a more humane
treatment of all who are dealt with within that system.
A further shared concern about research practice is the power differential which is
reflected in the process of knowledge creation (Mies, 1983; Croft and Beresford, 1986).
This is particularly pertinent to social work when the knowledge base is extrapolated from
the experience of selected individuals. When notions of power and legitimacy are
explored in relation to epistemology then the desire to know is a desire for power, but
knowledge of itself does not give power. On the contrary it is those who are authorized
to know and whose knowledge is afforded privilege (Worrall, 1990; p. 7).
Individuals or groups of clients of agencies who become the subjects or objects of
research are not empowered to know. It is the researcher whose position is most
validated, even though that position exists only because of the collection of data about the
experiences of others. Those who are objects of knowledge are frequently described as
the subjects of research. Such a definition reflects the worst aspects of positivist practice.
The extremes of such practices are summarized by Guba and Lincoln thus:
In order to accomplish rendering the study as confounding or contamination-proof
as possible it is thought sometimes appropriate or even necessary to deceive the
subject (humans who are by that very term dehumanized and objectified, that is, made into
objects), to invade privacy without prior knowledge, to place them at physical or
psychological risk, or otherwise exploit them for the researchers or evaluators own
private and professional ends. (1989, p. 120)
Such practices, or even those which are less extreme, arise directly out of epistemological
claims to be objective and value free, and the assumption that there is an external reality
which can be identified and measured, independently of either the measurer or the
individuals or groups being measured.
As Pat Usher has noted, feminist researchers make a direct challenge to such
assumptions and it is this challenge which has resonance for social work. In acknowledging
that the
objects of social science research are also distinguished from those of natural
sciences by being subjects in their own right, indeed by producing their own
understanding and theories of their independent experiences, but also those which
involve researchers and their activities. (Stanley, 1990; p. 9)
This reverses the standard perception of the relationship between the researcher and the
persons researched and is consistent with another fundamental principle of social work
practice, respect for persons. Such a principle requires that in the methods chosen for the
collection of data, the conduct of the research, the interpretation and dissemination of the
results the rights of individuals should be respected (Orme and Forbes, 1991).
A further feminist critique of the positivist approach which echoes the concerns of
social work is that the process of research separates the knowledge constructed from the
origin of that knowledge,
by constructing what is known at a conceptual and categorical level, even if
reference is made to research findings, then how these are known are rendered
invisible. Their indexical properties denied, their contextually specific meaning
glossed in universalist terms. The result is alienated knowledge, a product apparently
complete, bearing no apparent trace of the conditions of its production and the
social relations that gave rise to this. (Stanley, 1990; p. 11)
For social work and feminism the construction of knowledge through research, and
separation of that knowledge from its source, is a way of controlling those who are the
very subjects of that knowledge creation (Croft and Beresford, 1986). The production of
large data sets of, for example, abortion broken down into sub-categories of nationality,
age and marital status have no connection with the personal and, in some cases, traumatic
decision to terminate a pregnancy. To reduce such an experience to a statistic which is
then used to speculate on the morality of women, and on which to argue future policy is
part of the alienation to which Stanley refers.
This form of reductionism has precipitated concerns about the involvement of
practitioners in the process of knowledge creation, and has, for social work, contributed
to the resistance identified at the beginning of this paper. The concerns relate not only to
the conditions in which the research is conducted, but also to the assumption that
acquiring the knowledge contributes to the professionalization of social work. Power
relations are evident in the professional relationship and professional ways of knowing in
our society have served to objectify and control others. Professionals are regarded as
knowledgeable: others are objects of this knowledge (Everitt et al., 1992; p. 18).
Reactions to the striving for professionalism have been associated with the search for a
knowledge base and the establishing of an academic discipline of social work. The concern
is that a knowledge base, or theory, of social work intervention will dictate the outcomes
of intervention, legitimating the performance of some tasks rather than others. Moreover,
this legitimation is controlled by the environment in which the research is generated, and
by factors such as who funds or commissions it. The role of the researcher and those who
disseminate research thus becomes critical in influencing what is actually done under the
name of social work. The validation of knowledge and its role in legitimizing certain kinds
of activity also presents specific problems for social work in the academy and this leads to
a questioning of the validity of the academic project. Within the current climate, for
example, funded research into the criminal justice system is framed by the current
ideology of just desserts, and the Home Secretary has been known to ignore outcomes
which are counter to his own ideology. In such an environment, Carlens work is unlikely
to get support, or at least financial support, from official sources, and validation.
Feminism and social work therefore become subject to a debate about their place in the
academy. For feminism, the validation of a knowledge base which is framed by men, or by
women using masculinist techniques for its construction, legitimizes a particular way of
seeing the world, which is further reinforced by the practitioner base (be that as teacher
or researcher) being predominantly male, framing female experience (Bowles and Klein,
1983). However, one criticism of academic feminism is that the discourse generated by
those involved has become too abstract (Segal, 1987). This is part of the conundrum for
feminist researchers, and one which they share with social work researchers. To have
their findings respected and legitimized they must abide by the rules and norms of the
academy, even when the project is to challenge those same rules. In response, Mies claims
that the potential of an approach which integrates praxis and research which is not
acquired in the sheltered world of academic institutions (1989, p. 124) is pertinent for
feminists and social workers.
Dissonance
The argument that there is congruence between social work research and feminist
research goes beyond the simplest definition of feminist research as being on women, by
women, for women. Such a position was a direct challenge to the process of
universalizing results, drawn predominantly from studies of males. For Pat Usher, the
post-modern challenge from feminism is much more complex. It is not just about who
does the research, nor on whom the research is done, but about that very power balance
which is inherent in the notion of research by a researcher on a subject. It
simultaneously transcends discussion about methodologies, but is central to them.
This is important because a limited view of the feminist contribution to research, seeing
it merely as a critique of positivist approaches, has led to assumptions that it merely
involves a questioning of methodology, and that feminist methodologies equate with
qualitative methodologies. However, Pat Ushers account questions the very dualities
which have operated in the debates about methodologies. The important shift here is that
feminist researchers do not just challenge the facts they theorise the relationship
between personal and the political, private and public, between subjective and objective
(Everitt et al., p. 15), and demonstrate the false dichotomies between these.
It is therefore not enough to argue that feminist research has to eschew the old
positivist, quantitative methodologies, replacing them with a new one (Mies, 1983).
Identifying a single epistemology is in itself positivistic and deterministic (Halberg, 1989)
and argues for a privileged position on the basis that it provides the correct version of how
things really are. Halberg argues, in particular, that post modernism challenges, when
taken seriously, undermine the feminist epistemological project, unless the idea of a new
cognitively privileged position can be defended (1989, p. 4). Her error is in assuming that
there is a unifying feminism and a unified project which is merely to uphold womens
experience as a legitimation of a grounding for knowledge, over and above mens rather
than recognizing the difference.
The contribution that feminist research makes can be more radical than this. It involves
the acknowledgment of unalienated knowledge (Stanley, 1990), achieved by attention to
the relationship between ontology and epistemology, the relationship between ways of
knowing and ways of being. Access to such knowledge can be achieved by a variety of
methodologies, but it is the very process of accessing which makes it a project with which
social work researchers can identify. What makes it a feminist task, for Stanley (1990), is
the identification of the social construction of women, and the acknowledgment of this
construction as oppressive. Here is the challenge for the coupling of social work with
feminism because social work itself has provided challenges for feminism, being seen as
contributing to a particular construction of women and to their oppression (Brook and
Davis, 1985; Wise, 1990).
Praxis
It is the notion of praxis which ensures that this challenge can be responded to positively.
Praxis is relevant to social work in that it stems from an understanding that there are
many ways of being, and many social conditions which lead to oppression. These have to
be addressed both in the process of research, but also in the way that social work is
performed.
Hence the counter argument to Halbergs criticisms is that the feminist project is not
merely a perspective nor an epistemology, but an ontology, a way of being in the world.
What emerges is a deconstructed and reconstructed feminist standpoint epistemology,
one which rejects the successor science label and insists on the existence of feminist
standpoints, i.e. promoting academic feminist pluralism (Stanley, 1990; p. 47).
Such a notion can be open to criticism for being subjective or politically motivated.
This may be so, but being open about it may be more honest than falsely claiming
neutrality and objectivity. As Guba and Lincoln point out,
the value-free claim of positivism papers over the realization that science, like every
human activity, is a political act. Science, by asking only certain questions maintains
(or reinforces) the status quo; it asks those questions that have been formulated by
its own theories, and never takes account of the emic formulations of its subjects.
We would argue that conventional science is, as a result, a force for
disenfranchisement and disempowerment, for maintenance of the status quo
(1989, p. 125).
Feminism outside the academic mode has insisted on the crucial need for useful
knowledge, theory and research as practice. Social work makes similar demands of those
engaged in its pursuit. The essence of praxis is that it is knowledge for rather than any
one static concept of knowledge production. The purpose of the knowledge is to change
the world, not only to study it (Stanley, 1990; p. 15). In undertaking this project the
theory/research divide is erased, or rather united into a symbiotic activity in which the
need for theory to have a research base is acknowledged. More powerfully, the how and
the what of generating the theory are indissolubly interconnected and the shape and
nature of the what will be a product of the how of its investigation. Hence praxis
involves attention to every aspect of the process that undertaking research is based upon,
and contributes to the theoretical understanding. For researchers on practice, and these
predominate in social work, there is a much more specific connection because such
research is to do with affecting change. There is an imperative that the change to be
affected is beneficial, or at least not harmful.
Managerialism
There is a further set of concerns for both practitioner and researcher when considering
the impact of research, whatever the methodological approach. The validation of
knowledge can be used to legitimize certain activities, or alternatively to outlaw practices.
The critique of feminist approaches that they are associated with political correctness
implies rigidity and censorship. While this is rarely the intention of feminist
methodologies, it illustrates how outcome does not necessarily accurately reflect intent.
Similarly, the acceptance of research findings relating to professional practice can lead to
the research becoming part of a regulatory discourse, whether or not that is the intention
of the researcher. Having said that, there is no doubt that some researchers consider it
their responsibility to provide the template, or the yard stick, for professional practice.
Hence MacDonald (1994), reporting on the results of research into the use of behavioural
methods, questions why all managers in the probation service are not requiring staff to use
such methods. Such a stance contributes to the resistance of practitioners discussed earlier
in this chapter; it also raises questions about the appropriateness of privileging certain
findings. If there is an explicit assumption that research findings, however they are
generated, should automatically contribute to the regulation or management of professional
practice, then the concerns about access to publishing outlets and the constitution of the
gatekeeping bodies have to be addressed. Women and black and ethnic minority writers
have challenged the control exercised over publishing by white male academics.
More significantly for practitioners who are intervening in the lives of others, the
assumption is that there is a linear progression in terms of causal factor, intervention and
outcome and that there is absolute certainty about the effects of interventions. So, for
example, the development of risk scales in certain areas of social work draw upon the
accumulated findings of certain research to attempt to determine who might offend
against the law, commit an act of child abuse or be prone to a depressive illness. That such
research might draw social workers attention to situations is perhaps no bad thing, but if
the outcome of such scales is that resources are targeted only on those individuals, or that
punitive intervention or surveillance is imposed upon individuals, before they have
committed an anti-social act, or an act which puts themselves or others at risk, then the
role of research in the process of social control becomes open to question.
The dilemma, therefore, is between the need for practitioners to be aware of, and open
to, social sciences research, while remaining alert to the limitations of both the findings
and the process. To ensure this, it is necessary for practitioners to receive appropriate
level education which prepares them to critically evaluate knowledge which is generated.
It also requires that they become involved themselves in the process of research and
contributing to the analysis and evaluation of their own practice, and that of their fellow
professionals. Moreover, there is a parallel imperative on the part of those undertaking
research. It is not enough to assume that the presentation of detail about methodology,
findings and analysis allows for the replication of findings which, in themselves validate
and legitimize the findings. Just as there is a need for practitioners and educators to be
reflective, so too the researcher needs to acknowledge that the conclusions reached are
indeed only one perspective, arrived at at a certain moment in time when a certain set of
circumstances prevail, and that there is the potential for refutation or development of the
findings by others. This dialectical approach is a continuation of the research process,
which ensures that the knowledge generated is not assumed to be absolute and alienated,
but part of a wider canvas or, in a metaphor used elsewhere in this text, a different way of
telling a story.
Conclusion
Plotting the course of feminist research highlights the perils which are inherent in challenging
any orthodoxy. Similar perils exist for social work researchers. The unintended and
harmful consequences of intervention or analysis, or indeed of the interpretation of
research findings which are constructed as alienated knowledge, without attention to the
process of knowledge construction, can have significant impact on the understanding of
the individual within their environment. It can lead to research being undertaken in a
vacuum, with the results being challenged, or even worse, ignored. Alternatively, political
agendas may lead to some research findings being privileged and incorporated into
regulatory discourses.
The contribution that properly conducted research makes to social work is that it can
empower those who participate in the research, both workers and service-users. It can
make individuals and groups visible and it ensures the evaluation of practice which leads to
more effective interventions and outcomes. In order to satisfy the criticisms and
overcome resistance the research process must reflect good practice and be congruent
with the value base of social work.
This paper has sought to demonstrate that feminist writings on epistemology and
methodology, have provided a clarification and rationale for the ways that research
practice can be congruent with a value base which respects and validates those who become
the subjects of, or more accurately the sources of knowledge creation. It is for these
reasons that feminist approaches are relevant, not only to social work but to all research
practice.
References
AUDIT COMMISSION (1989) Making a Reality of Community Care, HMSO, London.
BAILEY, R. and BRAKE, M. (1975) Radical Social Work, London, Edward Arnold.
BARCLAY, P. (1982) Social Workers: their role and tasks, Bedford Square Press, London.
BOWLES, G. and DUELLI KLEIN, R. (1983) Introduction: theories of womens studies and the
autonomy/integration debate, in BOWLES, G. and DUELLI KLEIN (Eds) Theories of Womens
Studies, R., RKP, London.
BREWER, C. and LAIT, J. (1980) Can Social Work Survive? Temple Smith, London.
BROOK, E. and DAVIS, A. (1985) Women, the Family and Social Work, Tavistock, London.
CARLEN, P. (1989) Feminist Jurisprudence or Women-Wise Penology? Probation Journal 3, 3,
pp. 110114.
CROFT, S. and BERESFORD, P. (1986) Whose Welfare: Private Care of Public Services?, Brighton, L.
Cohen Urban Studies Centre.
DOMINELLI, L. and McCLEOD, E. (1989) Feminist Social Work, Macmillan, Basingstoke.
EVERITT, A., HARDIKER, P., LITTLEWOOD, J. and MULLENDER, A. (1992) Applied Research
for Better Practice, BASW/Macmillan, Basingstoke.
GOLDBERG, E.M. (1972) The Use of Research in Social Work Education, in New Themes in Social
Work Education, International Association of Schools of Social Work, New York.
GOLDBERG, E.M. and FRUIN, D.J. (1976) Towards Accountability in Social Work, British
Journal of Social Work, 6, 1, pp. 422.
GRIFFITHS REPORT (1988) Community Care: Agenda for Action, London, HMSO.
GUBA, E. and LINCOLN, Y. (1989) Fourth Generation Evaluation, Sage, Newbury Park.
HALBERG, M. (1989) Feminist EpistemologyAn Impossible Project?, Radical Philosophy, 53,
pp. 37.
HANMER, J. and STATHAM, D. (1988) Women and Social Work: Towards a Women Centred Practice,
Macmillan, Basingstoke.
HANVEY, C. (1990) Through a glass clearly, Community Care, 24 March, pp. 234.
MACDONALD, G. (1994) Developing Empirically-based Practice in Probation, British Journal of
Social Work, 24, 4, pp. 40527.
MIES, M. (1983) Towards a methodology for feminist research in BOWLES, G. and DUELLI
KLEIN, R. (Eds) Theories of Womens Studies RKP, London.
NOTTAGE, A. (1991) Women in Social Services: a neglected resource, HMSO, London.
ORME, J. and FORBES, I. (1991) Equal Opportunities in Higher Degree Research in ALLAN, G.
and SKINNER, C. (Eds) Handbook for Research Students in Social Sciences, Falmer Press, London.
PARTON, N. (1994) Problematics of Government, (Post) Modernity and Social Work, British
Journal of Social Work, 24, pp. 932.
PLANT, R. (1970) Social and Moral Theory in Casework, RKP, London.
SCHON, D. (1987) Educating the Reflective Practitioner, Josey Bassey, San Francisco.
SEGAL, L. (1987) Is the Future Female? Troubled Thoughts on Contemporary Feminism, London, Virago.
STANLEY, L. (Ed) (1990) Feminist Praxis: Research, Theory and Epistemology in Feminist Sociology,
Routledge, London.
STANLEY, L. and WISE, S. (1983) Breaking Out: Feminist Consciousness and Feminist Research, RKP,
London.
UNGERSON, C. (1987) Policy is Personal, London, Tavistock.
WILSON, E. (1977) Women and the Welfare State, London, Tavistock.
WISE, S. (1990) Becoming a feminist social worker in STANLEY, L. (Ed) Feminist Praxis: Research,
Theory and Epistemology in Feminist Sociology, Routledge, London.
WORRALL, A. (1990) Offending Women: Female Lawbreakers and the Criminal Justice System, London,
Routledge.
10
Scientific and Statistical Hypotheses:
Bridging the Gap
David J.Hand
Introduction
Mathematical statistics, if viewed as a branch of mathematics, is motivated partly by its
own internal structures and aesthetics. Statistical science in general, however, is driven
primarily by problems arising in applications. One might in fact define statistics in general
as the science of answering questions formulated in other disciplines. New statistical tools
are developed to answer new problems arising in the areas to which statistics is applied, as
well as to answer old problems more accurately. The ability to give better answers to old
problems occurs as restrictive and unrealistic assumptions are relaxed, so enabling more
appropriate models to be constructed and fitted to the data. Such models arise as a result
of progress in theoretical statistics and advances in computer technology.
The acronym GIGO, standing for Garbage In, Garbage Out, which is used in
computer science, applies equally to statistical science. Normally it is taken to apply to the
data to be analyzed: inaccurately measured data, biased samples, or incomplete data
necessarily raise questions about the validity of any conclusions. However, the acronym
applies equally to research questions themselves: poorly formulated research questions
inevitably raise doubts about the validity of the conclusions. The aim of this article is to
examine such issues of research question formulation in more detail.
It has often been said that the clear formulation of a question is half the battle of answering
it. Ambiguity in a question raises the possibility of different answers, each one applying to
a different interpretation of the question. The more precise a question statement, the
more confidence one can have in the answer. For statistical questions, however, ambiguity
can arise via two routes. The first is the initial scientific formulation of the question; the
statement in terms of the underlying scientific language. And the second is via the
mapping from the scientific question to the statistical version of it.
To illustrate, consider the question: Are the people in group A more intelligent than
those in group B?. This is really a pre-scientific statement of the question. Before we can
answer it we need to define what we mean by intelligent, a notoriously difficult exercise.
One strategy is to construct a measuring instrument, the value of which one believes
varies (preferably monotonically!) with intelligence (for example, the number of correct
scores in a mental test), and to adopt this as an operational definition of intelligence. People
who score higher (say) on this instrument are (subject to measurement error and other
sources of random variation) taken as more intelligent.
However, things do not stop there. Using the instrument, we can now reformulate our
initial question in a scientific form as Are the IQ scores of the people in group A larger
than those of the people in group B?. While superficially this is straightforward, problems
arise as soon as one tries to answer it. In particular, people have IQ scores, groups do not.
Somehow one has to introduce a definition by which the groups may be compared.
Obviously, to be meaningful, this has to be based on the comparison between individuals.
One might also require the definition to reduce to the between individuals comparison if
the groups each have a single member.
One way to approach such a problem is to summarize the scores in each group and then
compare the summaries. This is fine, but why should one summary (the mean, say) be
adopted rather than another (the median, say)? Any such summary statistic represents a
particular mapping of the scientific question to a statistical question. A strong justification
is surely needed for adopting one rather than another, especially if different statistical
versions lead to different conclusions.
A second way to approach such a problem is to compare individual scores between the
groups (for example, to compute all the pairwise differences), and then summarize the
comparisons. In many cases the two ways yield identical statistics but, as we shall see
below, this is not always the case. This distinction between these two approaches turns out
to be of fundamental importance, and yet quite often to be overlooked.
A slightly different class of problems arises with the question: Are the IQ scores of the
people in group A more variable than those of the people in group B?. Presumably
variable here refers to the magnitudes of the differences between the IQ scores, so that
now we need to (a) say what we mean by a difference between IQ scores and (b) how a set
of such differences is combined to yield a single measure of variability. The second of
these questions clearly poses the same problems as saying whether one group of scores is
larger than another. The first introduces issues such as whether or not it is meaningful to
speak of a difference between scoresas well as how to define such a thing (as the
simple arithmetic difference a-b, as the ratio a/b, as the difference between square roots,
etc). In general, the scientific question reflects the nature of the data: it is arguable, for
instance, whether it is meaningful to speak of the numerical size of the difference between
two preference scores. It would almost always not be meaningful to speak of the
arithmetic mean of numerically coded hair colours (red=1, blond=2, and so on).
Scientific (and hence statistical) data are not merely numbers. They are the numbers and
the context associated with those numbers: what the numbers mean. What the numbers
mean imposes constraints on what numerical manipulations can be performed on them in
scientifically meaningful ways.
This chapter illustrates the importance and difficulty of the mapping from the scientific
hypothesis to the statistical one via two examples. In the first, in Section 2, we explore two
group comparisons, briefly discussed above, in greater detail. In particular, we suppose
that a researcher wishes to make an inference about the difference between the means of
two populations based on samples from each and examine how this might be done. The
second example explores in more depth the distinction, also noted above, between
satisfy the assumptions made by the t-test. We can now legitimately proceed with the
test.
At least, the statistician will now be satisfied that we can now proceed with the test. But
what about the researcher? Recall that the researchers original aim was to study the
difference between the arithmetic means of the two groups, as measured on the original
scale. This is what a t-test does. However, by carrying out a log transform and then
calculating arithmetic means, one is no longer comparing the arithmetic means on the
original scale. In fact, the comparison is now between log(geometric means). Moreover,
instead of studying the difference between them, one is now studying their ratio. The
model required for the student t-test has been satisfied by the transformation, but the
hypothesis that is being tested has been altered.
Fortunately if, after the transformation, the two groups do satisfy the t-test assumption
then equality of means on the transformed variable implies equality of means on the
original variable (since it implies identical distributions). On the other hand, it might
happen that no suitable transformation can be found which simultaneously normalizes the
groups and yields equal variances. If such a transformation cannot be found, what should
be done?
There are versions of the t-test which permit different variances (Welch, 1938; Best
and Rayner, 1987). Unfortunately, however, these are inadequate for our purposes
because, if the variances are different, then it is no longer true that equality of the
expected values implies identity of the distributions and hence equality of all other
parameters. In particular, it is no longer true that equality of the expected value of the
transformed distribution (logically equivalent to equality of the geometric means of the
original distribution) implies equality of the expected value of the original distribution.
The model required for application of the Welch t-test is now satisfied, but the hypothesis in
question is not tested by this test.
What remains? One common alternative is to apply the Mann-Whitney-Wilcoxon test.
This tests the hypothesis that a randomly chosen score from one population will have a
higher value than one randomly chosen from the other population. It says nothing about
the size of the difference between the two scores, which is of the essence in the t-test,
where means are involved. The problem here, as with the t-test on transformed data, is
that although the model is satisfied (and for the Mann-Whitney-Wilcoxon test it is a much
less restrictive model) the test is testing a hypothesis other than the one the experimenter
really wished to test. We need to find some test which focuses on this hypothesis and
which also has a model which uses appropriate population distributions.
We can devise a suitable test if we have background knowledge about the underlying
distribution; if, for example, we assume that the skewness has arisen because the data come
from exponential distributions. Nowadays, for a large class of models, fitting them and
conducting tests is made simple by the existence of statistical software such as GLIM
(Aitken, et al., 1989). On the other hand, one may not be sufficiently confident in
postulating some alternative form, simply noting from the available data that the normal
models seem inappropriate.
One is in the position of having a highly specific hypothesisthe aim is to make a
statement about the difference between two meansbut a very loosely specified
distributional model. We need a test which uses the difference of means as a test statistic
but which requires no further distributional assumptions. There are various ways forward.
The Fisher randomization test, for example, satisfies these requirements and thus is one
possible solution. On the other hand, it is based on a fundamentally different underlying
philosophy from that of drawing a random sample from a population. Another approach
to such problems is based on quasi-likelihood, which focuses merely on the means and
variances and the relationships between the two, without requiring knowledge of higher
order aspects of the distributions. However, we shall not go into further details here.
The purpose of this, superficially straightforward, example was to focus attention on
the fact that both the model assumptionsthe things one was prepared to assume about
the distributions underlying the dataand the hypothesis one wanted to study have each
played crucial roles. It would be wrong to focus attention on only one of them. Ignoring
the model means that the results may not be relied upon. Ignoring the hypothesis means
that the researchers are studying something other than that which they want to study.
Individuals vs Populations
Suppose that the response of each individual to a stimulus of gradually increasing strength
is zero up to a certain threshold, after which it is a constant, k say, the same for all
individuals. That is, each individuals response is a step function, of the same height, but with
possibly different values for the threshold. Suppose also that we collect data for these
curves by measuring the responses at several values of the stimulus. Then there are two
ways of analyzing such data, corresponding to two different questions.
The first is simply to look at the average of the responses at each of the stimulus values
where measurements were taken. The results will be an average of n 0s and m ks, where
n individuals have thresholds greater than the stimulus at which the measurement is taken
and where m have thresholds less than the stimulus at which the measurement is taken. As
the measurement stimulus is increased, so the proportion of ks will increase, so that the
average score will increase monotonically. However, and this is the feature which is of
particular interest here, the curve will be relatively smooth. Steps will be of size 1/(n+m),
assuming that no two individuals thresholds are exactly identical. Certainly, the curve
will not be zero up to some value, and a constant non-zero value after that. That is, the
curve will not have the same form as the individual constituent curves. What we have
here is the pattern of average responses.
Figure 10.2 illustrates this kind of situation, but where the step functions have been
replaced by smooth curves, so that it is clearer what is going on. Each of the curves has the
same shape, but they are positioned at different points. This corresponds to step functions
with different thresholds. The dotted line shows the mean curve, the curve which results
by taking the average of the curves at each value of x. We see that it is flatter than the
individual constituent curves.
In contrast, the second way of analyzing such data studies the shape of each individuals
curveperhaps using measurements on each individual at several stimulus levels to
provide an estimate of the thresholds for each individual. These individual curves are then
combined to produce an average curve. For example, having established that the curves
have the same shape, although with different thresholds, the average curve is defined as
that which also has the common shape, and with threshold value equal to the average of
the individuals thresholds. We can describe this as the average pattern of responses.
In terms of Figure 10.2 and the smooth curves, the middle one of the five similar
curves shows where the mean curve, according to the definition, will lie. This mean curve
has the same shape as each of the constituent curves, and is positioned in the midst of
them. Of course, in order to undertake the second kind of analysis we need to be able to
model the shape of each individuals curve and to do that we need to have measurements
on each individual taken at several stimulus values. This is not always possible. The
obvious example is bioassay, in which the stimulus level is amount of toxic substance
above the animals threshold it diesbut less extreme examples can also be found.
Figure 10.2: The curve of the means versus the mean of the curves
We thus have two distinct kinds of question relating to pattern of response, and we
have seen that they can lead to quite different kinds of answer. Above we termed them
pattern of averages versus average of patterns. Other terms are also used. For
example, in the context of repeated measures data the terms population averaged and
subject specific are used respectively. More generally, we might characterize the
questions as relating to unconditioned questions and conditioned questions. Unconditioned
questions summarize the data as if it were a single set. Conditioned questions partition it
up according to the value of covariates (we might, for example, divide age up into five year
bands and model each age group separately, assuming that the models have the same basic
shape, but with different values of some parameter). We shall see examples of this below.
