Educational Psychologist

This article was downloaded by: [RMIT University]
On: 12 March 2015, At: 07:44

Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK
Educational Psychologist
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/hedp20
Generative Processes of Comprehension

Merlin C. Wittrock
Published online: 08 Jun 2010.
To cite this article: Merlin C. Wittrock (1989) Generative Processes of Comprehension, Educational Psychologist, 24:4, 345-376,
DOI: 10.1207/s15326985ep2404_2
To link to this article: http://dx.doi.org/10.1207/s15326985ep2404_2
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the
publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations
or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any
opinions and views expressed in this publication are the opinions and views of the authors, and are not the
views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be
independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses,
actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever
caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone
is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/
terms-and-conditions
EDUCATIONAL PSYCHOLOGIST, 24(4), 345-376
Copyright 8 1990, Lawrence Erlbaum Associates, Inc.
Generative Processes of Comprehension
Merlin C . Wittrock
Graduate School of Education
University of California, Los Angeles
This article presents a model of the generative processes of reading compre-

hension. The article begins with a discussion of the four parts of the model:
generation, motivation, attention, and memory. The discussion then reviews
laboratory smd classroom research relevant to the model. A series of
experiments by the author and his colleagues are presented to support the
Downloaded by [RMIT University] at 07:44 12 March 2015
instructional utility of the model. The article concludes with a discussion of

the model and its relation to the teaching of reading comprehension in
schools.
Within the last 20 years, many of the people who study reading have shifted
their perspective, and consequently their research interests, to focus on the
critical role that human cognitive processes, such as attention, motivation,
knowledge acquisition, encoding, learning strategies, and metacognition,
play in influencing comprehension (Wittrock, 1986b, 1989). The shift
meant that we became less interested in how environments directly and
automatically influence learning.
Instead, we became interested in how learners use background knowl-
edge, or schemata, and thought processes, such as verbal elaborations and
imagery, to construct meanings for the text. In short, reading became
reading with understanding.
At a deeper level, the conception of the learners' roles in reading and in
knowledge acquisition, as well as the teachers' roles in influencing both of
them, changed. The learners became the sources of plans, intentions, goals,
ideas, memories, strategies, and emotions actively used to attend to, select,
and construct meaning from stimuli and knowledge from experience. The
Requests for reprints should be sent to Merlin C. Wittrock, Division of Educational

Psychology, Graduate School of Education, University of California, Los Angeles, 321 Moore
Hall, 405 Hilgard Avenue, Los Angeles, CA 90024-1521.
teachers became the facilitators of these cognitive and metacognitive
processes of comprehension.
Gone then, or at least of less interest, were studies about how variables,
such as the length of practice and the amount of time spent on a task,
directly influence learning. In their places came studies about how these
variables influence the learner's thoughts and feelings that together result in
comprehension. In their places also came studies about how a teacher could
influence the learner to think about tasks differently, to construct different
meanings, to use different learning strategies, and to relate knowledge to
the material to be learned (Wittrock, 1981a). People began to ask about
which cognitive processes and learning strategies could be taught to learners
(Wittrock, 1979). How and to whom could they be taught? In what contexts
would they be used? Metacognition began to receive increased attention.
Many of these recent interests in cognitive processes fit well into the
historical mainstream of thought about learning, memory, and their
facilitation through teaching. Since the days of Aristotle, imagery has
played a central role in memory, and associations among ideas have played
a central role in retrieval. Since the days of the ancient teachers of rhetoric,
such as Cicero and Quintillian, constructing relations between experience
and new information, by generating interactive images between the old
ideas and the new events, has been a favorite and effective pedagogical
technique (Wittrock et al., 1977).
These interests also fit well with earlier but often forgotten 20th-century
findings. Thorndike (1931) largely abandoned his law of exercise or
frequency. He found that one could spend time practicing behavior and not
profit or improve as a result of the practice alone. Other conditions, such as
rewards and belongingness, were necessary for practice to produce learning.
Within the last two decades, we have again come to the view that reading
comprehension depends on the learners' thoughts, feelings, schemata, and
information-processingsystems. With the return to this ancient perspective,
new conceptions of reading, and of its learning and its teaching, arise in the
research literature on comprehension. In this article, I present one of these
conceptions -the model of generative reading comprehension. It focuses on
the constructive or, generative processes of reading usually considered
characteristic of written composition.
RELATIONS BETWEEN READING AND WRITING
Reading and writing differ from each other in the thought processes and
human behavior they represent. These well-known differences are summa-
rized best by the commonly accepted belief that writing is the process of
GENERATIVE PROCESSES OF COMPREHENSION 347
putting meaning on written pages, whereas reading is the process of getting

meaning from the written pages. This conventional wisdom implies that
writing is a constructive or generative skill but that reading is essentially an
imitative or r~eproductiveskill. I believe that this conventional and useful
conception of reading and writing leads us to misunderstand the generative
nature of readling comprehension. A more useful conception of reading can
lead to improvements in the learning and teaching of reading through
articulating the generative processes learners use to construct meaning for
text.
Good reading, like effective writing, involves generative cognitive pro-
cesses that crezite meaning by building relations (a) among the parts of the
text and (b) be!tween the text and what we know, believe, and experience.
The generation of those two types of relations is the essence of reading
comprehension. The meaning is not only on the page nor only in our
memories. When we read, we generate meaning by relating parts of the text
to one another and to our memories and knowledge.
When we write, we generate meaning by relating our knowledge and
experience to the text. Writing also involves building relations among the
words in sentences, the sentences in paragraphs, and the paragraphs in
texts. In these important ways, generative reading comprehensi~nand
effective writing relate closely to each other (Wittroek, 1983b).
READING AND WRITING AS GENERATIVE

PROCESSES
Young children generate spoken and sometimes written language before

they learn to read. In conversations, young children construct rule-
governed, syntactically correct sentences that communicate meaning to
others. In these social interactions, young children generate meanings for the
sentences they hear. We often take these remarkable generative, linguistic
abilities for granted. Perhaps because they are so commonplace, we
overlook one ad' their fundamental meanings -that language is a generative,
cognitive process beginning with very young children's conversations.
Equally remarkable is the finding that 4- to 5-year-old preschoolers, taught
to write letters fior sounds, generate written words and rudimentary sentences
(KEN I STA DAOON STERS) before they have learned to read (Chomsky,
1979). These invented spellings by children indicate again the generative
nature of language and some of the similarities that underlie reading and
writing. Try to make sense of the following sentence written by a 5-year-old
(Bissex, 1976): "EFUKANOPNKAZIWILGEWAKANOPENR."
THE MODEL OF GENERATIVE COMPREHENSION
Through their conversations and their invented spellings, preschool children

show that they approach speaking and writing as if they were generative
activities. Before these young children learn to read, they demonstrate
generative language abilities developed in social interactions involving
conversations and simple written compositions.
From our knowledge of the generative language abilities of young
children, it is reasonable to ask if they learn to read in much the same way
that they learn to speak, to invent spellings, and to write simple composi-
tions. Have we misunderstood an important part of the nature and
complexity of learning to read with understanding? Perhaps it is nearly as
difficult to become an excellent reader as it is to become an excellent writer.
Perhaps reading should be taught as a generative activity.
My model of generative reading comprehension explores these ideas,
building on an analogy between reading with comprehension and writing,
speaking, and listening with understanding. Although reading differs in
obvious and subtle ways from speaking, listening, and writing, it shares
important psychological and cognitive processes with them.
Briefly stated, the model includes four major components: generation,
motivation, attention, and memory. In the model, comprehension involves
the reader's active generation of two types of semantic and pragmatic
relations: (a) among the parts of the text and (b) between the text and
knowledge or experience. The active generation of these two types of
relations implies a motivation or willingness to invest effort in reading and
an ability to attribute success and failure in generating relations to one's
effort. Attention, the third element in the model, directs the generative
processes to relevant text, related stored knowledge, and memory of
pertinent experience. Memory, including preconceptions, metacognitions,
abstract knowledge, and concrete experience, comprises the fourth element
of the model.
The essence of the generative learning model (Wittrock, 1974a, 1974b,
1978a, 1978b, 1981b, 1983b) is that the mind, or the brain, is not a passive
consumer of information. Instead, it actively constructs its own interpre-
tations of information and draws inferences from them. People ignore some
information and selectively attend to other information. In other words,
although our brains often respond reflexively to incoming stimulation, our
minds are much more than "blank slates" that passively learn and record
incoming information (Wittrock, 1980). The stored memories and infor-
mation-processing strategies of our cognitive systems interact with the
sensory information received from the environment, selectively attend to
this information, relate it to memory, and actively construct meaning for it.
Generation is a fundamental cognitive process in comprehension. Gen-
eration is not the same as semantic processing, an often cited cognitive

