Professional Documents
Culture Documents
Tip-Of-The-Tongue States - Phenomenology, Mechanism, and Lexical Retrieval
Tip-Of-The-Tongue States - Phenomenology, Mechanism, and Lexical Retrieval
Phenomenology, Mechanism,
and Lexical Retrieval
Tip-of-the-Tongue States
Phenomenology, Mechanism,
and Lexical Retrieval
Bennett L.Schwartz
Florida Intemational University
Schwartz, Bennett L.
Tip-of-the-tongue states : Phenomenology, mechanism, and lexical
retrieval/Bennett L.Schwartz.
p. cm.
Includes bibliographical references and index.
ISBN 0-8058-3445-1 (cloth : alk. paper)
1. Memory disorders. 2. Memory. 3. Recollection (Psychology).
I. Title.
BF376 .S39 2001
153.1′2—dc21
2001023110
CIP
Preface ix
1 Introduction: What is a TOT? 1
Descriptions and Anecdotes 3
Definitions 5
Brown and McNeill (1966): Seminal Research on TOTs 6
The Doctrine of Concordance 14
Summary 17
2 The Properties of Naturally-Occurring TOTs 19
TOTs are Naturally-Occurring Experiences 21
The Universality of TOTs 21
TOTs in the Field: TOT Diaries 28
TOT Phenomenology 36
Summary 47
3 Theories of TOT Etiology 48
Direct-Access Theories 50
Inferential Views 63
Summary 70
4 TOTs as a Window on Retrieval 71
Models of Lexical Retrieval 71
TOTs and Theories of Bilingual Semantic Representation 85
The Relation of Gestures, TOTs, and Word Retrieval 89
The Tip-of-the-Pen Phenomenon in Chinese Writing 90
Summary 91
vii
viii CONTENTS
5 Theories of Metacognition 92
Monitoring and Control 93
Summary 108
6 Functional Aspects of the TOT 109
Monitoring and TOTs 110
Monitoring and Control Theory and Inferential Models of TOT 111
Etiology
TOTs as Control Mechanisms 114
TOTs and Other Control Features 118
Summary 125
7 TOTs, Development, and Neuropsychological Issues 126
TOTs and Aging 127
Decrement Theory Versus Incremental Knowledge Theory 133
TOTs in Children 139
TOTs in Patient Populations 142
Summary 150
8 Conclusions and Directions for the Future 151
A Word on Consciousness 153
Related Phenomena 153
A Final Word on TOTs 155
References 157
Author Index 167
Subject Index 171
Preface
ACKNOWLEDGMENTS
In the course of compiling this manuscript, I had to acknowledge the fact that no
one person can write a book by themselves. Only one name appears on the
cover, but that does not mean that many others have not had a hand in its
conception and its execution (although all errors are my responsibility). Help is
not only desirable but imperative. I thank Leslie Frazier and Richard Berg, in
particular, for reading through drafts of the entire manuscript. Their reading and
rereading were essential to making this volume a reality. I also thank reviewer
Richard Carlson of Penn State University for his commentary and criticisms.
Other people, although they did not read drafts, also deserve my thanks.
These people were important influences on the way I think about the issues
described in this volume. In particular, I thank Janet Metcalfe for putting me on
the right track in examining these issues years ago and for getting me back on
that track from time to time during the ensuing years. I thank George Wolford
ix
x PREFACE
for handing me a copy of Brown and McNeill’s paper to read in graduate school
and first putting me on this journey. I thank Thomas Nelson and Asher Koriat
for their encouragement over the years. I thank my chairperson Marvin Dunn for
encouraging me to take on the book even if it meant delaying other writing
projects usually thought more suitable for a researcher at the current stage of my
career. Finally, I thank Judi Amsel, Bonita D’Amil, and Bill Webber at
Lawrence Erlbaum Associates. Without them, I would not have had the
confidence to begin the project, and the patience to see it through to the end.
The greatest thanks goes to my family, in particular to my parents Foster and
Carol Schwartz and my wife’s parents, Jack and Nancy Frazier. I also thank my
daughter Sarina for being a full-term baby, which allowed me to complete the
first draft of this manuscript before she arrived on the scene. Lastly, I thank my
wife, Leslie Frazier, who did it all while pregnant and applying for tenure, to
make this book a reality.
—Bennett L.Schwartz
December 11, 2000
1
What is a TOT?
Several years ago, I was in a bookstore looking for a birthday present for my
aunt, a clinical psychologist in Boston. A few months earlier I had read an
interesting novel set in the exotic landscape of New Zealand. As I entered the
bookstore, I was thinking of the complex interactions between the two main
characters, a young boy and an old woman. I started toward the information
booth to ask where I might be able to find the book, but realized that I could
recall neither the title of the novel nor the author’s name. Moreover, I was stuck
in a horrible tip-of-the-tongue state for both names. I stammered to the clerk that
I was looking for a book written by a woman from New Zealand; not much for
the store clerk to go on. I could remember the plot and the names of some of the
characters in the book, but I could not remember the two critical pieces of
information that I needed—the title or the author. In vain, I simultaneously
searched the bookstore shelves and the recesses of my memory, but to no avail.
After a half an hour or so of this pursuit, I gave up and left the bookstore.
Approximately 1 hour later, while playing basketball with friends and not
thinking about the incident in the bookstore, the title popped into my mind with
the overwhelming confidence of complete certainty. It was The Bone People by
Kerri Hulme (1983).
The tip-of-the-tongue experience itself has not escaped literary reference. The
Anton Checkov short story, “A Horsey Name” is about a tip-of-the-tongue
experience (Pitcher, 1999). In the story, a servant is trying to recall the name of
a particularly painless dentist because his employer is suffering from a
monstrous toothache. Try as he might, Ivan Yevseich cannot remember the
name of the celebrated dentist. All he can remember is that the name of the
dentist in some way sounded like something related to horses. Ivan Yevseich has
a strong feeling of knowing combined with partial information, a hallmark of the
tip-of-the-tongue state. Only after a more painful dentist has pulled the tooth
does Ivan spontaneously retrieve the dentist’s name, a certain Dr. Oates.
At one time or another, everyone has probably experienced a tip-of-the-
tongue state. These stories often cause us a great deal of frustration as we cannot
remember something, which we are sure that we know. On other occasions, they
may fill us with excitement that we will remember something important soon.
Older adults often complain that they experience more tip-of-the-tongue states
now than they did when they were younger. Regardless of our age, we all
1
2 CHAPTER 1
experience tip-of-the-tongue states, and many of us may wonder how and why
we cannot remember something we know so well.
This volume concerns what contemporary cognitive psychology knows about
the tip-of-the-tongue phenomenon, which its not much, and that is why the book
is so slim. However, I think what we do know about the tip-of-the-tongue
phenomenon will both surprise and intrigue the readers. Moreover, what we
know about the tip-of-the-tongue phenomenon has applications beyond
understanding this rather annoying but persistent quirk of the human mind.
Investigating the tip-of-the-tongue phenomenon may help us to understand how
the human mind works.
The tip-of-the-tongue phenomenon (henceforth, TOT) straddles the line
between what we think of as memory and what we think of as language, two
closely related cognitive domains that have been studied somewhat
independently of each other. Throughout the volume, a border dispute will
become apparent—is the TOT a failure of memory retrieval or is it a failure of
language production? Obviously, language production involves a memory
retrieval process, and so the processes are intertwined, but scientists who study
memory retrieval are not the same scientists who study language production.
The implications of whether the TOT is memory-related or language-related has
different implications. Consider the following example.
Political pundits used to make fun of former President George H. Bush
because of his frequent word-finding failures. Despite his obvious depth of
knowledge and expertise, his speech was sometimes characterized by pauses
suggesting a failure to recall a known word. His deficit was usually attributed to
absent-mindedness, rather than a lack of clear thinking. In other words, it was
dismissed as a language-production failure, not a more consequential memory
failure. His son, President George W. Bush, suffers from a similar affliction.
However, the son’s speech errors (e.g., “Kosovarians,” “subliminable”) are often
interpreted as a lack of knowledge, and therefore, a learning deficit; a more
consequential one for a president. Speech errors occur for both of the Bushes,
but the inference of knowledge by observers is different. A small problem with
language (George H.) is considered less detrimen-tal than a small problem with
memory (Georg W.). Research into TOTs may tell us much about the nature of
speech errors (see Levelt, 1989).
Frequently, people find themselves in a position in which they are sure that
they were supposed to remember something, but cannot remember what. We
often label this absent-mindedness. As a professor, I sometimes get to the end of
my lecture and realize I forgot to cover a topic, but I cannot recall what topic I
omitted. Most readets will recognize this experience. As you can imagine, this
experience would be difficult to study in the lab because it happens so rarely and
would be difficult to induce. TOT states may serve as a good model for these
situations because the memory and speech error can be easily induced and
studied under laboratory conditions.
WHAT IS A TOT? 3
the night and looking up the name of the first man on the moon, after a
particularly frustrating TOT. I have been told about experiencing a TOT for a
word in one language, and being able to recall its translation equivalent in
another language. Others describe that anticipating a TOT can often prompt
them to use an alternate word instead. Some people complain that now that they
are older, the number of TOTs they experience has risen dramatically. Their
observation is correct; many studies (discussed in chap. 7) suggest that TOTs are
more common among older adults than they are among younger adults.
Common to all of these descriptions is the frustrating sense of failure combined
with the certainty that the elusive word is known. Usually, a sense that
eventually the word will be remembered also accompanies a TOT.
The scientific literature on TOTs is replete with colorful descriptions of them.
The most famous and oft-quoted passage from William James captures the
feeling of the TOT particularly poetically. James (1890/1964) wrote
Brown and McNeill (1966) described TOTs in equally colorful terms: “The
signs of it [TOT] were unmistakable; he would appear to be in mild torment,
something like the brink of a sneeze, and if he found the word his relief was
considerable” (p. 326).
These descriptions of TOTs all appear to share a number of characteristics.
each of which will be examined. First, the TOT is a strong subjective experience
that is hard to confuse with any other experience. Second, TOTs are generally
accompanied by emotional feelings. Emotional frustration may occur when the
TOT is first experienced, and relief may be experienced when the TOT is
resolved. Third, in general, people have some access to characteristics of the
TOT target word. Perhaps they can recall a first letter, what the word sounds
like, or how you might say the word in French or Spanish. Fourth, people
usually do know the word and eventually retrieve it. Fifth, there is a strong
motivation to resolve the TOT, that is, to search and find the elusive target word
for yourself.
WHAT IS A TOT? 5
DEFINITIONS
the feelings associated with that retrieval. In these third-person studies, the
emphasis is usually on the word retrieval process (Kohn et al., 1987). Whereas
studying impaired retrieval of known words is important to understanding the
retrieval process, I argue it is misleading to call this condition a TOT because
the TOT should refer to a subjective state of the organism. Another term, like
availability without accessibility, would be preferable (e.g., Tulving &
Pearlstone, 1966). In this book, I draw on research using this third-person
definition and discuss it at length in chapter 4, but at the outset I emphasize that
in my view, the TOT should be conceptualized as a subjective phenomena that
only the rememberer can decide has occurred.