Another way of describing the distinction that is sometimes used is to describe the
questions as referring to, respectively, the between and within aspects of variation.
Clearly it is vital that any investigation should sort out precisely which of the two questions
is intended.
That the distinction can have far reaching consequences can be seen from the following
simple example. Suppose that, in place of the step function response above, we assume
that each individual has a linear response to the stimulus, y=ai +biX, with intercept and
slope which vary independently from individual to individual. (This is a random effects
model, where the parameters defining each response curve are randomly sampled from a
distribution. The step function example, in which the position of the threshold varied
randomly from individual to individual, was also a random effects model.) Then, for a
fixed value of x, we see that , where and are, respectively, the variance of the a and b
distributions. It is immediately obvious that this variance changes with x. The assumption
of homoscedasticity, so common in introductory accounts of simple regression, does not
hold. In contrast, if one assumed that all individuals had the same response pattern (apart
from random measurement error) then constant variance would not be an unreasonable
assumption. The difference depends on whether one postulates a model for the average of
the response patterns or for the response pattern of the average. The above example also
illustrates how the difference can be described in terms of conditioning. Formulating a
model for individuals and averaging these to produce an overall model is, in essence,
formulating a model conditioning on individuals.
Here is another example showing how confusion can arise unless the conditioning is
made quite explicit. In a study of the effectiveness of trimmed means (variants of the
ordinary mean which ignore extreme values) Efron (1992) used jackknife-after-bootstrap
approaches (jackknife and bootstrap methods are sophisticated modern statistical tools for
squeezing information from data. They are based on re-examination of multiple subsets of
the available data and so are very computationally intensive). He applied the ideas to the
following problem. A subatomic particle, called the tau particle, decays soon after
production into various collections of other particles. Some of the time the decay produces
just one charged particle, and this can happen in four major and various minor ways. The
proportion of time that a single charged particle is produced can be estimated and
similarly the proportion of time that each of the four major types is produced can be
estimated. The primary aim of the study is to estimate and find a confidence interval for
d= d0d1d2d3d4 where d0 denotes the proportion of experiments producing a single
charged particle (of whatever kind) and di denotes the proportion of experiments
producing a particle of type i, i=1,, 4. Efron wanted to use trimmed means T(x) to
estimate d and the associated confidence interval.
However, things are not as straightforward as the above suggests. As Efron (1992, p.
956) says: because of certain physical constraints, any one experiment provides only
one estimate, either an estimate for the composite rate decay [d0 above], or for one of
the four modes [d1 to d4]. Because of this, instead of using T(d), Efron used T(d0)T(d1)T
(d2)T(d3)T(d4). That is, instead of computing the trimmed mean of individual statistics
of interest (the d values for individuals) he computed the statistic of interest for the
trimmed means. These two things will generally not be equal. Again the distinction is one
of conditioning: whether we condition on individuals, so finding the average individual
response pattern, or do not, so finding the pattern of average response. This example also
leads to another way of looking at things. In general, if f is a nonlinear function, then E[f
(x)] will not be equal to f[E(x)]. Here a trimmed mean is nonlinear.
Sometimes we want to study the average pattern of responses but are unable to do so.
The bioassay situation mentioned above represents an example of this. Such a situation
generally arises when it is impossible to take more than one measurement on each subject.
For example, consider the case when we want to compare two treatments or conditions
but where only one can be applied to each individual respondent. This would be the case,
for example, if the treatment was a method of teaching children to read or if it was a
medicine for extending life. What we really want to do in such situations is study some
summary statistic of the (zAzB) differences, where zi represents the response the subject
would give to treatment i, but we cannot do this since for any given individual we will
have only one of zA and zB. We might try to overcome the problem by matching subjects
but this can at best increase ones confidence and never completely remove doubts. After
all, one can never be certain that the covariates used for the matching are really the
important ones.
If the summary statistic one adopts is linear then no problem arises since L(zBzB)=L
(zA)L(zB), where L(x) represents a linear summary statistic of components like x. Then one
can estimate the two right-hand elements separately, from independent samples. In
particular, this is true of the mean, E. (We return to this issue in the next section.) In
fact, this strategy is often adopted for other, non-linear, summary statistics. That is, given
that it is impossible to summarise the (zAzB) differences directly, researchers often
summarize the zA and zB values separately (based on independent samples) and then
subtract the summaries. This means that the researchers have shifted from the typical
response pattern question to the pattern of typical responses question. However, as we
have seen repeatedly above, these two approaches will not generally yield the same result:
they are answering different questions.
One can ensure that the two questions are logically equivalent (the truth/falsity of one
implies the truth/falsity of the other) by making extra assumptions. What these should be
will depend on the summary statistic involved. However, they inevitably involve the
unknown and unknowable relationship between the two measurements on each subject,
one of which must remain latent. (Further discussion of this problem is given in Hand,
1992).
A similar example arises in Simpsons paradox, though here the conditioning is not taken
as far as the level of individuals and simply arises by classifying cases into strata or
subgroups. To take a hypothetical example, suppose we have two treatments, A and B,
and that we are interested in both the total response and the response of two age groups,
those under thirty and those thirty or over. Suppose the data are as given in Table 10.1,
where x/y means that, of y people in the category, x lived and (yx) died. We see that for
each of the two age groups separately treatment A is superior, but for both age groups
combined treatment B is superior. The apparent contradiction arises because of the
weights used in combining the separate proportions (the weights are the denominators, so
that the 20/100 and 39/100 have a greater influence on the combined result than the
other two ratios). However, the important thing for us is not why the apparent contradiction
arises, but the fact that the two questions, one relating to overall ratios and one to ratios
categorized into subgroups are different. It is hardly surprising that different questions
yield different answers. The issue which the researcher must address is which of the two
questions is really the object of the study. A real example of Simpsons paradox, with a
more detailed explanation of why it arises, is given in Hand (1994, Section 3.5).
All of the above examples illustrate the difference between the two questions by
contrasting the results obtained by summarizing the individual response patterns with the
pattern of the summary responses. A less obvious example, but one which nevertheless
falls into the same class, is as follows.
Diagnosis describes the problem of classifying a person into one of two or more disease
classes, according to the values of a number of variables such as symptoms (which may
simply be binary). For our purposes we shall suppose that there are only two disease
classes: sick and healthy. In contrast, screening is the problem of identifying those people
within a population who are most likely to be suffering from the disease in question. Again
we shall suppose that interest lies in a single disease, contrasted with healthy.
Superficially these two problems are identical: in both a classification of an individual into
class 1 (sick) or class 2 (healthy) is achieved on the basis of a vector of scores for that
individual. However, suppose that it is known that the disease prevalence is about 1/10,
so that about 1/10 are ill and 9/10 healthy. Suppose also that, no matter what further
covariates we split the population on, there is a majority of healthy people in each
subgroup. For example, categorizing the population by age may lead to fewer ill people at
younger ages and more at older ages, but in every age group there are more healthy
people than ill people. Then the most probable class for any individual, whatever their
values on other variables, is healthy. That is, if our aim is to diagnose, to classify an
individual into the group from which they have the highest probability of coming, we
should classify everyone as healthy.
This is fine from the individuals perspective. It is clearly the right thing to do since
each such individual is much more likely to be healthy than sick. But what about from the
overall population perspective? From this perspective it is absurd: we know that about 1/
10 are ill, so how can we possible justify concluding that none of them are ill? From the
population perspective, surely what we want to do is find that 1/10 who are most likely
to be ill. This is the objective of screening. The way to achieve this is not simply to classify
each individual into the class from which they have the highest probability of coming, but
to rank people according to their probability of being ill and classify the top scoring 1/10
as sick. Entirely different results have emerged from two superficially identical but in fact
quite distinct objectives. One relates to properties of individuals and the other to
properties of populations.
A similar example, in which the interests of the individual differs from the interests of
someone concerned with overall population properties arises in Rose (1985). The data are
reproduced in Table 10.2. Mothers aged less than 40 have a much lower chance of having
a Downs syndrome child. Therefore, to reduce an individuals probability of giving birth
to a Downs syndrome child we should encourage her to have the birth under the age of
40. However, we also see that 87 per cent (51+20+16 per cent) of such births occur in
mothers aged less than 40. Therefore, to reduce the number of Downs syndrome births
in the population we should focus research effort on ways of reducing the risks for mothers
aged less than 40. Again there is an apparent paradox: an individual should be encouraged
to have children at a younger age, but research aimed at reducing the number of Downs
syndrome children should be focused in the low risk groups!
Diagnosis and screening involve multivariate statistical ideas (or, at the very least,
bivariate: a predictor and the disease class) and multivariate statistics in general is
susceptible to problems of the kind described in this section. Multivariate statistics deals with
properties of multivariate distributions: their correlation structure in factor analysis, their
grouping structure in cluster analysis, and so on. However, often the questions of interest
relate not to the population structure but to the behaviour of individuals. In one of the
earliest papers on these sorts of issues, Robinson (1950), who introduced the term
ecological correlation to describe correlations between (rather than within) groups, said:
ecological correlations are used simply because correlations between the properties of
individuals are not available. The risk, however, is that, in answering different questions,
they may lead to incorrect conclusions. Robinson gave a now classic example of the
correlation between colour and illiteracy for the American Census Bureaus nine
geographical divisions of the United States in 1930. This was 0.946, apparently showing a
very strong relationship. However, the correlation coefficient computed at the level of
individuals turned out to be only 0.203, giving quite a different story.
A more modern example is Quality of Life assessment, where factor analysis of groups
of individuals is used to develop scoring instruments. In fact this merely tells us about the
structure of the population and not about how individuals are likely to behave under
different conditions. We may, as noted above, make additional assumptions relating the
two, but they are assumptions.
Conclusions
Precise specification of the research question is fundamental to scientific progress. To the
extent that the question is ill-defined, so the answer will be ambiguous. In the above, we
have looked at two examples of this. Further examples may be found in Hand (1994).
Whatever ones philosophical school of statistics, all will surely agree that the most
fundamental component is data. Moreover, although there are statistical schools
emphasizing model-free interpretative styles (such as exploratory data analysis (Tukey,
1977)), optimal scaling methods (Gifi, 1990), and biplots (Gower and Hand, 1996)),
modern statistics places a heavy emphasis on formal models (see for example, McCullagh
and Nelder, 1989, and Bernardo and Smith, 1994). In view of this, a suitable way to end
this article is to cite two passages from key researchers in statistical science, the first
contrasting the role of question formulation with that of the data, and the second
contrasting it with that of models.
The most beneficial result that I can hope for as a consequence of this work is that
more attention will be paid to the precise statement of the alternatives involved in
the questions asked. It is sometimes considered a paradox that the answer depends
not only on the observations but on the question; it should be a platitude. (Jeffreys,
1961, preface to the first edition)
A major point, on which I cannot yet hope for universal agreement, is that our
focus must be on questions, not models. Models canand willget us in deep
trouble if we expect them to tell us what the unique proper questions are. (Tukey
in discussion of Nelder, 1977)
References
AITKEN, M., ANDERSON, D., FRANCIS, B., and HINDE, J. (1989) Statistical Modelling in GLIM,
Clarendon Press, Oxford.
BAKER, B.O., HARDYCK, C.D., and PETRINOVICH, L.F. (1966) Weak measurements vs
strong measurements: an empirical critique of S.S.Stevens prescriptions on statistics,
Educational and Psychological Measurement, 26, pp. 291309.
BERNARDO, J.M. and SMITH, A.F.M. (1994) Bayesian theory, Chichester, John Wiley and Sons.
BEST, D.J. and RAYNER, J.C.W. (1987) Welchs approximate solution for the Behrens-Fisher
problem, Technometrics, 29, pp. 20510.
BONEAU, C.A. (1960) The effects of violations of assumptions underlying the t-test, Psychological
Bulletin, 57, pp. 4964.
Box, G.E.P. and Cox, D.R. (1964) An analysis of transformations (with discussion), Journal of the
Royal Statistical Society, Series B, 26, pp. 21146.
DALY, F., HAND, D.J., JONES, M.C., LUNN, A.D. and MCCONWAY, K.J. (1995) Elements of
Statistics, Wokingham, Addison-Wesley.
DAVISON, M.L. and SHARMA, A.R. (1988) Parametric statistics and levels of measurement,
Psychological Bulletin, 87, pp. 56467.
EFRON, B. (1992) Jackknife-after-bootstrap standard errors and influence functions (with
discussion), Journal of the Royal Statistical Society, Series B, 54, pp. 83127.
GIFI, A. (1990) Nonlinear multivariate analysis, Chichester, John Wiley and Sons.
GOWER, J.C. and HAND, D.J. (1996) Biplots, London, Chapman and Hall.
HAND, D.J. (1992) On comparing two treatments, American Statistician, 46, pp. 19092.
HAND, D.J. (1994) Deconstructing statistical questions (with discussion), Journal of the Royal
Statistical Society, Series A, 157, pp. 31756.
JEFFREYS, H. (1961, 1983) Theory of probability, Oxford, Clarendon Press.
LABOVITZ, S. (1967) Some observations on measurement and statistics, Social Forces, 46, pp.
15160.
MCCULLAGH, P. and NELDER, J.A. (1989) Generalized linear models, London, Chapman and Hall.
NELDER, J.A. (1977) A reformulation of linear models, Journal of the Royal Statistical Society,
Series A, 140, pp. 4876.
ROBINSON, W.S. (1950) Ecological correlations and the behaviour of individuals, American
Sociological Review, 15, pp. 35157.
ROSE, G. (1985) Sick individuals and sick populations, International Journal of Epidemi-ology, 14,
pp. 328.
TUKEY, J.W. (1977) Exploratory data analysis, Reading, Addison-Wesley.
WELCH, B.L. (1938) The significance of the difference between two measures when the
population variances are unequal, Biometrika, 34, pp. 2835.
Part 3
Research Practice
This final section, with its focus on practice, builds upon many of the philosophical
questions posed in the previous chapters. The emphasis is upon interpretative and
qualitative research methods. Only the last chapter deals with some of the practical
problems associated with quantitative methods. This was not a conscious decision taken by
the contributors or by the editors. This imbalance has occurred because no one
researching within a positivist tradition felt able to contribute to a volume which was
largely reflexive in nature. However, all the chapters in one way or another adopt the
empiricist tradition advocated by Francis Bacon in the seventeenth century. The research
studies discussed in Part 3 involve the collection of information about particular
individuals whether in the context of the school, the office or a hospital and then the
drawing of inferences from that information. Bacon felt that such investigations could be
objective. However, all of the chapters in this section are sensitive to the subject/object
dichotomy, namely that it is impossible for the investigator (the subject) to be independent
of the investigated (the objects). The process of data collection cannot be independent of
those collecting the data. Each of these chapters investigates the implications of breaking
down the subject/object dichotomy and transforming it into a complementary and
interactive relationship between the researcher and the researched.
Powells personal account of research practice emphasizes the importance and
relevance of the researchers biography in the research process. In exploring approaches
that enable a closer association between research, theory and social work practice, she
raises issues around different conceptions of knowledge and the need for critical reflection
and openness to uncertainty in undertaking research. This theme is also explored in
Scotts chapter, which describes how qualitative methods can be used to examine school
practices. By focusing on the processes of data collection and analysis, he illustrates the
ways in which the researchers knowledge and experience are brought to bear in a context
where change, realignment and modification is an integral part of the research process. In
setting out the rationale and historical development of grounded theory and detailing
many of the procedures adopted, Bartlett and Payne also refer to the need to maintain an
attitude of scepticism as a means of examining their own assumptions and expectations.
Grounded theory, they argue is a powerful and robust method for generating theory and,
as an approach in the post-positivist tradition, makes a significant contribution to the
search for justified belief in knowledge claims.
133
Both Chapters 14 and 15 examine the use of particular research methods in the context
of information systems. Avison argues that the value of action research lies in its ability to
engage both the researcher and the researched in a joint enterprise which is ultimately
directed at solving real-world problems. Drawing on his personal experience as a
participant observer in a large manufacturing company, Nandhakumar presents a critique
of this method. He concludes that despite the many practical difficulties associated with
this role, the direct access to participants own interpretations of events and the
understanding gained through immersion in the research context provide valuable insights
into the social and organizational practices associated with information systems.
In the concluding chapter, McKenzie discusses the conditionality of statistical enquiry
highlighting well known problems which are usually neglected in the reporting of
quantitative data. It seems appropriate that this chapter, as the final one in the overall
text, should end with a series of questions aimed at promoting inter-disciplinary research.
It takes us back to the origins of this enterprise and usefully reminds us that our
endeavours to stimulate dialogue and debate amongst social researchers must remain ongoing.
11
Researching Social Work and Social Care
Practices
Jackie Powell
tradition in social work was lost, or at least obscured, as social work and social
administration took different routes in their development of ideas and practices.
The history of social work as an emerging profession has been intimately bound up with
both the development of other caring professions, most notably those allied to medicine,
and the organizational contexts in which it has been located (Hugman, 1991). These
several strands have, over time, influenced both the nature of social work as a practice and
the type and range of research undertaken. A strong empirical tradition in social work
research reflects its long association with health, and the influence of medical research,
where positivistic approaches are more evident. For example, the use of experimental and
quasi-experimental designs to examine the effectiveness of social work practice has been
well documented, most notably by Sheldon (1986). A strong advocate of this approach,
Sheldon (1986) argues that experiments remain the strictest test of therapeutic intent
(p. 239).
This approach to evaluating social work practice raises wider questions related to the
very nature of social work itself, as well as about the most appropriate ways of studying it;
both in terms of its effectiveness in practice and as to means of generating knowledge and
understanding. Evidence of these wider debates can be found from time to time in the
academic journals (see Sheldon, 1978; Jordan, 1978; and Sheldon, 1984; Raynor, 1984;
Smith, 1987). For example, Smiths (1987) paper, entitled The Limits of Positivism in
Social Work Research, argues that such an approach is inadequate on epistemological
grounds, and that the procedures characteristic of positivism, particularly the
experimental method, are unlikely to prove generally feasible or useful.
A more recent debate about the nature of social work, and in particular the concept of
anti-oppressive practice (Webb, 1991a; 1991b; Dominelli, 1991; Smith, 1992) raises
similar issues, not least about the relevance of the social sciences to the study of social
work. The different theoretical traditions and their alternative methodological approaches
within the social sciences are evident in this discussion, or rather dispute, along with the
possibility and appropriateness of detachment, or, put another way, the inevitability of
commitment (conscious or by default).
For those who regard social work practice as essentially a set of skills and
competencies, the relationship between theory, research and practice is relatively
unproblematic. Both theory and research are given the status of real or universally
applicable knowledge, based on facts which are gathered independently of how people
interpret them. Practice is the application or translation of this knowledge into skills. This
differential status is also evident in other areas of professional practice, most notably
nursing and education. Theory and research are held in high esteem by practitioners,
although frequently criticized for being irrelevant and not located in the real world. In
response, greater attention is given to ways of disseminating knowledge in order to bridge
this gap (see Paley, 1987).
The view of theory as objective knowledge of the external world, i.e. a world
independent of the knower, somehow separate from practice has been challenged by a
number of writers, most notably Schon (1983; 1987). He argues that in many kinds of
practice, nursing and education as well as social work, it is not a straight-forward matter
of using acquired skills to undertake routine tasks. Given the complexity of task and
some educational researchers (Parlett and Hamilton, 1977; Buist, 1988) recognizes the
limitations of positivistic methods and adopts an interpretative perspective in studying the
implementation of the innovation in context. The progressive focusing on key issues in the
particular situation is central to illuminative evaluation.
Such an approach is common to many qualitative approaches, including ethnography. The
researcher engages in participant observation, meeting the research subjects in their
everyday contexts, and observes these at first hand. As Hugman (1988) comments, when
discussing the use of ethnography in the study of social care:
data are obtained through watching, listening and talking. However, this is not
to imply that anything goes. Rather it requires that the researcher makes a conscious,
deliberate and systematic use of ordinary skills in engaging with a chosen social
context. (p. 12)
Data are generated by the researcher in the form of notes which record events and
perceptions. As part of this activity the researcher includes herself or himself in the data
gathering and analysis. Whilst this continual process of reflexivity is applicable to all forms
of research, it is central to the conduct of this type of approach and the way in which data
is generated and analyzed.
This principle of reflexivity, alternatively described as the acknowledgement of the
human dynamic in the research process, is a much contested area in research
(Hammersley and Atkinson, 1983), although strongly and persuasively argued by many
researchers (Steier, 1991). Drawing on her own experience of research and practice,
Munn-Giddings (1993) makes the following observation:
Using a reflexive framework to reflect on practice seems to me to be the most
constructive way forward for practitioner researchers. Starting from the viewpoint
of those being studied, knowledge is not conceived abstractly but is viewed as a
dynamic process in which theory develops from practice. It is a perspective that
facilitates the understanding, undertaking and use of research by practitioners. (p.
283)
In arguing the importance of attention to process and the centrality of reflexivity on the
part of the researcher, an appropriate emphasis is placed on the use of qualitative data.
However, the use of quantitative methods are not thereby excluded. The combining of
several sorts and sources of data, often including qualitative and quantitative material, can
yield a more informed understanding for some aspects of a social phenomenon, rather
than relying on any single method or data source. This multi-method approach,
commonly referred to as triangulation (Denzin, 1970), is also discussed in relation to
case studies by Bryman (1989). However, there are epistemological issues as well as
technical considerations involved in bringing together qualitative and quantitative data,
and these are more fully explored in Brannens (1992) edited collection of papers on Mixing
Methods: Qualitative and Quantitative Research. Also, the case studies described in the
second part of this book usefully highlight the several ways in which qualitative and
quantitative approaches can be brought together in a critically reflective manner.
Exploring Differences of View
In the inherently political context of policy research, (within which social care research is
located,) there is considerable value in bringing together a variety of data, not least as an
attempt to recognize the multiplicity of views and often competing interests of those
involved in the process of policy implementation. In the specific context of evaluating
services, exploring the various legitimate definitions of success and the means by which
the various interested parties seek to sustain their definitions is a key area of focus in our
work (Powell and Lovelock, 1992).
Organizations, including those which provide social and health care services are not
homogeneous entities with clear and unitary goals; rather they are made up of a
variety of sub-groups with differing interests and concerns. In addition, social and
health care services typically have a variety of users with differing needs: not only
direct users and their families or other carers, but in a different sense professionals
providing services complementary to that forming the focus of the evaluation
concerned. (pp. 89)
Thus, the assumption of consensus and unity among all parties within an organization is
discarded and a rather different understanding of the social world, based on
acknowledging the complex nature of organizations is employed. This pluralistic
approach to evaluation (Smith and Cantley, 1985 and 1988) is founded on the view that
organizations have a diverse range of goals and are made up of a diverse range of subgroups with often competing concerns and interests, an understanding of which needs to
be incorporated into the research methodology.
Ambiguity and lack of agreement in perception between parties of the policy
shaping community would then be a central feature of the research, rather than an
embarrassment as is the case when the presumption of consensus prevails. (Smith
and Cantley, 1985, p. 9)
Smith and Cantleys approach seeks to overcome the conceptual, methodological and
political difficulties of the more traditional experimental strategies employed in
programme evaluation. They argue that a research strategy which acknowledges an
evolutionary process and a diversity of views, rather than one which seeks to minimize or
circumvent these features, has wider applicability. Thus:
pluralistic evaluation offers an ethnography of the way the service develops and
an explanation of this development (in terms of the pluralistic interests of
participating groups) as well as (somewhat complex) conclusions about the
success of the service on a range of criteria interpreted in different ways. Our claim
is that the attempt to link the problems of the organization as perceived by its
constituents, to the performance of the institutions programme of service,
constitutes evaluative material. (Smith and Cantley, 1988; p. 149)
What might be seen as both an advantage and a potential disadvantage of this approach lies
in its ability to identify several different constituent groups and to attend to the ways in
which they are able to marshal power and resources to pursue their interests, and at what
cost to other groups. The revealing of the processes, frequently exposed as jockeying for
power, opens up the way for change. Certain groups can be assisted or supported in pursuit
of their interests, not least those whose interests have received, until relatively recently,
insufficient attention in developing health and welfare programmes: especially users, front
line practitioners and other less powerful groups.
For example, the central focus of our study of the implementation of a policy of
community care for people with mental health problems was on exploring the impact of
this process on the multidisciplinary team providing the day-to-day services and on the
people using them. In seeking to explore the different perspectives of those involved in
this process, we actively sought to put onto the agenda the views of those who may
otherwise not have been heard, most notably those involved in front-line services as either
providers or recipients of the evolving service.
Whilst attention can be broadened to include a wide range of interested parties, the
risk of exclusion remains. There is always a danger of selecting key groups and justifying
their inclusion on grounds which reflect a particular set of interests, not made explicit.
For example, client or user views have become an integral part of many evaluative studies
of social work and social care services. However, as Sainsbury (1985; 1987) has noted, an
uncritical consumerist approach may present problems. Service users frequently have
very low expectations about both the availability and quality of services. It is important,
therefore, to ensure that such evaluations are not simply used to ration services or
preserve the status quo. The task of the researcher, like the social worker, is to negotiate
the meanings of what the client says. As Howe (1990) argues:
Client study researchers have tended to be stronger on moral prescription than on
sociological theorizing. Their wish to give the client a good hearing has had a
healthy and welcome impact on social work research and practice. But without an
equal measure of social theory there is always the danger that understanding is
sacrificed to reforming zeal. Asking clients to speak remains a fundamentally
important research task. Adding theory to the clients view not only tells us how
practice is experienced but also helps us to understand something of the nature of
social work itself and the times in which it was formed. (p. 76)
Wallace and Rees (1988) however, argue strongly for the highest priority being given to
clients evaluations.
Social works major task is to respond to problems which have largely been thrown
up by an economic and cultural system which operates for the considerable benefit
of some and results in the exploitation and suffering of others. In response to these
problems, social workers responsibility is never to lose sight of the needs of the
most powerless people (Wallace and Rees, 1988; p. 58)
Making this view explicit does not in their view hinder the conduct of the research.
On the contrary, the pretence that values do not influence the choice of research
interests and methods is more likely to produce bias than the explicitness that we
have documented. (Wallace and Rees, 1988; p. 59)
This is a contentious topic which, once again, demonstrates the importance of reflexivity
on the part of the researcher. It also raises issues about the purposes and use of research
and the role of the researcher. These are further explored in discussing the possibility of
collaborative partnerships between the researcher and those being researched.
Research as Partnership
The work of Guba and Lincoln (1981; 1989) in the field of educational evaluation in the
US, in several respects parallels the pluralistic approach of Smith and Cantley (1985) and
our own work. However, it also explicitly addresses the issue of organizational change
through working in partnership with the several stakeholders involved. This more active
involvement in a process which seeks to contribute to improvements in service quality
was adopted in our study of the development of community-based mental health service.
We worked most closely with front-line workers and managers, offering structured
feedback throughout the study period to those most closely concerned with providing the
service (Powell and Lovelock, 1989). This contact provided an opportunity to report on
work in progress and to offer a reflective account of the continuing process of moving out
and its impact on some of the interested parties. It also enabled us to check out our
understandings of developments with those of some of the direct participants, mainly
those engaged in front-line service delivery (McKeganey and Bloor, 1981). The final
report on the evaluation offered an assessment of the development of the service over a
critical two-year period of transition; issues needing further attention were identified and
discussed.