explanation of meaningful learning. Generation is more than the fitting of
information into slots or schemata, an explanation of understanding
frequently presented by schema theorists. Generation is the active construc-
tion of relations among parts of the text, and between the text and
knowledge and experience. Generation can result in assimilative learning,
that is, schema fitting. Generation can also result in accommodative
learning, leading to the construction of new schemata.
Relations between text and memory, text and experience, and among the
parts of the tavt facilitate comprehension because they are generated, not
just because they are semantic in nature or because they involve effort or
because they involve schema fitting. In addition to these important factors,
generation is il process of constructing relations that contributes to com-
prehension and that can occur in reception learning and discovery learning,
in laboratories and in lectures. This part of the model, which I first
published in 1974 (Wittrock, 1974b) does not occur in other cognitive
models of meaningful learning, semantic processing, information process-

ing, or schema fitting.
According to this model of generative comprehension, to learn with
understanding a learner must actively construct meaning. To comprehend
what we readl, we invent a model or explanation that organizes the
information selected from the experience in a way that makes sense to us,
that fits our logic or real-world experiences or both. People retrieve
information from long-term memory and use their information-processing
strategies to generate meaning from the incoming information, to organize
it, to code it, 'and to store it in long-term memory.
Motivation
The model of generative reading comprehension leads to several implica-

tions about understanding and facilitating the teaching of reading. In the
area of motivation, the model implies that students should become mentally
active, generative learners who hold themselves accountable and responsible
for constructinlg verbal and imaginal relations between what they know and
what they read.
Teachers can facilitate this active student role in comprehensi~nby
attributing learning to student effort. Only when the learners attribute
successful comprehension to their own effort at generating relations among
the text and knowledge or experience will the instructor's actions enhance
motivation in the sense of persistence and sustained interest in learning.
Success, reward, praise, reinforcement, and feedback do not necessarily
facilitate generative learning. Success and teacher approval should be
attributed by the learners to their efforts, not to the activities of other
350 WITTROCK
people nor even to the students' own abilities. When students attribute
learning to other people or to factors external to themselves, the effort they
invest in learning, that is, their motivation, tends to decline.
These ideas about motivation derive from recently developed models of
academic achievement, for example, as discussed by Bernard Weiner
(1979). These ideas are neither original nor unique to generative learning.
They imply that the learners' attributional processes must sometimes be
modified before reading comprehension can be facilitated. The meaning the
learners generate about the causes of learning influences their motivation
and their willingness to become active in generative learning.
These ideas contrast sharply with commonly practiced principles of
accountability only for teachers, of mentally passive learners acquiring
knowledge with little or no effort, of time to learn, or of practice and
reinforcement as adequate to produce learning. These principles are not
wrong. They are useful, provided that you understand how they operate.
For example, rewards do not invariably bring about learning or feelings of
success. Instead, rewards produce these effects when students value them
and attribute them, at least in part, to their own actions. Success is not
enough. Its effects depend on students' interpretations of its causes and
meaning.
From this perspective, how can one change student attributions? Richard
deCharms (1976) taught teachers and students in St. Louis to perceive
themselves as causes of their own actions and to take responsibility for their
own success or failure as teachers in inner-city schools.
decharm's origin training program continued over 2 years. As a result,
both teachers and students changed. They each increased in the responsi-
bility they took for their own actions. The children's achievement in
mathematics (arithmetic) and language also increased. The gains in reading,
unfortunately, were less than those in arithmetic and language, suggesting
that to increase reading comprehension improve more than motivation.
From a different perspective, Barbara McCombs (1983) taught military
personnel to modify their learning strategies and motivation to match the
requirements of the learning tasks. The motivational skills training program
improved performance in technical training.
Attention
Another old but recently revived concept, attention, offers some insights
into the problems of teaching reading and of understanding some learning
disabilities (Wittrock, 1986a). Recent research has analyzed mental retar-
dation and learning disabilities into cognitive processes, such as attention,
that lead to a deeper understanding of the practical problems and their
possible remedies.
Attention develops slowly in many children. Most normally developing
children from about age 5 years to age 15 years show a gradual increase in
ability to learn and to remember centrally important information, but they
show no increise in ability to remember incidental or irrelevant informa-
tion. Attention consists of several components, among them are a short-
term, orienting response, and a long-term or sustained voluntary response.
These two types of attention, along with other types, have been studied
extensively in neuropsychology (see Wittrock, 1980) and in special educa-
tion.
Krupski (1975) studied short-term involuntary reactions and long-term
voluntary reactions of mentally retarded children. She found that mentally
retarded children react normally to short-term attention tasks but abnor-
mally slowly to sustained attention tasks in an academic setting. In
nonacademic settings, however, mentally retarded children improved their
performance. In other studies (see Hallahan & Reeve, 1980), mentally
retarded children have also shown deficiencies in short-term voluntary
attention. Apparently, their attentional problems primarily involve voli-
tion, or self-control, something that might respond to appropriate training.

Consistent with Krupski's findings are Hallahan and Reeve's (1980)
results derived from over several years of study of attentional processes in
learning-disablled and mentally retarded children. They found that many of
these children lag about 2 to 3 years behind in their ability to ignore
incidental information and to selectively attend to relevant information.
These children usually do not use rehearsal strategies as well or as
frequently as do normal children. An important educational implication of
this research is that these deficiencies in attention, rehearsal, and self-
control might be remediated, at least partially, by appropriate cognitive
training.
Hyperactivity is a disorder that seems to involve attention. Hyperactive
children are distracted primarily by task-related, but sometimes by task-
irrelevant, stimuli. An attentional hypothesis explains the so-called para-
doxical effect of stimulant drugs reducing hyperactivity among learning-
disabled children by suggesting that the drugs steepen the gradient of
attention. Further, if hyperactivity involves a flat gradient of attention,
then an attentional hypothesis suggests that cognitive intervention strategies
that stress self-control over attention might ameliorate the problem, at least
under some conditions.
Following this reasoning, Meichenbaum and Goodman (1971) taught
second graders, who were either hyperactive or had poor self-control, to use
a strategy that asked them to control their impulsive behavior and to "stop,
look, and listen" before they acted. The strategy reduced the learners'
impulsivity on itbe Matching Familiar Figures Test, but the effects did not
generalize to the classroom.
Camp (1980) used training procedures modeled after Meichenbaum's
self-instructional techniques. She taught impulsive elementary school-aged
boys to ask themselves what problem they faced, how they were to try to
solve it, whether they were following their plan, and how well they did.
After 30 training sessions, she showed gains in IQ, reading, and social
behavior. In addition, these gains generalized to the classroom.
Douglas, Parry, Martin, and Garson (1976) also developed a self-
instructional cognitive training program to teach 7- and 8-year-olds to use
self-verbalization to control their impulsive behavior. After 3 months of
training, scores increased on the Matching Familiar Figures Test, on tests of
planning, and on tests of reading comprehension, although reading was not
taught in this program.
Among children who are not learning disabled, selective attention
sometimes presents a problem that might be remediated by training with
verbal strategies. Paris, Lindauer, and Cox (1977) found that 7- and
8-year-old "normal" children can, but rarely do, spontaneously construct
inferences about sentences they read as a way to achieve a goal, such as
remembering what they have read. In this same study, children learned to
construct stories from the sentences they read. The strategy of constructing
stories facilitated comprehension and increased memory of the sentences.
The goal apparently influenced selective attention. The construction of
stories probably enhanced comprehension by facilitating the construction
of relations across the sentences in the stories.
Malamuth (1979) used cognitive techniques, similar to methods devel-
oped by Meichenbaum and Goodman (1971), to improve self-management
among poor readers who were not hyperactive. Sustained attention in-
creased as a result of the training, and reading scores also improved,
indicating the utility of cognitive training programs for some normal
children with reading problems.
These cognitive training programs show that children, including hyper-
active students, can show improvements in reading. Most effective training
programs teach the children strategies of self-control that provide an
explicit set of steps for them to follow that usually involves analyzing visual
stimuIi carefully before responding to them. Most of the successful
programs also involve extensive training.
Generalization of these training programs is a problem, but the cognitive
training programs usually provide the greatest amount of generalization,
compared with the behavioral and drug programs (Keogh & Glover, 1980).
Memory
This massive topic includes preconceptions, abstract knowledge, everyday

experience, domain-specific knowledge, learning strategies, and meta-
cognition. This topic is too extensive to expand on here (see Wittrock et al.,
1977). The essential point is that generation functions by creating relations
between this vast store of organized information in long-term memory and
the text to be comprehended. The construction of these relations leads to
comprehension.
Generation
Essentially, reading comprehension is the process of actively generating