In the 1960s, researchers Roger Brown and David McNeill became interested in
the frustrating failure to retrieve known but inaccessible words. For several
months, they observed these TOTs in themselves. In their informal observations,
they noted some interesting features of TOTs. They observed that when
experiencing TOTs, they tended to be able to recall some of the characteristics
of the inaccessible target word. For example, one of them experienced a TOT
for the name of a street. When he experienced the TOT, he was able to retrieve
the words; Congress, Corinth, and Concord, which he was sure sounded like, but
were not, the name of the street. Investigation revealed that the actual street
name was Cornish Street. Brown and McNeill noted that they could sometimes
recall first letters, syllabic stress, and number of syllables when they were
experiencing a TOT, although they could not recall the actual target word. They
thought both the experience of a TOT and the observation that they usually
retrieved related information was important.
When Brown and McNeill began to examine TOTs, there was scant
published research on the topic. James (1890/1964) wrote eloquently about
TOTs, he had not carried out any specific research on the topic. Thus, Brown
and McNeill’s 1966 article achieved many significant landmarks in the study of
TOTs. First, Brown and McNeill introduced the first empirical methodology for
studying TOTs in the laboratory. Second, the study established the first
empirically verifiable observations about TOTs. Third, the article introduced
many of the important theoretical issues that have dominated the study of TOTs
over the ensuing four decades.
Because it is the seminal article in the area, and because it set the tone of
research on TOTs for the next 35 years, I will devote much of the remainder of
the first chapter to Brown and McNeill’s methodology, findings, and theory. The
strong points of their paper have led research in several productive directions,
although the weaknesses of their paper have, to some extent, delayed research
WHAT IS A TOT? 7
Methodology
Brown and McNeill’s (1996) “prospected” for TOTs. They did this by reading
definitions of words to a class of 56 Harvard and Radcliffe students. The words
were difficult and uncommon words, but Brown and McNeill suspected that
they might be in the participants’ recognition vocabulary, although they might
not be able to generate the words on demand. Examples of their stimuli were the
definitions for such words as “nepotism,” “ambergris,” and “cloaca.” The
definitions were adapted from a dictionary, but Brown and McNeill did not
provide any examples of the definitions, just some of the words used. Thus, the
definition for “cloaca” might have been something like “the cavity into which
both the intestinal and genitourinary tracts empty in reptiles, birds, amphibians,
and many fish” (Guralnik, 1984, p. 267). The participants were instructed to
raise their hands if they were experiencing a TOT for any particular word and
then answer a series of questions pertaining to that word. While the students
who were experiencing TOTs were reporting on their experiences, the remaining
students waited for the next item. Brown and McNeill defined a TOT as follows:
“If you feel you are unable to think of the word but feel sure that you know it
and that it is on the verge of coming back to you then you are in a TOT state …”
(p. 327). Today, participants are more likely to report their TOT on a computer
keyboard isolated from others, but in Brown and McNeill’s case, these
instructions were given to a large group. This minimizes any effects of the social
milieu. However, Brown and McNeill conducted all of their data gathering in
one group setting.
8 CHAPTER 1
Results
Brown and McNeill (1966) showed that it was possible to study TOTs in an
experimental setting. Participants readily reported TOTs, and these turned out to
mostly correspond to cases in which the person later correctly identified the
WHAT IS A TOT? 9
target word. Moreover, they found that when rememberers are experiencing a
TOT, they usually have a wealth of partial information, ranging from the
retrieval of similar sounding words to correctly estimating the correct number of
syllables of the missing target word. All of their findings turned out to be
replicable. In retrospect, it is a shame that Brown and McNeill moved onto other
issues after the publication of this article, and they never wrote another article on
TOTs. The organization of this section loosely follows the manner in which they
reported their data in their article.
Brown and McNeill (1966) collected 360 TOTs from 56 participants. Of
these, 233 were considered to be “positive TOTs.” Positive TOTs were those in
which participants resolved their TOT or recognized the experimenter’s word as
correct. TOTs were scored as positive even if the participant retrieved a word
that was not technically correct but for which the participant was sure that it was
the word at which the TOT was directed. Negative TOTs were those in which
the participant did not judge the actual word to be the one for which they were
experiencing a TOT or that they could not recall the actual word for which they
had experienced a TOT. Negative TOTs were far outnumbered by positive
TOTs, but it is unclear due to the inclusion of relatively difficult words in the
study whether this is a general feature of TOTs.
This distinction between positive and negative TOTs has persisted in more
recent investigations (e.g., Perfect & Hanley, 1992; Rastle & Burke, 1996). In
their results, Brown and McNeill (1966) specify some analyses in which they
only addressed positive TOTs, and therefore, presumably in others, both positive
and negative TOTs were included. Some researchers do not think of negative
TOTs as TOTs at all, and exclude them from analysis (e.g., Rastle & Burke,
1996). This exclusion is generally motivated by an interest in using positive
TOTs to study the retrieval process. Brown and McNeill’s reason for
distinguishing positive and negative TOTs was probably to address how they
differed.
Brown and McNeill (1966) also introduced the distinction between
phonological access and semantic access in TOTs (see Burke et al., 1991;
Caramazza & Miozzo, 1997). This distinction is that TOTs may elicit two kinds
of retrieved information; (a) partial information about the phonology of the word
or phonologically related words, and (b) semantic information, such as related
information or associatively related words. Phonological information includes
remembering the first letter, the first syllable, or a word that sounded similar to
the target (e.g., “croquet” for “cloaca”). Semantic information, such as related
information, entails remembering details of the functioning of the digestive
system, and associated words might include “intestine” or “crop.” Brown and
McNeill examined this distinction by asking participants to report words that
sounded similar to the target word and words that meant something similar to
the target word.
10 CHAPTER 1
TABLE 1.1 Actual and Guessed Number of Syllables for all TOTs in the Main
Experiment
In total, Brown and McNeill’s (1966) data paint a rather revealing portrait of
the TOT. TOT experiences are highly correlated with eventual recognition of the
target word. These experiences are often accompanied by the retrieval of words
that are either phonologically or semantically related, or both, to the missing
target word. Moreover, when a person is in a TOT they can accurately estimate
the number of syllables in the word, the stress pattern of those syllables, and the
first and last letter of the missing word. Thus, Brown and McNeill demonstrated
that the TOT is not an idle subjective experience; rather, it is accompanied by a
tremendous amount of objectively verifiable information.
Theory
Brown and McNeill (1966) realized that their study raised more questions than it
answered. However, they drew some tentative theoretical conclusions based on
the implications of their data. Their theoretical speculation became as influential
as their pioneering methodology. Brown and McNeill conducted their study just
as information-processing theory was rapidly developing, and therefore, they
drew heavily on the information-processing revolution in cognitive science to
explain TOTs. In this approach, of course, the emphasis is on hidden cognitive
processes, and like much cognitive psychology of the time, does not wish to
invoke conscious or phenomenological entities in explanation. Thus, their model
is essentially a serial information processing view of the TOTs, in which a
system initiates a search that yields only incomplete information.
Brown and McNeill (1966) introduced a concept that they called generic
recall, that the general sense of the target item is recovered without the specific
target word being retrieved. In their view, TOTs occur because the rememberer
accesses generic information, but not specific information. In this sense, Brown
and McNeill’s model is a direct-access model (see chap. 3) because it is the
failure of retrieval combined with access to the unretrieved word that elicits the
TOT experience. Generic recall drives TOTs, and TOTs are accurate predictors
of both recall and recognition performance.
Brown and McNeill (1966) also couched their explanation in terms of extant
models of word recognition. Generic recall is generally functional because the
letters specifying words in English, as in other languages, usually overspecify
the word. By overspecification they mean that, in most cases, the change of a
single letter changes a word from a real word to a nonsense word. For example,
changing one letter of the word “sextant” yields nonwords such as “textant,”
“sixtant,” and “sektant” (p. 335). Thus, generic recall is beneficial to the
rememberer even when specific word knowledge cannot be accessed because
often the generic information is all that is required to produce a comprehensible
utterance. Thus, if the listeners know I am talking about a nautical navigation
tool, saying “sektant” instead of “sextant” probably does not compromise the
comprehensibility of the sentence. Theoretically, the influence of this model,
WHAT IS A TOT? 13
and feeling-of-knowing became judgments is intuitive and natural, but the two
literatures have evolved largely independently (Schwartz, 1994, 1999b). Indeed,
as the TOT became mainly studied by psycholinguists, and the feeling-of-
knowing mainly studied by memory researchers interested in metamemory, the
two literatures seldom overlapped (for exceptions, see Gardiner, Craik, &
Bleasdale, 1973; Yaniv & Meyer, 1987; for a review, see Schwartz, 1999b). Had
Brown and McNeill discussed Hart’s work, it is possible that the two literatures
would not have become so segregated from each other. Given that Hart
published his article just a year earlier in the journal of Educational Psychology,
it is likely that Brown and McNeill simply had not seen or read it.
Despite these problems, Brown and McNeill (1966) set the stage for
subsequent research in the area, and no paper on TOTs has been more widely
referenced or had a greater impact than theirs. I recently surveyed my collection
of cognition and memory textbooks and found that roughly 75% discuss the
TOT, and most of these pay homage to the seminal work of Brown and McNeill.
Alan Brown reported a similar result in a survey he did a few years ago with
similar high marks for the Brown and McNeill study (Brown, 1991).
phenomenology of TOTs are the same process. When exceptions to the doctrine
are acknowledged, they have been considered to be unimportant (e.g., Astell &
Harley, 1996; Perfect & Hanley, 1992; Rastle & Burke, 1996). There are two
major consequences of this assumption. First, researchers assume variables that
affect retrieval also affect TOTs. Therefore, they claim that one can infer the
nature of retrieval by studying TOTs. Second, until now, most research studying
TOTs has implied that phenomenology is equivalent to retrieval because both
tap the same underlying processes. Therefore, little research has been directed
toward TOT phenomenology. I claim that new TOT research refutes the doctrine
of concordance (thereby supporting Tulving’s view) by demonstrating that the
processes of retrieval are not always identical to the processes that elicit TOTs.
There is growing evidence to suggest that the etiology of TOTs differs from the
processes that effect retrieval, leaving open the possibility for a breakdown in
concordance. In chapter 3, I present research from my own laboratory and from
others suggesting that the TOT phenomenology is dissociable from the
processes of word retrieval.
Despite these strong claims, I present the overwhelming evidence that TOTs
and retrieval are inextricably linked. Thus, although equivalence between TOTs
and retrieval cannot fully explain the phenomenology, there is evidence to
suggest that, to a large extent, the process of retrieval does guide TOTs;
however, it is not the only causal agent of TOTs. TOT etiology is a complex
process, presumably consisting of both retrieval-driven processes and
nonretrieval-driven processes. I consider models in which the TOT is based on
retrieval processes (chap. 4) and models in which the TOT is largely based on
nonconscious inference (chap. 5).
Returning to the classic work of Brown and McNeill (1966), who marveled at
the strength of the TOT experience, these researchers commented on the effects
of TOT on them and their participants. Brown and McNeill were clearly
interested in the phenomenological experience, but their empirical focus was on
the types of retrieval processes engaged in during a TOT which has remained
the focus of TOT research. This book’s goal is to add to that focus and
concentrate on why the phenomenological experience is equally important to
understand. Nonetheless, to understand the TOT, it is important to relate both its
role in language and memory retrieval, in addition to its metacognitive origin.
SUMMARY
Brown and McNeill (1966) demonstrated that TOTs could be studied in the lab
and are usually associated with accurate retrieval of partial information. I also
reviewed Tulving’s (1989) doctrine of concordance and explained how the
doctrine’s general failure can be used to understand the complex relation
between TOTs and word retrieval.