As indicated earlier, this piece of work was commissioned by those involved in policy
planning and implementation at a local level as a means of informing themselves and
others of the processes involved in developing a community based service. There have
been many attempts in recent years to consider the potential contribution of social
research to social policy and the various roles which researchers do or might play. Bulmer
(1982) identified three models of the relationship between research and policy: the
empiricist model, the engineering model and the enlightenment model. An alternative
role, that of the advocate, where the research seeks to represent the perspectives of its
subjects, has been proposed by Finch (1986). This opens up the possibility of a more
participative partnership between the researcher and those being researched; an
approach we have sought to adopt in our own practice.
In a subsequent study of implementing a community care policy for older people with
mental health problems, a more explicit stakeholder approach was adopted through the
setting up of a stakeholders representative group in conjunction with the service
managers (Powell and Goddard, 1996). Membership was drawn from all the principal
interest groups. In examining the value of this group as a forum for debate we concluded:
Both the setting up of the stakeholder representative group and the seeking of
stakeholders views via interviews contributed to the opening up of the decisionmaking processes within the service. Although these approaches had limitations,
the use of both mechanisms together proved successful in ensuring a wide range of
views were taken into consideration. (Powell and Goddard, 1996)
However, attention was also drawn to the way in which this approach can deny or accept
some of the built-in inequalities of power among, for example, managers, different
professional groups and service users.
This model of the researcher as actively engaged in the process of change is a well
established research tradition in education, but is nowhere near so developed in social care
research. The researcher is involved in the process of reflexivity and collaboration in
bringing about change. However, many forms of participatory research, which seek the
active involvement of research subjects in the research process, have been criticized for
failing to confront social inequality at whatever level it occurs (Holman, 1987; Oliver,
1990). Rather, the researcher retains the role of expert, with overall control of the
methodology. For example, in the field of poverty research, Holman (1987) argues:
The researchers have established not only the extent and nature of poverty, but can
claim some credit for bringing it to public notice. [these] studies have done
something to improve services for the socially deprived. May such efforts continue.
But valuable as it is, it is largely research about, on or for the poor, not by or with
the poor. (p. 670)
What is needed, Oliver (1992) and others argue, is a fundamentally new approach,
changing the relations of research production, going beyond partnership and adopting an
emancipatory stance. The importance of this lies:
in establishing a dialogue between research workers and the grass root people
with whom they work, in order to discover and realize the practical and cultural
needs of those people. (Reason, 1988, p. 2)
Once again, issues concerning the relationship between values and research practice are
raised. As Ravn (1991) states, social research is not a neutral medium, it is: an activity
recognized by many not just as unveiling the facts but as constructing them, and the
researcher plays a major role in this. (p. 112)
Mies (1983) takes this further by arguing that:
The postulate of value free research, of neutrality and indifference towards the
research objects, has to be replaced by conscious partiality, which is achieved through
partial identification with the research objects. (p. 122)
Conscious partiality recognizes the wider social context and the research participants
(including the researcher) place in it. This view has much in common with those
expressed by Wallace and Rees (see earlier), where the researchers own values are
acknowledged as part of the research.
Theorizing Social Work Practice
In bringing this chapter to a conclusionwhilst the discussion continuesI want briefly
to return to the contribution that research into practice might make to social work
theory. Much of the research into social work and social care practices has centred around
studies of effectiveness. This ever-growing demand to evaluate effectiveness comes from a
long-standing professional commitment to improve and enhance practice and from a more
recent need to demonstrate a careful utilization of public resources. Thyer (1993)
identifies a growing body of empirical research which demonstrates the value of social
work services in a variety of settings. He argues that these developments are largely due to
the use of research methods located within the positivistic tradition. However, he also
acknowledges that this approach has a number of limitations.
For example, conventional scientific methods do not provide us with much in the
way of guidance in developing new theories themselves, or in formulating original
research questions to be investigated. Such activities may be guided by ones
practice experience and clinical wisdom, serendipitous findings, curiosity,
professionally-informed speculation, reasoning processes, intuition and sometimes
(the history of science tells us) by our dreams! (p. 20)
Rather more contentiously, he continues:
A number of so-called qualitative research methods may be useful for this purpose,
investigatory strategies which in themselves do not produce facts in a positivistic
sense, but certainly may be valuable to generate hypothesis to be subsequently
tested in controlled studies. (p. 20)
Alongside this more explicitly evaluative research firmly located within a positivistic
tradition, other studies, whilst equally concerned with developing and improving
practice, have focused less on outcome and more on developing an understanding of the
processes involved. As was indicated earlier, our study of an evolving community-based
mental health service was seen as making a contribution to a wider understanding of the
changing nature of professional practice in the context of contemporary legislation. Other
studies (see Fisher et al., 1984; Howe, 1986), which have also sought to explore different
patterns of service delivery and to link these with the wider organizational context of
practice, offer a rich source of data for understanding more about the nature of social
work and social care practices.
Working within a broadly conceived interpretative framework, such understandings
can generate new ways of conceptualizing practice, thereby contributing to the
development of practice theories. As Curnock and Hardiker (1979) state:
We hope our work has helped us to understand in some small ways the complex
filtering processes in which social workers are engaged as they work with clients.
Traditionally this has been referred to as practice wisdom, but we think that it can
make claim to a higher theoretical status than this. This is why we have been
moving towards an understanding of practice theories too. (p. 172)
This reconceptualizing of theory and practice is further explored by Usher and Bryant
(1989). Using the exploratory metaphor of the captive triangle, where theory and
research are conventionally represented as the foundation or base which is then applied to
practice at the apex, they argue the need to re-think this conventional representation and
to open up this triangle to allow a new way of thinking about these three elements
without privileging any one at the expense of the other. They argue for the greater
awareness on the part of practitioners to convert their practice into praxis:
a form of practice which is both reflective and reflexive. The essence of being
reflexive is that theory and practice are dialectically interrelated. Implicit theory is
brought to consciousness and continually open to change in the light of practice,
which itself changes as informal theory is modified. This process, therefore, clearly
utilizes the hermeneutic circle of mutually-interactive backward and forward
movement between understanding and action. (p. 92)
They then go on to elaborate the nature of the relationship between formal theory and
informal theory using the notion of review. This involves accepting that the purpose of
formal theory is representation and explanation, and that of informal theory is judgment,
interpretation, and understanding (p. 93).
What is important here is that both formal and informal theory are acknowledged as
different and equally valid ways of knowing about the social world. This raises an issue of
central concern in the development of social work theory, namely the way in which
power operates through ways of knowing. It is also an area where feminist researchers
have made a growing contribution. As Worrall (1990), in her study of female offenders,
points out:
The relationship between knowledge and power is crucial to any attempt to
theorize womens experiences. The desire to know is a desire for power but
knowledge of itself does not give power. On the contrary, it is those who have
power who are authorized to know and whose knowledge is afforded privilege.
(p. 7)
Henkel (1995) discusses similar issues around different conceptions of knowledge in the
context of social work education, and identifies the need for critical reflection and
openness to uncertainty (p. 68) in the pursuit of reflective practice. She argues the
relevance of both pragmatism and hermeneutics which, although derived from the
different traditions of North America and continental Europe, have much in common.
Both challenge the Cartesian paradigm, at the centre of which are dualistic forms of
thought (mind and body; subject and object; theory and practice; knowledge and
action) and the ideal of the individual knowing subject contemplating an external
object. Instead they assert the active and collective nature of knowledge acquisition
and development, together with the role of language and disciplinary tradition in
shaping knowledge and confounding the clear cut division between subject and
object. At the same time, they facilitate a review of the relationship between
conceptions of knowledge as understanding and dialogue. (p. 68)
Critical Theory in particular, with its emphasis on dialogic communication, offers a
conceptual framework for social work (and social work research) practice (Blaug, 1995).
However, particular attention to either Critical Theor(ies) or feminist perspectives, both
with emancipatory aims, have remained outside the scope of this paper. The main puprose
of this paper has been to provide an account, to tell a story, about the development of an
approach to researching social work and social care practices; an approach located within
an interpretive framework and seen as one attempt at encouraging closer links between
research, theory and social work practice. It has been written at a particular time and for a
specific purpose. It has used autobiographical detail as a means of emphasizing the
importance of self-awareness in making choices between an array of divergent approaches
to social research.
In his tracing of the promise of biography, Rees (1991) confirms:
the value of developing questions over time, as part of a disciplined approach to
personal and professional development. A student can be encouraged to develop
questions about the present, not merely to solve an immediate problem but as a
means of developing a philosophy for practice, and skills in so doing. (p. 12)
In this context, I am the student, and the process of developing a philosophy for research
practice continues. Reflecting upon this account, key questions about the nature of
theory, and social work theory in particular, remain unresolved. The development of
feminist theory and its contribution to the formulation of feminist social work (Dominelli
and McLeod, 1989) provides alternative ways of knowing about the social world.
Feminist approaches to research, only touched upon here, need closer consideration in the
development of research practice relevant to social work and social care. However, in my
view, this takes me beyond an interpretive framework (and therefore the scope of this
paper) and into areas which identify the limitations of this tradition. It also begins the
telling of another story, currently in the making.
References
BLAUG, R. (1995) Distortion of the Face to Face: Communicative Reason and Social Work
Practice, British Journal of Social Work, 25, 4, pp. 42339.
BRANNEN, J. (Ed.) (1992) Mixing Methods: qualitative and quantitative research, Avebury, Aldershot.
BRYMAN, A. (1989) Research Methods and Organisational Studies, Unwin Hyman, London.
BUIST, M. (1988) Illuminative Evaluation, in LISHMAN, J. (Ed.), Evaluation. Research Highlights
in Social Work 8, 2nd edn, Jessica Kingsley Publishers, London, pp. 7281.
BULMER, M. (1982) The Use of Social Research: Social Investigation in Public Policy-making, George
Allen and Unwin, London.
CURNOCK, K. and HARDIKER, P. (1979) Towards Practice Theory: Skills and methods in Social
Assessments, Routledge and Kegan Paul, London.
DENZIN, N.K. (1970) The Research Act in Sociology, Aldine, Chicago.
DOMINELLI, L. (1991) Whats in a name? A comment on Puritans and paradigms, Social Work
and Social Sciences Review, 2, pp. 23135.
DOMINELLI, L. and MCLEOD, E. (1989) Feminist Social Work, Macmillan, London.
EVERITT, A., HARDIKER, P., LITTLEWOOD, J. and MULLENDER, A. (1992) Applied Research
for Better Practice, Macmillan, London.
FINCH, J. (1986) Research and Policy: the uses of qualitative methods in social and educational research,
Falmer, Lewes.
FISHER, M., NEWTON, C. and SAINSBURY, E. (1984) Mental Health Social Work Observed,
George Allen and Unwin, London.
GIBBONS, J.S., Bow, I., BUTLER, J. and POWELL, J. (1979) Clients reactions to task-centred
casework, British Journal of Social Work, 9, 2, pp. 20315.
GUBA, E.G. and LINCOLN, Y. (1981) Effective Evaluation: Improving the Usefulness of Evaluation
Results Through Responsive and Naturalistic Approaches, Jossey Bass, San Fransisco.
GUBA, E.G. and LINCOLN, Y. (1989) Fourth Generation Evaluation, Sage, Newbury Park,
California.
HAMMERSLEY, M. and ATKINSON, P. (1983) Ethnography: Principles in Practice, Tavistock
Publications, London.
HENKEL, M. (1995) Professional Competence and Higher Education, in YELLOLY, M. and
HENKEL, M. (Eds), Learning and Teaching in Social Work: towards reflective practice, Jessica
Kingsley Publications Ltd., London.
HOLMAN, B. (1987) Research From the Underside, British Journal of Social Work, 17, 6, pp. 669
83.
HOWE, D. (1986) Social Workers and their Practice in Welfare Bureaucracies, Gower, Aldershot.
HOWE, D. (1990) The Clients View in Context, in CARTER, P., JEFFS, T. and SMITH, M.
(Eds), Social Work and Social Welfare Yearbook 2, Open University Press, Milton Keynes.
HUGMAN, R. (1988) Ethnography and social care, Social Services Research, 17, 6, pp. 66983.
HUGMAN, R. (1991) Power in the Caring Professions, Macmillan, London.
JORDAN, B. (1978) A comment on Theory and practice in social work, British Journal of Social
Work, 8, pp. 235.
JORDAN, B. (1990) Social Work in an Unjust Society, Harvester Wheatsheaf, Hemel Hempstead.
MCKEGANEY, N.P. and BLOOR, M.J. (1981) On the retrieval of sociological descriptions:
respondent validation and the critical case of ethnomethodology, International Journal of
Sociology and Social Policy, 1, 3, pp. 5869.
MIEs, M. (1983) Towards a Methodology for Feminist Research, in BOWLES, G. and KLEIN, R.
(Eds), Theories of Womens Studies, Routledge and Kegan Paul, London.
MUNN-GIDDINGS, C. (1993) A Different Way of Knowing: social care values, practitioner
research and action research, Educational Action Research, 1, 2, pp. 27585.
OLIVER, M. (1990) The Politics of Disablement, Macmillan, London.
OLIVER, M. (1992) Changing the Social Relations of Research Production, Disability, Handicap
and Society, 7, 2, pp. 10115.
PALEY, J. (1987) Social Work and the Sociology of Knowledge, British Journal of Social Work, 17,
2, pp. 16986.
PARLETT, M. and HAMILTON, D. (1977) Introduction to Illuminative Evaluation, Pacific Soundings,
Cardiff-by-the-Sea, California.
POWELL, J. and GODDARD, A. (1996) Cost and Stakeholder Views: a combined approach to
evaluating services, British Journal of Social Work, 26, 1 (forthcoming).
POWELL, J. and LOVELOCK, R. (1989) The development of community based adult mental
health servicesa research contribution, Bulletin of the Royal College of Psychiatrists, 13, 12,
pp. 6626.
POWELL, J. and LOVELOCK, R. (1992) Changing Patterns of Mental Health Care: A case study in the
development of local services, Avebury, Aldershot.
RAVN, I. (1991) What should guide reality construction?, in STEIER, F. (Ed.), Research and
Reflexivity, Sage, London.
RAYNOR, P. (1984) Evaluation with one eye closed: the empiricist agenda in social work, British
Journal of Social Work, 14, pp. 110.
REASON, P. (Ed.) (1988) Human Inquiry in Action: developments in new paradigm research, Sage,
London.
REES, S. (1991) Achieving Power: practice and policy in social welfare, Allen and Unwin, London.
SAINSBURY, E. (1985) The contribution of client studies to social work practice, Social Work
Research into Practice, Proceedings of the First Annual JCC/BASW Conference, September 1985,
London.
SAINSBURY, E. (1987) Client Studies: their contribution and limitations in influencing social
work practice, British Journal of Social Work, 17, 6, pp. 63544.
SCHON, D.A. (1983) The Reflective Practitioner: how professionals think in action, Temple Smith,
London.
SCHON, D.A. (1987) Educating the Reflective Practitioner, Jossey-Bass, San Francisco.
SHELDON, B. (1978) Theory and practice in social work: a re-examination of a tenuous
relationship, British Journal of Social Work, 8, pp. 122.
SHELDON, B. (1984) Evaluation with one eye closed: the empiricist agenda in social worka
reply to Peter Raynor, British Journal of Social Work, 14, pp. 6357.
SHELDON, B. (1986) Social work effectiveness experiments: review and implications, British
Journal of Social Work, 16, pp. 22342.
SMITH, D. (1987) The limits of positivism in social work research, British Journal of Social Work,
17, pp. 57386.
SMITH, D. (1992) Puritans and paradigms: a comment, Social Work and Social Sciences Review, 3,
pp. 99103.
SMITH, G. and CANTLEY, C. (1985) Assessing Health Care: a study in organizational evaluation, Open
University Press, Milton Keynes.
SMITH, G. and CANTLEY, C. (1988) Pluralistic Evaluation, in LISHMAN, J. (Ed.), Evaluation.
Research Highlights in Social Work 8, 2nd edn, Jessica Kingsley Publishers, London, pp. 11836.
STEIER, F. (Ed.) (1991) Research and Reflexivity, Sage, London.
THYER, B. (1993) Social Work Theory and Practice Research: the approach of logical positivism,
Social Work and Social Sciences Review, 4, 1, pp. 526.
USHER, R. and BRYANT, I. (1989) Adult Education as Theory, Practice and Research: the captive
triangle, Routledge, London.
WALLACE, A. and REES, S. (1988) The Priority of Client Evaluations , in LISHMAN, J. (Ed.),
Evaluation. Research Highlights in Social Work 8, 2nd edn, Jessica Kingsley Publishers, London,
pp. 5871.
WEBB, D. (1991a) Puritans and paradigms: a speculation on the form of new moralities in social
work, Social Work and Social Sciences Review, 2, pp. 14659.
WEBB, D. (1991b) A stranger in the academy: a reply to Lena Dominelli, Social Work and Social
Science Review, 2, pp. 22341.
WORRALL, A. (1990) Offending Women: Female Lawbreakers and the Criminal Justice System, Routledge,
London.
12
Qualitative Approaches to Data Collection
and Analysis: Examinations and Schools
David Scott
With regard to the study of education, it is possible to identify four research paradigms:
positivism, post-positivism, post-modernism and post-structuralism (Denzin and Lincoln,
1994). The first of these, positivism, for so long the dominant influence, can be
characterized in the following way: values can be separated from facts, with the
researchers task designated as uncovering and reporting those facts. They are not
considered to be a part of the research and their presence is not expected to affect the
quality or type of data that are gathered. Positivists further argue that there are no
meaningful differences between natural and social science methodologies. Bhaskar (1989)
develops this argument, though his solution is radically different from other social
theorists in the research tradition under discussion. He suggests that both social scientific
and scientific methodologies should be understood as one: both the dominant naturalist
tradition, positivism, and its anti-naturalist hermeneutical foil rest on an ontology
rendered obsolete by these developments (new collateral theories of philosophy and
ideology). The time is therefore overdue for a sublation of their historic confrontation
(p. ix). Without a solution of this sort, social scientists who introduce reflexive and
interpretative elements into their methodologies are deemed to have acted unscientifically
and therefore without objectivity. The textual account that eventually emerges is
therefore objective, separate from the value positions of the researcher and representative
of an intransitive reality. Researchers working within this tradition would thus seek to
privilege their accounts by ignoring the problem of reflexivity, by denying that these
accounts are essentially constructed, and by minimizing through the use of certain textual
and methodological devices the place of the researcher in that construction.
However, seeking to decontaminate the research experience by separating the
researcher from the object of their study, involves the adoption of a particular value
position, principally that data which are collected in this way effectively and adequately
represent reality. Furthermore, it is difficult to imagine that those decontaminating
devices do in fact separate off the researcher from what is being researched. All types of
research involve selective and thus value-laden interventions of different types during
their conduct. It is in this sense that the data are value-impregnated. The researcher,
tacitly or overtly, has already mapped the area that they are studying, based on experience
and knowledge. But within this, there is a place for change, realignment and modification.
The researcher finds out things during the course of the research which they did not know.
Reflexive practices are therefore considered essential elements within post-positivist,
each school. They were chosen to give a gender balance, where this was appropriate; and
to allow a variety of responses from different ability levels within each school. In four out
of the six schools one of these pupils was observed throughout their timetable during a
full day. Other relevant lessons (assessed practicals in the sciences, oral work in English)
were also observed where I felt that this would contribute to further understanding of
coursework processes. Interviews were conducted with the parents of the sample of
pupils, so that the key issue of parental contributions to coursework assignments could be
investigated. Each of these thirty-six pupils kept a diary over five weeks to record the
extent of the work they were doing in all their subjects outside the classroom. Within
each school a number of pupils were identified as failing to complete coursework
requirements in some or all of their subjects and subsequently interviewed. Finally, all
year 10 pupils in each setting completed a short questionnaire, which asked them to list their
homework (including coursework completed at home) during one full week in the
summer term of the second year of the project.
Choosing Cases
Data collection methods were not predetermined. Research processes were guided
strategically by the developing theory. This is at the heart of what Glaser and Strauss
(1967) call grounded theorizing. A dialectical relationship exists between theory
building and data collection. As fieldwork proceeds, the researchers initial hunches,
hypotheses and conjectures are gradually refined and reformulated, and this acts
progressively to focus analysis and reorganize data collection methods.
Prior to this, there is what Malinowski (1922) calls the foreshadowed problem. This
involves clarifying and developing research ideas before fieldwork begins. Glaser and
Strauss (1967) in opposition to this approach, argue that the researcher should avoid
presuppositions, hypotheses and previous research studies, though they have recently
modified this position. Bulmer (1979) has criticized this tabula rasa approach because it is
difficult to achieve and it ignores the way researchers conceptualize research problems.
Furthermore, all data and data collection methods include theoretical assumptions, the
adoption of which occurs prior to fieldwork (Harris, 1979).
These initial hypotheses and suppositions that informed the initial stages of my research
were influential in guiding the selection of cases. They were:
a) Children in rural schools may have problems with completing history and geography
coursework projects because they may have limited access to primary sources of
information.
b) Parental interventions in coursework processes would be differentially distributed
and were likely to be of greater intensity in middle-class localities.
c) The amount of work students actually do may have gender implications. Coursework
processes which are assessed over the two years of the course may favour the more
persistent and hardworking pupils. Girls are therefore likely to benefit.
d) Coursework processes are of benefit to the taught curriculum, but reduce
examination reliability and comparability. (Fieldnotes, 6.6.1988)
proceeded and as it became clear that departmental processes were of greater significance
than school processes.
Negotiating Access
Gaining access to research settings involves far more than simply being granted permission
to begin research. It is a continuous series of negotiations and renegotiations, with
different personnel at different levels within the organization. Indeed, though access may
be granted by the principal gatekeeper, this may not reflect the interests and wishes of all
members of that organization. Having been granted permission to conduct research at
Lampton Independent School by the headteacher, I was enjoined to:
Go where you want; go into any classroom. We are a completely open school. We
have nothing to hide. Interview who you want. We will find the time for you.
Though I subsequently encountered little opposition, it was clear from one tentative
enquiry that I would not be welcome in her classroom. Having successfully negotiated
access with other members of staff, I chose not to pursue the matter. Gaining initial access
therefore is only the beginning of a process (Woods, 1986) which stretches throughout
the fieldwork and has consequences for both the research methods employed and the
subsequent account that the researcher eventually submits. They therefore, cannot assume
a neutral value-free position, but are fundamentally a part of the written account that is
finally completed.
The initial means of gaining access to the six case-study schools was by letter. Due to
lack of time this was changed to telephoning and three schools were approached only by
this method. There was no correlation between method and success rate. Two schools
that I telephoned, without sending an initial letter, subsequently granted access. Another
did not reply, so I telephoned to ask for an interview with the headteacher, which was
granted. Access was subsequently refused. Once an interview had been arranged, and it was
not necessarily with the headteacher, it was usual for permission to be granted at that
interview. Clearly the decision had been made in those cases before the interview took
place, using the information available in the initial letter. In other cases headteachers
initiated consultation procedures with their staffs. This resulted in successful access in some
cases, but failure in others.
The reasons given for my failure to gain access (in four schools) provide insights into
the social organization and operational state of particular settings. The headteacher of one
of the schools refused entry because of pressure from the unions. This was a single-sex
girls school in an urban area of high unemployment and extensive social deprivation. It
had just gone through a period of sporadic union action, and was threatened with closure
due to falling numbers. My presence as a researcher would have increased the pressure on
participants.
Undergrove School, on the other hand, suffered from none of these disadvantages,
except that teacher unions were taking industrial action in all schools at this time. I
received a prompt letter explaining that: It would take me several hours to explain to you
in detail the reasons, but if I say to you time and other commitments of colleagues, I am
sure you will understand.. The Headteacher of Grove Priory failed to reply to my letter.
The school secretary of Abbotsbury telephoned to say that it was inconvenient and
unnecessary. Teachers may fear outsiders, because they could expose shortcomings in
their professional practice. As Woods (1986) puts it: Teachers may fear disturbance of
their own delicately balanced survival equilibrium. Outsiders can only be perceived as a
threat to this (p. 28). The extent of this threat is determined by the perception of the
headteacher in each of the schools. However, as Wolcott (1973) acknowledges, a school
study can not be completed openly without seeking permission from the headteacher. But
this has repercussions, since the level at which the researcher enters will influence the
conduct of the fieldwork and the type of data that is gathered. Though Simons (1987)
advocates a democratization of research methods where research programmes are
constructed to meet the needs of all the participants in what are inevitably stratified
situations, in the initial stages of gaining access democratic goals cannot be met.
As a researcher, I used a number of familiarity devices to facilitate access. What I said,
what I looked like, how I behavedthe ephemera of rolewere all designed to establish
trust and thus ease the process of entry (cf. Hammersley, 1979; Delamont, 1984). I
deliberately used my extensive experience as a teacher as a device to reassure. I claimed
privileged knowledge of the research setting because I had taught in comprehensive
schools for thirteen years. There are, though, dangers of over-familiarity. My fieldnotes
record such an incident:
I notice two girls sitting outside the headteachers study, reading books. At my last
school pupils sent out of classrooms for misbehaviour were sometimes picked up by
the head and made to work for the rest of the day outside his study. I assume these
two girls have fallen foul of some teacher. (Lorton School for Girls)
Later I am disabused of such a notion:
The Deputy Head tells me that they try to welcome visitors by always having two girls
on duty outside the heads study, whose task it is to welcome visitors and ask them
politely who they would like to see.
Burgess (1984) argues that the degree of familiarity or strangeness found in research
settings has been polarized in some of the literature. He suggests that in his research he
found within the same social setting a series of situations which were both familiar and
strange. Whatever the degree of real familiarity with the six case-study schools where I
did my research, I assumed a persona that was familiar with the mores and codes of school
life. Since gatekeepers operate with expectations about the researchers identity and
intentions (Hammersley and Atkinson, 1983), role-playing that assumes empathetic
understanding of a schools culture is more likely to persuade gatekeepers to grant access.
In the case of Austell School, an independent boys school, I deliberately mentioned the
fact that I had been educated in a fee-paying boarding school. The headteacher asked me
about its headteacher whom he knew. Knowledge of particular settings has to be carefully
and sensitively reached; gaining access therefore requires the adoption of strategies
which are able to penetrate the protective skin which surrounds them.
One device I used to gain entry was the submission of a research brief to the gatekeeper
or gatekeepers. Commentators (cf. Burgess, 1984) have argued against this approach for a
number of reasons. Since it is usual to have only a limited understanding of the complete
research programme at the beginning of the research and since methodology and datacollection operate in a dialectical fashion, the researcher is rarely in a position to provide a
full account of their purposes to the relevant gatekeeper. Furthermore, as Hammersley
and Atkinson (1983) suggest, it may not be prudent to give a full picture because unless
one can build up a trusting relationship with them relatively rapidly, they may refuse
access in a way that they would not do later in the fieldwork (p. 71). On the other hand,
providing minimal information about the research design simply puts in writing what is
said anyway at initial access meetings. The site brief that I used included a clear
specification of the ethical safeguards I was prepared to build into the investigation. In all
but one of the cases they were accepted without argument. In my fieldnotes after my
initial interview with the headteacher of St Thomas School for Girls, I wrote: she
seemed to trust that I wouldnt be insensitive to confidences, that I would build in a
system of anonymity to the reporting. The degree of frankness, even indiscretion, was
high. The headteacher of Lorton School described her deputy as someone who should
have retired years ago. The Headteacher of Carseley High School described his staff as
willing, hard-working, but essentially apolitical; but he did say that there were one or two
members of staff who were very outspoken, and gave an indication of who they were.
In one school, the issue of confidentiality and anonymity assumed a position of central
importance in my initial discussions. Austell is an independent boys school, and the
headteachers fear was that unsubstantiated comparisons would be made between state
and independent schools. In order to gain access, I had to negotiate further ethical
safeguards, ones which I did not initially want to make. This involved agreeing to
submit to you (the headteacher of the school concerned) before publication of any
article or paper that I write, so that you may check for inaccuracies or bias; and so
that we can, if it proves to be necessary, reach a compromise about what should go
in the article. The article will of course not mention your school by name, nor any
of the teachers in it by name; but the above stipulation will apply to any part of that
article which concerns itself with your school.