relations among the parts of the text and between the text and one's
memories, knolwledge, and experience. I believe that to comprehend text we
must invent a model or explanation for it that organizes its parts in a way
that makes sense to us, that fits our logic and our world knowledge. The
invention or generation of a sensible model of the text and its relations to
us does not preclude our learning from direct instruction or from someone
else, such as a teacher. Even when a teacher tells us an answer, or when we
look it up in the back of the book to comprehend it, we must still discover
its intended meaning. Being given the answer does not necessarily aid or
deter comprehension. It depends on what we do with that information,
what we think about it, and how we relate it to our knowledge. Also, being
given an explanation by a teacher does not preclude our need to invent that
same explanation for ourselves to comprehend the intended meaning.
Generation, not discovery, is the process of comprehension.
From the perspective of my model, the teaching of reading and the
teaching of writing share subtle and important generative processes.
Writing is more: than the construction of text for meiming, and reading is
more than the construction of meaning for text. Writing is also a process of
constructing m~eaning,which gets revised and made more precise as one
edits, revises, and generates (Wittrock, 1983b). Reading involves recon-
structing the te.xt in familiar terms, examples, and experiences that allow us
to relate our knowledge and memory to the message and to the perspective
of the author. l[m generative learning, students' knowledge, experience, and
learning strategies are crucially important because, as strange as it may
seem, answers given to learners must still be generated by them (i.e., related
to context and to knowledge and experience) before they can comprehend
them.
A teacher facilitates these generative processes of comprehension. The art
of generative teaching is knowing how and when to facilitate the learners'
construction of relations among the parts of the text and their knowledge.
In addition, teachers can teach the learners how to increase their ability to
control their generative processes so that reading comprehension becomes
increasingly independent for them. An important goal is for the learners to
learn to controll their own generative processes.
354 WITTROCK
There are fundamental ways to stimulate generative processes. Table 1

lists some of these techniques. In addition to learners' voluntary control,
teachers can use the characteristics of the text itself, the familiarity of the
words, stories, analogies, metaphors, diagrams, headings, underlined
words, and pictures to stimulate generative processes (Kourilsky &
Wittrock, 1987).
Second, the teacher can give explanations, main ideas, inferences, sum-
maries, advanced organizers, questions, metaphors, analogies, examples,
paraphrases, and blanks to be completed either before or after the students
read the passages.
Third, the teacher can ask the students to construct headings, subheads,
inferences, summaries, questions, metaphors, analogies, answers, pictures,
flow charts, tables, rebuttals, alternative explanations, and critiques while
they are reading and after they have read the passage.
Fourth, the teacher can teach metacognitive strategies useful for solving
problems (e.g., Wittrock, 1967) or for comprehending story grammars and
types of discourse.
With this introduction to the model and some of its educational
TABLE 1
Ways to Stimulate Generation
Teacher Given Learner Constructed
Among Concepts Presented in Instruction
Titles Compose titles
Headings Compose headings
Questions Write questions
Objectives State objectives
Summaries Write summaries
Graphs Draw graphs
Tables Prepare tables
Main ideas Construct main ideas
Between Instruction and Prior Knowledge
Demonstrations Student demonstrations
Metaphors Compose metaphors
Analogues Propose analogues
Examples Give examples
Pictures Draw pictures
Applications Solve problems
Interpretations Develop explanations
Paraphrases Put into own words
Inferences Draw inferences
applications, I turn next to the empirical research on generation. First I

discuss laboratory research on generation. Then I discuss applied research
on generative reading comprehension.
LABORATORY RESEARCH ON GENERATION
Laboratory studies, usually with college students learning word lists and
word pairs, haire investigated the generation effect and have tried to explain
it. Wittrock and Carter (1975) gave college students hierarchially arranged
words and asked the learners to process them generatively, that is, to relate
them to one mother and to construct the proper hierarchical relations
among them. Whether the words were conceptually related to one another
or were chosen at random from a dictionary, generative processing of them
sizably increase~d,usually doubled, their retention.
In two experiments, Wittrock and Goldberg (1975) studied children's and
college students' learning and memory of words that they were asked to put
into sentences, a story, or an interactive image. The high-imagery words
were best remembered by the college students and the children. The
directions to group the words into stories, sentences, or images facilitated
recall for the children but not for the college students. The results imply that
the effectiveness of the directions depends on the learners' developmental
level and on the cognitive processes they have learned to use. Instructional
aids, such as ~directions,can help learners only if they are not already
spontaneously performing the generative activities mentioned in the instruc-
tions.
In three expe~imentswith fourth and fifth graders, the type of instruction
necessary to teach the children how to generate a concept depended heavily
on the previously learned relevant concepts (Wittrock, 1978b). When the
training focused on a relevant but not highly salient concept, transfer
increased quickliy and markedly.
In five experiments with college students, Slameeka and Graf (1978)
compared memory for words that were either read or generated. For all
measures in all five studies, the generation of words enhanced memory. The
generation phenomenon held across different rules and materials, across
timed or self-]placed learning, and across different types of tests and
different experimental designs.
In the third experiment, they tested selective attention as the sole
explanation for the generation effect. The responses, which were generated
by the learners, but not the stimuli, which were nlot generated by the
learners, were better recognized later. If selective attention were the sole
explanation of the generation effect, then the stimuli would have been as
well recognized i3s the responses, because the learners selectively attended to
both of them. However, the data from the fifth experiment are less clear cut
on this issue, making it difficult to establish any firm conclusions about the
role of attention in generation. The authors, however, concluded that their
data eliminate overt responding, frequency, exercise, time to learn, and
levels of processing as explanations of the generation effect. Therefore,
both distinctive encoding and effort are still possible explanations.
Jacoby (1978) compared reading with generating solutions to simple word
problems. Compared with reading the words, generating the response
words increased their recall when the problems were seen only once earlier
or were seen repeatedly in a spaced pattern. In massed practice, reading or
generation made no difference in recall, apparently because massed practice
trivializes generative reading. The learner can recall the answer directly
from short-term memory. When the interval between presentations of a
problem is longer than short-term memory, the learner must construct
rather than recall the solution. Jacoby concluded that overt responding is
not enough to facilitate learning. Reading the answer immediately prior to
the problem reduced performance. He concluded that effort was critical to
improving recall in this study. These two results seem incongruous.

Apparently something more than effort is necessary for generation to be
effective, otherwise overt responding, which is effortful, would be a
sufficient explanation.
Generation has also been studied with sentences. Slamecka and Katsaiti
(1987) argued that the common use of within-subjects designs in the studies
of the effects of generating words may have led the learners to rehearse
selectively the generate items more than the read items. Hirschman and
Bjork (1988) tested this hypothesis. They showed that with across-subjects
designs, the word-generation effect occurs in cued recall by enhancing the
relations between the stimulus and the response terms. Wittrock and Carter
(1975) found that the generation of hierarchical relations among words also
facilitates free recall. The word-generation effect seems to be robust and
does not seem to be due to procedural artifacts.
In five experiments with college students, Graf (1980) compared gener-
ating meaningful and anomalous sentences from words given to learners
with simply reading the same sentences contrasted by the experimenter.
Generation of the sentences enhanced word recognition with meaningful
and anomalous sentences. Generation enhanced word-pair recognition, a
measure of sentence organization, with meaningful sentences but not with
anomalous sentences. The results indicate that generation enhances
intraword organization with both types of sentences but enhances integra-
tion of sentences only if they are meaningful. He concluded that reading
should stimulate bringing words to mind simultaneously and constructing
meaningful relations among them.
McFarland, Frey, and Rhodes (1980) compared the effects on memory of
learner-generated words missing in sentences with experimenter-given

words in the s,ame sentence. The generation of the missing words by the
college undergraduates consistently increased their free recall, even across
different types of congruity between the word and its context, and across
different levels of processing, such as a phonemic level or a semantic level.
McFarland et al. (1980) concluded that generation is distinct from the
phenomena called elaboration and levels of processing, such as
phonemically processed and semantically processed words. In their study,
phomenically encoded words, as well as semantically encoded words,
evidenced a generation effect. For example, learner-generated rhymes
(phonemically processed) were remembered better than experimenter-given
semantically rellated words, a result not predicted by levels-of-processing or
elaboration theory, which predicts that semantic processing, but not
phonemic processing, enhances long-term memory.
Another inteiresting finding in the McFarland et al. study is that the usual
encoding congruity effect, in which a word must be well-integrated with its
context, was reversed for the learner-generated words. The generation of
semantic incongruities facilitated recall sizably. These authors disagree with