2
The Properties of
Naturally-Occurring TOTs
19
20 CHAPTER 2
Another important reason justifying the study of TOTs is that they appear to be a
universal experience across cultures and languages. If TOTs are constant across
cultures and languages, then what we find out about English-speaking
Americans may indeed generalize to other populations. This is a strong
statement to make given that most of the research on TOTs has taken place in
Western societies among educated peoples, but I will try to make an argument as
to why I think the statement is true. The rationale comes from a handful of
studies done on TOTs in different languages, and a recent survey that I
conducted (Schwartz, 1999b).
Brown (1991), in his review of TOTs, lamented the paucity of cross-
linguistic research on TOTs, and he called for more. At that time, there was one
published study with Polish-speaking people (Bak, 1987) and several with
Japanese-speaking people (e.g., Murakami, 1980). Since then, TOT research has
been conducted in German (Priller & Mittenecker, 1988), Japanese (Iwasaki,
Vigliocco, & Garrett, 1998), Italian (Miozzo & Carmazza, 1997; Vigliocco et
al., 1997), Spanish (Gonzalez & Miralles, 1997, as cited in Vigliocco et al.,
22 CHAPTER 2
1999), Farsi (Askari, 1999), Hebrew (Faust, Dimitrovsky, & Davidi, 1997;
Silverberg, Gollan, & Garrett, 1999), and American Sign Language (Marsh &
Emmorey, 2000). Ecke and Garrett (1998) looked at TOTs in translating from
one language to another. Taken together, the studies demonstrate striking
parallels. In each language, people are able to access semantic, phonological,
and, in some cases, syntactical elements of the missing target word.
I choose a different method to address whether the TOT represents a
universal experience across language and culture. Rather than run the Brown-
McNeill (1966) paradigm in an assortment of languages (which clearly is a
desirable strategy), I began by conducting a survey of languages (Schwartz,
1999b). The survey attempted to assay TOT universality by tracing the way that
the experience is described in each language. The survey simply asked people to
identify the term that they would use to describe a feeling that a word is known,
but cannot be recalled at that instant.
At issue here is whether there is a phenomenological basis for referring to an
instance of a feeling of temporary inaccessibility by the term “tip of the tongue.”
Possibly other languages have a literal expression that means “feeling of
temporary forgetting” and do not need to employ the metaphor that we use in
English. I suspected that it is likely that the term “on the tip of the tongue” is a
metaphor that was derived to capture the phenomenological experience.
Temporarily forgotten words feel like they are on the tips of our tongues.
Metaphors, in memory speech, are actually quite common. For example, you
may not be sure of a person’s name, but his face may “ring a bell.” If the TOT
experience is a universal one, it is likely that the “tongue” metaphor is common
to many languages and to examine the existence of the expression “on the tip of
the tongue” in different languages may be an effective research tool.
In the survey (Schwartz, 1999b), I contacted as many fluent, preferably
native, speakers of as many languages as I could. I tried to contact more than
one speaker for each language but for some languages, like the Cheyenne
language (a Native American language), that proved impossible. I asked each
person if they would provide for me the expression that most closely matched
what we (English speakers) mean in the expression “on the tip of the tongue.” I
made clear that I did not want a literal translation. Rather, I wanted the
expression that meant the same thing in that language, that is, how one expresses
the feeling of temporary inaccessibility in the non-English language.
There were many obvious flaws in the survey. The sample consisted only of
people who were at least partially fluent in English. How knowledge of English
influenced their reporting of their other language is anybody’s guess. The
majority of the informants were professors, students, or otherwise well-educated
people, although some languages were provided by a chance encounter, for
example, with an Amharic-speaking cab driver or a Hebrew-speaking waiter.
Moreover, the sample of languages is not representative. Dozens of Indo-
European languages are represented, but only one Native American language,
PROPERTIES OF TOTS 23
one native Australian language, and one sign language (American Sign
Language) are represented. All of the languages are currently living languages. I
was unable to find a scholar of ancient Latin, ancient Greek, Aramaic, or
Sanskrit who could confidently provide if the term existed in these extinct
tongues. Despite all these difficulties, the results of the survey are intriguing.
Across 51 languages known by the informants, 45 languages employed an
expression that used the “tongue” metaphor to express the feeling of not being
able to retrieve an accessible word (see Table 2.1). Amharic, American Sign
Language, Icelandic, two sub-Saharan African languages (Kiswahili, Kalenjin),
and Indonesian do not use the metaphor of the tongue to refer to the TOT. The
most common term was an equivalent of the English expression, “on the tip of
the tongue” (see Table 2.2). For example, the expression in Italian is “sulla
punta della lingua,” and in Afrikaans, “op die punt van my tong,” which both
translate to “on the tip or point of the tongue.” Other languages use slight
variants. For example, Estonian (“keele otsa peal”) translates to “at the head of
the tongue,” whereas Cheyenne (“navonotootse’a”) translates as “I have lost it
on my tongue.” Irish (“ar flaen fy nhafod”) refers to the “front of the tongue,”
whereas Welsh (“ar bharr theanga agam”) translates to “on the top of my
tongue.” Marathi (“jeebhe-war”) simply uses “on the tongue.” Perhaps the most
poetic is Korean, in which the expression (“Hyeu kkedu-te mam-dol-da”) means
“sparkling at the end of my tongue.” The general similarity of the expressions
suggests that the “tongue” metaphor captures a common phenomenology across
languages.
French
German
Gujarati
Haitian Creole
Hindi
Irish
Italian
Marathi
Norwegian
Polish
Portuguese
Russian
Serbo-Croatian
Slovak
Spanish
Swedish
Urdu
Welsh
Yiddish
Non Indo-European Non Indo-European
Arabic American Sign Language
Chamarro Amharic
Cheyanne Indonesian
Chinese (Cantonese) Kalenjin
Chinese (Mandarin) Kiswahili
Estonian
Finnish
Hausa
Hebrew
Hungarian
PROPERTIES OF TOTS 25
Ibo
Japanese
Jaru
Korean
Tamil
Turkish
Vietnamese
Note. On the left is a list of languages surveyed that use an expression equivalent to
“on the tip of the tongue” to express a state of temporary forgetting. On the right are the
languages that do not use the “tongue” metaphor. Adapted from Schwartz, 1999b.
Swedish
“On Tip/Point/Head of the Tongue”
Afrikaans
Arabic
Armenian
Bulgarian
Chamarro
Czech
Danish
Dutch
English
Estonian
Farsi
Finnish
Flemish
French
German
Gujarati
Haitian Creole
Hebrew
Hungarian
Italian
Polish
Portuguese
Russian
Spanish
Tamil
Turkish
Urdu
Vietnamese
PROPERTIES OF TOTS 27
Yiddish
“On the Top of the Tongue”
Irish
Serbian
“On the Front of the Tongue”
Welsh
“Sparking at the End of the Tongue”
Korean (1)
“In the Mouth and Throat”
Cantonese Chinese
Mandarin Chinese
Hausa
Hindi
Ibo
Japanese
Korean (2)
Note. Adapted from Schwartz, 1999.
Nonetheless, these data suggest that the term tip of the tongue derives from
the experience of being unable to remember a known word. In a
phenomenological sense, the word genuinely feels as if it is on the tip of one’s
tongue, regardless of whether one speaks English, Arabic, or Chamarro. Because
these languages evolved in different parts of the world at different points in
time, and yet each uses the same phrase, it argues for the universality of the
experience of the tip of the tongue. Unlike many laboratory phenomena, the
TOT exists in most cultures, in most languages, and presumably, in the past
languages as well as present ones.
I also wondered what the direction of causality was in the use of this
metaphor. Does the feeling precede our choice of words, or does the existing
expression “on the tip of the tongue” suggest to us how to describe the
experience. In other words, would a person who had never heard another person
use the expression “on the tip of the tongue” spontaneously use the expression to
describe the frustrating feeling of temporary inaccessibilty? Or, would a person
unfamiliar with the expression describe this feeling in a completely different
way? It is possible that the existing expressions affect our experience, causing us
to attribute the feeling to our tongues. On the other hand, it is possible that the
28 CHAPTER 2
metaphor is driven by a feeling that the word is right there on the tongue.
Originally, this question seemed to be quite unanswerable. However, some
recent research suggests that the term “tip of the tongue” may reflect actual
phenomenology. Marsh and Emmorey (2000) recently looked for “tip-of-the-
finger experiences” in deaf users of American Sign Language. Although the
term “tip-of-the-finger” does not exist in American Sign Language, the signers
indicated that they had often had such experiences, and that the term had some
appeal to them. This suggests that the experience of the word being on the
tongue (“finger”) may exist even in those not familiar with the term TOT.
Based on this survey data and the ready comprehension of the term TOT by
the participants, I have become convinced that the TOT represents a universal of
human experience, common across languages, cultures, ages, and backgrounds.
The expression is much the same from language to language. Thus, it is highly
likely that the phenomenological experience of the TOT is similar from person
to person and culture to culture. As such, it represents a genuinely unique aspect
of cognition that is not specific to literate readers, or rememberers in a particular
cultural milieu. It may, therefore, represent a unique lens through which human
phenomenology can be studied.
records various features of the TOT on an entry form (e.g., Burke et al., 1991).
These features may include partial semantic and phonological information,
phenomenological experiences, attempts or successful resolution, and if
resolved, what the word is (see Burke et al., 1991; Schwartz, 1999b).
There have been five reported diary studies (Burke et al., 1991; Ecke, 1997;
Heine, Ober, & Shenaut, 1999; Reason & Lucas, 1984; Schwartz, 1999a). From
these studies, I have extracted a general portrait of the properties of a TOT. To
give an initial summary, most TOTs are strong feelings with accompanying
emotion. They are typically accompanied by a feeling of imminence, that is, the
feeling that the word will be recalled soon. TOTs are generally experienced for
proper nouns, although other nouns also make up a sizable portion of the TOTs.
Many TOTs are also accompanied by words that seem to block the retrieval of
the correct word. Almost all TOTs are resolved, usually within a few minutes of
the original experience. TOT resolution is achieved most often by spontaneous
retrieval, although active search and consultation with others are often
important. Finally, older adults experience more TOTs than younger adults.
This general description should accord with readers’ experiences of TOTs
because these results are based on ordinary naturally occurring TOTs recorded
by diarists. As researchers, of course, we want more than just a description of
the experience. We want to know why and how these phenomena occur. Thus
we discuss each of these results.
Schwartz (1999a) asked diarists to record whether each TOT experienced was
(a) strong or weak, (b) accompanied by a feeling of emotion, or (c) accompanied
by a sense of imminence. Strength was defined as the intensity of the TOT
feeling; emotionality was defined as a feeling of frustration at being unable to
retrieve the target; and imminence was defined as the feeling that retrieval was
about to occur. Diarists recorded their answers on an answer sheet when they
recorded the TOT.
Schwartz (1999a) found that 66% of TOTs were judged to be strong, 61%
were judged to be emotional, and 65% were judged to be imminent. Furthermore
strength, emotionality, and imminence were correlated with each other, but not
perfectly so. In fact, the correlations were between .20 and .30 for each
correlation. Imminence was predictive of successful recognition, but oddly,
emotionality was predictive of an inability to resolve the TOT. I speculated that
emotionality in TOTs might be indicative of the presence of a retrieved word
that blocks or interferes with the retrieval of the correct target.