In order to maintain momentum, I was prepared to relax the principle articulated in the
site brief of the researcher owning the data. In the end, for reasons not connected with the
negotiated process of gaining access, I decided that the school would not be an appropriate
place in which to work. Thus intra- and extra- sampling judgments reflect both the
original research design and the contingencies of agreeing access.
Ethical Concerns
Having successfully negotiated initial access to each of the six case study schools and
having done so with the aid of a site brief which contained certain ethical safeguards, it
was important to put these into effect during the fieldwork. In my fieldnotes, I listed
those areas of the project which involved me in ethical undertakings:
1 Since I am dealing with children as well as adults, I need to set up mechanisms which
allow teachers and students to say what they feel.
2 It is important that I do not misrepresent individual contributions in my final account
of coursework processes.
3 I have a responsibility to protect particular individuals because insensitive handling of
their data may harm them. This goes further than simply agreeing to negotiate the
release of data, because such negotiations are always going to take place on an
unequal basis.
4 I have an obligation to the public: to place in the public domain information I have
gathered about coursework processes.
5 I have an obligation to the school to protect its interests. Anonymity can partly fulfil
this responsibility.
6 I have been commissioned by an external sponsor. They have certain expectations.
7 Finally I have a responsibility to certain ethical principles of conduct; that I should
tell the truth for instance. (Fieldnotes, 5.5.1988)
This list was not able to serve as a working model for a number of reasons. Some of these
obligations are in conflict with each other. Telling lies may be necessary to maintain
confidentiality. In granting anonymity to schools and research participants, I was involved
in a number of small deceptions and evasions. When asked, for instance, by the
headteacher at Lampton about the identity of a particular teacher whose practice he
disapproved of, I declined to answer. Moral obligations are always conditional; that is they
have consequences outside of themselves which cannot be subsumed absolutely under the
aegis of any moral prescription. Given the fluid nature of the principles that researchers
can use to guide them in their endeavours, as a general rule therefore, ownership of the
data resides with the researcher. This can be contrasted with more democratic approaches
to data ownership in which research participants retain rights of veto over publication, and
the narrative that is eventually produced is in effect a negotiated account between
researcher and participants.
There are a number of problems with this. First, because the needs of different
participants may not be known at the beginning of the research process, there are bound
to be initial decisions made about methodology which will structure the type of data that
is collected. Second, there is a temptation on the part of the researcher to present their
negotiated account as a neat and coherent view of reality because they are operating in the
public domain. Third, negotiating with participants (especially with children) can never be
negotiating between equals. The researcher understands the consequences of release and
publication better than the participant. Finally, the manner of negotiation may be
determined by the implicit structures of the institution. Burgess (1984) reminds us that
people respond to the structured situations in which they are located. Negotiation of the
release of data in non-democratic research situations therefore serves to generate more
data, whereas for democratic researchers it serves to construct the reality of the situation.
From the outset, I was determined that my research should be as open as possible.
There was never any intention on my part to engage in secret or covert research, where
participants are unaware their behaviour is the subject of scrutiny (Homan, 1980). Even
when researchers feel it is necessary to adopt covert stances, they may still express
reservations (Bulmer, 1979). The British Sociological Association in 1992 clarified their
position on covert research:
As far as possible sociological research should be based on freely given informed
consent of those studied. This implies a responsibility on the sociologist to explain
as fully as possible, and in terms meaningful to participants, what the research is
about, who is undertaking and financing it, why it is being undertaken and how it is
to be disseminated. (p. 2)
However it has to be acknowledged that all research necessarily entails a hidden element
(Roth, 1962). Since researchers are unlikely to have a clear idea of what exactly they want
to research at the beginning of the investigation, they can only provide participants with a
limited account of their purposes. The giving of consent by participants is always subject
to the power/knowledge differential that exists between participant and researcher. As the
British Sociological Association (1992) acknowledges:
Wherever possible they (researchers) should attempt to anticipate, and to guard
against, consequences for research participants which can be predicted to be
harmful. Members are not absolved from this responsibility by the consent given by
research participants. (p. 2)
Protecting the interests of participants in research projects though, is difficult. Individuals
and locations cannot be disguised completely, though it is easier to deceive outsiders than
insiders. Anonymizing individual responses enabled me to protect participants from
external scrutiny, and in a project such as this, this was important. For instance Sarah, a
fifth year pupil at Tidehill, would not have wanted her remarks about her well-developed
work avoidance techniques to be heard by her teachers:
You can get away with not doing homework. I have got away with not doing a load
of homework for certain subjects. They have a look at my homework diary and
they see that I am getting work. Usually the teachers give you work and I write it in
my book, and then in my homework diary I would add another extra amount; so
they were thinking I was getting enough. So after that if I didnt have any, I would
put in something that I had been doing in lessons.
There are always tensions between confidentiality and portrayal. Burgess (1984) writes:
it is evident that whatever precautions are taken to protect those involved in a field
study, nothing is foolproof (p. 206).
Fieldwork Relations
Though I used a site brief which was read by the key gatekeepers in each school, this did
not prevent a temporary breakdown in relations half way through the initial fieldwork period
in one of them. This hiatus served to underline the need to negotiate and re-negotiate
access at every level of the organization. Indeed, access may need to be negotiated with
representatives from staff bodies, who may not even play a significant part in the research
project.
Having initially been granted access by the headteacher, to whom I had outlined a
programme of action, I was passed across to the deputy head (Curriculum) who
proceeded to organize a schedule of interviews with key personnel in the school. These
were timetabled during free lessons and were located in the library. They included both
staff and pupil interviews. Having assumed that the site brief would at this stage have been
read, I accepted this pattern of working, and decided that I would use these interviews
both as data-collection sessions and as opportunities to negotiate further access to
classrooms and laboratories. I also, as I had done with the other schools, asked pupils to
keep a diary of activities that related to coursework processes. Allport (1942) has
categorized the use of diaries in three ways; as intimate journals, as memoirs and as logs.
Though these three uses are not mutually exclusive, I was interested in their capacity to
document accounts of processes which I could not observe. Coursework completion
involves extra- as well as intra-classroom and school processes. I was not able to be
present at completion processes which occurred outside the school. I was therefore only
able to gain access to them by dint of retrospective interviews and personal diaries.
Having completed a week of interviews which included gaining permission to visit a
number of classes and the promise by my cohort of pupils to keep a diary over a six week
period, I was then taken aback at the beginning of the next week to be told that I could
not continue with my fieldwork, because the unions are hopping mad and I cant afford
to upset them (Deputy Head [Curriculum]). He went on to argue that in my initial
planning interview with him I had not told him I wanted to visit classrooms, and I had not
told him that I wanted to snoop behind teachers backs by getting pupils to spy on
teachers. Considerably shaken by these events, I was on my way out of the school when I
was stopped by a teacher whom I had already interviewed who told me, This has been
brewing for months. He is using you to get the unions on his side. Hes under a lot of
pressure at the moment. Lack of time meant that I was not able to pursue the
micropolitical implications of this, but it does point to the potentiality of research projects
to be hijacked for overt political purposes. There is within any research project a
momentum for exploitation of the researcher by teachers within the school. Indeed in
many cases the researcher is quite happy to accept a degree of exploitation. In this case it
threatened to prevent further fieldwork in the school. I was only able to resume data
collection by going back to the headteacher and re-opening access procedures. In the end I
was able to complete my fieldwork, when the headteacher was persuaded to transfer
gatekeeping duties from the original deputy to another.
Interviews
The primary data collection method that I used during fieldwork was interviewing.
Burgess (1983) offers three reasons for concentrating upon interviewing at the expense of
making further observations. Interviews can allow the researcher access to past events.
They can allow access to situations at which the researcher is not able to be present (see
above). Thirdly, they can allow access to situations where the teacher refuses permission
for the researcher to be present. In a project which sought information about fieldwork
trips, parental contributions to coursework projects, pupil experiences over a two year
period, it was never going to be possible to be present at more than a limited number of
these important events. Interviewing and diary keeping were therefore used as substitutes
to direct observations.
I did not use schedules during the interviews, but relied on the ebb and flow of
conversation with the occasional use of pointers to focus the interview. This technique
more closely corresponds to Stenhouses (1982) notion of a conversation rather than
interrogative questioning. My interviews though, were not unstructured. My specific
time-bound concerns and perceptions of coursework served to structure the interviews
and impose an agenda on them. The way I was perceived by interviewees, the role they
perceived me as playing, acted to give shape to their answers. Finally, data collection
methods operate in a dialectical fashion with data analysis, and are therefore constantly
changing. Different themes and different areas of interest were pinpointed at different
phases of the fieldwork.
Furthermore, interviewer and interviewee do not operate on a level plane. Despite
arguments that the gap between adult and child can be bridged if the researcher adopts
particular roles, it is doubtful whether the differences in age, size and status can be
successfully overridden. It would be false to assert that the interviews I conducted with
pupils were entirely open. Though some of the pupils took me into their confidence and
told me things they would not have told their teachers (see above), in the main they were
only too well aware of the proper relations they should maintain with guests to the
school. Even with teachers, the asymmetrical relationship between interviewer and
interviewee, identified by Ball (1983), acts to structure the type of data that emerges.
This is an inevitable consequence of the outsider seeking to discover meanings and
understandings shared by practitioners within particular institutions.
Observations
I attended a variety of lessons which were relevant to coursework processes, and this
served to supplement data gathered from the interviews. My role in these lessons could
usefully be described as participant-as-observer (Gold, 1958). Gold identifies four types
of observer field roles. The complete participant conducts covert observations of
participants and settings. The participant-as-observer accepts the inevitable
contamination of natural settings as a result of their presence; but develops relationships with
informants and makes no attempt to conceal their purposes. Researchers, conducting
classroom ethnographies anyway, are unable to conceal their role, and thus cannot make
complete participant observations. A third type, identified by Gold, is observer-asparticipant. Here the researcher formalizes their role and sets limits to the amount and
type of contact they have with participants. In classroom situations, the researcher makes
little effort to interact with pupils and teachers, though they may use interview techniques
to gather data about the processes they have just observed. Finally, there is the complete
observer role, in which interaction between researcher and researched is limited to
gaining and sustaining access. In some classroom situations it is difficult for the researcher
to maintain an observer-as-participant role, because teachers and pupils naturally and
without being asked invade the researchers space.
My preferred style corresponded more closely to the participant-as-observer role, as
I listened, watched and took notes. I also at appropriate moments walked round the class
and talked to pupils about their work. Though formal and systematic observation category
systems have been developed (Flanders, 1970; Delamont, 1984) and used (Galton, Simon
and Croll, 1980), commentators (Woods, 1986; Burgess, 1984) have suggested that predefining category instruments limit the data that can be collected, pre-specify the types of
meanings that can be developed, and may easily misrepresent the complex social setting
that is being explored. For these reasons, ethnographers (Fuller, 1984; Woods, 1986)
either use such instruments in limited ways or abandon them altogether. I did not employ
formal and explicit category systems, but simply took detailed notes of observations. This
does not mean that my observations were unstructured. As theory was developed during
the fieldwork, I focused on different aspects of classroom activities. My time-and spacebound presence influenced intra-sampling judgements, which acted to structure
observational data. At Tidehill for instance, some parts of the geography coursework
programme preceded my period of fieldwork. Finally my direct interventions in lessons
(talking to pupils when they were engaged in writing up coursework projects) served to
change the natural setting, and acted as a means of organizing the experiences that I was
seeking to understand.
Theory Development
I have already suggested that data are value-impregnated, and therefore, as Harris (1979)
argues, are never free of the concerns, interests and conceptual frameworks of the
researcher. If this is so, then reflexive processes are essential and would involve
examination of both the research processes by which data were collected (see above), and
as importantly, the epistemological status of those data. It is with the latter that we are
now concerned.
What follows is a description of theory-development in the project. Examination
syllabuses can be read to support the following agenda: examinations with coursework
elements are fair and equitable, not arbitrary, guarantee some measure of predictive
validity, enable reasonable comparisons to be made between students by positing similar
assessment environments, and allow improvements in curriculum strategies which
maximize learning opportunities. Teachers in the case study schools conformed to this
agenda in full, or rejected parts of it. Table 12.1 contrasts belief or non-belief in this
agenda with teachers attitudes towards examination rule-following. Six situational
stances were formulated, and each is briefly described below. This use of typologies is
common in qualitative research (Woods, 1986).
Table 12.1: Typology of attitudes about examinations amongst teachers teaching GCSE
Conformist teachers believe that public examinations with coursework elements are
useful and meaningful devices. Conformists attach great importance to observing rules and
procedures, and are likely to treat texts in readerly ways (Barthes, 1975). They will
accept meaning frames as given and will not seek to discover alternative interpretations.
They are likely to want to adjust their practice in line with such textual readings. Adaptive
stances are similar in that the teacher has confidence in formal examination systems, even
if they do include coursework elements. They are different because such teachers are
prepared to reinterpret examination rules so that they conform to their notion of good
practice. They are likely to treat texts in writerly ways (Barthes, 1975), and to protect
pedagogical strategies if they feel they may conflict with examination rules. Oppositional
stances combine belief in an examination agenda which implies that certain rules have to
be followed with an apparent disregard for those rules in practice. The stance therefore
can be said to be oppositional because it counter-positions belief and rule following.
Examinations, including those with a coursework element, can provide fair, accurate and
predictively valid accounts of pupil aptitudes, but if through non-observance of those rules
other ends may be achieved, an oppositional stance is sustained.
Those teachers adopting ritualistic forms of behaviour attach more importance to
observing rules and procedures than in achieving the purposes for which those rules exist.
This has been referred to as a form of goal displacement (Merton, 1957). Juxtaposing nonbelief with rule following may create tensions at the level of practice and disregard for those
rules. But the textual reading is still writerly (Barthes, 1975), even if the practitioner is
not prepared to translate such a stance into elastic rule-following or disregard for those
rules. Those teachers who do not accept that examinations can be fair and equitable, and
who furthermore believe that they may damage the curriculum, may reinterpret but not
reject the rules that underpin that agenda. This transformative stance combines a sceptical
attitude towards formal examinations with an elastic approach to rule-following. Finally,
there are non-conformist practitioners who are prepared to ignore examination rules.
Since they do not believe in the examination agenda, they do not see the need to follow its
rules anyway.
They are also subject to the mediating effects of the research process. The researcher
offers a time and place specific perspective on that slice of reality which concerns them.
Their positioned account implies a gap between those sets of rationalized perspectives and
the researchers narrative report, in the same way as we have argued here that there is
necessarily a gap between agents accounts of processes and what actually happened.
Conclusion
The central focus of this chapter has been to document how qualitative research methods
may be used to examine school practices. By focusing on the processes of data collection
(foreshadowing the research problem, constructing a research plan, gaining access to
fieldwork sites, negotiating access within those sites, choosing which data collection
methods to adopt, sampling within each case, adopting ethical and thus by implication
epistemological frameworks, adapting the design to fit the burgeoning theory and
analyzing the data), and on epistemological questions concerning those data, it is possible
to show that these practices are essentially social and rooted in relationships between
researcher and researched. This is in contrast to those working within research traditions
which argue that data can be theory-free and independent of the values of the researcher.
Qualitative research practitioners on the other hand, would argue that the adoption of
such a stance involves a profound misunderstanding of the real relationship between data,
methods and theory.
References
ALLPORT, G.W. (1942) The Use of Personal Documents in Psychological Sciences, Bulletin No. 49, New
York: Social Science Research Council.
ARCHER, M.S. (1982) Morphogenesis versus Structuration, British Journal of Sociology, 33, 4, pp.
45583.
BALL, S. (1983) Case study research in education: some notes and problems, in HAMMERSLEY
M. (ed), The Ethnography of Schooling, Driffield: Nafferton.
BARTHES, R. (1975) S/Z, London, Jonathan Cape.
BHASKAR, R. (1989) Reclaiming Reality, London, Verso.
BRITISH SOCIOLOGICAL ASSOCIATION (1992) Statement of Ethical Principles and their application
to Sociological Practice, London, British Sociological Association.
BULMER, M. (1979) Concepts in the analysis of qualitative data, Sociological Review, 27, 4, pp.
65177.
BURGESS, R.G. (1983) Experiencing Comprehensive Education: A Study of Bishop McGregor School,
London, Methuen.
BURGESS, R.G. (1984) In the Field: An Introduction to Field Research, London, Allen and Unwin.
DELAMONT, S. (1984) The old girl network: Reflections on the fieldwork at St Lukes, in
BURGESS R.G. (ed) (1984), The Research Process in Educational Settings: Ten Case Studies, Lewes,
Falmer Press.
DENZIN, N. and LINCOLN, Y. (1994) (eds.) Handbook of Qualitative Research, London, Sage.
DEPARTMENT OF EDUCATION AND SCIENCE (1978) Part 1 School Examinations: Report of the
Steering Committee established to consider proposals for replacing the GCE O level and CSE
examinations by a common system of examining (Waddell Report), London, HMSO.
FLANDERS, N.A. (1970) Analyzing Teaching Behaviour, New York, Addison-Wesley.
FULLER, M. (1984) Dimensions of gender in a school: reinventing the wheel, in BURGESS R.G.
(ed), The Research Process in Educational Settings: Ten Case Studies, Lewes, Falmer Press.
GALTON, M., SIMON, B. and CROLL, P. (1980) Inside the Primary School Classroom, London,
Routledge and Kegan Paul.
GLASER, B.G. and STRAUSS, A.L. (1967) The Discovery of Grounded Theory: Strategies for Qualitative
Research, Chicago, Aldine.
GOLD, R. (1958) Roles in sociological field observation, Social Forces, 36, 3, pp. 21723.
HAMMERSLEY, M. (1979) Data Collection in ethnographic research, block 4, part 3 of course
DE304, Research Methods in Education and the Social Sciences, Milton Keynes, Open University
Press.
HAMMERSLEY, M. (1985) From Ethnography to Theory: A programme and paradigm for case
study research in the Sociology of Education, Sociology, 19, 2, pp. 187211.
HAMMERSLEY, M. and ATKINSON, P. (1983), Ethnography: Principles in Practice, London,
Tavistock.
HAMMERSLEY, M., SCARTH, J. and WEBB, S. (1985) Developing and testing theory: the case of
research on examinations and pupil learning, in BURGESS R.G. (ed), Issues in Educational
Research, London, Falmer Press.
HAMMERSLEY, M. and SCARTH, J. (1986) The Impact of Examinations on Secondary School Teaching,
School of Education, Open University.
HARRIS, K. (1979) Education and Knowledge, London, Routledge and Kegan Paul.
HOMAN, R. (1980) The Ethics of Covert Methods, British Journal of Sociology, 31, 1, pp. 4659.
HONIGMANN, J.J. (1973) Sampling in ethnographic fieldwork, in BURGESS R.G. (ed), Field
Research: A Sourcebook and Field Manual, London, Allen and Unwin.
JOSEPH, K. (1984) Speech to the North of England Conference, Sheffield, January, quoted in
Secondary Examinations Council (1985), Report of Draft Grade Criteria Working Party: English,
London, SEC.
MALINOWSKI, B. (1922) Argonauts of the Western Pacific, London, Routledge and Kegan Paul.
MERTON, R.K. (1957) Social Theory and Social Structure, Chicago, Free Press.
MITCHELL, J.C. (1983) Case and Situation Analysis, The Sociological Review, 1, pp. 187 211.
ROTH, J.A. (1962) Comments on Secret observation, Social Problems, 9, 3, pp. 28384.
SHARP, R. and GREEN, A. (1975) Education and Social Control, London, Routledge and Kegan
Paul.
SHIPMAN, M. (1981) The Limitations of Social Research, London, Longman.
SIMONS, H. (1987) Getting to know schools in a democracy: the politics and process of evaluation, Lewes,
Falmer Press.
STENHOUSE, L. (1982) The conduct, analysis and reporting of case study in educational research
and evaluation, in MCCORMICK R. (ed), Calling Education to Account, London, Heinemann.
WALKER, R. (1974) The conduct of educational case studies, in DOCKWELL B. and
HAMILTON D. (Eds), Rethinking Educational Research, London, Hodder and Stoughton.
WOLCOTT, H. (1973) The Man in the Principals Office: An Ethnography, New York, Holt, Rinehart
and Winston.
WOODS, P. (1986) Inside Schools: Ethnography in Educational Research, London, Routledge and
Kegan Paul.
WOODS, P. (1987) Ethnography at the crossroads: A reply to Hammersley, British Educational
Research Journal, 13, 3.
13
Grounded TheoryIts Basis, Rationale and
Procedures
Dean Bartlett and Sheila Payne
Introduction
Grounded theory is most commonly thought of as a method of qualitative data analysis and
is therefore distinguished from more traditional quantitative methods which, as Berg
(1989) points out, are often given more respect due probably to the exalted view of
science held by the educated public and its tendency to regard science as related to
numbers and implying precision. Qualitative researchers often appear, therefore, to be
suffering from an inferiority complex as they feel it necessary to defend their enterprise,
often by attacking the perceived alternative of quantitative research, referring, for
example to Kaplans famous statement if you can measure it, that aint it! (cited in Berg,
1989; p. 2).
However, grounded theory is much more than simply a method of data analysis and
legitimizing its use by reference solely to the quality/quantity debate does scant justice to
the approach which actually adopts a far more sophisticated and persuasive rationale. In
order to fully understand the rationale behind the grounded theory approach and the
power it confers we must examine more closely the theoretical basis of research in the
social sciences.
The Theoretical Basis of Research
Psychology and other social sciences have conventionally adopted the scientific metatheory of empiricism and have therefore relied almost exclusively on the hypotheticodeductive logic of statistical inference which may be graphically represented as in
Figure 13.1, below.
Starting from a body of theory, research hypotheses are formulated by a process of
deductive inference and are then operationalized in an experimental design where the
appropriate dependent and independent variables are measured and controlled for in a
particular representative sample of the target population.
This formulation of the research process leads to a natural emphasis on the accurate
measurement and recording of valid and reliable empirical data, while the task of the
researcher is crystallized in the host of methodological problems such an endeavour
engenders. As Agar (1986) points out, the questions which need to be addressed in
formulating a research problem are the product of the researchers received view of
science: What is the hypothesis? How do you measure that? How large is your
sample? Did you pre-test the instrument?. According to Levy (1981), psychologists
have become preoccupied with methodological subtleties to the neglect of more
substantive issues concerning the meaning of what researchers do:
The organism we study is a complex system. It is capable of presenting us with
many faces postures, strategies, rule systems or whatever, depending upon what we
do with it.
measurement is not another methodological problem: it is a matter of substance
of the meaningfulness of the assignment of numbers and of their selected
properties to behaviourand that requires some theorizing about behaviour and
not reference to some handbook of methodology. (p. 269)
Similarly in 1953, C.Wright Mills criticized the abstracted empiricism of quantitative
research for being atheoretical (cited in Silverman, 1993) and the critique of positivism
has continued since this time, becoming ever more impassioned.
Denzin (1989) lists five characteristics of the positivistic paradigm which, he claims,
divorce the researcher from the world under study (p. 23) and which assume: (1) that
objective reality can be captured, (2) that the observer can be separated from what is
observed, (3) that observations and generalizations are free from situational and temporal
constraints, (4) that causality is linear and (5) that inquiry is value-free. Because of
concerns about the applicability of these positivistic assumptions to human mental and
social life, and as a result of the loss of meaning in an effort to maintain scientific rigour,
there has, in recent decades been a shift by many psychologists and other social scientists
towards an alternative qualitative and interpretive school of thought.
Silverman (1993) contrasts two schools of thought which he labels positivism and
interpretive social science respectively, the former utilizing concepts of social structure,
facts and quantitative hypothesis-testing while the latter is concerned with social
construction, meanings and qualitative hypothesis-generation. However, other authors
reject the supposed distinction between qualitative and quantitative approaches (e.g.
Dabbs, 1982; Bryman, 1988). In any case, the reactionary nature of the debate does not,
in and of itself, provide a convincing argument for the value of qualitative methods such as
grounded theory because, as Silverman goes on to note, it became increasingly clear that
positivists were made of straw since very few researchers could be found who equated
the social and natural worlds or believed that research was properly theory-free (p. 21).
Similar criticisms were made by Levy (1981) who, in his presidential address to the British
Psychological Society, said that:
we have a simplified view of science, perhaps duefor understandable sociohistorical reasonsto our anxious desire to receive the accolade of being scientific.
Perhaps the amount of scientific progress in psychology is inversely proportional to
the time we spend arguing about it. I believe that science is much more a process of
social exchange than many of us are prepared to admit. We shun the thought that
scientific knowledge may be ideologically flavoured, yet ideologies as an aspect of
the structure of personal and social knowledge, are properly part of psychologys
interests. We prematurely separate data and theory, sometimes calling data fact,
when both data and theory are obviously intertwined in the processing of
experience and ideas. We employ barely defensible notions of objectivity and
subjectivity. We speak of validating rather than discovering the meaning of. And
we are tempted to speak of such things as objectivity, truth, proof and
methodology where I believe we mean to refer to the more human and social
qualities of communicability, generalizability, plausibility and interpretability. (p.
269)
Despite the criticisms of philosophical naivety which have been levelled against
psychologists and other social scientists it is still possible to distinguish between
traditional, positivistic, quantitatively oriented research and research which adopts an
alternative paradigm. Thus there has developed a paradigm dialog (Guba, 1990) in the
social sciences between traditional positivism, on the one hand, and a host of qualitative
and interpretive alternatives that have emerged to challenge it, on the other. As Phillips
(1987) writes, at the meta-level, expressions such as crisis in the social sciences,
uneasy social science, paradigm clash, hotbeds of pluralistic activity, garrulous and
vacuous contributionsand worseare common, and indicate that important issues are
being hotly debated (pp. viiviii).
Alternative paradigms are often referred to broadly as qualitative methods, naturalistic
approaches or any other of several ill-defined collective terms, however, despite the
gamut of alternatives which exist, they do share some common features, as noted by
Marshall and Rossman (1989), for example. They characterized qualitative research as
involving immersion in the everyday life of the chosen setting for the study, valuing
participants perspectives on their worlds and seeking to discover those perspectives,
viewing inquiry as an interactive process between the researcher and the participant,
being primarily descriptive and relying on peoples words as the primary data. While
Marshall and Rossmans characterization of alternative paradigms gives a useful indication
of the nature of qualitative research, it does little to illuminate the underlying
philosophical issues which lie at the heart of the different approaches to social research.
These issues revolve around two, or possibly three, sets of questions: ontological,
epistemological and, according to Guba (1990), methodological. That is, the researcher
needs to know what is knowable, the nature and status of knowledge claims and how to
discover them. In order to highlight the different positions held on each of these issues by
researchers in each of the paradigms and also to avoid the criticism of philosophical
naivety, let us examine briefly recent developments in the philosophy of science.
Positivism adopts a realist ontology, that is it assumes that there exists a reality driven
by immutable natural laws, and the object of science is to discover those laws and thereby
describe the true nature of reality by the process of empirical experimentalism. This
method grapples with the complexity of the natural world by controlling for confounding
variables and it is committed to an objective epistemology; if there is one true world out
there, then it is possible to know about it (discover the truth) by eliminating subjective
experimenter bias.
This represents the scientific world view commonly held by many psychologists but, as
Phillips (1987) has noted, stemming from the work of Popper, Kuhn, Lakatos and others,
a new understanding of the logic of science has been forged in recent decades. Popper
showed that, because empirical support for theoretical propositions relied upon statistical
sampling and inductivist principles, the logic of experimentation must rest on
falsificationism, rather than verificationism. The classic example is that of the white swan.