Slamecka and cGraf (1978), who posited that generation enhances the
congruity between a word and its context. McFarland et al. (1980) suggested
that congruity is important for experimenter-given words but nat for
learner-generated words. They also posited that Jacoby's (1978) and
Slamecka and Grafs (1978) data can be explained by an effort hypothesis,
an explanation that is difficult to defend.
In sum, the iinteresting interpretation given by McFnrland et al. (1980) is
that reading, a different from generation, does not ensure a phonemic or
semantic analysis of words nor an integration between words and their
context. I would add that reading also does not ensure an integration
between words a d the learner's experience and knowledge.
McFarland et al. (1980) stated that generation is independent of depth or
type of processing. In this experiment, generation produced a bigger effect
than did semantic encoding, although the semantic-encoding effect did ocur
with experimenlter-given words. They concluded that the actual operations
performed on the words, more than their phonemic or semantic character-
istics, determined learning from generation. What are these operations?
They conjectured (a) personal relevance and (b) effort used in connection
with episodic memory. They offered little explanation for the processes
involved in generation from semantic memory, as it differs from episodic
memory.
In summary, McFarland et al. (1980) agreed with many earlier studies in
the classroom aind in the laboratory when they concluded that "an individ-
ual's memory will improve dramatically if he provides some of the
to-be-remembered information himself' (p. 224).
Stein, Morris, and Bransford (1978) also found that semantic processing
of sentences is not always more effective than other types of processing at
enhancing comprehension and retention. In this study, enhancement de-
pended on the relation between the learner's elaboration and the rest of the
sentence. When the elaboration made the sentence meaning more precise,
comprehension increased. Other types of semantically congruous elabora-
tions also debilitated retention. In a related study, Stein and Bransford
(1979) compared learner-generated and experimenter-given elaborations.
Again, when the elaboration contributed to the development of the tested
meaning, or when learners were led to generate relevant questions that lead
to precise relations among the parts of the sentence, comprehension
increased.
Dee-Lucas and DiVesta (1980) gave college students, or asked them to
generate, headings, related sentences, or topic sentences as they read a text
about minerals. These instructional aids enhanced most types of learning
only when they were generated by the learners. Generation of topic
sentences produced the greatest enhancement of learning from the text.
Giving learners these instructional aids enhanced learning of subordinate

information but not of the structure of the passage. The learning of
structural information was enhanced or reduced by generation, depending
on the relevance and appropriateness of the generative activity to the
learning measured on the test. Fluent learners sometimes learned as much
from given information as from generated information. The results de-
pended on what the learners knew and what they did spontaneously. In
Dee-Lucas and DiVesta (1980); generation seemed to facilitate selective
attention. Generative activities worked primarily when learners would not
have spontaneously performed them. The learning of material already
salient to the learner did not profit from generative activities that enhanced
selective attention.
Dee-Lucas and DiVesta reported important findings about the teaching
of generative activities. They showed that an instructional intervention
enhances learning or comprehension only when it induces learners to
perform activities they would not otherwise perform or not perform as well.
It is not enough then to induce generative processing. In addition, one must
enhance the learner's generative processes appropriate for the situation, in
this case to read with comprehension.
From this review of laboratory studies on the generation effect, at least
two findings are supported. First, the generation effect is a real one. It is not
synonymous with semantic processing or elaboration, although it produces
a sizable impact on reading with comprehension and on retention. Second,
the nature of the generation effect is only partially understood. It seems to
involve a number of cognitive processes in much the same way that
language production, as in speaking or writing, involves them. Reading does
not always involve these same constructive processes. When we can make
reading become more like written composition, and engage generative
activities, it seems that we can increase reading comprehension.
APPLIED RESEARCH ON GENERATIVE READING

COMPREHENSION
The data collected by my students and me on the model of generative

comprehension relates closely to research on learning strategies, reading
comprehension, and students' cognitive processes. Far example, see Claire
Weinstein's work on learning strategies (Weinstein, Goetz, & Alexander,
1988; Weinsteiin & Mayer, 1986); Richard Mayer's research on problem-
solving and leruning strategies (e.g., Mayer, 1980); and the recent research
on teacher and student cognitive and metacognitive processes, conducted by
Walter Doyle (1980), Penelope Peterson (e.g., Peterson & Swing, 1982),
and Donald Darnsereau (1978). The findings of these studies often parallel
the results of our studies.

Over the lasit 20 years, my students and I have completed a series of
empirical studies designed to test implications for teaching of the model of
generative reading comprehension. In separate studies (all except one were
experiments involving individual random assignment of the learners to the
treatments), we have asked elementary school children, junior high school
students, and college students, as they read a text, to generate paragraph
headings, sumnnaries, interpretations, images, and pictures that related the
parts of the text to one another and to their knowledge and experience. We
have also tried giving children familiar words to induce the generation of
sentence meanings and familiar stories to induce the generation of meanings
for unfamiliar and undefined vocabulary words.
In one experiment (Doctorow, Wittrock, & Marks, 1978), 488 public
school, sixth-grade students were individually randomly assigned to eight
treatments. In this study, we asked the students in one group to generate a
summary sentence for each paragraph they read in stories taken from a
commonly used, commercially published reading textbook appropriate in
difficulty for their reading ability. Another group was also given paragraph
headings to use in the summary sentences they were asked to construct. As
seen in Table Z!, with time to read and to learn held constant across all
treatments, the students who generated the paragraph summaries sizably
and statistically significantly increased their retention and comprehension
of the text, from a mean of 35 for the control group to a mean of 51 for an
experimental group. The group given the paragraph headings and asked to
generate surnmiuies of the paragraphs doubled their retention and compre-
hension, from an average of 35 for the control group to an average of 68 for
TABLE 2
Treatment Means (and Standard Deviations) for Comprehension and
Recall Test Scores
High-Ability Readers Low-Ability Readers
Treatment Comprehension Retention Comprehension Retention
Generate summary
using two-word heading
Generate summary
using one-word heading
Generate summary '
Two-word heading
One-word heading
Control
Note. From "Generative Processes in Reading Comprehension" by M. J. Doctorow, M. C.

Wittrock, and C. B. Marks, 1978, Journal of Educational Psychology, 70, p. 115. Copyright
1978 by the American Psychological Association.
the experimental group. The control group read the same stories, but they
were not asked to generate summaries. The active generation of relations
among the sentences in a paragraph sizably facilitated comprehension and
retention.
In another study (Wittrock, Marks, & Doctorow, 1975), we gave 468
fifth- and sixth-grade public school students familiar stories in which we
embedded unfamiliar and undefined vocabulary words. Compared with the
control groups given the same vocabulary words embedded in unfamiliar
stories, all the generative groups increased, from 50% to 100%, the number
of new vocabulary words they could correctly define on a vocabulary test.
The study implies that students use familiar story contexts to generate
meanings for vocabulary words.
In a related study (Marks, Doctorow, & Wittrock, 1974) with 230
sixth-grade students, we substituted one familiar vocabulary word per
sentence (e.g., boy for lad) for one unfamiliar word per sentence in stories
from commercially published reading stories. Comprehension of the stories
increased at least 50%, and sometimes 100010, with time to learn held
constant across the treatment groups. The generation of sentence and story
meaning seems to depend heavily on understanding all of the parts of the
sentences, as the model of generative learning implies.
With 87 fifth-grade students, we examined whether generating pictures
for vocabulary words would enhance memory of definitions, compared
GENERATIVE PROCESSES OF COLMPREHENSION 361
with usual teaching procedures that emphasize memorizing definitions (Bull

& Wittrock, 1973). The drawing of pictures statistically, but not sizably,
enhanced the memory of the definitions.
Next, I report the results of four related studies on generative reading
comprehension. The first study, reported in Table 3, used college students.
The next two studies (see Tables 4 and 5) were done with junior high school
students. The last study (see Table 6) used elementary school students.
In the study reported in Table 3, we increased reading retention and
comprehensioln by asking college students to construct either verbal analo-
gies or sumrnasy sentences (that consisted only of their own words) as they
read a chapter in Rachel Carson's book The Sea Around Us (Wittrock &
Alesandrini, in press). Self-generated analogies or self-generated verbal
summaries facilitated reading. Scores on the Street Test (a measure of
spatial-holistic ability) correlated positively with reading scores in the
imagery condition but not in the verbal analogy condition. Scores on the
Similarities sulbscale of the Wechsler Adult Intelligence Scale (a measure of
verbal-analytic ability) correlated positively with reading scores in the

verbal analogy condition but not in the imagery condition. Scores in the
summary concljltion correlated positively with analytic and holistic abilities.
Although these data are tentative, they imply that different cognitive
processes can be stimulated and used to construct meaning for text. Their
effectiveness also relates to the learner's relative ability to use the verbal or
imaginal strategies.
In the study reported in Table 4 (in preparation), several hundred junior
high school students read the stories entitled "Conductor Moses" and "The
Mirror." In this experiment, all generative procedures facilitated compre-
TABLE 3
Treatment lkleans (and Standard Deviations) for Recall, Time to Learn,
Analytic Ability, and Spatial Ability
Treatment Read Texta Summarizationb Analogy"
Recall 22.4 29.8 27.2
(9.21) (7.72) (6.25)
T i e (minutes) 76.1 86.3 85.2
(23.51) (18.6) (15.3)
Holistic ability 8.2 8.4 7.5
(2.0) (2.2) (2.6)
Analytic ability 16.9 16.6 18.0
(2.8) (4.0) (3.2)
'n = 21. bn = 19.
Note. From "Generation of Summaries and Analogies and Analytic and Holistic Abilities"
by M. C. Wittrock and K. A. Alesandrini, in press, American Educational Research Journal.
Copyright by Amt:rican Educational Research Association. Adapted by permission.
TABLE 4
Treatment Means for the Comprehension and Retention Tests
Generate Generate
Generate Generate Generate Experimental Experimental
Heading Generate Generate TS TS Experimental Sentence Sentence
Given, Ts, TS & Using Relating Sentence, Using Using
Control Heading Generate Generate Heading Paragraph Given Test to Head Paragraph Given
(Read) Generate Given Summary TS Given Head Head Experience Given Head Heading
Treatment n = 129 n = 63 n = 59 n = 60 n = 66 n = 69 n = 67 n = 60 n = 63 n = 62 n = 63 n = 64
Comprehension
Story A 10.4 14.1 13.8 12.9 13.0 12.1 12.6 12.7 13.4 12.0 12.3 12.4
Story B 9.0 12.2 11.6 12.4 12.2 11.9 11.6 11.9 12.4 11.8 11.6 11.1
Retention
Story A 8.7 13.0 11.9 11.2 12.0 12.1 11.8 11.9 12.4 11.7 11.8 11.4
Story B 8.3 19.6 17.2 19.3 19.2 21.0 17.1 17.0 19.5 18.5 18.0 17.0
-- -
Note. TS = topic sentence.