30 CHAPTER 2
TOT Words
Diary studies show that TOTs are usually experienced in everyday life for
proper nouns. Burke et al. (1991) found that across age groups, 65% of TOTs
occurred for proper nouns (e.g., “Chuck Berry”). Abstract words (e.g.,
“idiomatic”) followed, with the fewest TOTs occurring for object names. These
findings have also been replicated by Schwartz (1999a), who found a 68% rate
for proper names (see Table 2.3). However, Ecke (1997) found only a 33% rate
for proper names, but found that another 45% were for nouns. None of the
researchers speculated as to why proper nouns are so common TOT inducers.
However, it is likely that proper nouns are often important to remember and are
frequently discussed. However, a name is often an arbitrary associate to the item
it names. For example, names of people in American culture are somewhat
arbitrary designations, such as “Richard Berg.” That name tells us almost
nothing about the person, except perhaps some of his ancestors may have
spoken German. But even people’s names are not as seemingly arbitrary as the
names of American pop music groups, which were a particularly common TOT
target in the Schwartz (1999a) study. The relation of “Afgan Wings,” “Counting
Crows,” “10,000 Maniacs,” and “Limp Biskit” to groups of musicians is very
arbitrary. I wonder if those societies that name people by salient characteristics
of the person (e.g., “Scarface,” “Dances with Wolves”) show lower rates of
proper noun TOTs. This issue could be studied in a laboratory study, in which
nicknames are assigned to faces. Those that are based on salient characteristics
should be less likely to generate TOTs than those nicknames that are purely
arbitrary.
Blocking
One diary study (Reason & Lucas, 1984) explored an issue called blocking,
which means the retrieval of words known to be incorrect that then interfere
with the retrieval of the correct word. For example, somebody might be trying to
think of the word “rhinoceros,” but failing to retrieve that word might keep
retrieving “hippopotamus” (a semantic blocker) or “rhinovirus” (a phonological
blocker), which are known to be incorrect (see Table 2.4 for more examples).
Some theories of TOT etiology argue that TOTs arise when an incorrect word is
retrieved and temporarily blocks retrieval of the correct target. Reason and
Lucas (1984) were interested in what percentage of resolved TOTs would be
reported as resulting from blocking. They found that over a 4-week period,
participants averaged 2.5 resolved TOTs. More specific to their interests,
Reason and Lucas found that over 50% of resolved TO Ts were preceded by
intrusive blocking words; words that kept resurfacing, but were recognized as
incorrect. In a second study, with fewer participants and fewer TOTs, they found
that 70% of resolved TOTs were preceded by intrusive blockers. However, their
sample size was low (only 115 TOTs between two studies), and thus, may be
less than representative.
Burke et al. (1991) also looked at blockers, although they called the blockers
by a different name (persistent alternates). For the young participants, 67% of
TOTs were accompanied by alternates. However, as people got older, the
number of alternates decreased. In the older group, only 48% of TOTs were
accompanied by alternates. Burke et al. found that nearly 90% of the alternates
were from the same syntactic category as the missing word, that over one third
shared the initial phoneme, and almost half shared the same number of syllables
(see Table 2.4). Burke et. al. suggested that the decreased number of persistent
alternates may be due to decreased phonological access in older adults (I explain
this view later). This would account for the decrease in alternates when many
theories of cognitive aging suggest that older adults become less efficient at
screening out alternative retrieved responses (e.g., Salthouse, 1991).
The interpretation of blockers is problematic, however. Both Reason and
Lucas (1984) and Burke et al. (1991) demonstrated a correlation between TOTs
and blockers, but it is unclear whether the presence of blockers drives the TOT,
32 CHAPTER 2
whether the TOT creates the search for alternates, or whether each is a product
of a slowed retrieval process. Nonetheless, the perseverance of these blockers is
interesting. Rememberers know that they are not the right word, yet they have
difficulties preventing their repeated retrieval when trying to get the target word.
The topic of blocking will be revisited in chapter 3.
Resolution
Burke et al. (1991) found high rates of resolution for TOTs; roughly 95% across
the three age groups. They also found a trend toward higher accuracy with age
(see Table 2.3). The young adults resolved 92% of TOTs, whereas the middle-
aged adults resolved 95%, and the older adults resolved 97% of TOTs. This
PROPERTIES OF TOTS 33
trend was also replicated by Heine et al. (1999) who found 91% resolution in the
younger adults, 95% in the older adults, and 97% resolution in the oldest. Ecke
(1997) found 89% of TOTs were resolved in one’s native language, but
interestingly, 96% of TOTs in a foreign language were resolved. Schwartz
(1999a) found an 89% resolution rate when sampling with a young-adult
population. Although ceiling effects prevent the definitive conclusion that older
adults are more likely to resolve their TOTs than younger adults, the data
certainly point in that direction. I delay discussion of why this interesting age
difference might occur until chapter 7.
Resolution Time
Diary studies also allow for an analysis of how long it takes people to remember
the target word of naturally occurring TOTs. In the laboratory, participants are
constrained by the nature of the experiment as to how long they can spend
attempting retrieval, but in the natural world, the person may return to the search
at any time. Unfortunately, unlike the other measures, retrieval times have
differed substantial0ly among the existing diary studies. For example, Schwartz
(1999a) found that when deliberate search was the method of resolution, it took
participants 14.8 min to resolve the TOT. However, Burke et al. (1991) found
that it took participants only 4.3 min. It is unclear why these differences are so
large. Alternately, Burke et al. found that spontaneous retrieval took 22.9 min,
whereas Schwartz found that it took only 15.3 min. Hopefully, future diary
studies will address this issue and find more reliable methods of measuring
resolution time.
Resolution Method
Another important issue that can be addressed in diary studies is how people
naturally resolve their TOTs. In the laboratory, participants are tested under
relatively tight time constraints. Participants know that they have to go through a
large number of items, and that they should not delay too much on any particular
item. They also do not have access to outside sources. Thus, their ability to
resolve TOTs is limited. In diary studies, people can resolve TOTs in a variety
of ways over a variety of time frames. The diary studies identify three main
methods of TOT resolution. First, participants can engage in active and
deliberate search. Second, they can consult with other people or look up the
answer in books, on Web sites, or through other outside means. Third, they can
wait until spontaneous retrieval or “pop-up” occurs. The data from the three
diary studies is presented in Table 2.3. The two main findings are that
spontaneous retrieval is the method most often resulting in TOT resolution, and
that the reli ance on spontaneous retrieval increases with age.
34 CHAPTER 2
Age
Age differences are perhaps the most studied issue in TOT research. The general
finding is that older people experience more TOTs than younger people. This
basic finding has been shown in the diary study by Burke et al. (1991) and in the
diary study by Heine et al. (1999). In the study by Burke and colleagues, 130
participants were each asked to carry a TOT diary for 4 weeks. The participants
were divided into three age groups: college-age adults, middle-aged adults, and
older adults. The older (mean age=71.0) and middle-aged (mean age=38.7)
participants scored higher on the WAIS (Wechsler Adult Intelligence Scale)
vocabulary than did the younger adults (mean age=19.4), but the younger adults
had higher digit spans. Education levels were roughly equivalent.
TOTs varied significantly with age. The younger participants experienced
only 3.92 TOTs over the 4-week period, for an average of slightly less than 1
per week. Middle-aged participants experienced 5.4 TOTs over the 4-week
period, and older adults averaged 6.56 TOTs. Thus, their notion that natureally
occurring TOTs would increase with age was confirmed. Heine et al. (1999)
found an even higher TOT rate (9.33) for the oldest adults (participants over 80).
Thus, these findings confirm an a priori idea about TOTs. First, older adults
do experience more TOTs than younger adults. However, because a longitudinal
study has not yet been conducted, it is not quite definitive that this is really a
function of age or other factors that might be correlated with age. Because many
older people complain they experience more TOTs now than they did when they
were younger, such a longitudinal study would be interesting. Alternately,
contrary to expectations, it is the older adults who are actually better at resolving
their TOTs. There are a number of reasons why this may be so. I return to this
issue in chapter 7.
Diary studies, although few in number, have already generated some interesting
and replicable findings. First of all, the diary studies show that proper names are
the kind of word retrieval most likely to induce a TOT. Second, most TOTs are
eventually resolved (roughly 90%). Third, resolution of TOTs derives from a
combination of methods such as deliberate search, spontaneous retrieval, and
consulting outside sources. These conclusions were originally shown by Burke
et al. (1991) and then replicated more recently by Heine et al. (1999) and
Schwartz (1999a). Finally, the various studies converge on the broad
generalization that the typical person, on average, experiences a TOT on an
almost weekly basis. This rate increases for older individuals.
The diary methodology offers several advantages and disadvantages over
laboratory tests of TOTs. I will outline the advantages first. In a diary study, the
PROPERTIES OF TOTS 35
TOTs originate from everyday life. Therefore, they directly reflect the
researcher’s interest in the natural phenomena. Analysis, therefore, of the kinds
of words for which people experience TOTs is illuminating. Thus, for example,
it gives us an accurate reflection of how often TOTs occur for proper and
common nouns. In contrast, laboratory studies rely on lists of general-
information questions or word definitions. Thus, the ratio of proper nouns to
common nouns to other kinds of words may be distorted because the
experimenter provides the stimuli, and they are not the kinds of information that
people are likely to discuss normally. Thus, the diary methodology gives a more
ecologically valid picture of the kinds of materials for which TOTs occur.
Second, diary studies also allow researchers to better address issues of
resolution. Because laboratory studies are usually conducted on relatively tight
time frames, most researchers are only able to look at immediate resolution, that
is, if the TOT is resolved within a few minutes of the onset of the TOT.
However, diary studies show that many TOTs are resolved hours, if not days,
after the original experience. Diary studies allow for that extended window on
resolution. Related to the time course of resolution is the manner in which a
TOT is resolved. TOTs may be resolved by an active memory search,
spontaneous retrieval, consulting with other people, and by consulting other
sources. Some TOTs, of course, may never be resolved.
One of the main drawbacks to the diary study is that it does not allow
comparisons between TOTs and n-TOT s (unrecalled words not in TOT states).
N-TOTs, or “don’t know” states, in which people cannot recall the target but are
not experiencing a TOT, provide a crucial comparison to TOTs. If we wish to
know what is unique about a TOT, we must compare it to n-TOTs, regardless of
whether we are looking at resolution rates, partial information retrieval, or
recognition. For example, we should expect more target knowledge during
TOTs than during n-TOTs. However, because the diarist can only record exp-
erienced TOTs and not equivalent control items, the diary studies are lacking in
the kind of controls that laboratory studies can utilize to clarify results. For
example, in a laboratory, one can examine differences in resolution rates for
unrecalled targets for both TOTs and n-TOTs. However, in the diary study, it is
impossible to get people to record items that they do not know and did not try to
recall. Thus, the diary study methodology is most successful when used in com-
bination with laboratory studies (e.g., Burke et al., 1991; Heine et al., 1999).
Another drawback of diary studies is, of course, that by gaining ecological
validity, one loses experimental control. The researcher no longer has control
over input parameters, the situations in which the diarists find themselves, or the
kinds of words that they need to retrieve in ordinary life. Nor is there a sufficient
database on TOTs for the researchers to covary any of these variables in a
meaningful way. With the exception of age, it is unclear whether any kind of
individual difference (e.g., IQ, spatial ability, verbal ability, etc.) is correlated
with TOT rate (but see Dahlgren, 1998; Frick-Horbury & Guttentag, 1998).
36 CHAPTER 2
Thus, diary studies do not necessarily give us a complete picture of what a TOT
is, but they do provide us with some interesting ideas for laboratory testing and
allow us to validate lab findings in the real world.