As it is never possible for someone to observe all swans, they must select a sample of
swans. If all the swans in this sample are white, it does not necessarily follow that all
swans are white; this particular observation has not verified the proposition that all swans
are white, it has merely not refuted or falsified it. Later Hanson (cited in Phillips, 1987)
demonstrated the theory-laden nature of observation claiming that the background
knowledge held by an investigator strongly influences what is observed. Thus positivism
was shown to be fallible in that the accumulation of evidence permits a host of alternative
theories to coexist, but does not compellingly disqualify any.
A further development in the philosophy of science stemmed from the growing respect
for the historiocity of science in the work of Kuhn (1962) for example. Kuhns central thesis
was that the key concepts of a science and the methodological rules to be followed are
dependent upon the particular paradigm within which a scientist works and therefore
when a paradigm shift occurs, the old concepts and rules are no longer valid. Thus, within
any particular science, the knowledge accrued within any particular paradigm at any given
time can not really represent absolute knowledge or truth, as it is necessarily relative
and changes when another paradigm shift occurs. While Kuhns work has been criticized,
largely because of his postulate concerning the incommensurability of different paradigms,
it certainly highlighted the important dynamic quality of ongoing scientific programmes
and this theme was taken up by Lakatos who acknowledged the dynamic nature of science
and proposed that it is guided by a positive heuristic under which it is improved as it
increases its scope to account for new facts.
Thus it can be seen, from our brief excursion into the philosophy of science, that the
psychologists belief in a positivistic ontology and epistemology as the basis of scientific
research is indeed naive as we are now in a post-positivistic era. Post-positivism maintains
that scientists employ exactly the same type of fallible reasoning as other types of
The socio-cognitive model of human functioning draws on theory and data from the
fields of both cognitive and social psychology and assimilates the two areas into a new,
hybridized whole. The paradigm relies on no single theoretical framework, but sociocognitive research shares a set of assumptions about the nature of human beings as social
thinkers, naive scientists and cognitive misers. Social cognition is mentalistic and processoriented, focusing on elements such as attribution, attention, inference, memory and
schemata and, from its origins in both phenomenology and experimental cognitive
psychology, is able to relate the private and subjective life-world of the individual to the
general cognitive processes common to all people.
Given such an account of human functioning and the aims of science to search for
justified belief in knowledge claims, the hypothetico-deductive method of empirical
experimentalism, while useful as part of the tool-kit of the social scientist, does not provide
the sole solution to method in the social sciences. In the words of Lincoln and Guba
(1985), its neglect of the subjective life-world of individuals has produced research with
human respondents that ignores their humanness, a fact that has not only ethical but also
validity implications (p. 27).
Grounded theory adds to the tool kit of the social scientist by allowing the investigator
to enter the life-world of participants own understandings while maintaining the search
for justified belief central to a scientific enterprise. In the words of Corbin and Strauss
(1990),
grounded theorists share a conviction with many other qualitative researchers that
the usual canons of good science should be retained, but require redefinition in
order to fit the realities of qualitative research and the complexities of social
phenomenasignificance, theory-observation compatibility, generalizability,
consistency, reproducibility, precision and verification. (p. 4)
In contradistinction to the emphasis on verification of the hypothetico-deductive system
of logic, grounded theory utilizes induction directly in an attempt to build theory from
observation. In this sense, it is not incompatible with the deductive theory-testing of the
hypothetico-deductive system as depicted in Figure 13.1, but it is truly driven by
observation of the phenomenon under study and turns the chain of logic on its head with
inductivism taking the primary role and deductive inferences following on from this. In
the preface to their seminal work The Discovery of Grounded Theory, Glaser and Strauss
(1967) explained how grounded theory was complementary to the theory testing
prevalent in most social science research and aimed towards the generation of theory:
Attempts to close the gap between theory and research have concentrated
principally on the improvement of methods for testing theory Our book is
directed toward improving social scientists capacities for generating theory
What is required, we believe, is a different perspective on the canons derived from
vigorous quantitative verification on such issues as sampling, coding, reliability,
validity conceptual formulation, construction of hypotheses and presentation of
evidence. We need to develop canons more suited to the discovery of theory.
These guides, along with associated rules of procedure, can help release energies
for theorizing that are now frozen by the undue emphasis on verification
Although our emphasis is on generating theory rather than verifying it, we take
special pains not to divorce these two activities, both necessary to the scientific
enterprise. (p. vii)
Having thus examined how grounded theory fits in with the theoretical underpinnings of
social science research, we are now able to appreciate the usefulness of such an approach
in the post-positivistic scientific endeavour of the search for justified belief in knowledge
claims. We now turn to examine the emergence of grounded theory, not only as a
method of data analysis, but as a method of investigation from the inception of a research
project through decisions about sampling and data collection to analysis and finally the
presentation of results. In order to achieve this, we firstly take a brief look at the
historical development of grounded theory and then examine in more detail some of the
procedures it adopts.
The Discovery and Development of Grounded Theory
Grounded theory was first presented to the research community in Glaser and Strausss
1967 book which described how theory could be constructed from data by the
development of in vivo categories and their properties using the method of constant
comparison (see below). Glaser and Strauss also described the process of theoretical
sampling and how one moves from substantive grounded theory, relating to the particular
context under study, to formal grounded theory which is more general in its scope.
Finally, they discussed how insight on the part of the analyst aids theory construction.
Glaser and Strauss aimed to legitimize and encourage the generation of theory, set out
the general method of comparative analysis which they advocated and finally, to stimulate
others to publish their own methods for generating theory (Glaser and Strauss, 1967).
Interestingly, in a later article written with Juliet Corbin, with whom he has published
much subsequent work, Strauss claims rather different aims of the book, namely that it
aimed to offer a rationale for theory that was grounded, suggest the logic and specific
procedures for generating grounded theories and to legitimate qualitative research in
general (Strauss and Corbin, 1994). In fact, since the publication of their seminal work in
1967, when they collaborated on research in the School of Nursing at the University of
California into American health institutions and related topics such as dying, there has
been something of a split between the two researchers. Stern (1994) reports that even
back in the 1960s and 70s, students of Glaser and Strauss knew the two had quite
different modus operandi, but Glaser only found out when Strauss and Corbins Basics of
Qualitative Research came out in 1990, whereupon Glaser wrote his second solo book on
grounded theory (p. 212), however, the title of a recent (1991) article by Glaser, In
Honor of Anselm Strauss: Collaboration, would suggest that the two were not quite as inimical
as Stern implies.
In any case, following the publication of Discovery, Glaser (1978) produced a volume
entitled Advances in the Methodology of Grounded Theory: Theoretical Sensitivity to give it its
full name, or just Theoretical Sensitivity for short. This volume went beyond the scope of
Discovery by developing ideas concerning the social psychology and creativeness of the
analyst which merely surfaced in the final chapter of their original work. Theoretical
sensitivity is a quality of the analyst which allows him or her to think analytically and
transcend the data to produce theoretically dense grounded theories and will be explained
more fully in the following section. In this volume, Glaser also explains the nature and
process of memoing, theoretical coding, sampling and sorting and the write-up and
generation of formal grounded theory.
Nearly a decade after Glasers solo volume, Strauss (1987) published a solo volume of
his own entitled Qualitative Analysis for Social Scientists which discussed in detail the basic
analytic procedures of grounded theory. In the preface to this book Strauss wrote of
Barney Glaser that he teaches and uses [grounded theory] in research essentially as I do.
There are some differences in his specific teaching tactics and perhaps in his actual
carrying out of research, but the differences are minor (p. xiv).
Since this time, Strauss and Corbin have collaborated on several publications about
grounded theory (e.g. Strauss and Corbin, 1994; Strauss and Corbin 1990; Corbin and
Strauss, 1990) which document the development of the grounded theory approach, outline
its theoretical basis and assumptions and provide clear guidelines on how to actually
perform and evaluate grounded theory research. In the preface to their book on the
subject, Strauss and Corbin (1990) write that:
This is the fourth in a series of books on grounded theory [referring to Glaser and
Strauss (1967), Glaser (1978) and Strauss (1987) as the other three] If you read
and study each of these books, which we strongly advocate, you will find that some
of their terminology and specific recommended procedures are not always
identical Basically, however, all of the books express an identical stance towards
qualitative analysis and suggest the same basic procedures. (p. 8)
It wasnt until the publication of Glasers second solo book in 1992 that the differences
between the two researchers referred to by Stern (1994) became glaringly obvious, with
Glaser suggesting that the Glaserian and Straussian methods should be distinguished and
even given different names. While these differences are significant, it is still possible to
refer to a corpus of information which one may label grounded theory and which is set
out in the following section.
As Turner (1981) noted, the initial formulation of grounded theory by Glaser and
Strauss met with only limited interest, while it took a further two decades before those doing
qualitative research showed much appreciation for the explicit and systematic
conceptualization of grounded theory. Notwithstanding the differences in opinion about
grounded theory between the original authors, along with an increase in popularity comes
the diffusion of a method and, as highlighted by Strauss and Corbin (1994), the diffusion of
grounded theory methodology seems recently to be increasing exponentially in numbers
of studies, types of phenomena studied, geographical spread and disciplines (education,
psychology, nursing and sociology, for example) (p. 277). Strauss and Corbin go on to
note the risks attending such diffusion such as the tendency to label a study as adopting
grounded theory merely because it has become fashionable to do so. This leads to a
bastardization of the original method which compromises its integrity and therefore
potentially weakens it, as evidenced by the title of Sterns (1994) recent article, Eroding
Grounded Theory. Stern claims that one reason for the erosion of grounded theory stems
from what she calls the menace of minus mentoring, that is the learning of grounded
theory from a book or book chapter rather than the face to face training seminars
advocated by Glaser (1978). Thus she ironically concludes her article about grounded
theory with the injunction Get thee to a mentor.
Strauss and Corbin (1994) also go on to highlight a second problem of the rapid
diffusion of grounded theory which derives from its piecemeal adoption and which means
that it may be used inappropriately when other methods may serve the purpose better.
Indeed the piecemeal adoption of different methods is rife in qualitative research as
pointed out by Morse (e.g. 1994) who refers to the practice of muddling methods and
Baker et al. (1992) who suggested that method slurring often occurs between grounded
theory and phenomenology.
A further problem associated with the widespread diffusion of grounded theory stems
from the way it was initially presented in Discovery and relates to its inductive nature.
According to Glaser and Strausss original formulation, the investigator can go out into
the field without having reviewed the existing literature. This is a corollary of the
inductive nature of grounded theory and the desire not to contaminate ones effort to
generate concepts from the data with preconceived concepts that may not really fit The
danger is to force the data in the wrong direction if one is too imbued with concepts from
the literature (Glaser, 1978; p. 31). This is a very attractive proposition for the novice
researcher or post-graduate student as it legitimizes doing the fun bit of research (going
out into the field) without having to first do the boring bit (sitting in the library),
however, Glaser goes on to note that when reading for his research, the analyst should
read for ideas [which] of course, make one theoretically sensitive, andmake the
analyst sensitive to what he may discover in his data (p. 32). There is indeed some
controversy over when is the correct time to review the literature. However, in the
current authors opinion it is necessary to perform a literature review of previously
published material in the area; if this were not performed, how would the researcher
know that his or her project had not already been undertaken? The skill of the grounded
theorist lies in reading around the project in order to become theoretically sensitive,
while at the same time phenomenologically bracketing specific theoretical knowledge
while actually performing the analysis. The necessity to perform at least some degree
of literature review before collecting data is supported by the comments of Strauss and
Corbin (1994) who write:
because of the partly rhetorical purpose of that book (Glaser and Strauss, 1967) and
the authors emphasis on the need for grounded theories, Glaser and Strauss
overplayed the inductive aspects. Correspondingly, they greatly underplayed the
unquestionable fact (and advantage) that trained researchers are theoretically
sensitized Many people still get their conceptions of grounded theory from the
original book, and have missed the later more realistic and balanced modifications
of that books purposeful rhetoric. (p. 277)
Thus, since its original formulation in 1967, grounded theory has evolved and matured
with each successive generation of researchers and there are those who work more or less
closely to its original canons as expounded by Glaser and Strauss. The original grounded
theory derives its theoretical underpinnings from Pragmatism and Symbolic Interactionism
(Corbin and Strauss, 1990), although others have labelled it as an interpretive method
which has phenomenology as its philosophical basis (e.g. Stern, 1994). While it is true
that grounded theory and phenomenology do share several features in common such as
exploiting the rich, thick description of human experience and seeking understanding of a
problem through the participants own words and frames of reference, as Baker et al.
(1992) point out, the two approaches are based on different intellectual assumptions and,
flowing from these, have clear differences in purpose and methodological prescriptions
(p. 1355). The legitimacy of adopting only parts of the grounded theory method, or any
other method for that matter, must ultimately rest upon the justifications put forward by
individual researchers, however perhaps there are a core of procedures and stages of
analysis which define grounded theory as implied by Glaser (1978) who wrote, in
reference to the production of theoretical memos, that if the analyst skips this stage by going
directly from coding to sorting or to writinghe is not doing grounded theory (p. 83).
For this reason the current authors reserve the term grounded theory for those analyses
which closely resemble the original definition of the method in that they include all of the
stages advocated by both Glaser and Strauss with only minor modifications, while analyses
which include only some of the stages are labelled grounded analyses. Alternatively,
Strauss and Corbin (1994) consider that the definitive features of the grounded theory
method consist of the grounding of theory upon data through the process of data-theory
interplay, the making of constant comparisons, the asking of theoretically oriented
questions, theoretical coding and the development of theory, however they go on to
comment that:
no inventor has permanent possession of the inventioncertainly not even of its
nameand furthermore we would not wish to do so. No doubt we will always prefer
the later versions of grounded theory that are closest to or elaborate our own, but a
child once launched is very much subject to a combination of its origins and the
evolving contingencies of life. Can it be otherwise with a methodology? (p. 283)
Having given a brief outline of the history of grounded theory and discussed some of the
issues its subsequent evolution has raised, we go on, in the next section, to examine the
procedural details of how to do grounded theory by describing and explaining a number
of processes through which a grounded theory study progresses. However, before this can
be done, a word of warning is in order. As we have seen, different authors have used
various interpretations of grounded theory in different ways, even to the extent that there
is disagreement between the original authors, and it must be noted, therefore, that the
following guidelines represent the current authors interpretation of the grounded theory
method, based primarily on reading the work of each of its two inventors, Barney Glaser
and Anselm Strauss.
The Procedures of Grounded Theory
The procedures used in grounded theory may be succinctly summarized in the form of a
table such as that which may be seen in Table 13.1. While the table implies a series of
stages, it is not our intention to imply that these stages are discrete, or that they follow
each other in a strict linear sequence. As noted by Glaser (1978) as one moves forward,
one constantly goes back to previous steps (p. 16) and similarly, Strauss (1987) writes
that the essential procedures for discovering, verifying, and formulating a grounded
theoryare in operation all through the research project andgo on in close relationship
to each other, in quick sequence and often simultaneously (p. 23). Obviously some stages
must logically precede others, for example one cannot develop any categories unless one
has gathered some data, however, the table below represents merely a heuristic device
which may be used to describe the family of inter-related procedures used to generate
grounded theory and, while we may talk of stages, progression and previous and
subsequent procedures, the dynamic and iterative nature of grounded theory is hereby
noted. Having made our apologies, we now go on to discuss the first stage of any
grounded theory, that of data collection.
Stage 1.
Data Collection
As previously mentioned, grounded theory derives its theoretical basis from symbolic
interactionism. Symbolic interactionists suggest that individuals order and make sense of
their world according to shared meanings which they develop through a process of social
interaction and which are conveyed by the reflexive use of symbols. Thus, according to
Morse (1992), grounded theory provides a means for eliciting these meanings and for
describing the psychological and social processes that have been developed to assist people
make sense of their world (p. 257). Grounded theory is therefore able to exploit the
richness and thick description of human experience through the participants own words
and frames of reference and, for this reason, the primary source of data in a grounded
theory is the semi-structured interview, although the actual method of analysis can be
used to analyze any form of textual data, such as observations, newspaper reports or diary
entries. Thus with respect to both the source of data and how to actually go about
gathering it, grounded theory offers little advice, other than to suggest one refers to the
ethnographic interview and fieldwork literature in order to follow the usual procedures to
avoid biasing responses and to assure credibility (Corbin and Strauss, 1990).
Stage 2.
Transcription of Data
It is very important to produce full transcriptions of your data in the initial stages of data
collection, however, once analysis has begun on the first few interviews, the researcher
already has the beginnings of an emerging theory. Just as theoretical sampling (see Stage 6
below) guides what kind of data you sample, so your emerging theory may guide exactly
what data should be transcribed. Strauss and Corbin (1990) advise that one should be
selective and transcribe only as much as is necessary, however this merely reformulates
the question of how much to transcribe. There is no point in transcribing reams of data
which is irrelevant, but until enough data has been collected to allow the theory to
emerge, one must transcribe everything as, until one has an idea of the story which the
data tells, one does not have the criteria necessary for selectivity. Finally, Strauss and
Corbin advise that if this is your first grounded study, or you are still quite inexperienced,
then it is wiser to transcribe all the data. Ultimately, however, the decision rests with the
individual researcher who is under an obligation to justify why some bits of data were not
transcribed.
We will just mention briefly the format in which the data is transcribed, which involves
using only the left hand side of the page. This leaves the right hand side of the page free to
write your codes on. Also, it is useful to number the lines of text for future reference and
using double spacing makes the process of reading and coding the transcripts easier.
Finally, because the process of coding involves taking chunks of text and sorting them
according to the categories which are developed, it is necessary either to have several
photocopies of each transcript which are cut up into segments of text at various stages or
to use a computer which is able to search and gather chunks of text for you and then print
them out. The current authors preferred aid to analysis consists of a qualitative analysis
computer software package called The Ethnograph (Seidel et al., 1988). This package
allows one to produce as many soft or hard copies of the data as necessary and greatly
speeds up the mechanics of analysis, however packages such as these cannot do anything
that could not be done with several copies of your data, a pair of scissors and gallons of
glue; they do not do the analysis for you, they merely make the mechanics of analysis
easier and quicker, thereby freeing up your intellectual capacities to concentrate on the
actual analysis itself.
Stage 3.
The Development of Categories
The development of categories lies at the heart of the grounded theory method and begins
as soon as the first interview has been recorded and transcribed. Thus data collection and
analysis go on simultaneously in grounded theory, with the analysis producing the
emergent theory which in turn guides data collection. Categories are developed using the
method known as open coding which consists of reading through the transcripts line by
line and breaking down the data into discrete parts known as chunks or segments. A
segment may consist of an idea, an event, a name or something that represents a particular
phenomenon, but it generally consists of a concept. Thus open coding involves reading
through the data and conceptualizing it into meaningful units of analysis. As for what
constitutes a meaningful unit of analysis, this depends on the research questions you are
asking. In open coding it is important to analyze the data minutely and the current authors
therefore generally take the smallest meaningful unit of analysis, by which we mean a
chunk of text which stands on its own and which is meaningful and therefore interpretable
in isolation from its context. An initial reading through of a transcript may generate
literally hundreds of such chunks, segments, or concepts.
The next stage of open coding consists of categorizing the concepts which means gathering
similar concepts, i.e. concepts which relate to the same phenomenon, together and
labelling this group of concepts. The conceptual categories are at a higher level of
abstraction than the individual concepts, or exemplars, of a category. A quick note about
terminology here: the terms concept and category are often used interchangeably by
grounded theorists as a category is merely the collection of specific ways in which a
concept has appeared in the data.
Coding is an art and is very difficult to learn how to do properly, however as Strauss
(1987) points out, the excellence of the research rests in large part on the excellence of
the coding (p. 27). In open coding it is important to think analytically, that is one must
not merely describe the data or paraphrase it, but one must see with analytic depth what
is there (Strauss and Corbin, 1990; p. 76) and in order to do this, one needs theoretical
sensitivity. Theoretical sensitivity opens up our thinking about the phenomenon we are
studying and is a characteristic of the researcher which consists of having insight into the
data and the capacity to understand and see what is actually there in the data. It enables us
to shake loose the blinkers which we habitually wear and which are composed of
assumptions, experience and immersion in the technical literature.
Paradoxically, our personal and professional experience, as well as a knowledge of the
technical literature, can provide the source of theoretical sensitivity in that they make us
aware of the sort of assumptions which we may hold. We must step back from and
question those assumptions, bracketing them in order to maintain an attitude of scepticism
towards anything which is not directly suggested by the data themselves. It therefore
allows us to formulate theory which is faithful to the reality of the phenomenon under
study. Strauss and Corbin (1990) offer a number of techniques for stimulating and
enhancing theoretical sensitivity which aim to steer your thinking out of the confines of
both the technical literature and personal experience, avoid standard ways of thinking
about phenomena, stimulate the inductive process, focus on the data and not take it for
granted, prevent rushing past diamonds in the rough when examining data, and discover
properties and dimensions in the data. These techniques consist of asking a series of
fundamental questions of the data such as who does it relate to, who is it about, who else
is involved, what is it about, what is going on, how and why is it occurring, when does it
occur, who controls itand a whole series of questions about the particular phenomenon
of the form who?, when?, where?, what?, why?, how? and how much? Questions such as
these are asked of the word, phrase or sentence and it can be an eye-opening and
fascinating experience asking these questions, formulating tentative answers to them and
checking out these provisional answers by referring back to the data.
A second technique for enhancing theoretical sensitivity consists of the making of
comparisons. As has been mentioned, grounded theory is often called the constant
comparative method because one is constantly making comparisons between a particular
concept and other concepts and categories, looking for similarities and differences. This
assists the researcher to guard against bias and produce valid and reliable analyses because
as each category is being developed, it is constantly being verified and refined or else
rejected according to whether or not it is present in the data, how pervasive it is and what
different forms it may take. Thus, as Glaser (1978) points out, categories must earn their
way into the theory by systematic generation from the data. Similarly, by asking questions
of the data and coming up with provisional answers, we are going beyond the data to
produce provisional hypotheses which must then be verified by returning to the data.
Thus our tentative answers are only ever provisional and must be checked out and either
rejected or confirmed by the data, or else revised to fit the reality of the situation under
study.
As well as comparing concrete segments of data with other segments of data, the
researcher must also think abstractly about it and make comparisons against hypothetical
categories and concepts. For example, using the flip-flop technique, one compares a
concept which has emerged from the data with its hypothetical opposite along any
particularly salient dimension. It will be useful here to consider an example piece of text
to see how it is coded. The following piece of text was taken from a study by one of the
current authors (DTB) about minor forms of stress:
I: Right and how do you react when this sort of thing happens?
S: Um, well, basically if I get annoyed I feel like I could tell them to put it right and, um,
if somebody else is not doing their job and they should be doing their job and sorting it
out, for instance we just got to Waterloo station on our way here today and when you
get there, theres no information you can find which is relatively simple and so I just
said to somebody else, I said the way they run this station, you know, sort it out its just
ridiculous.
Now, looking at this piece of data, we can see that the participant mentions someone doing
their job, so we might suggest the category meeting responsibilities. Now, by examining
the data closely and thinking about the data analytically using the questioning and
comparison techniques that have been discussed, one is able to pick out the properties or
attributes of the category and its dimensions. By thinking about this example and looking
at other examples in the same category, we may notice that it appears to have the
attribute of frequency, that is how often are responsibilities met which may appear to vary
along the dimension always-never. Another property of meeting responsibilities may be
extentthat isto what extent are responsibilities met, and this may vary along the
dimension not at all to exceedingly well. This process is known as dimensionalizing the
data and is used to help define the categories in Stage 5 of the analysis where the
provisional properties and dimensions are either confirmed or disconfirmed by their
presence or absence in actual data.
Now, in the flip-flop technique of making comparisons, we compare an incident with
a hypothetical extreme along a salient dimension, so for example in thinking about how
the fact that the station administrators at Waterloo were not meeting their responsibilities
we can see that this made the participant angry and made him want to complain. Now, if
we compare this to a hypothetical case where the responsibility of providing clear and easy
to understand information was at the other end of the dimension of the property extent
and was performed exceedingly well by, for example, an efficient, user-friendly
interactive computer terminal, how would this make the participant feel? Would he still
feel annoyed? What is it about whether or not an individual meets his or her
responsibilities that causes annoyance? Is it a matter of principle, that the participant feels
a moral code is being violated? Or does it only annoy him when it affects him or his plans
directly? I recall that I have another category called planning, so how does this category
relate to the plans the participant has made in catching the train? How would meeting
responsibilities exceedingly well in this case affect him? Would he still want to complain?
Would he congratulate the station staff? What motivates him to complain? We can see
that by making this sort of comparison, we have managed to think deeply and analytically
about the data, discovered some of its salient properties and dimensions, uncovered
possible relationships between other categories and generated further questions which
need verification. Open coding, in which the data is analyzed minutely, enables us to
create conceptually dense theories, while the techniques described to enhance theoretical
sensitivity aid the generation of theoretically dense explanations because they open up our
creative thinking abilities, giving us new avenues to explore and therefore new insight into
the data.
The types of comparisons which are made may be far-out or close-in so that, for example,
in comparing one category with another, we may compare catching a train with catching a
bus. This is a close-in comparison because each activity has much in common and shares many
of the conceptual properties, however, we could also compare catching a train with
swimming at the local swimming pool. This is a far-out comparison because it seems far
removed from the phenomenon under study. Upon initial consideration the making of
such comparisons may appear to be quite ridiculous and a waste of time, but if we think
about the sort of issues and questions such a comparison raises we can see that it is actually
a useful technique to encourage analytic thought, an integral part of theoretical sensitivity.
For example, swimming at the local pool is a leisure activity, but catching a train is not. Or
is it? It can be; what does this depend on? We often feel relaxed while swimming at the
local pool, but not when catching a train. Why is this? When we catch a train we are
bound by the timetable but this is not so when we are swimming; is this anything to do
with it? We have control over when we start or stop swimming, but not over when the
train runs; is this relevant? There exists a literature about control and its relation to stress.
Does it have anything to say about this particular incident? Have any of the concepts it
refers to already emerged from the data? This may be interesting to look at during later
stages of analysis, but for now, I will bracket my knowledge about control theory and
concentrate on what my data says. You can see that the use of such comparisons really
stimulates the train of thought and allows us to go beyond an individual segment of data
and think creatively, whilst always making sure that we return to our data in order to
ground our emerging theory in it.
The techniques described enable the researcher to think analytically rather than
descriptively about the data and help to make the analysis theoretically and conceptually
dense by helping to generate provisional categories and their dimensions and think about
generative questions and the inter-relationships between categories. One can see that
open coding is time-consuming but is also fascinating, triggering off creative, stimulating,
exciting and complex trains of thought. By scrutinizing the document very closely, line by
line or even word by word, and producing concepts which seem to fit the data, but which
are only, as yet, provisional the analyst quickly produces enormous numbers of questions
and amounts of information relating to the data and in order to help us keep track of our
thoughts and the progression of the analysis, we need to use the adjunctive procedures of
memos and diagrams. Memos are simply notes to ourselves and may be code notes,
theoretical notes or operational notes. In open coding we use only code notes which are
memos about concepts and categories and their properties and dimensions. They consist of
a dated memo with a heading to indicate what type of memo it is and to what concepts
and categories it pertains. They can contain short quotes or phrases in the form of
segments of text but are conceptual in nature in that they are not about specific people or
incidents, but about abstractions of incidents, events or happenings to which particular
concrete examples pertain. This is a result of the nature of grounded theory which may be
represented by the concept-indicator model (Strauss, 1987). This model may be seen in
Figure 13.2, below and represents the way in which conceptual coding is guided by a set
of empirical indicators.