hension and retention, sometimes by large amounts (100% or more) but
usually by 25%)or so.
The study reported in Table 5 (Wittrock & Roberts, in preparation) tested
whether the geineration sf relations between text and experience increases
literal compreh~ension.The control group of junior high school students
wrote the same sentences generated by their yoked counterparts in the free
generation experimental group. The generation effect occurred again,
indicating that comprehension is influenced by the active process of
generating representations for the meanings in the text, not only by effort
or by practice, or by writing, or by the additional information in the
inserted sentences, all of which were controlled by this design.
In a study with 58 fourth- and fifth-grade public school students (Linden
& Wittrock, 1981), we compared conventional reading instruction, where
the children's regular reading teacher taught reading in her usual fashion,
with two experimental conditions in which the children were taught to
generate and t~odescribe aloud (a) interpretations of the paragraphs in their
readings and (b) images relating to sentences of the stories to one another.
These children were also taught to generate relations between the stories
they read and their own experience. As indicated in Table 6, on a test of
comprehensioin of the stories they read, the two experimental classes scored
means of 31 and 29, whereas the control class taught by the children's
regular teacher averaged 22. A second control class, taught by the teacher
of the experimental classes but without using the generative exercises,
averaged 18 on this same test of comprehension. These mean differences are
statistically significant and indicate an increase in comprehension of about
50% due to generative activities, again with time to learn held constant
across all treatments. In this study, the teaching occurred in groups in
realistic classmom settings using conventional reading materials. The
teacher variable was also controlled by using the children's regular teacher
in one control group and the teacher of the experimental or generative
TABLE 5
Means (and Standard Deviations) of the Treatment Groups
Facts Inferences
Treatment (20 Items) (M Items)
Free generationa 14.2* 10.8**
(3.1) (3.0)
Written generationb 13.1 9.3*
(3.2) (3.7)
Yoked controlc 12.1 7.8
(3.2) (3.0)
"n = 30. bn = 35. 'n = 25.
*p < .05, higher than the control group. **p < .01, higher than the control group.
364 WITTROCK
TABLE 6
Means (and Standard Deviations) of the Generation, Retention, and
Comprehension Scores of the Treatment Groups
Number Fact Retention Comprehension
Treatment of Generations (35 Items) (49 Items)
Imaginal to verbal 13.0 27.6 28.6

generationa (4.8) (3.3) (6.3)
Verbal to imaginal 10.7 23.3 31.3
generationsb (2.5) (3.2) (5.7)
No instructions to 0.0 25.1 17.7
generateb (0) (3.1) (8.6)
Classroom teacher 21.57 21.6
taught controlb (7.12) (12.7)
an = 16. bn = 14.
Note. From "The Teaching of Reading Comprehension According to the Model of
Generative Learning" by M. Linden and M. C. Wittrock, 1981. Reading Research Quarterly,
17, p. 52. Copyright 1981 by the International Reading Association. Adapted by permission.
groups in a second control group. The results of this study indicate that
generation of relations between the text and experience, and among the
parts of the text, facilitates comprehension in realistic classroom group
teaching situations.
A series of three applied studies was conducted over a period of 2% years
in basic skills classes taught at four Army bases in California and in Hawaii
(Wittrock & Kelly, 1984). Soldiers with low reading ability, regularly
enrolled in these classes, learned in 9 hr to increase their reading compre-
hension of Army manuals and texts as well as nontechnical materials.
In the first and second of the three studies, five different curricula, each
using paragraphs from Army manuals and from magazines and newspa-
pers, were used in basic skills classes. A comparable control group of
soldiers given customary Army basic skills instruction, taught by regular
basic skills teachers, offered on the same Army bases in the same class-
rooms for the same amount of classroom time, was used to evaluate the
effectiveness of the new materials and teaching procedures. Pretests and
posttests were also given to all experimental and control classes to provide
a second measure of effectiveness of the experimental curricula and the
generative teaching methods.
The soldiers in each of the experimental groups showed statistically
significantlygreater (p < .01) improvement in reading comprehension than
did the soldiers in the control group. The second measure of effectiveness,
the gains from the pretest to the posttest, also showed the same statistically
significant increase ( p < .01) in reading comprehension only in the
experimental treatments. The control treatment showed no gain (see Tables
7 and 8).
TABLE 7
Means, Standard Deviations (in Parentheses), and Gain Scores of the
Experimental and Control Groups of Study 1
Experimental Groups
% of Gain
Treatments Pretest Posttest Gain t P Over Pretest
Imagery 22.2 26.7 +4.5 5.8 .001 20.3
strategya (4.6) (2.7)
Verbal 20.6 23.8 3.2 3.7 .01 15.5
strategyb (5.6) (4.5)
Verbal 24.2
strategy (4.1)
(posttest only)"
Control Groups
Control: Conventional Instruction

Pretest and posttest 21.1 20.8 -.3 -.4 ns -
(5.4) (5.4)
Posttest only 18.2
(7.4)
"n = 23. bn =: 24. 'n = 19.
Note. ns = nonsignificant.
TABLE 8
Experimental and Control Groups of Study 2
% of Gain
Treatment Pretest Posttest Gain t P Over Pretest
Summaries, 18.9 22.6 +3.7 5.9 .oOol 19.6
headings, and (4.1) (3.9)
inferences
strategya
Metacognitive 20.5 23.1 +2.6 4.2 .001 12.7
strategyb (3.6) (4.2)
Metacognitive ancl 17.9 22.1 4.1 6.7 .oOol 22.9
examples (5.7) (4.6)
strategy"
Control groupC 21.1 20.8 - .3 .4 ns -
(5.4) (5.4)
"n = 29. bn = 26. 'n = 16.
Note. ns = nonsignificant.
366 WITTROCK
In the third study of the series, the curricula and teaching procedures
developed in the first and second studies were rewritten for use with
microcomputers. The rewritten materials were presented by microcom-
puters to regularly enrolled basic skills students at two Army bases in
California. Statistically significant gains in improvement in reading com-
prehension again occurred (see Table 9). In addition, the time needed for
the instruction dropped sizably, from an average of 450 min (nine 50-min
class hours) to an average of 251 min (about five 50-min class hours).
The research indicates that, through generative learning procedures,
reading comprehension can be sizably increased, at no added cost or time,
in typical basic skills classes offered on Army bases by regularly employed
basic skills teachers working with normally enrolled soldiers. Under these
realistic and typical conditions, the reading comprehension instruction and
curricula should teach soldiers how to generate meaning by relating their
knowledge and background to the manuals and other texts they read and by
generating relations among the sentences in the text. These two basic
principles may lead to inexpensive but useful modifications in teaching and
to increases in ability to read with comprehension. The research also implies

that, with microcomputers and appropriate curricula, reading comprehen-
sion can be increased, and training time can be markedly reduced, at least
with some soldiers.
THE FACILITATION OF READING COMPREHENSION
From this review of laboratory and applied research, there is substantial

support for the model of generative-comprehension generation and for its
practical utility in improving reading comprehension among school children
and young adults. There are also good reasons to believe that effective
generative activities interact with the developmental level and ability of the
learners, the type of reading material, and the teachers and curriculum
writers' goals in the design of the instruction. For example, prior to about
TABLE 9
Experimental Groups of Study 3
% of Gain
Army Base Pretest Posttest Gain t P Over Pretest
Fort A" 19.38 23.83 4.25 4.7 .005 21.9