TOT PHENOMENOLOGY
The final issue to be addressed in this chapter is the clear and strong feeling that
defines the TOT. Phenomenological salience can be thought of as being the key
element of TOTs. When people enter a TOT, they experience a feeling that they
will be able to remember a missing target. Informally, the feeling has been
described by numerous researchers; the most well-known being James, quoted
in chapter 1. Brown and McNeill (1966) described TOTs in the following way:
“The signs of it [TOT] were unmistakable; he would appear to be in mild
torment, something like the brink of a sneeze, and if he found the word his relief
was considerable” (p. 326). A.S. Brown wrote, in reference to James and to
Brown and McNeill that, “Such descriptions are congruent with one’s personal
introspections of inner turmoil when grappling for an elusive word” (p. 205).
However, no empirical research has been directed at whether torment, turmoil,
or relief is involved in the experiencing of TOTs. Most research has been
content to note the phenomenology but not explore it in any systematic way.
Because the phenomenology of TOTs has been largely ignored, there is little
existing research examining the experiential components of TOTs, or how such
components might relate to objective retrieval behavior. Therefore, the
following section represents a preliminary sketch as to what the important issues
in TOT phenomenology may be, based primarily on some of my recent research
(Schwartz et al., 2000; see also Burke et al., 1991, p. 560). The focus of my
studies over the past few years has been the experience of strength, imminence,
and emotionality that appear critical to TOTs. I have also addressed, albeit only
descriptively, feelings of relief that may follow TOT resolution and also stability
and fluctuation in TOTs over time. Feelings of relief refer to the emotion that
occurs once a TOT has been resolved. Stability and fluctuation refer to how long
a person actually experiences a TOT once it has been induced.
Research has used the retrieval of target information to divide TOTs into
substates. For example, Koriat and Lieblich (1974) divided TOTs into nine
different substates depending on whether the target was recognized or not
recognized, recalled or not recalled, and whether partial information was
recalled or not recalled (see their article for the specifics of the classification
scheme). Jones and Langford (1987) divided TOTs into objective TOTs, those
for which target information was retrieved, or subjective TOTs, those for which
no target information was retrieved (also see Perfect and Hanley, 1992). Burke
et al. (1991) defined “proper” TOTs as those followed by successful recognition.
Vigliocco et al. (1997) divided TOTs into positive or negative TOTs, depending
PROPERTIES OF TOTS 37
on whether the participant could correctly verify the TOT target. To the
rememberer, however, it is the intense feeling that marks the TOT. Because the
TOT is a subjective state and not simply a marker of partial accessibility, the
phenomenology of TOTs, separate from the phenomenology of retrieval, is
important. Indeed, the TOT feeling is not uniform across all instances and
occasions. I suspect that TOTs vary with respect to intensity, emotional content,
and feeling of imminence. These three dimensions come from my own
experience and empirical studies. It is likely that they do not capture all of the
differences in TOT phenomenology, but do provide us with a place to start. The
current section addresses whether people can distinguish among
phenomenological substates of the TOT, and whether these substates are
predictive of memory performance.
One of the most critical aspects of TOT phenomenology is exactly what the
role of emotion is in the formation and experience of TOTs. Most descriptions
of the TOT include its frustration, or the “turmoil” it generates. Indeed, when we
experience TOTs, it is the frustration of not being able to remember that may be
most salient. In contrast, at other times, TOTs may be accompanied by a feeling
of excitement that the answer will be eventually retrieved.
Emotionality
Many researchers report that rememberers often feel a surge of emotion when
they experience a TOT. They report that TOTs do not simply elicit a feeling that
a word can be recalled, but that the feeling is tinged with emotionality.
Sometimes, this emotionality may be experienced as frustration with not being
able to retrieve a word that is so obviously in memory. Alternately, the
emotionality may take the form of excitement because of the feeling of
imminent retrieval.
In turn, the experience of emotion may influence whether one experiences a
TOT or may affect the rate at which people experience TOTs. When people are
experiencing emotional stress, they may be more likely to experience TOTs.
Anecdotal reports often describe a stressful situation that precedes a TOT.
Indeed, in my own case, I often experience TOTs when a student from a
previous semester approaches me while walking on campus. When I see the
student, and the pressure to retrieve his or her name is there, I cannot do so.
Usually a minute or so after the student has walked past, my retrieval system
supplies the name. Oddly, research on emotionality and TOTs, or stressful
conditions and TOTs, is rare. In this section, I describe research that has been
done and its implications.
38 CHAPTER 2
The first issue to address is whether anecdotal reports and researchers’ intuitions
have any basis in reality. With this in mind, research in my lab addressed how
often participants ascribe an emotional state to TOTs, and whether that
emotionality judgment correlates with objective measures of performance
(Schwartz et al., 2000). College students answered a series of general
information questions such as,” Who was the first person to walk on the moon?”
If participants knew the answer (Armstrong), they typed it in and were presented
with the next question. If participants could not recall the target word, they
indicated whether they were experiencing a TOT. Furthermore, if they did
indicate that they were experiencing a TOT, they made a decision as to whether
the TOT was emotional or not. Emotionality in TOTs was described by
Schwartz et al. (2000) as the following: “Sometimes, you may feel frustrated or
emotional that you cannot recall a word that you are sure you know. If you are
in a TOT state that is accompanied by this emotional content, indicate a TOT
with emotion by pressing the E key” (p. 22). After completing of the questions,
they were shown the set of unrecalled questions again. They were given a
chance to produce the answers a second time. If resolution was unsuccessful,
they then attempted recognition. Rememberers quite readily made the distinction
between emotional and nonemotional TOTs, and the judgments were predictive
of performance. Roughly half (55%) of TOTs were judged by participants to
involve an emotional component. TOTs judged to be emotional were more
likely to be resolved and to be recognized than TOTs not judged to be
emotional. Nonemotional TOTs were more likely to be resolved and recognized
than n-TOTs. Thus, the study suggested that TOTs can be divided into
emotional and nonemotional substates, and that these substates are predictive of
performance.
In the diary study described earlier (Schwartz, 1999a), participants also
decided whether their TOTs were emotional. With the naturally-occurring
TOTs, 61% were judged to be emotional, slightly higher than in the laboratory
study. Interestingly however, emotionality was counter-predictive of whether the
TOT would be resolved. In fact, 89% of unresolved TOTs were judged to be
emotional, whereas only 58% of eventually resolved TOTs were judged to be
emotional.
In the laboratory study (Schwartz et al., 2000), however, emotionality of
TOTs was positively correlated with resolution and recognition. Emotionality
was less predictive of resolution than were other phenomenological variables
(e.g., imminence, strength). Thus, it is not contradictory with the diary study.
One possible explanation, as yet unexplored, is that emotionality in TOTs is
associated with blocking, that is, that the retrieval of related but incorrect items
interferes with the resolution of the TOT. This interference creates both
difficulties in retrieving and induces the emotional state characteristic of TOTs.
PROPERTIES OF TOTS 39
is precluding retrieval of the target word. Thus, when we are forced to say that
we do not know the target, our TOT is frustrating. However, when we are
retrieving the blocker or other words that we suspect are close, but not quite
there, then the TOT may be experienced as exciting because we feel the correct
target is on its way.
However, the direction of causality is unknown from these studies. It is
possible that the emotionality is a product of the processes that drive the TOT.
On the other hand, it is also possible that processes that produce emotional
responses may, to some extent, be also responsible for eliciting the TOT in the
first place. Moreover, the relation of emotional excitement to the feeling of
imminence of retrieval is only a hypothesis. It is possible that the feeling of
imminence drives the emotional excitement or vice versa. These topics are areas
for future study.
Gradations of TOTs
Some TOTs are particularly strong overriding feelings whereas other TOTs may
be fleeting sensations. Intuitively then, some TOTs are stronger than others.
Typically however, following Brown and McNeill (1966), most researchers ask
participants to make dichotomous judgments, such as “yes, I’m in a TOT,” or
“no, I’m not in a TOT.” For data analysis purposes, one can look at resolution,
recognition, or reported partial information as a function of whether a TOT was
reported. Dichotomizing the judgments, however, may cause people to neglect
some important features of TOTs. There is no a priori reason to believe that the
TOT is an all-or-none state.
An exception to the tradition of dichotomizing comes from a study by
Kozlowski (1977), who used a 5-point Likert-type scale to measure TOTs for
general-information questions (see also Gardiner et al., 1973). Kozlowski asked
participants to answer the following question: “Would you say that the correct
PROPERTIES OF TOTS 43
Imminence
A frequent experience during a TOT is the feeling that the word is about to be
retrieved. This sense of the closeness of the retrieval has been labeled
imminence (Smith, 1994). James (1890), for example, wrote that TOTs are
characterized by “making us at moments tingle with the sense of our closeness
and then letting us sink back without the longed-for term” (p. 251). Smith
(1994) argued imminence may be the key to understanding TOTs, and that some
TOTs are accompanied by imminence and others are not. Potentially,
rememberers might be able to distinguish between “close” TOTs, those TOTs
that are subjectively likely to be resolved soon, and “far” TOTs, those TOTs that
are subjectively not likely to be resolved.
44 CHAPTER 2
Two important but neglected aspects of the TOT experience are the “aha”
experience that accompanies a spontaneous pop-up and feelings of relief that the
TOT has been resolved. Neither of these phenomena has received but the
scantiest attention in the scientific literature. Therefore, what follows is almost
exclusively conjecture, backed up by only a sliver of data (Schwartz, 1999a).
The “aha” experience occurs when the TOT target is retrieved, and the person
feels a strong emotional surge in conjunction with strong confidence in the
correctness of the retrieval. The “aha” experience therefore, may be analogous to
the “insight” phenomena observed in the problem-solving literature (Metcalfe &
Wiebe, 1987; Smith, 1994). According to Metcalfe and Wiebe (1987), certain
problems are solved all at once, and only after they are solved does the persons
metacognitive system become aware of the solution. Therefore, confidence
comes all at once in a sudden “aha.” Indeed, Metcalfe and Wiebe (1987) found
that when confidence slowly increased for insight-type problems, confidence
mispredicted a correct solution to the problems. However, despite speculation
concerning the relation of TOTs to insight problem solving (e.g., Smith, 1994),
there has been no empirical data collected to date on the “aha” experience with
TOTs.
In a diary study (Schwartz, 1999a), rememberers were to indicate whether, on
the resolution of a TOT, they felt a feeling of relief. Unfortunately, the diary
PROPERTIES OF TOTS 45
instructions were somewhat unclear as to what was meant by the term relief, but
most rememberers interpreted it to mean a sense of emotional relief had been
brought about by remembering the forgotten word. In fact, 79% of resolved
TOTs were accompanied by feelings of relief. Although it is highly possible that
feelings of relief did not mean the same thing to all participants, the high rate
coincides with the findings that many TOTs are emotionally frustrating
experiences. Therefore, it should not be surprising that the resolution of the
frustration is accompanied by relief.
questions, the participants were again shown the set of unrecalled questions.
They were given a second chance to produce the answers, and if they could not,
they were asked again whether they were in a TOT for that question. Each
participant made an initial TOT judgment after the initial questioning and a
subsequent TOT judgment after the second round of testing.
The results demonstrated two phenomena. TOTs tended to be stable across
time, but there was some fluctuation. Of the targets that were not resolved on the
second recall attempt, 60% remained TOTs during the second round. This
means that 40% fluctuated from TOTs to n-TOTs. This 40% change represents
waning, that is, items that no longer induced a TOT. Interestingly, waxing from
a n-TOT to a TOT was less common but did occur. After successful resolutions
were accounted for, 8% of the questions that originally did not induce a TOT
succeeded in inducing a TOT during the second round of questioning.