The model shows how events or happenings observed in the data serve as empirical
indicators of a concept which the analyst derives from them, initially only provisionally,
but later with more certainty. The indicators are compared to each other and from these
comparisons, those that share similarities are taken to be indicators of an emergent
abstract concept or category, the nature of which is indicated by the nature of those
features which the indicators have in common. The indicators are then compared directly
to the category along with additional indicators which occur in the data and the category is
thereby refined as further properties of it are generated and the code becomes verified and
saturated (see Stage 4). Thus a further question to ask which increases theoretical
sensitivity when doing open coding is what category does this incident indicate? (Glaser,
1978).
Code notes refer, then, to the abstract category, but may contain examples of
empirical indicators. They may also include lists of and questions about the attributes and
dimensions of a particular category and its possible relationship to other categories.
Operational memos on the other hand are simply reminders about things you have to do,
such as look up a particular reference, make a list of codes, sort and order previous memos
or consider a particular sample to interview next, for example. Theoretical memos and
diagrams are discussed more fully below under Stage 7 of the analysis. Having gone
through the basic procedures associated with open coding we now turn to look at the next
stage of analysis, that of the saturation of categories.
Stage 4.
Saturate Categories
Open coding is the predominant type of coding in the initial stages of data collection and
analysis, but may come to a temporary stop when the categories have been saturated.
Glaser and Strauss (1967) and Strauss (1987) called this theoretical saturation and
describe it as the point at which additional analysis no longer contributes to discovering
anything new about the data and thus it is not really a discrete stage of the research or
something which you actually do; it just happens that at some time in the analysis you are
not learning anything new about a category. At this point, new instances of the category
are not coded since this only adds bulk to the coded data and nothing to the theory
(Glaser and Strauss, 1967; p. 111) and the researcher feels confident that he or she is
fully aware of what is meant when any new phenomena encountered are classified into the
category in question (Turner, 1981; p. 236). Glaser and Strauss (1967), in their original
formulation, also used the term theoretical saturation as a threshold criterion which
determined how many groups should be sampled for each theoretical point, stopping
theoretical sampling when data from the different groups did not lead to any development
of a particular category. Saturation will occur at different times for different categories
and is ultimately a decision for the individual researcher. Once a category is saturated, the
researcher may progress on to the next stage of the research.
Stage 5.
Abstract Definitions
This stage of research involves formulating an abstract definition of each of the categories
in terms of their properties, dimensions and subcategories. By this stage, the researcher
will have saturated the categories and constant comparisons will have generated an
accumulated knowledge about each of the categories in the form of a number of code
notes relating to them and their properties and dimensions. At this point the researcher must
formulate an explicit abstract definition of each category by integrating those properties in
terms of the qualities which are being implicitly recognized each time a new incident of
the category is recognized. The process of defining categories is exacting and demanding
but, as pointed out by Turner (1981), often develops a deeper and more precise
understanding of the nature of the phenomenon being studied and may cause the
researcher to decide that theoretical saturation has not been reached and further open
coding or theoretical sampling is therefore required.
Stage 6.
Theoretical Sampling
Theoretical sampling refers to the process of deciding on analytic grounds what data to
collect next and where to collect it by sampling particular groups of people according to
their significance to the development and testing of the emergent theory. Theoretical
sampling is perhaps the least discrete of all stages in the grounded theory research process
as it occurs right from when the first open coding is complete up until the final stages of
filling in the gaps. It refers to the types of questions which are asked of participants, and
the types of participants who are asked, and particular questions or samples of people are
chosen because of their proven theoretical relevance in terms of their repeated presence
or absence which has been noted in the process of constant comparison. They are also
chosen to permit comparison between more or less similar groups which not only
determines the generalizability of the emerging theory, but also increases theoretical and
conceptual density by maximizing the possibility of discovering new relevant categories
and also of finding differences at the dimensional level between properties of a particular
category. Sampling may cease only when one has achieved full theoretical saturation not
only of each of the categories, but of the theory as a whole and this occurs in the final
stage of filling in the gaps.
Stage 7.
Axial Coding
Axial coding refers to the process of making connections between categories by use of a
coding paradigm. The coding paradigm is a heuristic device used to generate questions
and provisional answers and hypotheses about the relationships between categories and
may be represented graphically as in Figure 13.3, below.
The diagram shows a set of possible relationships that may exist between categories or
between categories and their subcategories. The paradigm is therefore also used in
Stage 5 of the grounded theory process, abstracting definitions. It consists of a model of
the conditions, interactions and consequences surrounding the particular phenomenon
under study and has several components. Causal conditions refer to the events or
incidents that lead to the occurrence or development of a phenomenon, while the context
represents both the specific properties of the phenomenon and the specific conditions
under which the sequences of action and interaction it stimulates take place. Intervening
conditions refer to the broad and general conditions which influence the sequences of
action and interaction. Action and interaction refer to the ways in which an individual or
group of individuals respond to the phenomenon in terms of the purposeful, goal-oriented
strategies they adopt to manage or deal with the phenomenon. Finally, the consequences
refer to the outcome of the occurrence of the phenomenon and the action and
interactional strategies that were adopted to deal with it.
Categories are related to each other and to their subcategories by means of the same
techniques that were discussed under Stage 3, the development of categories, namely the
making of constant comparisons and asking of questions. The questions that are asked
derive from the coding paradigm and relate to the potential relationships which it depicts,
thus to return to the example above, we may ask is the category of meeting responsibilities
related to the category of planning, and, if so, how? In this case, the phenomenon is
planning and the collection of information may be an action/interaction strategy in the
context of others having control over the availability of that information. Others meeting
their responsibilities constitutes an intervening condition which leads to the consequence
of the experience of stress when you try to obtain the information. We then return to our
data to see if these hypothesized relationships are borne out by evidence in the form of
incidents and events that support our hypotheses. In cases where the hypothesized
relationships are not supported we ask why not and consider additional factors, such as
extra intervening conditions which may explain the anomaly. In this way we gain depth of
understanding and our emerging theory gains variation and density.
The paradigm is also used to link categories at the dimensional level by examining the
relationships that exist between the various categories at different dimensional locations,
thus we may examine how the occurrence and experience of stress varies according to the
nature and degree of planning and the extent to which others meet their responsibilities to
the individual actors or group of actors. This allows us to make further connections
between various categories at the dimensional level and increase theoretical density and
sophistication by elaborating upon the processes that occur.
These ideas are recorded in the form of theoretical memos, some of which may have
already been accumulated during the process of open coding when certain questions or
possible relationships began to emerge. Theoretical notes suggest relationships and verify
them with pieces of evidence consisting of actual segments of text from the data. These
memos link categories to other categories and their subcategories, explore and describe
the nature of these relationships and the conditions under which they hold and pull other
theoretical notes together as the emerging theory takes form. Some of the ideas that we
come up with are better represented graphically in the form of either flow charts or
conditional matrices and these products of analysis constitute one of the major ways in
which the results of a grounded theory study are communicated (see Miles and Huberman
(1984; 1994) for more details on how to present data graphically).
You will note that in axial coding, as indeed in open coding, one is constantly moving
between inductive and deductive logic, using concrete examples or empirical indicators
from the data to suggest possible categories and their relationships and then verifying
these against other data. Thus to label grounded theory a wholly inductive method is
slightly misleading as it does make use of deductive logic, however, it is inductive at the
level of analysis depicted in Figure 13.1.
As axial coding progresses the researcher moves towards greater and greater
theoretical density and theoretical integration until, at some point, it is time to progress
onto the stage where one actively attempts to integrate the emerged theory.
Stage 8.
Theoretical Integration
Theoretical integration is the actual formulation of a grounded theory in its entirety. It
involves identifying the story line of your research, that is identifying the central
phenomenon of interest and relating it to all of the categories that you have identified and
validating those relationships. At this stage you will have identified and defined the
categories and noted the most important ones which are significant to your storyline.
Theoretical integration is achieved by the process of selective coding. This involves firstly
selecting a core category, that is the one category which identifies what your research is
about. This is achieved by asking questions such as: what are you studying and what are
your findings, and a useful technique is to produce a short (few lines) general descriptive
overview of the research which is then compared to a list of the important categories that
have emerged and the one which is abstract enough to encapsulate your story line is
chosen as the core category. Only one core category must be chosen, which is related to
all the other subsidiary categories that have emerged. This is achieved by use of the coding
paradigm shown in Figure 13.3.
The categories are arranged and rearranged in terms of the paradigm until they seem to
fit the storyline. The ordering of the categories is dependent upon the logic of the story
line and the patterns and relationships that emerged during axial coding. One then begins
to selectively code by hypothesizing relationships between the core category and all the
other categories according to the coding paradigm and then validating these relationships
by going back to the data. Thus, selective coding uses exactly the same techniques as axial
coding, but relates all the categories to the core category and the categories are then
related not only at the conceptual level, but also at the property and dimensional levels as
discussed in the previous section.
Stage 9.
Grounding the Theory
The theory is then grounded by applying it as a whole to the data. This is achieved by
laying it out in the form of theoretically sorted memos and integrative diagrams and then
validating it against the data by individually validating all the statements which support the
individual relationships in the theory and by testing the theory through looking at
individual cases in their entirety. If one comes across cases which do not fit the theory,
one asks why and looks for additional intervening conditions which may explain the
variance, or else adjusts the theory to accommodate such cases. It is unlikely that the
researcher will achieve a perfect fit with every individual case and thus the aim is to
produce the best fit for the highest number of cases; if the analysis has been done well,
there should be a good fit to most cases.
Stage 10.
Filling in the Gaps
Once the theory has been integrated and grounded, thereby satisfying the researcher that
the theoretical framework holds up to scrutiny, the analyst is able to go back to the categories
and fill in any missing detail. This involves identifying those aspects of the theory for
which not much data exists and going back into the field to collect this data and analyze it
in the usual way in order to give the final theory conceptual density and specificity. Once
this stage of analysis has been complete, the researcher may finally write up the study (see
Meloy (1994) for advice on this).
Conclusion
In conclusion, we have seen that grounded theory provides a powerful and robust method
for constructing theory. In contrast with assumptions to the contrary based on a naive
conception of science, it is consistent with the search for justified belief in knowledge
claims and constitutes not only good, but valuable science. It enables researchers to
generate theory which is true to the data and, because of its emphasis on inductive logic,
closes the theory-data gap. Finally, we should like to give two warnings to the potential
grounded theorist. Firstly, grounded theory is incredibly time-consuming and secondly it
is incredibly complex. Those of you who are not daunted by hard work will have no fear
of taking up the grounded theory method; those of you who are should avoid it at all
costs. In relation to its complexity, we may take comfort in the words of its originators,
Glaser and Strauss:
Not everyone can be equally skilled at discovering theory, but neither do they need
to be a genius to generate useful theory.
References
AGAR, M. (1986) Speaking of Ethnography, London, Sage.
BAKER, C.B., WUEST, J. and STERN, P.N. (1992) Method Slurring: The Grounded Theory/
Phenomenology Example, Journal of Advanced Nursing, 17, pp. 13551360.
BENNER, P. and WRUBEL, J. (1989) The Primacy of Caring: Stress and Coping in Health and Illness,
NY, Addison-Wesley.
BERG, B.L. (1989) Qualitative Research Methods For Tthe Social Sciences, London, Allyn and Bacon.
BRYMAN, A. (1988) Quantity and Quality in Social Research, London, Unwin Hyman.
CORBIN, J. and STRAUSS, A. (1990) Grounded Theory Research: Procedures, Canons and
Evaluative Criteria, Qualitative Sociology, 13, 1, pp. 321.
DABBS, J.M. (1982) Making Things Visible, in VAN MAANEN, J. (Ed) Varieties of Qualitative
Research, California, Sage.
DENZIN, N.K. (1989) Interpretive Interactionism, Volume 16, Applied Social Research Methods
Series, London, Sage.
FISKE, S.T. and TAYLOR, S.E. (1984) Social Cognition, NY, Random House.
GLASER, B.G. (1978) Advances in the Methodology of Grounded Theory: Theoretical Sensitivity.
California, Sociology Press.
GLASER, B.G. (1991) In Honor of Anselm Strauss: Collaboration, in MAINES, D.R. (Ed), Social
Organization and Social Process: Essays in Honor of Anselm Strauss, New York, Aldine de Gruyter.
GLASER, B.G. (1992) Basics of Grounded Theory Analysis, California, Sociology Press.
GLASER, B.G. and STRAUSS, A.L. (1967) The Discovery of Grounded Theory: Strategies for Qualitative
Research, London, Weidenfield and Nicolson.
GUBA, E.G. (1990) The Paradigm Dialog, London, Sage.
KUHN, T.S. (1962) The Structure of Scientific Revolutions, Chicago, Chicago University Press.
LEVY, P. (1981) On the relation between method and substance in psychology, Bulletin of The
British Psychological Society, 34, pp. 26570.
LINCOLN, Y.S. and GUBA, E.G. (1985) Naturalistic Inquiry, London, Sage.
MARSHALL, C. and ROSSMAN, G.B. (1989) Designing Qualitative Research, London, Sage.
MELOY, J.M. (1994) Writing the Qualitative Dissertation: Understanding by Doing, Hove, Lawrence
Erlbaum.
MILES, M.B. and HUBERMAN, A.M. (1984) Qualitative Data Analysis: A Sourcebook of New Methods,
Beverly Hills, CA, Sage.
MILES, M.B. and HUBERMAN, A.M. (1994) Qualitative Data Analysis: An Expanded Sourcebook, (2nd
edition), London, Sage.
MORSE, J.M. (1992) Qualitative Health Research, London, Sage.
MORSE, J.M. (Ed). (1994) Critical Issues in Qualitative Research Methods, London, Sage.
PHILLIPS, D.C. (1987) Philosophy, Science and Social Inquiry: Contemporary Methodological Controversies
in Social Science and Related Fields of Research, Oxford, Pergamon Press.
PHILLIPS, D.C. (1990) Postpositivistic Science, in GUBA, E.G. The Paradigm Dialog, London,
Sage.
SEIDEL, J.V., KJOLSETH, R. and SEYMOUR, E. (1988) The Ethnograph, California, Qualis
Research Associates.
SILVERMAN, D. (1993) Interpreting Qualitative Data: Methods for Analyzing Talk, Text and Interaction,
London, Sage.
STERN, P.N. (1994) Eroding Grounded Theory, in MORSE, J.M. (Ed), Critical Issues in
Qualitative Research Methods, London, Sage.
STRAUSS, A. (1987) Qualitative Analysis for Social Scientists, Cambridge, Cambridge University
Press.
STRAUSS, A. and CORBIN, J. (1990) Basics of Qualitative Research: Grounded Theory Procedures and
Techniques, London, Sage.
STRAUSS, A. and CORBIN, J. (1994) Grounded Theory Methodology: An Overview, in
DENZIN, N.K. and LINCOLN, Y.S. Handbook of Qualitative Research, London, Sage.
TURNER, B.A. (1981) Some Practical Aspects of Qualitative Data Analysis: One Way of Organising
the Cognitive Processes Associated With the Generation of Grounded Theory, Quality and
Quantity, 15, pp. 22547.
14
Action Research in Information Systems
David E.Avison
Introduction
Chapter 7 outlines many research methods appropriate to information systems: conceptual
study, mathematical modelling, laboratory experiment, field experiment, surveys, case
studies, phenomenological research/hermeneutics, ethnography, grounded theory,
longitudinal study and action research. Other chapters have looked at some of these in more
detail, for example, grounded research is discussed in Chapter 13 and participant
observation in Chapter 15.
Traditionally, information systems research has been based on laboratory-based
experimentation or field surveys, with much emphasis on statistical analysis. However,
there are only a limited number of factors that can be studied under laboratory conditions
and such an approach does not simulate the real-world environment which is the domain
of information systems. Further, it is difficult, if not impossible, to allocate meaningful
values on people factors, the domain of information systems and any of the social
sciences. The numbers that are produced from such statistical analysis might seem
accurate, however, and have credence that they do not deserve. Such research work
might lead to publications, but is unlikely to make a real contribution.
In this chapter we look at a research approach which I have found particularly useful in
information systems because I am interested in problems in their organizational context.
However, as Trist and Murray (1990) show, it is well suited to research in the social sciences
generally, including inter-disciplinary work. This paper describes action research and then
shows how action research was used in my own research in the description of Multiview,
a framework for developing information systems, and then a particular action research
project as part of this research. Potential advantages and disadvantages of action research
are discussed and some issues related to action research for debate are introduced.
What is Action Research?
In action research, described in Checkland (1981), researchers test and refine principles,
tools, techniques and methodologies to address real-world problems. It is characteristic of
action research that the practitioners as well as the researchers participate in the analysis,
design and implementation processes and contribute as much as the researchers in any
decision-making. Thus there is a synergy between the researchers and practitioners, the
researchers building up theories and modifying them on the basis of practical experience
and the practitioners using and modifying research ideas for solving real-world problems.
Action research is often confused with case study research, but, as Benbasat et al. (1987)
have shown, whereas case study research examines phenomena in its natural setting with
the researcher an independent outsider, in action research the researcher might be a
participant in the implementation of a system and simultaneously evaluate a particular
approach. In action research, the researcher ought to be useful as well as an observer.
The justification for action research is based on the acceptance of the subjective processes
and the possibility that knowledge is socially constructed.
The history of the development of action research has been traced by Warmington
(1980), who attributes the first explicit use of the method as that applied to social
problems by the social psychologist Kurt Lewin in the 1940s (Lewin, 1946). He used
action research in a number of projects, for example, relating to the promotion of lessdesirable cuts of meat in wartime. It was used by Collier (1945) in the area of race
relations. In Warmingtons account, the implicit use of the method by researchers such as
F.W.Taylor and Elton Mayo in solving practical organizational problems predates Lewins
work.
Other versions of the method were used in the 1960s at the Tavistock Institute in
developing a socio-technical view, and that at Lancaster in soft systems and these have
been particularly influential in information systems.
Hult and Lennung (1980) argue that:
Action research simultaneously assists in practical problem-solving and expands
scientific knowledge, as well as enhances the competencies of the respective actors,
being performed collaboratively in an immediate situation, using data feedback in a
cyclical process aiming at an increased understanding of a given social situation,
primarily applicable for the understanding of change processes in social systems and
undertaken within a mutually acceptable ethical framework.
Action research therefore attempts to link theory and practice, thinking and doing,
achieving both practical and research objectives. Gaining this knowledge is seen as an
active process, so that beliefs may be redefined in light of the outcomes. A representation
of reality is not so much desired, but rather is a means for dealing with reality. Action
research is a pragmatic approach which desires to come to terms with the world. This
emphasis on the pragmatic is further reflected by Susman (1983), who states that action
research builds on a learning cycle. His learning cycle is diagnosis, action planning, action
taking, evaluation and the specification of learning. The people involved in action
research, be they called researchers, actors, participants or subjects, begin in a real
situation and return to it at the end of the learning cycle.
This type of learning represents the enhanced understanding of a complex problem.
Information about a particular situation and a particular environment have been obtained,
which gives a contingent value to the truth learned. The researcher expects, however, to
generate knowledge which will further enhance the development of models and theories.
The aim is the understanding of the complex human process rather than a universal
prescriptive truth.
The mutually accepted ethical framework mentioned in the quote from Hult and
Lennung (1980) may cause some concerns. If the goals of the researcher and client differ
drastically, there is tension. As Willcocks and Mason (1987) show, coercive situations are
not uncommon in information systems work. Where these tensions do occur, some
method for satisfying each of their goals needs to be found (Warmington, 1980; Jackson,
1987).
Finally, in the process of learning, an explicit clear conceptual framework which is
acceptable to the researcher and the organizational actors in the action research study
must exist. This is needed so that the explicit lessons will emerge from the research cycle.
In comparing action research to other research methods, three important distinctions
can be brought out:
1 The researcher is actively involved, with expected benefit for both researcher and
organization.
2 Knowledge obtained can be immediately applied. There is not the sense of the
detached observer, but that of an active participant wishing to utilise any lessons
learned based on an explicit clear conceptual framework.
3 The research is a cyclical process linking theory and practice.
The above discussion can be seen within Checklands intellectual context: the organized
use of rational thought (Checkland, 1985). This context is based on a simple model of a
cycle of continuous inquiry where theory interacts with practice.
Depicted in Figure 14.1 is the use of a theory: an intellectual framework of linked ideas
(F), through a methodology or an intervention process (M), to an application area (A). In
going through a cycle, learning can be generated about F, M and A.
The implications can also be seen in comparison to the other methods and the
comparison of two extremes, laboratory experiments (positivist science) to action
research, is shown in Figure 14.2.
Action Research in Information Systems development
I illustrate action research through research concerned with the development of
information systems undertaken by the author with others. The whole research cycle is
Figure 14.2: A comparison of positivist science and action research (source: Susman and
Figure 14.3: Learning cycle of action research (adapted from Avison and Wood-Harper, 1991)
techniques and tools can be made within a loose methodology or framework (4). At least,
that is one of the theoretical bases of Multiview.
As a result of this research, the methodology framework known as Multiview (5) was
refined and defined. This is partly a synthesis of many of the themes discovered in the
research, and includes techniques and tools (6) and their later modification. Multiview is
described in Avison and Wood-Harper (1990 and 1991) An important aspect of this
research has been to use Multiview in a number of problem situations (7) in different
types of organisation. Six such projects are described in Avison and Wood-Harper (1990).
The next stage is evaluation. There were a number of lessons learned from the experience
of using Multiview in various problem situations and these have led to modifying the
Multiview framework. The conclusions about the methodology framework, its use in
various problem situations and of action research in information systems, is carried out in
the last formal stage (8) (see also Schn, 1983) but there is feedback to stages 3, 4 and 5,
so that the whole process is cyclical in nature. This feedback process of theory, action,
reflection, modification of theory and further action is described in Avison (1990).
An Action Research Project
Having looked at a cycle of research which is centred on the action research approach, I
will look at an example of the approach used in a problem situation. This is discussed in much
more detail in Avison and Catchpole (1988) and is one of the six cases discussed in Avison
and Wood-Harper (1990).
The project was carried with the Darlington Health Authority and concerns
information systems provision in the community care sector. It involved the analysis,
design and prototyping of an information system which fulfils some of the local needs of
fourteen nursing and para-medical professions working in the community in a district
health authority, whilst satisfying the statutory requirements of the UK National Health
Service (NHS) Krner steering group for those professions. The work leading to the
implementation of a prototype computer information system was achieved through the
participation of those professionals who would eventually use the system. This prototype
system covers the chiropody and school nursing staff groups. The case study illustrates the
contingency aspect of Multiview, that is, the choice of tools and techniques within the
framework and the adaptation of the structure of the framework itself to suit the needs of
a particular problem situation.
The health care activity of the community health services creates tens of thousands of
health care interactions annually. These relate to a whole range of activities, from the
domiciliary midwifery service to the domiciliary care of a terminally ill elderly patient.
The previous system in the health authority had involved several discrete manual record
systems. There was little comprehensively organized information available and it was
difficult to evaluate how effective the services were. The need for the systematic
collection, collation and analysis of information would be a valuable information base for
planning and organizing health care activities, providing feedback to the professionals
working in the area and fulfil the statistical requirements of the central government in
order to control the National Health Service (the Krner requirements). Data relating to
the total community care services could provide valuable epidemiological and
demographic information.
In current discrete information systems which have operated for many years, data has
been collected on forms relating to the provision of services in the community. Individual
practitioners submit data on their activities to managers, usually on a monthly basis. A
characteristic of this data is that it is of high volume, which means that it is very difficult if
not impossible to process manually into meaningful information. Data collection is often
duplicated by different staff groups. Returns made by staff are frequently stored away for
many years without use being made of them. The recording of this data is often carried out
merely to fulfil statutory requirements or for no good reason at all.
An increasing emphasis is being placed on caring for clients in the community situation
and the potential of community information systems is large. At this particular health
authority the number of beds in the district general hospital is 750. One community
service, school nursing, has 20,000 registered clients. This data could provide valuable
epidemiological and demographic information. The opportunity is presently being lost.
Our remit was to analyze, design and implement a prototype of an appropriate
information system for the community health services of the health district, whilst fulfilling
the related statutory requirements of the NHS Krner steering group.
The purposes of this particular action research project were many. From the
researchers point of view, it was a further opportunity to test the Multiview approach to
information systems development in action so that problems with using the approach
could be looked at and, perhaps, refined Multiview could bed so that it would be more
suitable for other applications in this area. The differences in the description of the
approach found, for example, in Wood-Harper, Antill and Avison (1985) and in Avison
and Wood-Harper (1990) stem from the tuning processes that came from using the
approach in action research projects such as this. From the users point of view, the purpose
was to develop a prototype information system for community health workers in a local
area health authority. From senior managements point of view, it was to fulfil the
requirements of government to provide statistics related to community health.
One of the principles embodied in the Multiview approach is that stakeholders in any
information system ought to have control over their work environment. This is at odds
with more conventional approaches to information systems development which tend to
preclude real participation (as against lip-service participation) of the stakeholders, the
methodology being a straight-jacket preventing them from contributing effectively. User
groups, for example, participated at all stages in the development of the information
system (not only in the design of the human-computer interface, which is now the limit of
even the more enlightened conventional approaches). Another requirement of the
researchers was to enable prototyping in a natural environment (Avison and Wilson,
1991). This facilitates the demonstration of aspects of a potential information system to be
made without any irreversible work being done to the operational system.
The first phase of Multiview is an analysis of human activity systems. Our
familiarization process had two strands. We carried out a survey of computer applications
for the community aspects of the NHS and also talked to the professionals working in
community health for the health authority and went on visits with them to clients. It was
obvious from our survey that there was no existing information system that could be used
off the peg, we would have to develop our own.
The Multiview framework used to develop the system lends itself to participation of
user groups and at all stages of the development of the information system, not only the
design of the human-computer interface. A second and related requirement was to enable
prototyping. This facilitates the demonstration of aspects of a potential system to be made
without any irreversible work being done on the real system. There are a number of
application generators and fourth generation languages which facilitate prototyping. A
quick delivery of a skeletal working system can be made which will test out design principles
of a system with users, who can then be involved participatively in the construction of the
system. A prototype can be used as a tangible starting point for discussions. It may be
possible to use the prototype as a basis for the operational system, it becomes the
operational system after many iterations when users regard it as being satisfactory. A final
requirement of the methodology was to enable the setting-up of the database in such a
way that it could be useful to the various professional groups and managers at the health
authority and fulfil the Krner requirements. As authors of Multiview, we claimed that it
would support these requirements. The claim needed to be tested.
Two staff groups, chosen at a meeting by the representatives of all community care
groups, carried out the initial pilot studies. One nursing group (school nurses who carry
out group session type activities) and one para-medical group (chiropodists as they carry
out face-to-face contact type activities) were particularly suitable as they tested different
parts of the system. The way the system was developed led to modifications of the
Multiview framework. This is almost inevitable if the researchers are using action research
methods: the practitioners have as much influence on decision-making as the researchers
and therefore are likely to change the researchers theories.
We were fortunate to receive the full co-operation from user groups. These included
management grades, such as the director of nursing services for the community, the six
assistant directors of nursing services, the senior clinical medical officers, and district
heads and non-management grades such as health visitors, school nurses, community
midwives, district nurses and dietitians. These formed design groups meeting regularly to
decide on overall strategy as well as detailed designs.
Users were also happy to try new tools and techniques such as portable handheld data
recorders as an alternative to the redesigned forms. The portable data recorders, no
bigger than hand calculators, enable direct data entry into the computer system. This
experiment was not imposed on staffwe as facilitators made the staff groups aware of the
various ways of collecting data and the staff groups chose to try these and the redesigned
forms in our prototypes.
The applications were produced with reports and screens providing information about
the community care service. The prototype was evaluated from the point of view of
effectiveness, that is how the system met objectives and goals, and efficiency, that is some
measure of the mechanical aspects of the system such as the accuracy, timeliness, and
speed of access to information. We also wished to review the training and design process,
essentially reviewing the participative and prototyping approach to information systems
development, and opinions were solicited about the system from the senior managers,
user groups, individual users and clients. The methods used for this purpose included
questionnaires, individual interviews and group discussions. This feedback included
comments on the methods used to develop the application, the training and education
aspects, as well as the information system itself.