(6.57) (5.15)
Fort B~ 19.36 20.77 1.41 2.2 .025 6.8
(3.46) (2.83)
*n = 8. bn = 22.
age 8, childrem do not often increase their reading comprehension by
generating pictures appropriate for the text; however, these children do use
pictures and high-imagery words quite effectively to improve their reading
comprehension (Levin, 1981). We also know that when they are asked,
children 5 years and older can construct their own verbal aids- headings,
summaries, inferences, and answers to questions- to enhance their reading
comprehensioin. In addition, we know that children can use these same
verbal organizational aids, when they are given to them in a text or by a
teacher, to facilitate their reading comprehension.
To facilitate reading comprehension, the research on the model of
generative comprehension implies that we stimulate the learners to con-
struct relations (a) among the parts of the text and (b) between the text, on
the one hand, and the readers' knowledge and experience, on the other. To
facilitate com~prehension,these constructed relations should have the
following chamcteristics. First, they must be relations that the reader would
not equally well construct without our intervention. Second, the relations
must not trivialize comprehension. They must involve more than the
learner's short-term memory and more than the surface structure of the
text. They shaillld involve the learner's long-term memory of experience or
the learner's knowledge or both of these. They should involve the text's deep
significance in the construction of one or more of its legitimate meanings.
We have lecwrned through our research studies discussed in the previous
section that we can stimulate the construction of relations having these two
characteristics by designing the reading materials i%ppropriatelyfor the
interests and abilities of the learners and by directing them to generate
meaning for the text as they read. Whether we should make the relevant
relations explicit or ask the learners to construct them is not the central
issue. In either case, so long as we do not trivialize learning, the learners can
and should be actively engaged in the understanding of the relations and in
the text's meaning. When the learners can attend to the task and can
construct the text's meaning or meanings, then they should be givien the
instructions and the direction appropriate for their developmental level,
knowledge, and background. When the learners cannot adequately attend
to the task or cannot construct important meanings from it, they should be
helped to attend to the meaning of the text, and they should be given the
relations to be learned, which they can elaborate on in an attempt to
understand and to remember them.
More precisely, the model implies that when the readers cannot attend to
the text, then training in self-managementtechniques, in cognitive strategies
of focusing attention, and in rehearsal techniques are appropriate ways to
improve the learners' attention, as are appropriate goals and objectives. The
model also imlplies that the text can be changed to provide more interesting
or more appropriate reading materials for the learners' objectives.
When the readers can attend to the text, have the appropriate focus and
knowledge, decoding skills, and vocabulary, but cannot generate the
significant relations discussed in the previous section of this article, then the
relations should be given to the learners in verbal statements or in pictures,
diagrams, graphs, and the like. To avoid trivializing learning, the readers
should not be asked to memorize these verbal statements nor should they be
asked to copy these spatial representations. The readers should be actively
engaged in testing these relations against the parts of the text or against their
knowledge and experience.
When the readers attend to the text, have the appropriate background
and knowledge, decoding skills, and vocabulary, and can but do not
generate legitimate meaning for it, then teachers should teach them learning
strategies and metacognitive processes; question them about its meaning;
give them objectives to attain; ask them to draw pictures; make predictions;
look for main ideas, sequences, cause-and-effect relations; draw inferences;
infer conclusions; construct summaries; make comparisons; and evaluate
the significance of the text. In addition, the text should include metaphors,
similes, analogies, pictures, questions, objectives, inferences, comments,

and comparisons that do not state the relations to be learned but that do
lead to their construction by the learners.
Last of all, when the readers attend to the text, have the appropriate
background knowledge, decoding skills, and vocabulary, and do generate
legitimate meaning for it, then the teacher, or the directions in the text,
should subtly suggest alternative, deep meanings that might otherwise be
omitted.
The abilities and background knowledge of the readers influence the
choice of generative activities, as this progression of teaching methods
clearly indicates. Other considerations also influence the choice and the use
of generative reading tasks. First of all, the type of text or discourse will
influence the choice of generative activities. For a technical manual, the
generative activities will sometimes be different from those chosen for an
expository text. The type of comprehension of the skills of comprehension
emphasized in a given problem will also influence the choice of generative
activities.
Even more important than the choice of a generative task is the
understanding of the nature of a generative task. Almost any of the
just-mentioned tasks or activities can be trivialized. Take metaphor for
example. In a well-written chapter on generative metaphor, Schon (1980)
showed the significance for social policy of metaphors that lead to a new
understanding of old social problems. An urban area often described as a
slum receives quite different treatment from a city government that has
learned through generative metaphor to view it as home to people whose
dislocation because of urban renewal would destroy stable social relations
and disrupt individuals' lives.
Not all constructions are generative for every reader. A generative
activity should be chosen to lead to a new or better relation. In a nontrivial
sense, a generative task must lead to the construction of a relation among
the parts of the text or between the learner's knowledge and the text.
PRINCIPLES OF THE TEACHING OF GENERATIVE

COMPREHENSION
Our studies auld the related research discussed in the previous sections
support the folllowing inferences about the teaching of comprehension.
1. The teaching of comprehension is the process of getting the learners to

generate relatimu (a) among the parts of the text and (b) between the text
and their experience.

2. Learners' knowledge, preconceptions, and experience are critical to
the design of generative instruction (Osborne & Wittrock, 1983).
3. The relations constructed by the learner must be relevant to the
comprehension that is taught and tested. When learners underline irrelevant
words in a text, construct inappropriate comments, or view misleading
pictures about the text, their comprehension usually declines markedly.
This finding supports the belief that we are working with a powerful
variable for influencing comprehension. When learners are taught to use it
appropriately, generation sizably and regularly enhances comprehension.
When learners are led to generate inappropriately, it distracts from their
comprehensioin .
4. The generation of summaries, analogies, and related constructions
functions by increasing learners' construction of relations among the
meanings in the text and between the text and learners' knowledge and
experiences. Effective summaries and related constructions involve learners'
own words and experiences.
5. To be effective, generative teaching activities induce learners to
construct relewmt representations that they would not compose spontane-
ously.
6. There is a developmental progression in children's ability to learn from
teachers' elaborations and from their own generations. Children learn from
teacher-given elaborations before they can learn as well from their own
generations.
7. Generative teaching can be either direct or indirect, structured or less
well structuredl,,depending on the learner's background knowledge, ability,
and learning strategies. Discovery is not the issue. Generation of appro-
priate relations is the issue.
8. Metacognitive strategies of comprehension can be directly taught to
learners to facilitate their ability to organize, monitor, and control their
generative thought processes.
9. Learners' thought processes can be adapted to instruction, just as
instruction can be adapted to learners. Analytic and holistic abilities can be
differentially employed by different generative activities to facilitate com-
prehension.
FOR THE FUTURE
We need further research on the following issues in the teaching of the

generative processes of comprehension.
1 . Types of main ideas and how to generate them. There are different
types of main ideas. In a narrative, the main ideas often involve under-
standing relations about characters. In other types of discourse, the main
ideas are of a different nature.
2. Story endings. What are the different categories of story endings that
children generate? How do the story endings follow from the earlier text?
How can learners be taught to generate a good story ending?
3. Summaries. The type of summaries that I teach learners to generate
differs from the conventional types of summaries taught in other research
studies or in other curriculum studies. I do not have the readers select or
modify a sentence in the text to construct a summary. Instead, I have them
use only their own words to generate summary sentences that do not appear
in the text but that do relate information from different sentences in the text
and that relate the text to their experience. We need research on the
generation of this type of summary in narration, exposition, and other types
of discourse. We need to study how to teach what a summary does in each
of the different types of discourse.
4. Titles. We need to learn the characteristics of a good title, how to
construct it, how to recognize it, and how to use it.
5. Paragraph headings and subheads. Generation of both of these types
of information facilitates comprehension. We need exercises to teach
students the characteristics of good paragraph headings and subheads and
of ways to construct them.
6. Plot structure. We need to study how to teach students the types of
plot structure, how to recognize them, and how to generate them. We need
more explicit teaching of the types of plot structure, their differences from
each other, and how to construct them.
7. Types ojfparagraphs. Descriptive paragraphs, summary paragraphs,
and the like differ in their functions in a passage. We need to teach students
how to recogruze the function of a paragraph and to generate its contribu-
tion to understanding the text.
8. Functions of similes, metaphors, and analogies. We teach similes and
metaphors, b t ~wet need to teach their essential functions in comprehension,
one of which is to relate previous experience to new events and situations in
order to transfer knowledge from familiar to new situations.
9. Relating past experience to text. This fundarnental process in the
teaching of ca~mprehensionrequires research. It needs to be approached in
a multitude of ways, including explicit teaching of how to adopt different
perspectives, to think of examples, to apply knowledge, and to test
information against what one knows or believes.
10. Generating examples and applications. This is a subhead of 9, but it
deserves speciid attention because of its importance in the teaching of
comprehension^. Many students lack the ability t~omove between the
abstraction and the specific applications, which we should try to develop
through instruction.
11. Revisin,g stored concepts by reading text. Accommodation of orga-
nized memory is different from assimilating text to memory. It is important
to comprehension of new paradigms and new systems.
12. Relatingpictures to text. The interactive qualities of pictures, that is,
their ability to show relations among the parts they juxtapose (the charac-
ters or the elements of a story or diagram), need to be taught to children.
We also need )to study how to teach them the construction of interactive
images as a way to facilitate memory and comprehension.
13. Asking questions as one reads. Recently, this topic has been widely
studied. It is an important area that involves metacognitive research and the
goals of readiing. It leads into the teaching of metacognitive strategies that
help answer learners' questions about cognition during reading: For exam-
ple, What questions should you ask yourself before you read a text? while
you are reading it? after you finish reading it?
METACOGNITIVE STRATEGIES OF GENERATIVE