Therefore, TOTs can both wax (here defined as a transition from n-TOT to
TOT) and wane, at least during repeated questioning.
This study is merely preliminary because it has yet to be replicated with other
stimuli. However, it documented both waning from a TOT to a n-TOT and
waxing from a n-TOT to a TOT. However, it does not indicate any mechanism
for why some TOTs would remain stable whereas others changed. What could
be going on? The following is a speculative account of these data. It is possible
that TOTs are based on the amount and rate of information that the rememberer
can retrieve (e.g., Koriat, 1993). If so, as long as information is being retrieved,
even if it is repetitive, the TOT may persist. Thus, TOTs will remain stable
across repeated questioning if the rememberer again retrieves related
information (even if it is the same partial information as in the earlier attempt).
However, if the rememberer no longer retrieves any information, the TOT may
not reoccur. In contrast, a n-TOT that transforms into a TOT may result because
partial information is retrieved now that was not retrieved before. This
speculative account is consistent with some mechanistic accounts of TOT
formation (Schwartz & Smith, 1997).
This study did not address whether TOTs remain stable across a concerted
retrieval effort. It only shows whether a TOT will reoccur when the person is
reminded of an otherwise forgotten retrieval attempt. The methodology of an
experiment that addressed stability and fluctuation across a single TOT
experience would be relatively straightforward. One would “prospect” for TOTs
using the Brown-McNeill procedure (Brown & McNeill, 1966). On finding a
TOT, the rememberer would be asked to make a concerted effort to retrieve the
target word. During this retrieval effort, the rememberer would be asked to make
a TOT judgment every 15 sec or so. Metcalfe and Wiebe (1987) used a similar
strategy to track “warmth” judgments in problem solving. Using this basic
methodology, one could track whether TOTs persisted over the retrieval
interval, or whether they faded if the person was unable to recover the target. If
the basic methodology worked, the researchers could also track the amount of
PROPERTIES OF TOTS 47
SUMMARY
In this chapter, a case was made for the universality of TOTs. A survey was
discussed in which I found a literal translation of the metaphor “on the tip of the
tongue” in 90% of living languages. Several diary studies presented a portrait of
what a TOT looks like in naturally occurring instances. The nascent literature on
TOT phenomenology was reviewed to begin to explore the important aspects of
the TOT experience. The chapter ended with some experimental findings
describing the paradox of illusory TOTs. This will set the stage for the
discussion of issues of TOT etiology in chapter 3.
3
Theories of TOT Etiology
The research discussed in the first two chapters was mainly descriptive.
Descriptive research is of crucial importance because it defines the phenomena
under investigation and suggests avenues of understanding. Description should
always accompany explanation. However, as scientists, our chief interest lies in
what is the cause of the phenomena we study. Thus, good science must always
move back and forth between description of the phenomena and experimentation
on the phenomena’s causation. In this chapter, we begin to discuss the theories
that have been advanced and tested to understand why TOTs occur.
Following Brown (1991), I use the term etiology to refer to the causes of
TOTs. Not normally used in cognitive research, the word etiology captures what
is meant by the causes that underlie the TOT. Etiology is the set of
psychological circumstances that cause a person to experience a TOT for a
known (or unknown) target word. Understanding the etiology of TOTs will
inform research in lexical retrieval, interference in memory, and the nature of
metacognition.
The question as to the mechanisms that drive the TOT may be the most
interesting. Answering this question may resolve many of the other debates
concerning TOTs. What might such a psychological mechanism entail? The
mechanism must be able to monitor word retrieval because the TOT experience
is essentially a retrieval experience. Moreover, the mechanism ought to be
accurate, that is, when a TOT occurs, it should correlate with the likelihood of
subsequent retrieval. Finally, there may be some reason for the conscious
experience known as the TOT. In essence, what function does the TOT
experience serve the organism? It has been hypothesized that the func-tion of
TOTs is to guide retrieval, to influence the choice of retrieval strategies, or to
“buy time” in front of others when knowledge is not forthcoming.
The explanations advanced in this chapter are psychological in nature. In
chapter 7, I discuss the limited neurological evidence on TOTs and proffer some
conjectures about the biological causes of TOTs. Chapter 6 discusses control
functions of TOTs. In this chapter, however, I focus exclusively on explanations
of TOT etiology at the psychological level. Indeed, there are several compelling
psychological theories for which substantial evidence has been gathered. Unlike
many sections in this volume, in which I lament the lack of research, TOT
etiology has been well studied.
48
THEORIES OF- TOT ETIOLOGY 49
Nelson, Gerler, and Narens (1984) categorized theories concerning the bases
of metamemory into two main classes. Recently, these classes have been applied
to TOTs (Askari, 1999; Metcalfe, Schwartz, & Joaquin, 1993; Miozzo &
Caramazza, 1997; Schwartz, 1994; Schwartz & Smith, 1997; Smith, 1994). The
two classes of theories are labeled direct-access views and inferential views.
Direct-access views argue that TOTs arise from sensitivity to the unretrieved
target. Although items have insufficient memory “strength” to be recalled, they
are strong enough to signal their presence as TOTs. Inferential views claim that
TOTs are not based directly on an inaccessible but activated target. Rather,
TOTs arise from clues that the rememberer can piece together. These clues are
information that is accessible to the rememberer, such as information from the
retrieval cue or information related to the target that is retrieved. Although direct
access and inferential views are not mutually exclusive, they represent different
approaches to the etiology of TOTs, and have led to different research
emphases. The two classes of theories differ with respect to their position on the
doctrine of concordance. Direct-access views typically, but not always,
implicitly endorse concordance, whereas inferential views usually challenge the
doctrine (see Table 3.1).
Direct access views, in effect, do not really postulate any mechanism at all
for TOT formation. They simply posit that the retrieval mechanism
simultaneously works to retrieve target words from memory and produces a
TOT in the word’s absence. Thus, TOTs are simply a by-product of a faulty
retrieval mechanism. In contrast, the inferential view suggests that there are
separate processes involved in retrieval and in the formation of a TOT
experience. Thus, although the views are not mutually exclusive, and both types
of mechanisms may be partially responsible for some TOTs, the hypothesized
mechanisms in each view have radically different implications for what a TOT
is and how it functions in the organism. For example, if a TOT is simply a point
on a continuum from a weak representation to a strong representation, it may not
necessarily serve any useful function in cognition, and may only be a simple
failure in the retrieval process. On the other hand, if a complex inferential
50 CHAPTER 3
DIRECT-ACCESS THEORIES
Direct-access approaches fall into three basic theories. The first theory, the
blocking hypothesis (Jones, 1989), contends that TOTs occur when a retrieval
cue prompts retrieval of an incorrect but closely related word. The realization
that the retrieved word is incorrect prompts the experience of the TOT, although
the blocking word continues to interfere with the retrieval of the correct word.
The second theory, the incomplete activation hypothesis, states that the TOT
occurs when rememberers cannot recall the target word, but sense its presence
nonetheless. The third direct-access theory has been developed by Burke and her
colleagues (Burke et al., 1991; James & Burke, 2000; Rastle & Burke, 1996; see
also Askari, 1999; Harley & Bown, 1998; Heine et al., 1999). Burke’s theory,
the transmission deficit model, argued that TOTs arise from a multi component
memory represen-tation. TOTs occur when there is activation of the semantic
component of the word, but priming does not pass to the phonological level of
the word. For example, on hearing the definition, “a device for protection from
the rain or sun,” a semantic and perhaps a visual representation of the object
may be retrieved, but the definition does not activate the phonological
representation (e.g., the word umbrella). Related to Burke’s model are recent
models by Miozzo and Caramazza (Caramazza & Miozzo, 1997; Miozzo &
Caramazza, 1997). These models are similar to Burke’s in the etiology of TOTs,
although they differ in the mechanism of word retrieval. Like Burke, Miozzo
and Caramazza argued that TOTs arise when phonological information does not
follow the retrieval of semantic information, but they disagree on the number of
THEORIES OF- TOT ETIOLOGY 51
There is ample evidence for the view that TOTs are a product arising from the
same processes that are involved in retrieval. Three basic kinds of evidence have
been used to promote direct access theory: resolution or recall of TOT targets,
recognition of TOT targets, and retrieval of partial information. Retrieved partial
information may be phonologically, semantically, or syntactically related to the
target. Each methodology reveals strong associations between TOTs and
retrieval. The evidence to be reviewed here first supports each of the three
different direct-access models. After discussing the evidence concerning direct
access, I will break down the different versions of direct access and discuss
evidence that supports or refutes each model.
Resolution of TOT Targets. Support for the assumption that TOTs arise from
retrieval processes comes from studies that find strong positive correlations
between the likelihood of experiencing a TOT and the likelihood of resolving
the target. Resolution is the retrieval of a previously unrecalled TOT item. In
diary studies, better than 90% of recorded TOTs were reported as eventually
resolved (Burke et al., 1991; Heine et al., 1999; Schwartz, 1999a). This
remarkably high percentage indicates a strong correspondence between TOT
phenomenology and objective performance. Unfortunately, in diary studies, it is
impossible to compare TOT resolution with n-TOT resolution.
In the laboratory, resolution rates with definitions or general-information
questions are typically around 40% (Brown, 1991). Resolution is even lower
with episodic-memory stimuli (Ryan, Petty, & Wenzlaff, 1982; Schwartz, 1998).
Typically, however, laboratory data are based on studies that investigate
immediate resolution, whereas diary studies tap both immediate and delayed
resolution, so it is unclear whether these studies are comparable in this regard. In
fact, Smith (1994) found that immediate resolution of TOTs for fictional animals
was only 9%. Resolution rose to 43% after a 6-min delay. Furthermore, Smith
found that resolution was much more likely to occur after a TOT than a n-TOT.
Resolution rates for n-TOTs were 3% at the immediate test and 19% at the
delayed test. Schwartz (1998) also found lower resolution rates for n-TOTs.
Gollan and Silverberg (in press) found that resolution of TOT targets was more
than double the resolution of n-TOT targets for both Hebrew and English words.
These studies demonstrate a strong correspondence between TOTs and the
likelihood of successful resolutions, which supports the view that TOTs reflect
target retrieval processes.
52 CHAPTER 3
TOTs n-TOTs
Burke et al. (1991) 81% Not Reported
Kozlowski (1977) 73% 46
Schwartz (1998) Exp. 1 81% 63
Exp. 2 40% 11
Schwartz (2001) Exp. 1 46% 14
Exp. 2 55% 31
Exp. 3 65% 30
Exp. 4 49% 16
Schwartz & Smith (1997) Exp 1 81% 70
Exp. 2 83% 74
Exp. 3 84% 64
Schwartz et al. (2000) Exp. 1 33% 10
Exp. 2 35% 10
Exp. 3 32% 10
Smith et al. (1994) 78% 57
Note. Percent correct on recognition tests as a function of phenomenological state,
TOT Versus n-TOT (items not recalled and not in TOT). Higher recognition performance
for TOTs than n-TOTs indicate accurate performance. TOT=Tip-of-the-tongue
phenomenon.
THEORIES OF- TOT ETIOLOGY 53
strong correlations between the presence of TOTs and the likelihood of correct
target retrieval.