The results are discussed in detail in Avison and Catchpole (1988). Comments such as:
as it is the community health professionals who have to use the system, we are in the
best place to help design it; we know what information is important and relevant for
recording; It is good to be involved with developing new system; I feel I have been
able to make a valid input in the designing of a systems which meets our needs; and I
have been made aware of the difficulties in the systems design process were positive.
These comments are not surprising, it is their system, not that of the technologists alone.
They considered the alternatives, contributed to the design and implemented it with the help
of the researchers, who were seen by them partly as facilitators. However, the usefulness
of this type of response is open to debate. There are parallels with the meaningless
questions asked of athletes after winning a race or of a parent following the death of a son
in a car crash: What does it feel like? (Silverman, 1990).
It may be helpful to show the types of lessons and conclusions that have been drawn
from the action research that led to the definition and tuning of the Multiview information
system development framework. These are explored further in Avison (1990) and Avison
and Wood-Harper (1990 and 1991).
Lesson 1A methodology takes time to define and test: The initial formulation of the
Multiview framework, which is described in Wood-Harper, Avison and Antill (1985),
took some years to develop through practice and field work. The description in that
text is developed further in the 1991 text and is being further tuned through action
research projects undertaken since that time.
Lesson 2The traditional model is inappropriate: As evidenced by our field work,
information systems development does not, in practice, exhibit the step-by-step, topdown nature of the traditional waterfall model, sometimes referred to as the systems
development life cycle. None of the action research cases have exactly followed the
Multiview approach espoused in either text. Further, in the real-world cases
undertaken, some phases of the approach were omitted and others were carried out in
a different sequence from that expected.
Lesson 3The political dimension is important: The manipulation of power, that is, the
political dimension, is important in real-world situations. This transcends the rationale
of any methodology. Most of the case studies show decisions being made which were
influenced by considerations beyond those that are implied by the Multiview
framework.
Lesson 4Responsible participation is contingent: A high level of responsible participation,
where appropriate, is a positive ingredient of successful information systems
development. In other case studies, some potential users were apprehensive about the
proposed information system and did not cooperate fully.
Lesson 5the methodology is interpreted by users/analysts: The users/ analysts affect the
perception of the situation and they interpret the methodology. This effect is
highlighted in professional analysts and users, where cultures, education, background,
and so on, will greatly effect the interpretation of the methodology itself and the way
it is used.
Conclusions that might be drawn from the above lessons might include:
Conclusion 1The Multiview methodology is in a continuing state of development:
Information Systems is a comparatively new discipline, the diversity of approaches is
caused to some extent by the background and cultures of their authors and none is allinclusive. The methodologies address a moving target in that the technology, along
with tools and techniques supporting it, develop relentlessly. We therefore regard
Multiview as defined as part of a process of improving information systems practice.
Multiview is an exploration in the development of information systems rather than a
methodology, because the latter term implies a formal, fixed and inflexible approach.
Conclusion 2Defining an information system is contingent: Defining an information system
is contingent on the methodology, the situation and the information systems
development team. In some of the case studies not all of the stages of Multiview were
used because of the situation (in one case there was no computer system developed).
In any situation where an information system might be appropriate, there are factors
such as culture, language and education which have to be taken into consideration.
Sometimes the political and social climate is such that participation is difficult to
achieve. In other situations particular tools and techniques are not appropriate to the
problem situation. The use of an information system methodology in practice is a
hermeneutic process in which the situation is read by the problem solvers (Argyris,
Putman and Smith, 1986).
The above lessons and conclusions are not of the type that might be drawn from more
conventional research. They may seem disappointing to readers expecting some exact
correlation between variables or some other definite conclusion. But I would claim that this
research has made a valuable contribution to information systems research and practice.
Advantages and Problems of Action Research in
Information Systems
Action research allows the researcher great potential to utilize the ideas of users and
change concepts and methods as the work develops. Researchers and subjects co-operate
in solving a real-life problem. It is particularly useful that action research allows work to
take place in its natural setting and it gives the researcher an insight into real-life practical
areas. Feedback from the practical application of techniques can be used to refine and
improve those techniques and their description. Although the results of action research
are of a qualitative nature, they do offer a degree of external validity because the theory
developed can be interpreted and refined by others in other real-world situations.
Action research has proved helpful to the development of Multiview. The weaknesses
in the descriptions of some of the techniques were revealed when using them in one action
research project described in Avison and Wood-Harper (1990). The practicality of using
techniques and tools contingent on a particular problem situation as suggested by
Multiview, can only be revealed by its use in different situations. Major omissions in the
exposition of Multiview were revealed in a number of projects. Some assumptions in
Multiview, such as the users enthusiasm for participation and the low relative importance
of the technical dimension have also been questioned through this experience. The part
played by researchers as active players in the problem situation (not merely impartial
observers), along with users and analysts, has meant that both researchers and
practitioners have together influenced the practice (by implementing change in the
problem situation) and the theory (by changing the Multiview approach).
However, there are disadvantages of action research. The lack of impartiality of the
researcher has led to its rejection by a number of researchers and academic departments.
The lack of scientific discipline in such research makes it difficult for the work to be
assessed for the award of research degrees and for publication in academic journals. A
particular difficulty that researchers have is persuading research funding bodies that this
type of research is as valid and as useful as conventional methods of scientific research.
Further, although the researchers intent is to conduct research while effecting change,
the approach is sometimes branded with the description consultancy and not research.
The open-endedness of such research and the consequent flexibility necessary in writing a
research proposal also provide additional difficulties. Further, a major consequence of the
choice of the action research method is that the research is context-bound as opposed to
context-free. It is difficult to determine the cause of a particular effect, which could be
due to environment (including its subjects), researcher, or methodology. This has been
made explicit in some research, but it can mean that action research produces narrow
learning in its context because each situation is unique and cannot be repeated. This is a major
criticism of action researchit does not produce generalizable learning naturally.
However, there are attempts to reconcile the narrow learning that can come from
action research with the need for generalizable research (Warmington, 1980). The
findings from the Multiview research (discussed in this chapter) has resulted in other
researchers and practitioners applying the Multiview theory in other problem situations,
and by taking into consideration the learning that emerges in the complex process of
developing an information system, generalizing these findings.
Issues
In this chapter, I have provided an overview of action research, describing the action
research cycle in the context of my own research in information systems, then looking at a
particular action research project before considering some of the advantages and
disadvantages of the approach. Readers are warned that it has necessarily been simplistic:
there are many forms of action research (Elden and Chisholm, 1993), although there is a
broadly-shared definition of the approach. In this final section, I wish to raise some issues
which might be used to start a debate about action research. I will restrict my list to just a
few issues, though there are many. My list is not only limited but opinionated, so I trust
other issues will be raised and counter-views will be forthcoming! Questions are asked which
may also be used to start the debate.
1 Is there any alternative? There are a number of research methods appropriate to
information systems. The traditional, empirical research more suited to the natural
sciences, is not so consistently appropriate to the social sciences. I believe that the
less conventional action research approach proved to be appropriate for research in
information systems development and the definition of Multiview. In particular, it
proved appropriate to understanding information systems in organizations and
assessing their impact on people and organizations. However, I leave it to the reader
to decide or, preferably, debate with others whether other methods could have been
equal to the task. Case study research (Yin, 1994) is possibly the most plausible
alternative, and, along with the survey approach, a particularly well-used research
method in information systems.
2 Research or Consultancy? In my view, many people who say they are doing action
research are failing to adhere to its tenets. Typically, the researcher carries out a
consultancy project at an organization and writes it up afterwards as a paper. The
term action research is used because the researcher is involved in the process with
practitioners (and wants the work to be academically respectable), but the term is not
justified. First, the practice is not developed from thinking into the development of a
theory and is not used to test and modify this theory. There is, in short, action but
no research. Multiview was developed from ideas and, although a novel
development, is steeped in the literature (see Figure 14.3). Second, the basis of the
work is not compatible with action research. The researcher is working for and
paid by the organization in order to do a job. The employer expects the work to be
done as he or she wishes. In the Health Authority project discussed above, the managers
and other staff were aware that we were researchers wishing to test and modify our
theories, this has to be established at the beginning. However, through working with
us, practitioners could achieve some improvement of their work situation. You may
debate whether being paid for the work is incompatible with the ideals of research.
3 Differing world views: It is very difficult, if not impossible, to work in situations in
which researchers and practitioners have radically opposing views about the project.
If, for example, the managers wish to develop a computer information system to
dehumanize the workplace or to downsize, I would find it difficult to participate.
In the Health Authority project discussed above, the managers and other staff agreed
with the researchers that an emphasis would be placed on providing information to
and for the community care workers (though it was also providing data for
government). This stance may by regarded as naive, indeed impossible to sustain
absolutely, for it is unlikely that all ethical views will match neatly. Situations will
often be pluralist, and sometimes coercive (where one group dominates another),
and it is often difficult to establish common ground on objectives to be sought.
However, in such situations, particularly coercive ones, I am not sure how action
research can proceed.
Acknowledgments
Multiview was designed with Trevor Wood-Harper and the case study developed with
Paul Catchpole.
References
ARGYRIS, C., PUTMAN, C.R. and SMITH, D. (1986) Action Science, Jossey-Bass, San Francisco.
AVISON, D.E. (1990) A Contingency Framework for Information Systems Development, PhD thesis, Aston
University.
AVISON, D.E. and CATCHPOLE, C.P. (1988) Information Systems for the Community Health
Services, Medical Informatics, 13, 2.
AVISON, D.E. and FITZGERALD, G. (1995) Information Systems Development: Methodologies,
Techniques and Tools, 2nd edition, McGraw-Hill, Maidenhead.
AVISON, D.E. and FITZGERALD, G. (1985) Information Systems development: Current Themes
and Future Directions, Information and Software technology, 30, 8.
AVISON, D.E., FITZGERALD, G. and WOOD-HARPER, A.T. (1988) Information Systems
Development: A Tool-kit is not enough, Computer Journal, 31, 4.
AVISON, D.E., SHAH, H.U., POWELL, R.S. and UPPALL, P.S. (1992) Applying
Methodologies for Information Systems Development, Journal of Information Technology, 7, 2.
AVISON, D.E. and WILSON, D. (1991) Controls for Effective Prototyping, Journal of
Management Systems, SA-58.
AVISON, D.E. and WOOD-HARPER, A.T. (1990) Multiview: An Exploration in Information Systems
Development, McGraw-Hill, Maidenhead.
AVISON, D.E. and WOOD-HARPER, A.T. (1991) Information Systems Development Research:
An Exploration of Ideas in Practice, Computer Journal, 34, 2.
BENBASAT, I., GOLDSTEIN, D. and MEAD, M. (1987) The Case Research Strategy in Studies of
Information Systems, MIS Quarterly, February, 1987.
15
Issues in Participant Observation A Study
of the Practice of Information Systems
Development
Joe Nandhakumar
Introduction
In recent years, there has been increasing recognition of the importance of social and
organizational aspects of information systems. This has provided new insights into the
application of information systems in organizations, and has led to a growing interest in
interpretive approaches to information systems research. The interpretive approaches are
based on rather different assumptions about the nature of the world and the way we can
know it than traditional positivist research. A number of authors such as Boland, 1985;
Zuboff, 1988; Boland and Day, 1989; Orlikowski, 1992; 1993; Jones and Nandhakumar,
1993; Walsham, 1993; Jones, 1994; DeSanctis and Poole, 1994, have adopted
interpretive epistemology to seek to throw some light on the complex process of IS
development practice.
Several different methods of data collection are considered as being suitable for use in
interpretive information systems research. Interviews are perhaps the most widely used
method. Observation may also be carried out directly with the observer being either a
passive or active member of the research setting. Active involvement in which the
observer does not seek to influence the situation more than would be expected from
other participants is described as Participant Observation (Taylor and Bogdan, 1985;
Atkinson and Hammersley, 1994). It may be distinguished from Action Research
(Reason, 1994) in which the researcher seeks deliberately to intervene in the situation,
often by employing specific techniques, in order to achieve a particular outcome.
Rather than attempting to review the suitability of each of these methods, in this
chapter, I seek to identify a range of epistemological, methodological and practical issues
raised by the application of participant observation in interpretive research, since it
appears to be used relatively rarely in information systems research. In order to illustrate
this, I shall first describe my experience in conducting a participant observation study of
the Executive Information Systems (EIS) development process in a large manufacturing
company (LMCa pseudonym).
Mark spent most of the day assessing the performance of the new logon
procedure and then carried out extensive checks to ensure that the EIS was running
correctly. He found by chance that there were errors in the Daily Production
Report and corrected this. He also dealt with a technical problem with one of the EIS
PCs in accessing the external news services which he found to have been caused by
a loose cable
Phil continued writing the program for Personnel-EIS project while David was
on leave
Joe had to revise the Product Development-EIS prototype again as the data
providers wanted to test the system before a presentation to their managers. As a
result Joe could not carry out any further work on the Inventory-EIS prototype as
planned, and it was evident that it would not be ready that week. Joe then spent
time on catching up with office work, sorting out his PC, and learning to use the
new graphical capabilities of the upgraded software
The team members also debated who should carry out the upgrading of the
executive PCs in one of the overseas branches, which was situated in a holiday
resort.
At the start of the research period I attempted to record activities every hour. This was
found to be impractical because of the volume of data to be recorded and the time
available, so I resorted to writing up the log at the end of each day. Occasionally
discussions were specifically arranged to enable me to obtain other participants accounts
of their actions. Data were also sought from other key players such as support staff. Data
on the historical context were also gathered from the staff at LMC and from company
documents.
The study resulted in hundreds of pages of log containing a large amount of data which
was then written up in the form of a case study soon after the field study was completed.
This described the development process in the order that events happened, and recorded
many aspects of the real-life context. The case study was sent to William for comments.
Although analysis and interpretation of the data presented in the case was made during the
period of participant observation the final comprehensive analysis happened after the field
study was completed.
Discussion
The above briefly outlines how I applied participant observation to study the EIS
development at LMC. A number of issues may be identified from the use of participant
observation in this study. This section therefore seeks to examine a range of
epistemological, methodological and practical issues raised by the application of participant
observation in interpretive research.
Through participation in the EIS teams activities I was able to overcome, at least to
some extent, the above barriers to understanding. Thus the meanings of particular terms
and activities in the LMC context could be learned from daily observation of the
behaviour of the team members. For example, the term travelling-man was used by team
members to describe the new divisional vice president who took over responsibility for
the EIS. This was explained as meaning that he was rarely in his office and was therefore
thought unlikely to make as much use of the EIS as his predecessor. As a result, team
members became concerned about a possible loss of senior executive support for the EIS
and several meetings were held to discuss how they should respond to this development.
The term closed-door-meetings was also used in conversations to describe urgent
meetings in LMC, but this also had a wider meaning in that you should keep your
distance from the manager because something was wrong, such as someone was being
made redundant or adverse data on departmental performance was being published. The
actual door slamming was used by managers to convey this message to their staff. When
activities were delayed such as a demonstration of a new EIS project which was repeatedly
cancelled because managers were engaged in closed-door-meetings, its meaning was
understood and accepted by team members. By observing these rituals over time an
understanding of their meanings and their effects on the development process could be
developed.
Through participating in the activities I was able to appreciate for myself the subjective
experience of work such as motivation and time pressure in EIS development and to
understand the ways that team members responded to them. For example, when
restrictions were imposed on the LMC employees overseas travel in response to the
outbreak of the Gulf War, this contributed to a loss of motivation among the team
members. They therefore fulfilled only the minimum requirements and spent most of the
time on reading magazines/software manuals and occasionally engaged in disputes with
other team members (especially with junior members) on development standards.
By getting to know the team members personally I was also able to gain insights into
the reasons why individuals interpreted situations in a particular way. For example, when
William asked Mark to find out executives needs from their secretaries he complained
that they would not talk to the EIS team members and treated them like maintenance
workers. Luke, however, did not share this view. This may have reflected the fact that
Mark had made his way up from the shop floor while Luke was a graduate who had been
recruited from Finance division.
The problem of misrepresentation and non-disclosure was reduced by participant
observation in a number of ways. For example, it enabled me to check team members
statements against observed behaviour and also gave me access to internal memos,
company documents and other staff, which could be compared against observation and
initial interpretations. This was illustrated when the EIS team sought to spread a story that
the suspension of one of the EIS by the data providers division was to hide the adverse stock
performance figures resulting from a major strike. It subsequently transpired, however,
that the suspension was largely due to the inability of the EIS to support the changing
needs of the data providers division. The EIS team had seen this as bad for their image in
the company and had therefore encouraged an alternative story which cast the blame for
the suspension elsewhere. It was not possible, however, to sustain the deception with a
team member for any length of time.
Many of the practices of the EIS team were also confidential and could not have been
discovered without being involved in development work. For example, the EIS team did
not talk about the systems security to anyone outside the department because tight
security was essential to safeguard the confidential reports and win the support of data
providers. In practice, however, security was compromised by many issues such as the
inability of the software to provide the necessary password security and the reluctance of
executive users to use many levels of passwords. Team members therefore covered this
issue up to project an image of a secure EIS. For example, they followed an elaborate
password allocation procedure to satisfy other interested parties such as data providers
and users that the EIS complied with company practices. These informal practices were
not concealed from me, but rather they involved me in them. By comparison they usually
gave a well rehearsed version of the EIS story to outside researchers.
The influence of unstated cultural norms on EIS development could also be observed
through the sustained involvement offered by participant observation. For example, LMC
had a strong organizational hierarchy that prohibited direct contact between EIS team
members and executives. Thus when one of the EIS team members was seen in a Vice
Presidents office, the Systems Executive demanded an explanation from the EIS area
managers. This meant that team members had to guess what executives requirements
might be during the design of EIS and that it was not possible for them to question
executives about the usefulness of EIS applications.
Being an observer as well as a participant, however, I was able to notice and question
aspects of the social routine of which the participants were hardly aware unless
interruptions occurred. For example, expectations of timing and location of many
activities such as report updates and team meetings were embedded in the social context.
The deeper structure of these phenomena was understood only by noticing these activities
through involvement and inquiring about their origins.
Involvement of the researcher
Although the above discussion illustrates that participant observation seems to provide
the kind of access to actors interpretations implied by an interpretive research approach,
this method is often seen as intervening in the research context too much and hence
distorting the data. For example, my participation in the work of the EIS team clearly
meant that I influenced their activities. The most tangible evidence of this was the various
EIS projects in which I played a major development role. I also sought to introduce a
number of new design conventions for EIS projects. Although not all of these were
accepted, some helped to influence other team members practice.
While team members fairly rapidly accepted me as an ordinary colleague, and did not
treat me differently because of my status as a researcher, my presence may still have
affected their interpretations more than normal. My questions about team conventions
and practices may have encouraged members to pay particular attention to these. My
occasional absences to write up the research may also have served to remind them of my
dual status.
My perceptions of the development process were inevitably influenced by my prior
beliefs, values and interests and by my personal involvement in the activities under
investigation. By maintaining contact with my academic supervisor throughout the study,
The interpretive approach believes that interpretations are time and context dependent
and these may shift as meanings are formed, transferred and used (Orlikowski and
Baroudi, 1991). In order to understand how interpretations may change over time, the
researcher therefore needs to be exposed to the situation for an extended period rather
than obtaining occasional snapshots of the process. This implies a longitudinal study. The
context dependence of the phenomena also requires that the researcher gains a thorough
knowledge of the social setting to appreciate its influence (Geertz, 1973).
At LMC, through participating in the developers activities in the company setting over
a period of time I was able to get insights on how processes were changing and also how
they were sustained. For example, the way in which design standards used for EIS
interfaces and working practices were enforced could be observed as well as the way in
which they evolved. The six month research period also allowed the complete
development process for an EIS project to be studied from the initial concept through to
its implementation as a working component of the EIS. As the timing of work on this
project changed frequently in response to other events and pressures on team members
time, continuous participation was necessary to be able to observe the whole process. While
some potentially significant activities may occur too infrequently to be observed in a
period of six months, others may be happening every second. A decision had therefore to
be made about the level of detail in which to record and interpret the development
process. For example, micro-ethnographic techniques such as Discourse Analysis (Stubbs,
1983) could be applied to investigate the second to second process. However, these
techniques may be too detailed to be feasible for a longitudinal study of the processes over
the course of a whole information systems development project. Our experience with
participant observation demonstrated the complexity of dealing with a large amount of
data even from our level of detail. Further, such techniques make it even more difficult to
address the broader historical and social context, which we found to be an important element
in understanding information systems in organizations.
The data collection and interpretation may require the researcher to retain a
commitment to being a researcher while being an employee in the company. I had to
clarify my role in the company in advance because over-commitment in work might have
restricted my ability to collect data. I kept close contact with the academic base and
research supervisor throughout the field work period and also took time off to write up
data. My previous experience in the systems development enabled me to fit comfortably
in the employment role.
Dealing with constraints for data collection and presentation
Data collection was affected by the inevitable selection process during the study. For
example, at LMC the way data were gathered was affected by practical constraints such as
not being able to be in two places at the same time to observe some activities, especially
when the activities were carried out in a foreign branch. Because I could not record much
during office hours, the way data were recorded was affected by memory limitations to
recall events at the end of the day. I therefore had to use many mechanisms to get around
these practical constraints. For example, in order to overcome the practical limitations of
only being able to observe activities in which I was involved, I initiated conversations with
other team members, often during lunch hours or evenings, about their activities. During
formal team meetings the team members also provided a brief description of their
activities and problems. The team members daily task list, a list (informally and openly)
maintained to record outstanding tasks and their execution, was another source of data. In
order to recall events I also maintained a diary to jot down key events and activities when
I paused between tasks, and in conversations, I tried to remember a few key words, such
as boring design, crowded screen and difficult environment, to recall the discussion
later.
There were also ethical dilemmas such as writing about colleagues activities and
confidentiality agreements which limited which items could be recorded explicitly. Any
ambiguity on the companys requirements about the confidentiality was clarified before
starting the field work. This enabled me to work around this constraint from the
beginning to convey the level of details I intended.
Other selection processes also shaped the presentation of the data. For example, the
transcription of the log into the main case study and subsequently into a published version
involved considerable selection. In this process some issues were expanded while others
were filtered out. The way these data were interpreted was also influenced by our prior
assumptions, beliefs, values and interests. The whole process was further influenced by
our personal involvement in the project. The effect of this process is illustrated below.
Log entry:
Luke was again prevented from making progress with upgrading any EIS PCs and
spent most of the day again on reading general magazines and talking with Steve
(another analyst in the Systems Division). This was noted by William who had a lengthy
discussion with him about the situation.
Case Study entry:
Although many EIS PCs had been upgraded with the new software during that
period, Lukes progress in this was constantly affected. Because of the company
convention that development staff should not work on an executives PC while
they were in their office, Luke had to wait until one of the executives was away to
upgrade the software on his PC. He therefore spent his time on other activities and
was fulfilling only the minimum requirements.
Case entry for publication:
Because of the hierarchical norms Luke had to wait until one of the executives
was away to upgrade software. This was noted by William who had a lengthy
discussion with him about the situation.
Interpretation: (discussion on management control and team members autonomy).
CHUA, W.F. (1986) Radical Developments in Accounting Thought, The Accounting Review, 61, 4,
pp. 60132.
DESANCTIS, G. and POOLE, M.S. (1994) Capturing Complexity in Advanced Technology Use:
Adaptive Structuration Theory, Organization Science, 5, 2, pp. 12147.
EASTERBY-SMITH, M., THORPE, R. and LOWE, A. (1991) Management Research: An
Introduction, Sage: London.
GEERTZ, C. (1973) The Interpretation of Culture, Basic Books, New York.
GIDDENS, A. (1984) The Constitution of Society, Polity Press, Cambridge.
GILL, J. and JOHNSON, P. (1991) Research Methods for Managers, Paul Chapman, London.
JONES, M.R. and NANDHAKUMAR, J. (1993) Structured development? A structurational
analysis of the development of an Executive Information System, in AVISON, D. KENDALL,
J.E. and DEGROSS, J.I. (Eds), Human, Organizational, and Social Dimensions of Information
Systems Development, North-Holland, Amsterdam.
JONES, M.R. (1994) Learning the Language of the Market: Information Systems Strategy
Formation in a UK District Health Authority, Accounting, Management and Information
Technologies, 4, 3.
ORLIKOWSKI, W.J. (1992) The Duality of Technology: Rethinking the Concept of Technology
in Organizations, Organization Science, 3, 3, pp. 299326.
ORLIKOWSKI, W.J. (1993) CASE tools as organizational change: Investigating incremental and
radical changes in systems development, MIS Quarterly, 17, 3, pp. 309340.
ORLIKOWSKI, W.J. and BAROUDI, J.J. (1991) Studying Information Technology in
Organizations: Research Approaches and Assumptions, Information Systems Research, 2, 1, pp.
128.
REASON, P. (1994) Three Approaches to Participative Inquiry, in DENZIN, N.K. and
LINCOLN, Y.S. (Eds) Handbook of Qualitative Research, Sage, London, pp. 32439.
STUBBS, M. (1983) Discourse Analysis: The Sociolinguistic Analysis of Natural Language, Blackwell,
Oxford.
TAYLOR, S.J. and BOGDAN, R. (1985) Introduction to Qualitative Research Methods: The research for
meanings, (2nd Edition), John Wiley, New Work.
VAN MAAANEN, J. (1979) The Fact of Fiction in Organizational Ethnography, Administrative
Science Quarterly, 24, pp. 53950.
WALSHAM, G. (1993) Interpreting Information Systems in Organizations, John-Wiley, Chichester.
ZUBOFF, S. (1988) In the Age of the Smart Machine; The Future of Work and Power, Heinemann,
Oxford.
16
Enigma Variations: Uncertainty in Social and
Economic Research
George McKenzie
Introduction
One of the great works on the epistemology and methodology of social and economic
research is Joseph Schumpeters History of Economic Analysis (1954). Although the focus of
his volume is the detail of the nature and structure of economic thought from GrecoRoman times until the 1950s when he died, it is much more than an encyclopedia.
Schumpeter was ringing the warning bells for the increasingly specialized nature of social
and economic research that has characterized the latter half of the twentieth century. He
is keen to emphasize that research is openended, that it is highly subjective, that it is prone
to prejudice and that it generates discourse that may take on a life independent of the
social and economic problems being investigated. He warns that there is the danger that
sophisticated methods can become the tools of those seeking influence rather than
knowledge and understanding. In this chapter, I seek to use the History of Economic Analysis
as a launching pad to discuss recent trends in social and economic research. I shall
highlight some of the comments made by Schumpeter, Descartes, Weber, Marshall, Keynes
and Friedman in so far as they relate to issues concerning research practice. In doing so,
however, my ultimate aim is to focus upon the application of statistical methods in the social
sciences. When Schumpeter died, their use was still in infancy. However, many of the
problems which he discussed are particularly relevant to this area of practice which has
attracted widespread attention during the latter half of the twentieth century. The
question that is central to what follows is: what meaning(s) can we attach to the results
obtained from statistical investigations?
It is significant that Schumpeter first addresses one of the most fundamental
methodological questions: what is a science? Is it possible to apply the methodology of the
physical sciences to the social sciences? He deals with this by defining science in such a
way that it is all-encompassing:
a science is any kind of knowledge that has been the object of conscious efforts to
improve it. (p. 7)
but we can calculate the average outcome and its probability. In statistical analysis, by
contrast, a model is constructed which is designed to represent how a group of people
will behave on average. The model is not exact in the sense that errors are represented in
the model. These errors are assumed to be generated by a probability distribution in much
the same way that the outcomes of the roulette wheel are generated by a probability
distribution. For example, consider the following equation designed to represent the
conjecture that the money supply, denoted Ms, influences total expenditure on goods and
services, denoted Y.