READING COMPREHENSION
One of the most promising lines of research in generative reading compre-

hension is the teaching of metacognitive strategies. Our work with the
difficult reading comprehension problems of functionally illiterate young
readers indicates some encouraging possibilities. Learners need to be aware
of and to control their use of different generative learning strategies, whose
appropriateness depends on the task and its context. For example, a main
idea is different in a narrative than in an exposition. To get the main idea
and other important information, one should read a narrative story
differently from the way one should read an expository text.
It is important to teach learners the differences among types of text. It is
also important to teach children when to use different learning strategies
appropriate for different types of texts. Although it may not seem likely, it
is quite possible for children to learn a cognitive strategy in a rote or blind
fashion and to apply it without thinking.
It is important for readers to be aware of what they are doing, and why
they are doing it, as they learn to comprehend what they are reading. The
recent research on metacognition cited earlier indicates a facilitation in
learning due to awareness of and control over what one is doing while
reading. For this same reason, it is important to teach children how to
monitor and to evaluate their comprehension activities as they read. What
is their purpose in reading the passage? What comprehension strategy is
best for this type of text? What type of relation (sequence, cause and effect,
main idea, or relations among characters) across the parts of the text should
they be generating? What is the best strategy to use to generate the main
idea or another type of relation? What knowledge, schemata, perspective,
or experience is best related to the text? How well are they building relations
between the text and their background knowledge?
To teach them a strategy for reading a story, we need to teach them the
following information. A story has a grammar or a structure, including a
theme, a setting, characters, location, episodes, and a resolution. Some
stories have a structure that begins with a stable situation that is followed by
a change or disruption, a solution to the problem, and a restoration of a
desired state.
Next we need to teach them to ask themselves questions that would direct
their reading toward the progression of events in the story and to organize
its characters and events around the goals, obstacles, and outcomes that
involve the central characters in the development of the plot. We must also
teach them to see a more general significance to the story theme that would
enable them to extract its meaning or implication for their lives. Last of all
we should, if we think it is not too difficult, teach them to evaluate the form
of the story, the importance of its content, and the quality of the affect or
feelings it arouses about oneself, friends, and human emotions.
We also need to teach them to identify the type of discourse; to choose an
appropriate strategy for it; to generate appropriate main ideas, summaries,
themes, and relations between the text and their experience and knowledge.
In expository text, the pattern or relations among the parts is often
different from that found in a narrative. But some of the same generative
techniques are still useful- for example, constructing main ideas, cause-
and-effect relations, summaries, headings, contrasts, inferences, questions
to guide reading, purpose for reading, topic sentences, details, semantic

mapping, andl categorizing. All these devices take a new role in expository
text. We need to organize them, at least the most important ones, into a
coherent strategy that would teach the learners to be aware of their choice
among strategies, to monitor the strategy they were using, to construct the
structure of the expository passage, to direct their reading with self
questions, to construct appropriate main ideas and summaries, to apply the
abstractions in the exposition to their own experience, and to relate their
background knowledge to the text.
The advantages of putting comprehension activities, such as main idea,
summary, self-questions, and reader's purpose, into a coherent strategy is
that it organizes and sequences the individual generative activities that we
have studied imd found to be useful.
We also neexi research on the development and measurement of cognitive
and metacogruitive processes. Much of our research on individual differ-
ences and cognitive development deals with aptitudes and variables, such as
sex and IQ, that could be but are not process oriented. They do not help us
greatly to understand the different ways people arcquire knowledge by
organizing experience and by generating relations between experience and
memory. For example, we need to know more about the development of
imagery strategies and verbal strategies. We need tests of what learning
strategies children can use on request at different ages. We also need to
study the strategies we can teach to children at different developmental
levels to facilitate their ability to control attention and to acquire knowl-
edge.
SUMMARY
Comprehension is not the process of transforming a stimulus on a page into

a product. That digestive metaphor does not capture the generative nature
of comprehension. The neural system provides a better metaphor. Neural
systems, especially cognitive ones, are control systems. They do not
transform input into output. They generate signals, strategies, and plans
that relate events to one another and to memory to give them meaning and
significance important for understanding and survival. Comprehension
involves these generative processes.
REFERENCES
Bissex, G. (1976). DO NAT DSTRB GNYS AT WRK;.Invented spelling and beginning

reading development. Unpublished manuscript, Harvard Graduate School of Education,
Cambridge, M L ~ . .
374 WITTROCK
Bull, B. L., & Wittrock, M. C. (1973). Imagery in the learning of verbal definitions. British
Journal of Educational Psychology, 43, 289-293.
Camp, B. W. (1980). Two psychoeducational treatment programs for young aggressive boys.
In C. K. Whalen & B. Henker (Eds), Hyperactive children: The social ecology of
identification and treatment (pp. 191-219). New York: Academic.
Chomsky, C. (1979). Approaching reading through invented spelling. In L. B. Resnick & P. A.
Weaver (Eds.), Theory and practice of early reading (Vol. 2, pp. 43-65). Hillsdale, NJ:
Lawrence Erlbaum Associates, Inc.
Dansereau, D. F. (1978). The development of a learning strategies curriculum. In H. F. O'Neil,
Jr. (Ed.), Learning strategies (pp. 1-29). New York: Academic.
deCharms, R. (1976). Enhancing motivation: Change in the classroom. New York: I ~ n g t o n .
Dee-Lucas, D., & DiVesta, F. J. (1980). Learner generated organizational aids: Effects on
learning from text. Journal of Educational Psychology, 72, 304-3 11.
Doctorow, M.J., Wittrock, M. C., & Marks, C. B. (1978). Generative processes in reading
comprehension. Journal of Educational Psychology, 70, 109-1 18.
Doyle, W. (1980). Student mediation responses in teaching effectiveness (Fimal Rep. No.
0-0645). Denton: North Texas State University.
Douglas, V. I., Parry, P., Martin, P., & Carson, C. (1976). Assessment of a cognitive training
program for hyperactive children. Journal of Abnormal Child Psychology, 4, 389-410.
Graf, P. (1980). Two consequences of generating: Increased inter- and intraword organization
of sentences. Journal of Verbal Learning and Verbal Behavior, 19, 319-327.

Hallahan, D. P., & Reeve, R. E. (1980). Selective attention and distractibility. In B. K. Keogh
(Ed.), Advances in special education (pp. 141-181). Greenwich, CT: JAI Press.
Hirschman, E., & Bjork, R. A. (1988). The generation effect: Support for a two-factor theory.
Journal of Experimental Psychology: Learning, Memory, and Cognition, 14, 484-494.
Jacoby, L. L. (1978). On interpreting the effects of repetition: Solving a problem versus
remembering a solution. Journal of Verbal Learning and Verbal Behavior, 17, 649-667.
Keogh, B. K., &Glover, A. T. (1980). The generalizability and durability of cognitive training
effects. Exceptional Education Quarterly, 1, 75-82.
Kourilsky, M., & Wittrock, M. C. (1987). Verbal and graphical strategies in the teaching of
economics. Teaching and Teacher Education, 3, 1-12.
Krupski, A. (1975). Heart rate changes during a fixed reaction time task in normal and
retarded adult males. Psychophysiology, 12, 262-267.
Levin, J. R. (1981). On functions of pictures in prose. In F. J. Pirozzolo & M.C. Wittrock
(Eds.), Neuropsychological and cognitive processes of reading (pp. 203-228). New York:
Academic.
Linden, M., & Wittrock, M. C. (1981). The teaching of reading wmprehension according to
the model of generative learning. Reading Research Quarterly, 17, 44-57.
Malamuth, S. (1979). Self-managementtraining for children with reading problems: Effects on
reading performance and sustained attention. Cognitive Therapy and Research, 3,279-289.
Marks, C. B., Doctorow, M. J., & Wittrock, M. C. (1974). Word frequency and reading
wm~rehension.Journal o f Educational Research, 67, 259-262.
~ a ~ e r ,E.k (1980).
. labora at ion techniques that increase the meaningfulness of technical text:
An experimental test of the learning strategy hypothesis. Journal of Educational Psychol-
ogy, 72, 770-784.
McCombs, B. L. (1983). Motivational skills training: Helping students adapt by taking
personal responsibility and positive self-control. Paper presented at the annual meeting of
the American Educational Research Association, Montreal, Quebec.
McFarland, C. E., Jr., Frey, J. J., & Rhodes, D. 0. (1980). Retrieval of internally versus
externally generated words in episodic memory. Journal of Verbal Learning and Verbal
Behavior, 19, 210-225.
Meichenbaum, D., & Goodman, J. (1971). Training impulsive children to talk to themselves:
A means of developing self-control. Journal of Abnormal Psychology, 77, 115-126.