Partial Information of TOT Targets. In partial information reports, the
rememberer is asked to recall specific aspects of the target word. Usually, this
involves phonological information, such as the first letter, the number of
syllables, and the syllabic stress (Brown & McNeill, 1966). It may also involve
semantic information. Participants are asked whether they can retrieve words
that sound similar to or that mean something similar to the TOT target. The
rationale for this kind of report is the assumption that the TOT reflects activation
of a target word, but insufficient activation to achieve full recall. Indeed, almost
all studies that elicit partial information find that participants can accurately
retrieve some. TOTs have been shown to be correlated with knowledge of first
letters, last letters, number of syllables, knowledge of similar words, and
syllabic stress (see Brown, 1991, for a review). Koriat and Lieblich (1974)
showed that more partial information was retrieved when rememberers were in
TOTs than could be explained by a guessing strategy based on the frequency of
those letters as initial-position letters in the English language. Murakami (1980)
found partial recall of letters in Japanese, and Bak (1987) found partial recall of
first letters in Polish. Bak (1987) also documented that Polish rememberers were
able to retrieve words that sounded similar to or meant something similar to the
TOT word. Askari (1999) showed the same type of partial retrieval of
information in Farsi.
Other studies examined the relation between TOTs and syntactical features of
language. Italian- and Spanish-speaking rememberers have access to the gender
of words when in a TOT (Miozzo & Caramazza, 1997; Vigliocco et al., 1997),
although Hebrew speakers do not access gender dur-ing TOTs (Silverberg et al.,
1999). Iwasaki, Vigliocco, and Garrett (1998) found that Japanese rememberers
have access to parts of speech, such as whether a word was an adjective or an
adjectival noun (a feature in Japanese that indicates that an adjective derives
from a noun, as in “rocky”), when in a TOT. Furthermore, Vigliocco et al.
(1999) found that English-speaking participants have access to the type of article
used for mass (i.e., “some”) or “count” (i.e., “the”) nouns when in a TOT. For
example, the words peppermint or asparagus are never pluralized; we indicate
quantity by terms such as more, some, and less. These are considered mass
nouns. Count nouns, such as cucumber or photograph are pluralized by adding
an s. In the absence of being able to recall the target word, rememberers are
nonetheless able to categorize them as mass or count nouns at a better-than-
chance rate when experiencing a TOT.
TOTs have been shown to be accurate predictors of objective memory
performance. Moreover, when rememberers are experiencing TOTs, they often
successfully recall partial information about the target. This suggests that people
experiencing TOTs are retrieving target information. These phenomena are
consistent with the position that the retrieval process is causing the TOT for the
54 CHAPTER 3
unrecalled word. Thus, these findings support the view that TOTs reflect target
word activation.
To summarize, a large database exists demonstrating strong correlations
between TOTs and the likelihood of resolution, recognition, and retrieval of
partial information, whether that information is phonological, semantic, or
syntactic. This strong correlation is suggestive, but not conclusive, of direct
access mechanisms. Because the data only show correlations between retrieval
and TOT experience, causal mechanisms cannot be definitively determined.
Indeed, inferential mechanisms can also account for these correlations (i.e.,
Koriat, 1993). Nonetheless, I have included them in this section because of
general acceptance that the accessibility of partial information and the eventual
retrieval of TOTs support direct-access mechanism (i.e., Brown, 1991;
Vigliocco et al., 1997).
Direct-access approaches divide into three separate theories: the blocking
hypothesis, the incomplete activation hypothesis, and the two-stage transmission
deficit model (Burke et al., 1991). Although all three hypotheses agree that
TOTs arise from access to an unretrieved target, each postulates a different
mechanism. I now examine each of these three specific theories in greater detail.
Blocking Hypothesis
included. Perfect and Hanley, like Meyer and Bock, argued against any role for
blocking in the etiology of TOTs.
Although these studies cast doubt on whether Jones’s data supported
blocking, they really do not refute it. That Perfect and Hanley (1992) showed
that the materials themselves played a role in the number of TOTs is neutral
with respect to the blocking hypothesis. Meyer and Bock (1992) showed that
phonological priming influences correct recall, but left TOTs unchanged. If one
assumes that a TOT represents a criterion along a continuum of target
knowledge, and that primes can boost the strength of that target knowledge, then
Meyer and Bock’s data are ambiguous. Items just below threshold may be
boosted from a TOT state into a recalled state, but items further below threshold
may be moved from a n-TOT state into a TOT state. Thus, it is possible that
more difficult words or less obvious primes might leave recall unaffected but
increase TOTs. Indeed, Askari (1999) found exactly that pattern. Phonological
priming increased correct recall, but nonsignificantly so. Phonological priming
did significantly boost the number of TOTs reported. Because both increased, it
is likely that some blockers cued recall, but others induced TOTs via blocking
the correct tar-get. Thus, inhibitory or blocking mechanisms may still represent
viable explanations.
Smith (1994), moreover, found evidence that supports the blocking
hypothesis, using episodic-memory stimuli. Rememberers learned the names of
fictional animals (known as TOTimals, see Fig. 3.1). Drawings of the fictional
animals were combined with a brief description and the name of the kind of
animal. Later, at test, participants were presented with the picture of the animal,
along with a word that was phonologically similar to the name, the name of
another animal, or a word that was unrelated to the target name. Smith (1994)
found that the phonologically similar items increased recall, relative to the
control, without increasing the reported number of TOTs. The semantically
similar cues increased both recall and TOTs. This pattern of results is consistent
with the blocking hypothesis. Semantically similar items caused participants to
think that the items were recallable. It is not clear, however, that the TOTs
resulted from blocking of target retrieval with misleading information, or that
the TOTs were induced because semantically similar items serve as subjectively
better cues (see section on cue familiarity).
The contradiction between Smith (1994) and Askari (1999) is that Smith
found that including semantic blockers increased TOTs, but phonological
blockers did not, whereas Askari (1999) found that including phonological
blockers increased TOTs, but semantic blockers did not. It is not necessarily
clear why these data are incompatible with each other, other than the simple
difference in stimuli. I offer a clearly ad hoc explanation here, which is left to
the reader to evaluate. It is possible that the newly learned material in Smith’s
experiments did not have sufficient phonological strength to allow for
phonological priming to be helpful. The semantic blockers, however, provide
THEORIES OF- TOT ETIOLOGY 57
clues from which the rememberer might inferentially determine that they are in a
TOT. In Askari’s experiments, the target words and definitions are much more
familiar to the participants. As a consequence, the semantic blockers are
recognized as incorrect. Therefore, they are dismissed and do not enter into the
TOT equation. However, the phonological primes increase the strength of the
target, and through direct-access, increase TOT rates. This explanation is post
hoc, but does resolve the apparent contradiction between the two sets of data.
FIG.3.1 “Yelkey.”
This interpretation, however, cannot account for recent data from Smith’s lab.
Smith (2000) provided word definitions to participants and asked them to
retrieve the target word. For half of the items, he provided them with unrelated
words and for the others he provided related words. All of the participants were
required to provide definitions of the accompanying words. Smith reasoned that
the focusing on the meaning of the words would increase the likelihood that they
would block recall. Relative to the targets accompanied by unrelated words,
recall decreased and TOTs increased for the targets accompanied by semantic
associates. Thus, it may be more likely that participants must focus on semantic
aspects of accessible words for them to block retrieval and create TOTs.
58 CHAPTER 3
Blocking theory states that the rememberer retrieves blockers, and these
interfere with retrieval of the correct target. The blocking hypothesis is difficult
to evaluate experimentally because it is not clear whether the experimenter-
provided blockers serve as blockers or retrieval cues. Furthermore, it is also
unclear whether the experimenter-provided blockers actually mimic the role of
natural blocking in retrieval and TOT etiology. Therefore, despite the mixed
evidence on blocking, it should remain a viable hypothesis. Indeed, it is a
hypothesis waiting for a better, as yet undiscovered, methodology.
Incomplete Activation
The transmission deficit model is the most well-articulated and studied version
of direct-access theory. It has also motivated a great deal of recent research on
the TOT from a number of different labs (e.g., Askari, 1999; Heine et al., 1999;
James & Burke, 2000; Rastle & Burke, 1996; White and Abrams, 1999). I
consider it here in detail.
In the transmission deficit model (Burke et al., 1991; James & Burke, 2000;
Rastle & Burke, 1996), TOTs are induced whenever the semantic representation
of a word is activated, but the activation fails to prime the complete
phonological representation of the target word (see Fig. 3.2). The transmission
deficit model differs from the incomplete-activation model because it assumes
two separate target representations: semantic and phonological. For this reason,
THEORIES OF- TOT ETIOLOGY 59
Harley and Bown (1998) label the transmission deficit model a two-stage model
of TOTs and the incomplete activation a one-stage model. For now, we focus on
how the model accounts for TOT production. In chapter 4, we discuss its
predictions concerning lexical retrieval.
Burke et al. (1991) wrote that the “basic cause of TOTs…is a deficit in the
transmission of priming across critical connections required for producing the
target word” (p. 545). Therefore, although the model requires a dual
representation of the target, TOTs arise from direct access to these particular
representations via the same process that is used in retrieval. Because a single
process elicits both retrieval and the TOT phenomenology, the transmission
deficit model is a direct-access model.
FIG. 3.2
The explanation for the accuracy with which TOTs predict later memory
performance is similar to the account of accuracy discussed in the section on
incomplete activation. When a word definition is given, the semantic
representation of the target word is activated. Priming spreads to an existing
phonological representation. Changes in retrieval conditions will lead to changes
in activation and priming, which may result in resolution of TOTs. In other
words, accuracy of the TOT is assured because the same processes that elicit
retrieval (semantic plus phonological activation) elicit the TOT (which occur
when less phonological activation is present).
60 CHAPTER 3
originally recalled, measured recall and TOTs are not independent of each other.
If there are differences in recall, the subset of items for which TOTs can occur
may be different. If an experimental variable such as the phonological priming
variable in the Rastle and Burke experiments increases recall, then it decreases
the set of items on which TOT judgments can be made. In the Rastle and Burke
study, it is possible that the unrecalled words in the phonological prime
condition represented a harder set of items than the unrecalled unprimed words.
Thus, the lower rate of TOTs may be a function of the more difficult items and
not the manipulated variable (see Schwartz and Metcalfe, 1994). When
evaluating etiologic issues in TOTs, it important to equate recall or recognition
across experimental conditions.
These problems were corrected in a recent study by White and Abrams
(1999; see also James & Burke, Exp. 2, for a similar paradigm). They compared
three age groups: young adults (ages 18–25), older adults (ages 60–72), and
oldest adults (age 73–83). In their study, participants were given definitions of
words and asked to retrieve the word (e.g., “What word means to formally
renounce a throne?”—Abdicate). If they could not retrieve the word, they were
asked whether they were experiencing a TOT. After that, they were given a
phonological prime, and then an opportunity to resolve the TOT In this study,
TOTs were collected in advance of the priming, so the priming could not
directly affect the TOT judgments. The results were quite clear: Phonological
priming increased resolution of TOTs relative to n-TOTs for the younger
participants, and to some extent for the older participants. The oldest
participants could not resolve more TOTs after phonological priming. The study
represents a better test of transmission deficit theory because, unlike Rastle and
Burke (1996), who confounded TOT rates with TOT resolution, White and
Abrams kept TOT rates constant across the primed and unprimed conditions.
Moreover, White and Abrams looked at resolution of n-TOTs as well as TOTs.