(1)
a0 and a1 are the parameters of the model which the statistician wishes to estimate and is
the random variable designed to capture the influence of all other variables. Such errors
or deviations from the model structure are usually assumed to be generated by a normal
distribution such as that shown in Figure 16.1. The sources or reasons for the errors are
taken to be outside the scope of the investigation. But that does not mean that they are
unimportant. The flatter is the distribution, the weaker the ability of changes in the
money supply to explain changes in expenditure, and conversely.
Although at first glance this procedure appears different from Descartes method it
remains in the same spirit. Instead of focusing upon a unique outcome of the investigation
(i.e. perfect certainty), the emphasis shifts to the certainty in the probability of outcome.
The analysis appears to become richer but it also becomes problematic. If the assumptions
of the modelling procedure are valid, then it can be argued that there is a universal
representation of social and economic behaviour which can be used to predict, and
consequently can be used for, the purpose of formulating policy. Errors will be involved
but they are assumed to be unsystematic. Any systematic errors will be recognized and
corrected. Thus over time, it is claimed that the representation will be correct, on
average.
Two of the most influential economists of the twentieth century, Frank Knight and
John Maynard Keynes were both sceptical of this approach since they believed that the world
was clouded by uncertainty. That is, they believed that it was not possible to assume that
there existed a known, objective probability distribution driving the model. Consequently
it would not be possible to verify or falsify conjectures. For Keynes, there existed no
universal theories which could apply at all times. Rather, there existed many partial
theories which might be valid at some point in time but not for all time. A theory could be
used to inform and enhance understanding but it could not be used for prediction.
Following the Great Depression of the 1930s, Keynes had outlined a framework which he
believed would assist in enabling economic recovery. However, he did not believe that
this model would be valid during periods of rapid expansion and/or inflation. In terms of
equation 1, he would have argued that the parameter a0 was very small during a
depression although it might be quite large during an inflation. In addition, he would have
argued that an additional variable, government expenditure, should be added to the
equation, viz
(2)
According to his theory a2 would be very important during times of recession and/ or
severe depression but relatively small during times of inflation. These views were
particularly apparent following the publication in 1939 by the Dutch economist Jan
Tinbergen, of an econometric model of the UK economy. Tinbergen had sought to base
his investigation upon Keynes influential General Theory. However, Keynes was furious. In
correspondence with Roy Harrod, he wrote
In chemistry and physics and other natural sciences the object of experiment is to
fill in the actual values of the various quantities and factors appearing in an equation
or a formula; and the work when done is once and for all. In economics this is not
the case, and to convert a model into quantitative formula is to destroy its
usefulness as an instrument of thought. Tinbergen endeavours to work out the
variable quantities in a particular case, or perhaps in the average of several
particular cases, and he then suggests that the quantitative formula so obtained has
general validity. Yet in fact, by filling in figures, which one can be quite sure will
not apply next time, so far from increasing the value of his instrument, he has
destroyed it. He works out a figure for it and proposes to predict by using the
result, regardless of the fact that his own investigations clearly show that it is not
constant, in addition to the strong a priori reasons for regarding it as most unlikely
that it can be so.
The point needs emphasizing because the art of thinking in terms of models is quite
difficultlargely because it is an unaccustomedpractice. The pseudoanalogy
with the physical sciences leads directly counter to the habit of mind which is most
important for an economist proper to acquire.
I also want to emphasize strongly the point about economics being a moral
science. I mentioned before that it deals with introspection and with values. I might
have added that it deals with motives, expectations, psychological
uncertainties. One has to be constantly on guard against treating the material as
constant and homogeneous.
Many, indeed most, contemporary statisticians would not go as far as Keynes. However,
they too are seeking to inform about the limits and potential pitfalls of quantitative
modelling. For example, two prominent British econometricians, David Hendry and
Grayham Mizon, follow Weber in the sense that they recognize the subjective origins of
all conjectures. For example, they suggest that whim is a valid basis for conjecture. They
then suggest that competing theories should be compared. This involves the construction
of what is termed an encompassing model, i.e. a model that encompasses all the
characteristics of the competing theories. The performance of each theory is compared
against the general model. However, Hendry and Mizon emphasize that their modelling
method does not yield a true model. Their interpretation is much more limited. They
argue that the best that any statistical model can achieve is to capture a process generating
the data. It is not necessarily to be the true process. Indeed it is highly unlikely to be the
true process. All tests are conditional. Different theories can be compared but at the end
of the day, the acceptance of one structure over another is conditional upon the structure
of the model and the statistical assumptions.
Just as during the Methodenstreit of the nineteenth century social scientists were
propelled into a battle for the victory of one idea over another, so today statistical method
seems to invite the investigator to engage in exercises that will reveal one model to be
superior to another. Within the Hendry/Mizon perspective this either/or approach is
replaced by an and/both approach. It is possible that the best model on the basis of their
subjective criteria could include elements of all competing models. In a sense a new
model is constructed. The Hendry/Mizon approach recognizes that the search for truth is
not a valid objective. However, they believe that procedures designed to seek to identify
order within data series is a valid objective and can inform. However, at the end of the
day we must remain uncertain and sceptical about what our subjective attempts to
structure the data really mean. Let us seek to develop some of the themes relating to
uncertainty in social and economic research in greater detail.
States of Mind
For Keynes, the existence of uncertainty meant that decisions taken in markets could not
be thought of as objective. Consequently, it was not possible to model them as if they
were objective. Consider his views about the operation of financial markets:
the vast majority of those who are concerned with the buying and selling of
securities know almost nothing whatever about what they are doing. They do not
possess even the rudiments of what is required for a valid judgment, and are the
prey of hopes and fears easily aroused by transient events and as easily dispelled. This
is one of the odd characteristics of the capitalist system under which we live,
which, when we are dealing with the real world, is not to be overlooked.
But there is also a further reason why it may often profit the wisest to anticipate
mob psychology rather than the real trend of events, and to ape unreason
prolepticallyso long as the crowd can be relied on to act in a certain way, even if
it be misguided, it will be to the advantage of the better-informed professional to
act in the same waya short period ahead. Apart, moreover, from the calculations
of greater or less ignorance, most people are too timid and too greedy, too
impatient and too nervous about their investments, the fluctuations in the paper
value of which can so easily obliterate the results of so much honest effort, to take
long views or to place even as much reliance as they reasonably might on the
dubieties of the long period; the apparent certainties of the short period, however
deceptive we may suspect them to be, are much more attractive.
Because of this uncertainty, emotional influences would come to dominate and these
could not be modelled in any systematic fashion. His models were conditional upon a
particular emotional state of mind. Thus during an economic depression, people would
appear to be bearish, that is pessimistic and negative. Any optimism about a recovery is
only temporary. Consequently, Keynes argued that direct government intervention in the
form of increased spending would be needed to break the emotional negativism of a
depression. On the other hand, during periods of economic expansion, people would be
bullish, that is, euphoric. And in both states, because of uncertainty, herd behaviour
would come to dominate as people sought to identify in the behaviour of others the true
state of the world. Because of this uncertainty, any attempt to model behaviour over time
would itself be uncertain.
The Lucas Critique
A similar conclusion can arise from another line of reasoning. In 1976 Robert Lucas
argued that if a model is used to formulate policy, people will realize this and take it into
account in their own behaviour.
given that the structure of an econometric model consists of optimal decision
rules of economic agents, and that optimal decision rules vary systematically with
changes in the structure of series relevant to the decision maker, it follows that any
change in policy will systematically alter the structure of econometric models.
In other words, the structure of the economy would be continually changing so as to
render the use of empirical investigations virtually meaningless for policy purposes.
Superficially, it might appear as if Keynes and Lucas are saying the same thing. In point
of fact, their motivations are really quite different. Keynes was arguing that in certain
circumstances government had a role to play in stabilizing the economy. However, the
same set of policies could not be applied universally. Different policies were relevant to
different conditions. Lucas, on the other hand, is attempting to develop a neutrality
theorem. This is one of a large number of neutrality theorems in economics designed to
present a case that government policy will have no effect. For example, in my own area of
research, the behaviour of financial markets, it is noted that financial institutions often
change their behaviour so as to avoid the effects of new regulations and policies. This will
usually involve the design of new financial instruments which are not covered by the
regulations. As a result, the regulators are frequently in the position of having their
objectives thwarted. For example, if a central bank seeks to reduce the money supply so
as to reduce the rate of inflation, banks may seek to create new types of assets that are not
Table 16.1
subject to the control of the central bank. These assets cannot usually be used to carry out
transactions but they can easily be sold to generate funds as and when required. Thus they
are pretty close substitutes for money. As a result of this innovation, monetary policy
becomes less effective. To deal with this, the central bank begins to widen its definition of
money to include the new financial instruments. New data series are published and used
as targets for control. But this simply induces commercial banks to introduce further
innovations.
Verification and Falsification
The issues that we have recognized so far are well-recognized by statisticians under the
guise of Type I and Type II errors. These are explained in detail in any good introductory
statistics textbook. However, they are frequently overlooked in actual research practice.
Hence it is important to outline the nature of these errors. Any statistical test involves a
comparison between two alternative conjectures: the theory or model is true or it is not
true. The case where it is not true is referred to as the null hypothesis. If we knew the
truth there would be no problem. But since we are not privileged in this way, we run the
risk of committing errors. Thus there is the possibility of rejecting a model when it is true
(but we do not know when this is the case) and of accepting a model when it is false (but
we do not know this either). The possibilities are illustrated in Table 16.1. Note that
statisticians, unlike social scientists, are careful never to use the terms verify or falsify.
They use the very conditional terms accept and reject, respectively, so as to reflect the
subjective nature of the investigation.
For the purposes of this paper it is important to examine the detail of the problem.
Readers totally unfamiliar with statistical theory may wish to refer to a good introductory
statistics textbook and skip this section for the moment. As above we consider two
hypotheses, say one where the expected value of a variable, denoted x, is ten and the
other where it is zero (i.e. the null hypothesis). Thus in Figure 16.2, I have drawn two
identically shaped distributions, one centred on the mean of zero (H0) and the other on
the mean of ten (H1). The question is: if we observed a value for x of three or of eight,
how would we know from which distribution it was drawn. The answer is that we do not.
Standard practice involves creating a subjective decision rule. For example let us suppose
that any value of x greater than five is assumed to be drawn from H1 whereas any value
less than five is drawn from H0. The probability of making the two types of errors can
now be identified. The probability of making a Type I error, i.e. the probability that x is
greater than 5 (thereby leading to the acceptance of H1) when H0 is actually true, is equal
to the area denoted by . Similarly, the probability of making a Type II error, i.e. the
probability that x is less than 5 (thereby leading to the acceptance of H0) when H1 is
actually true, is equal to the area denoted by . We can of course increase or decrease the
benchmark criterion from 5 so as to change the ratio of a to but we have no scientific
basis upon which to make this decision. The tools of hypothesis testing do not provide us
with a basis for either falsification or verification.
It is here that the methodology of Karl Popper falls apart. Following Popper, many
economists believe the purpose of research is to falsify. Indeed, he argued that rules
should be established which only admit hypotheses than can be falsifiable (1965, p. 49).
However, Popper avoided discussions of probability and statistics like the plague. As I
discussed in Chapter 2, in a sense, he was applying the logic of Descartes. He was seeking
certainty in falsification. However, for the social scientist, only relatively uninteresting
propositions are capable of falsification. In social sciences, auxiliary hypotheses are added
to the main hypothesis of interest. For example, the errors are normally distributed or the
decision criterion is 5, as in the above example. Hence, any test is conditional upon these
auxiliary statements as well. Since the auxiliary assumptions are frequently untested and
the decision criteria are subjective, there can be no certain falsification.
Predictability
There is yet another variation on this theme. Milton Friedman (1953) has suggested that
the criterion for model selection is its ability to predict and that it does not matter if its
assumptions are unrealistic.
the relevant question to ask about the assumptions of a theory is not whether
they are descriptively realistic, for they never are, but whether they are
sufficiently good approximations for the purpose in hand. And this question can
only be answered only by seeing whether the theory works, which means that it yield
sufficiently accurate predictions.
So far we have acknowledged the fact that the assumptions of a model or theory may be
unrealistic. Because of uncertainty we will never know. Friedman, however, suggests that
we should persevere even if we know with certainty that the assumptions are incorrect.
But this is logically unsound. If we know that the assumptions of our analysis are false,
then we cannot claim any universality for the conclusions derived through a process of
deduction.
Statisticians recognize this problem under the heading spurious correlation. A good
example of this was discussed in a famous tongue-in-cheek paper by Jerzy Neymann
(1952). On the basis of his data set, Neymann showed that there existed a high correlation
between the number of storks per 10,000 women and the number of babies per 10,000
women. He concluded that although there is no evidence of storks actually bringing
babies, there is overwhelming evidence that, by some mysterious process, they influence
the birth rate.
Neymanns results would thus satisfy Friedmans criterion based upon prediction. This
simply shows again the futility of trying to create universal rules to guide our research.
But there are also some other issues lurking in this example. On the basis of his
research, Neymann has identified a question: how do storks influence the birth rate of
humans? Should this question now be the focus of investigation? Or is some lateral
thinking required? Perhaps we are really dealing with a case of reverse causality. Storks do
like to nest on chimney tops. In other words, they are attracted to areas where there are
people. Perhaps it is humans who influence the stork population, not vice-versa. We may
have a high correlation between two variables X and Y. However, this does not enable us
to identify whether X causes Y or Y causes X or indeed if the relationship is pure chance.
Such argumentation may seem ridiculous, but it does highlight the problems associated
with utilizing prediction as the criterion for model selection.
Reflections
When I have taught post-graduate econometrics, I set an applied project relatively early in
the course in order that students may become aware of the sorts of problems that are
involved in applied work, as outlined above. When it becomes clear to the students that
statistical work is highly subjective and hence highly problematic, I seek to dispel their
despair by reading to them the following passage from Alfred Marshalls Principles of
Economics.
If we shut our eyes to realities we may construct an edifice of pure chrystal by
imaginations, that will throw lights on real problems at all like our own. Such
playful excursions are often suggestive in unexpected ways: they afford good
training to the mind: and seem to be productive only of good, so long as their
purpose is clearly understood.
Social science research is really a library of stories. Each story represents the conjectures of
its author (or authors) at a particular moment of time. Those conjectures reflect the authors
history in terms of educational, family and work experiences. These experiences are, in
turn, dependent upon the histories of others. The conjectures may appear as a result of pure
whim, personal experience or derived from the ideas of others. They may relate to a
particular moment in time in which case they cannot form the basis of a universal theory
or explanation of what is going on. The story is based upon the conscious and subconscious
thought processes of its author. It can never be a complete and correct reflection of reality
which can only be viewed incompletely and with uncertainty. The aim of the story is to
inform readers as to how the author interprets the world.
Statistical and econometric practice involves turning the story into a cartoon.
Characters in a cartoon are representations of the characters in the narrative (or theory).
Simplification is introduced to highlight what the author believes to be the most important
issues and points adjudged to be unimportant are eliminated. For example, a cartoon is twodimensional, whereas reality is three-dimensional. Similarly, in statistical analysis the main
elements of the conjecture form the equation to be estimated (i.e. the cartoon).
Everything is assumed to be random and relegated to an error term, and like the narrative
upon which it is based, the cartoon is designed to inform about the artists (i.e.
researchers) thoughts. People will interpret the cartoon from their own perspective.
Perhaps their thoughts diverge from the authors intent, therefore, discussions take place,
workshops and conferences are set up. These enable researchers to inform each other
about their different stories and cartoons.
Research in the social sciences is indeed an enigma. The most optimistic of us believe
that the world can be made a happier and more enjoyable place not only for ourselves but
for others. Indeed these were the objectives of the Enlightenment. But as we seek to
achieve these objectives, we run the risk that we become more interested in the process
of drawing the cartoon than in the story that it was originally designed to represent.
Although workshops and conferences are designed to inform, they also provide the
incentive for participants to refine their language and cartoon representations to such an
extent that their meaning is only understood by themselves.
There is the danger that the cartoon becomes the reality rather than reality itself.
People who are not members of the club are unable to comprehend. This asymmetry in
understanding then provides a platform for the club to exercise control over the
uninformed, and because the language of the club is perceived by the outsiders to have a
higher, scientific status, the club is accorded prestige and trust. Councils of Wise Men are
established to inform elected politicians. The elected politicians may then seek to impose
the cartoon on reality. In a sense we have developed a generalization of Lucas critique,
discussed earlier. Research is part of a social process that feeds back on itself in complex
ways that the researcher must be consistently on guard to identify.
It is appropriate to end our discussion of uncertainty where we began. As Schumpeter
wrote,
a science is any kind of knowledge that has been the object of conscious efforts to
improve it. (p. 7)
This definition does not seek to privilege one particular research practice or methodology
over any other. Nor does it seek to exclude. Provided that we are prepared to accept the
universality of the uncertainty that conditions our work, then to my mind, this statement
represents an optimistic platform from which to launch our search for knowledge and
understanding. Uncertainty conditions our work in many ways: the practices which we
have inherited from the past, the environments in which we have lived, the nature of our
education and our perceptions of the problems faced in the world about us.
Some might claim that such reflexivity is a recipe for disaster since it can only
encourage inaction. I will conclude this chapter by arguing the contrary position and
suggesting that as researchers we could do well to adopt the following three principles:
1 No research practice or research methodology should view itself as superior to any
other. If progress is to be made in our uncertain world, then it is necessary to
recognize that there can only be a plurality of understandings and consequently no
such thing as a universal truth.
2 If one is prepared to adopt the first principle, then the logical implication is that any
research project should investigate the use of different research methods and
methodologies. For example, in an attempt to understand the causes and effects of
unemployment, we might seek to use the methods of both the economist and the
grounded theorist. The former conjectures a model with a particular structure and
then investigates whether that model captures the data generation process. The latter
seeks to develop a theory from the ground up by becoming involved with the
subjects of investigation, in this example, the unemployed. With high levels of
unemployment throughout the industrial world, it would be folly to restrict lines of
investigation.
3 This final principle follows from the previous two. If they are accepted, then a
logical implication is that we should seek to promote inter-disciplinary research.
Despite the good will that develops in common room and staff club discussions
concerning this objective, actual implementation has become more rather than less
difficult. In the 1980s and 1990s, universities throughout the world have been
suffering from funding cut backs. Departments must battle with university
management and with each other for adequate funding. The end result has been a
firming up rather than a weakening of disciplinary walls.
What is required before we can effectively implement these principles is a research
project into the nature of the discourses that characterize social and economic research.
Such a project might include in its agenda the following questions:
1 What are the rules of the discourse. That is, what practices do they privilege
and what practices do they exclude?
2 What rules decide when a new proposition or method is to be accepted
within the discourse?
3 What rules constrain the issues to be discussed. What topics are excluded?
4 How are the internal rules of a discourse repackaged so as to achieve
influence, control over others and perhaps elimination of others who do not
have allegiance to the discourse.
Notes on Contributors
Index
academic discipline 83
Academy 912, 13, 17, 19
action research 862, 189202, 202
Agar, M. 166
alienated knowledge 111
Allport, G.W. 158
American Census Bureau 129
anthropology 85, 92
anti-monopoly legislation 15
Antill, L. 195, 197
Archer, M.S. 163
Archimedes 28
architecture 26
Aristotle 10, 1114, 19, 41, 87, 170
Arrow, K. 53
assumptions 394, 50, 706, 783, 99
Atkinson, P. 155
Audit Commission 107
Avison, D. 8395, 106, 189202
axial coding 1843
Backhouse, J. 90
Bacon, F. 13, 14, 15, 16
Bailey, R. 107
Baker, C.B. 174
Ball, S. 160
Baroudi, J.J. 84, 85, 91
Barthes, R. 34
Bartlett, D. 16695
Baudelaire, C. 16
Baudrillard, J. 24
Becker, G. 673, 7079
Begg, D. 68
belief 11
Benbasat, I. 190
229
230 INDEX
Cicero 12
civil servants 62
coding 17894
cognitive psychological model 97
cognitivism 171
Collier, J. 190
commitments 97
common sense 14
community care 137, 142
comparative advantage 74
comparisons 1797
compensation principle 77
competition 15
computer science 84, 89, 90, 118
Comte, A. 15
concept-indicator model 182
conceptual study 86, 87
conditioning 125, 126
confidentiality 1574, 204, 20715
conformity 161
conspiracy theory 24
constant comparative method 101
constitutional political economy 54
consultancy 200
contextual issues 32
control theory 181
Corbin, J. 171, 17282, 1776
Corner, J. 103
cost-benefit analysis 78
coursework 1518
covert research 157
Cox, D.R. 121
criminal justice system 110
critical epistemology 85
critical theory 145
cultural anthropology 92
culture, definitions 92
Curnock, K. 144
Darlington Health Authority 194
data analysis 87, 89, 98, 14972
grounded theory 1007, 17694
statistics 11836
data collection 14972
data triangulation 102
death 100
deconstructivism 19, 35, 47
deduction
logic 58
methods 16
reason 1315, 66
definitions 1011
Denzin, N.K. 167
Department of Health 103
Derrida, J. 18
Descartes, R. 9, 1316, 19, 21332
deductive reasoning 66
rationality 58
deviance 19, 44
diagnosis 127, 128
disciplinary power 44
disciplines 806
discourse analysis 209
discrimination 20
dissonance 11219
diversity 91
division of labour 68, 718, 77
divorce 761
domination 19
Dornbusch, R. 68
Downs syndrome 128
dualism 416, 46, 48, 171
Eco, U. 2427, 31, 361
ecological correlation 128
econometrics 2165
economics 5165, 6679, 21332
economies of scale 71
Efron, B. 125
egoism 63
Electronic Revolution 8, 18
elites 10, 97, 98
emotion 79
empiricism 25, 26, 31
approach 14
investigation 58
knowledge 27
model 141
science 1665, 1697
engineering 84, 141
English Revolution 13
enigma variations 21332
Enlightenment 9, 1215, 17, 21
assumptions 50
INDEX 231
epistemology 42
knowledge-making 383, 40
model 141
epistemology 9, 12, 20
conspiracy theory 24
Enlightenment 42
feminism 4346, 108
interpretivism 10916, 202
knowledge-making 383
modernism 41
participant observation 20513
philosophy 54
pluralism 986
positivism 84
postmodernism 26
practice 106, 107
randomized control trials 104
rationalism 584
science 1697
social work 110, 113, 135, 1374
uncertainty 213
equality 20
ethics 517, 56, 58, 1564, 21019
ethnicity 110, 1141
ethnography 85, 86, 138
discourse analysis 209
education 150, 152, 160
grounded theory 178
European Community 15
Everitt, A. 107, 109, 136
examinations 14972
Examinations Board 150
executive information systems (EIS) 20218
existence 10
experimentation 78, 99
Halberg, M. 113
Hamlin, A. 5165
Hammersley, M. 152, 155
Hand, D.J. 11836, 129
Hanson 169
Hanvey, C. 108
Hardiker, P. 144
Harding, S. 46
Harris, K. 1618
Harrod, R. 216
Hawker, R. 104
232 INDEX
health policy 20
health professionals 98
Hekman, S. 20, 41
Hendry, D. 217
Henkel, M. 145
hermeneutics 18, 27, 42, 86
Hicks, J. 70
Holman, B. 142
Honigmann, J.J. 151
household management 67
Howe, D. 140
Hugman, R. 138
Hult, M. 190, 191
humanism 44
Hume, D. 1415, 61
Hngler, B. 99
hyper-reality 26
hypotheses 30, 11836
illuminative evaluation 138
income generation 97
individualism 54, 563
individualization 110
individuals 12335
induction
grounded theory 174, 186
investigation 58
methods 16
reasoning 14
science 169
inductive-deductive process 11, 14
Industrial Revolution 8, 13, 18
infinite regression 58
inflation 16
information superhighway 26
information systems (IS) 8391, 106
action research 189202
participant observation 20220
information technology (IT) 839
integration 186
interactionism 103
interpretation 18, 26
interpretivism 13, 416, 45, 10916, 202
information systems 8485
participant observation 209
social work 137
statistics 129
intertextuality 34
interviews 1596
investigator triangulation 102
invisible hand 617
jackknife method 125
Jameson, F. 26
Jordan, B. 1363
Joseph, K. 150
judges 62
jurisprudence 110
Kaplan, 166
Keen, P.G.W. 84
Kelman, 59
Keynes, J.M. 2132, 2166
Knight, F. 216
knowledge
community 30
empiricism 27
feminism 108
managerialism 114
Plato 10
power 24, 283, 44, 49, 1441
power differential 11017
praxis 107
sensory experience 8, 11
Krner 194
Kuhn, T. 17, 27, 34, 169
disciplines 81
paradigm shifts 42
laboratory experiments 86
labour force 718, 77
Lakatos 169
Land, F. 90
Lane, R. 79
language 2526, 327, 79
Lather, P. 20, 349, 37
Lawler, J. 102
Lazarus 97
leisure time 7576
Lennung, S. 190, 191
Levy, P. 167, 168
Lewin, K. 190
liberal feminism 42
Liebenau, J. 90
INDEX 233
monitoring 109
morality 51, 617
Morse, J.M. 99, 174, 177
motivation 59, 62, 63
Mulhall, A. 97
multivariate statistics 128
Multiview 89, 1937
Munn-Giddings, C. 138
Murray, H. 189
Nandhakumar, J. 20220
National Health Service (NHS) 95, 194, 195
Newton, I. 79
Neymann, J. 221
Nobel Prize 67, 70
normalization 44
North American Free Trade Area 15
Northern Education Association 150
numerical research 109
nursing research 9511
object-oriented methodologies 874
observation 26, 85
Occams razor 63
Oliver, M. 142
ontology 3, 9, 113, 149, 169
open coding 183, 185
opportunity cost 67, 738, 76
optimal scaling methods 129
organization theory 84
Orlikowski, W.J. 8485, 91
Orme, J. 20, 10623
orthodoxy 8, 397, 109
paradigm shifts 42
parents 151
Pareto efficiency 761
participant observation 85, 102, 138, 1607,
20220
participative methodologies 874
partnership 1419
patriarchy 20, 40, 43, 110
Payne, S. 9511, 16695
phenomenology 86, 97, 162, 175
Phillips, D.C. 168, 169
philosophy 54
Plato 1012, 13, 14, 20
234 INDEX
INDEX 235
paradigm shifts 42
philosophy 169
politics 114
postmodernism 39
Scott, D. 14972
Scott, J. 46
screening 127, 128
self-interest 61
semiology 83
Sen, A. 53
sensory experience 8, 11
sexism 394
Sheldon, B. 135
Shotter, J. 33
Silverman, D. 1675
Simons, H. 154
Simpsons paradox 126
single-minded model 60, 62
Smith, A. 15, 71
Smith, D. 135
Smith, G. 1396, 141
social anthropology 92
social care practices 13454
social choice theory 53
social control 44
social work 10623, 13454
socialization 38, 62
sociology 84, 85, 89, 92
Socrates 12
software 122, 178
soul 11, 19
specialization 717
Spinoza, B. 58
Stanley, L. 109, 113
statistics 109, 111, 11836, 169, 21332
status 97
Stenhouse, L. 159
Stern, P.N. 17281
story-telling 2437
strategic methodologies 87, 88
Strauss, A.L. 100, 151, 17186, 183, 187
structured methodologies 874
sub texts 338, 45
sufficiency aspect 63
surrogacy 40
surveys 86, 87
Susman, G. 190
symbolic interactionism 175, 177
236 INDEX
Wilson-Barnett, J. 103
Wiseman, J. 53
Wolcott, H. 154
womens studies 35
Wood-Harper, A.T. 87, 192193, 195, 1975
Woods, P. 151, 154
working women 7476
Worrall, A. 144
Wright Mills, C. 167
Yeatman, A. 48