Osborne, R. J., L Wittrock, M. C. (1983). Learning science: A generative process. Science
Education, 67, 8489-503.
Paris, S. G., Lindauer, B. K., & Cox, G. L. (1977). The development of inferential
comprehension^. Child Development, 48, 1728-1733.
Peterson, P. L., 81: Swing, S. R. (1982). Beyond time on task: Students' reports of their thought
processes during direct instruction. Elementary School Journal, 82, 481-491.
Schon, D. A. (1980). Generative metaphor: A perspective on problemsetting in social policy.
In A. Ortony md.),Metaphor and thought (pp. 254-283). Cambridge, England: Cambridge
University Press.
Slamecka, N. J., 1k Graf, P. (1978). The generation effect: Delineation of a phenomenon.
Journal of Bperimental Psychology: Human Learning and Memory, 4, 592-604.
Slarnecka, N. J., L Katsaiti, L. T. (1987). The generation effect as an artifact of selective
displaced reheausal. Journal of Memory and Language, 26, 589607.
Stein, B. S., & bansford, J. D. (1979). Constraints on effective elaboration: Effects of
precision and subject generation. Journal of Verbal Learning and Verbal Behavior, 18,
769-777.
Stein, B. S., Morris, D. D., & Bransford, J. D. (1978). Constraints on effective elaboration.
Journal of Verbal Learning and Verbal Behavior, 17, 707-714.
Thorndike, E. L. (1931). Human learning. New York: Century.

Weiner, B. (1979). A theory of motivation for some classroom experiences. Journal of
Educational Psychology, 71, 3-25.
Weinstein, C. E., Goetz, E. T., & Alexander, P. A. (Eds.). (1988). Learning and study
strategies: Zssuc~in assessment, instruction, and evaluation. New York: Academic.
Weinstein. C. E., & Mayer, R. E. (1986). The teaching of learning strategies. In M. C.
Wittrock (Ed.), Handbook of research on teaching (3rd ed., pp. 315-327). New York:
Macmillan.
Wittrock, M. C. (1%7). Replacement and nonreplacement strategies in children's problem
solving. Journal of Educational Psychology, 58, 69-74.
Wittrock, M. C. (1974). A generative model of mathematics learning. Journal for Research
in Mathematics Education, 5, 181 -197.
Wittrock, M. C. (1974b). Learning as a generative process. Educational Psychologist, 11,
87-95.
Wittrock, M. C. (1977). Generative processes of memory. In M. C. Wittrock, J. Beatty, J. E.
Bogen, M. S. Gazzaniga, H. J. Jerison, S. D. Krashen, R. D. Nebes, & T. J. Teyler (Eds.),
The human brain (pp. 153-183). Englewood Cliffs, NJ: Prentice-Hall.
Wittrock, M. C. (1978a). The cognitive movement in instruction. Educational Psychologist,
13, 15-29.
Wittrock, M. C. (1978b). Developmental processes in learning from instruction. The Journal
of Genetic Psychology, 132, 37-54.
Wittrock, M. C. (1979). Applications of cognitive psychology to education and training. In H.
O'Neil, Jr. & (2. Spielberger (Eds.), Learning strategies (pp. 309-317). New York:
Academic.
Wittrock, M. C. (1980). Learning and the brain. In M. C. Wittrock (Ed.), The brain and
psychology (pp. 371-403). New York: Academic.
Wittrock, M. C. (1981a). Learning and memory. In F. H. Farley & N. J. Gordon (Eds.),
Educational Ps.vchology @pa242-264). Chicago, IL: National Society for the Study of
Education.
Wittrock, M. C. (1981b). Reading comprehension. In F. J. Pirozzolo & M.C. Wittrock (Eds.),
Neuropsychological and cognitive process in reading (pp. 229-259). New York: Academic.
Wittrock, M. C. (1983a). Generative reading comprehension (Ginn Occasional Reports).
Lexington, MA Ginn and Company.
376 WITTROCK
Wittrock, M. C. (1983b). Writing and the teaching of reading. Language Arts, 60, 600-606.
Wittrock, M. C. (1986a). Education and recent research on attention and knowledge
acquisition. In S. L. Friedman, K. A. Klivington, & R. W. Peterson (Eds.), Brain,
cognition, and education (pp. 151-169). New York: Academic.
Wittrock, M. C. (1986b). Students' thought processes. In M. C. Wittrock (Ed.), Handbook of
rarearch on teaching (3rd ed., pp. 297-314). New York: Macmillan.
Wittrock, M. C. (1989). Educational psychology and the future of research in learning,
instruction, and teaching. In M. C. Wittrock & F. H. Farley (Eds.), The future of
educationalpsychologv (pp. 75-89). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.
Wittrock, M. C., & Alesandhi, K. (in press). Generation of summaries and analogies and
analytic and holistic abilities. American Educational Research Journal.
Wittrock, M. C., & Carter, J. (1975). Generative processing of hierarchically organized words.
American Journal of Psychology, 88, 489-501.
Wittrock. M. C., & Goldberg, S. M. (1975). Imagery and meaningfulness in free recall: Word
attributes and instructional sets. Journal of General Psychology, 92, 137-151.
Wittrock. M. C., & Kelly, R. (1984). Teaching reading comprehension to adults in basic skills
courses (Firnal Report, Vols. 1-3). Los Angeles: University of California, Los Angela,
Graduate School of Education.
Wittrock, M. C., Marks, C. B., & Doctorow, M. J. (1975). Reading as a generative process.
Journal of Educational Psychology, 67, 484-489.
Wittrock, M. C., & Roberts, R. (in preparation). Reading comprehension and the generation
of relations between text and memory.

Educational Psychologist

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Educational Psychologist

Uploaded by

Copyright:

Available Formats

This article was downloaded by: [RMIT University]

On: 12 March 2015, At: 07:44

Generative Processes of Comprehension

To link to this article: http://dx.doi.org/10.1207/s15326985ep2404_2

PLEASE SCROLL DOWN FOR ARTICLE

Generative Processes of Comprehension

This article presents a model of the generative processes of reading compre-

instructional utility of the model. The article concludes with a discussion of

Requests for reprints should be sent to Merlin C. Wittrock, Division of Educational

RELATIONS BETWEEN READING AND WRITING

putting meaning on written pages, whereas reading is the process of getting

READING AND WRITING AS GENERATIVE

Young children generate spoken and sometimes written language before

Through their conversations and their invented spellings, preschool children

eration is not the same as semantic processing, an often cited cognitive

models of meaningful learning, semantic processing, information process-

The model of generative reading comprehension leads to several implica-

tion, or self-control, something that might respond to appropriate training.

This massive topic includes preconceptions, abstract knowledge, everyday

Essentially, reading comprehension is the process of actively generating

There are fundamental ways to stimulate generative processes. Table 1

With this introduction to the model and some of its educational

applications, I turn next to the empirical research on generation. First I

LABORATORY RESEARCH ON GENERATION

improving recall in this study. These two results seem incongruous.

learner-generated words missing in sentences with experimenter-given

semantic incongruities facilitated recall sizably. These authors disagree with

Giving learners these instructional aids enhanced learning of subordinate

APPLIED RESEARCH ON GENERATIVE READING

The data collected by my students and me on the model of generative

the results of our studies.

Note. From "Generative Processes in Reading Comprehension" by M. J. Doctorow, M. C.

with usual teaching procedures that emphasize memorizing definitions (Bull

verbal-analytic ability) correlated positively with reading scores in the

Note. TS = topic sentence.

Imaginal to verbal 13.0 27.6 28.6

Control: Conventional Instruction

to increases in ability to read with comprehension. The research also implies

THE FACILITATION OF READING COMPREHENSION

From this review of laboratory and applied research, there is substantial

Fort A" 19.38 23.83 4.25 4.7 .005 21.9

similes, analogies, pictures, questions, objectives, inferences, comments,

PRINCIPLES OF THE TEACHING OF GENERATIVE

1. The teaching of comprehension is the process of getting the learners to

and their experience.

FOR THE FUTURE

We need further research on the following issues in the teaching of the

METACOGNITIVE STRATEGIES OF GENERATIVE

One of the most promising lines of research in generative reading compre-

to guide reading, purpose for reading, topic sentences, details, semantic

Comprehension is not the process of transforming a stimulus on a page into

Bissex, G. (1976). DO NAT DSTRB GNYS AT WRK;.Invented spelling and beginning

of sentences. Journal of Verbal Learning and Verbal Behavior, 19, 319-327.

A means of developing self-control. Journal of Abnormal Psychology, 77, 115-126.

Thorndike, E. L. (1931). Human learning. New York: Century.

You might also like