White and Abrams’ (1991) data support the transmission deficit model for
several reasons. In younger adults, TOT resolution was increased by
phonological priming. If TOTs are caused by activation of a semantic node, but
with insufficient activation of the phonological node, then priming the
phonology should increase access to the word. Additionally, Burke et al.
hypothesized that the connections between semantics and phonology weaken
with age. In this study, phonological priming benefitted the older adults, but
only when it was the first syllable that was primed. Finally, priming did not
benefit the oldest adults at all, presumably because their baseline phonological
activation was lowest of all. For the oldest adults, resolution rates were no
higher after priming than no priming. White and Abrams found that
phonological priming improved TOT resolution relative to n-TOT resolution
and did so more in the younger than the older adults. This pattern is logically
predicted by the transmission deficit model.
62 CHAPTER 3
Like White and Abrams (1999), James and Burke (2000) also introduced
phonological priming after an initial retrieval failure. Consistent with the
transmission deficit theory, more TOTs were resolved following phonological
priming than when the primes were phonologically unrelated. Moreover, for n-
TOTs (or “don’t know” items), phonological priming did not influence
resolution. This suggests that it may unactivated phonological information that
induces TOTs.
Smith et al. (1994), using their TOTimal paradigm, also found evidence that
supports the transmission deficit model. Participants were presented with
TOTimal pictures and the names of animals (e.g., see Fig. 3.1). In one
experiment, participants were asked to rehearse the stimuli by writing the names
of the fictional animals. In this experiment, writing practice increased the rate of
correct recall and reduced the number of TOTs, without affecting the number of
n-TOTs (“don’t know”) responses. Thus, the writing practice shifted TOTs to
correct recall, but did not shift the n-TOTs to TOTs. Smith et al. (1994)
reasoned that if the picture produces a semantic representation, a TOT or a
correct recall can occur. The writing practice only influences phonological
activation, thus it only shifts TOTs to correct recall, without affecting the n-
TOTs. Smith et al.’s findings support a view in which semantic and
phonological representations are separate, as suggested by Burke et al. (1991).
Brennan, Baguley, Bright, and Bruce (1990) conducted a study on TOTs,
paying particular attention to the roles of face cues in TOT resolution. They
presented participants with general-information questions such as,” Who is the
actor who played The Bionic Man?’” (Lee Majors). Participants who were
unable to retrieve the target answer were given one of two kinds of cues. Either
the participant was shown a photograph depicting the question, such as a
photograph of the actor, Lee Majors, or the participant was given the initials of
the person (L.M.). Brennan et al. then compared the rate of TOT resolution.
They found that TOT resolution was higher when the initials (phonological
cues; 46.6%) were given than when the faces (14.5%) were given as cues.
Brennan et al. were most interested in models of face recognition, but their data
strongly support the importance of phonology in TOTs, as subsequently
emphasized by Burke et al. (1991). The face cues might be considered visual, or
perhaps even semantic. It is possible that the participants experiencing TOTs
were visualizing the bionic man, running in slow motion, presumably at high
speeds. However, the TOT was for the name, a phonological entry. Thus, access
to the photo did not help in the resolution of TOTs. However, the first initials
provided the cues to overcome the retrieval problem and resolve the TOT. Burke
et al. argued that it is a lack of phonological activation that induces TOTs. Here,
phonological priming resolves TOTs, but face priming does not.
The transmission deficit model can account for much of the data on TOTs. It
succinctly accounts for frequency effects, neighborhood effects, phonological
priming on TOT rates and resolution, the accessibility of partial information,
THEORIES OF- TOT ETIOLOGY 63
INFERENTIAL VIEWS
In the inferential approach, the TOTs are not based on sensitivity to inaccessible
but activated targets. Instead, rememberers infer the target’s existence from a
host of clues that inform them that it is likely to be in memory. The clues that
are used to make the inference may include the cue, which may include the
question itself and information implicit in the question (e.g., Metcalfe et al.,
1993), retrieved partial information about the target (Koriat, 1993), or any
generated information related to the target name (Schwartz & Smith, 1997).
Cues can be considered anything in the environment that may help the
rememberer retrieve the sought-after target. Generated information may consist
of bits and fragments of the sought-after target or may consist of words,
pictures, and other kinds of information that are related to the target.
The inferential approach does not mean that the process producing a TOT is a
conscious one. Instead, the inferential process is assumed to be largely
nonconscious. The cues and clues are processed by a nonconscious mechanism
evolved to assess the likelihood of recall. If the nonconscious process yields a
“positive,” the result is the feeling of the TOT, which signals to the rememberer
that the target is in the memory store. I argue that the processes are hidden or
nonconscious although the product (the TOT) is by definition, conscious.
We may use a similar process consciously to make feeling-of-knowing
judgments. For example, if the question is “What is the planet closest to the sun?
(Mercury),” I may employ a host of conscious inferential strategies. First, how
much I know about astronomy may influence my judgment. Second, because I
64 CHAPTER 3
know that most of the planets are named after Roman gods and goddesses, I may
try to retrieve a few of these names and try to match them up with the right
planets. Third, if I have some phonological information, such as the first letter, I
may consider my feeling of knowing to be quite strong. The inferential view
stipulates that we unconsciously use a similar process to determine whether we
are experiencing a TOT. This unconscious inference makes us feel as if the
target word is actually right there—on the tip of the tongue—giving rise to the
phenomenology that feels like direct-access.
Inferential approaches to metacognition have been shown to be useful in
explaining the bases of several different metacognitive judgments, including
speeded feelings of knowing (Reder, 1987; Reder & Ritter, 1992), slower
feeling-of-knowing judgments for unrecalled targets (Koriat, 1993; Metcalfe,
1993; Metcalfe et al., 1993; Schwartz, 1994; Schwartz & Metcalfe, 1992), and
judgments of learning (Benjamin, Bjork, & Schwartz, 1998; Koriat, 1997).
Inferential explanations, developed within a metacognitive perspective, are
inconsistent with the doctrine of concordance. The inferential approach assumes
that TOTs are caused by a process in addition to the retrieval process. Therefore,
consistent with Tulving (1989), the phenomenology of TOTs is elicited by one
process, and target retrieval occurs due to another.
Unlike the straightforward manner in which direct-access views account for
the accuracy with which TOTs predict recognition, the inferential account of
accuracy requires an additional step. The inferential view claims that the
processes that produce TOTs and the processes that effect target retrieval are
generally correlated. Thus, when cues are familiar and considerable related
information is retrieved, the rememberer will usually have a representation of
the target word in memory. Unfamiliar cues that do not elicit related information
are, correspondingly, less likely to induce TOTs, but it is also likely that the
rememberer will not have the target representation. In this way, TOTs are
thought to be the product of a heuristic. The experience of a TOT generally
means that target retrieval is likely; however, because the TOT is not based on
target retrieval, occasions will arise when a TOT is elicited by certain factors
when no representation of the target is present. Thus, inferential mechanisms
can also account for the high correlation between resolution, recognition, and
access to partial information, if we assume that the processes that cause TOTs
are correlated with the processes that evoke retrieval. In addition, TOTs can
account for instances in which the correct target is not known, but the person
experiences a TOT anyway (Perfect & Hanley, 1992; Schwartz, 1998).
There are two main hypotheses within the inferential framework. These two
hypotheses are not mutually exclusive, but involve different cognitive
mechanisms. Cue familiarity theory suggests that some TOTs arise because of a
strong feeling elicited by recognizing a familiar cue (Metcalfe et al., 1993).
Koriat (1993) argued for an accessibility heuristic, which states that TOTs arise
from inferences made based on the amount and kind of information retrieved
THEORIES OF- TOT ETIOLOGY 65
from memory when the target is not recalled. In the next two sections, I consider
the evidence for each of these views.
Cue Familiarity
The cue familiarity hypothesis states that TOTs are based on an assessment of
the level of recognition of a particular cue or question (Metcalfe et al., 1993).
This approach dismisses the relevance of the unretrieved target altogether and
instead emphasizes the role of the cue. In its purest form, the cue familiarity
hypothesis suggests that a familiar cue should induce a TOT regardless of
whether the target is known. There is much data describing the importance of
cue familiarity in feeling-of-knowing judgments (Connor, Balota, & Neely,
1992; Metcalfe, 1993; Metcalfe et al., 1993; Reder, 1987; Reder & Ritter, 1992;
Schwartz & Metcalfe, 1992). However, less data exist linking cue familiarity
with TOTs.
Koriat and Lieblich (1977) examined the relation between TOTs and the cues
or questions that elicited the TOTs. They labeled the questions as pointers
because they point or cue the relevant target answer. Koriat and Lieblich
analyzed the pointers along several dimensions, but the most relevant finding is
that redundancy within the question led to more TOTs for unrecalled items than
did more succinct definitions. For example, definitions with repetitive elements,
such as “a circle, or any indication of radiant light, around the heads of
divinities, saints, sovereigns in pictures, medal, etc.” (nimbus) tended to produce
more TOTs than did more concise definitions such as, “the science of coins”
(numismatics). This trend toward more TOTs in questions with repetitive
elements in the cue was the same across TOTs that were resolved (i.e.,
eventually recalled or recognized) and those that were not. Thus, Koriat and
Lieblich’s results suggest that cue factors play a role in causing TOTs.
Metcalfe et al. (1993) directly tested the roles of cue-based sources in TOTs
in an episodic memory paradigm. Participants studied cue-target word pairs. In
one condition, both the cue and the target were repeated (“captain-carbon,”
“captain-carbon;” A-B, A-B). In a second condition, the cue was repeated, but
with a new unrelated target (“pasture-dragon,” “pasture-movie;” A-B, A-D). In a
third condition, neither the cue nor the target was repeated (“trouble-camel,”
“garden-miser;” A-B, C-D). In accordance with results from interference
paradigms, Metcalfe et al. found that the A-B, A-D condition showed the lowest
recall. Recall was highest in the A-B, A-B condition, presumably because both
cue and target are repeated. If TOTs are based on the strength of the
representation, as predicted by incomplete activation, the A-B, A-B condition
should show the highest number of TOTs. Cue familiarity, however, predicts
that because the cue is repeated in both conditions (A-B, A-B and A-B, A-D),
TOTs will be high for each when compared to the condition in which the cue is
presented only once (A-B, C-D). Indeed, Metcalfe et al. found that both the
66 CHAPTER 3
repetition condition (A-B, A-B) and the interference condition (A-B, A-D) each
showed a high percentage of TOTs relative to the once-presented condition (A-
B, C-D). Further supporting the cue-familiarity explanation, Metcalfe et al. also
found that recognition of the unrecalled targets mirrored the pat-tern observed in
recall and not the pattern observed in the TOT judgments. Thus, they found that
cue repetition led to more TOTs, regardless of the level of memorability.
Schwartz and Smith (1997) also found evidence that supports cue familiarity.
In this study, country names were used as cues to retrieve the names of
TOTimals that supposedly came from those countries. Prior to the test,
participants were asked to judge a list of country names for pleasantness. The
participants were asked if the countries would be good places to go for vacation
(Italy, by the way, was judged the most desirable vacation country). Some of the
counties were habitats (cues) for TOTimals, whereas others were controls.
Following the pleasantness rating task, participants attempted to retrieve the
names of TOTimals that came from the primed countries and to retrieve the
names of unprimed TOTimals. Relative to the unprimed countries, there were
more TOTs experienced for TOTimals cued by those countries that had been
primed in the unrelated task. Given that memorability of the TOTimal names
was equivalent between the primed and unprimed cues, this data also supports a
cue familiarity component to TOT etiology (cf. Harley & Bown, 1998; Rastle &
Burke, 1996). This means that there were no differences in memorability across
the conditions, and therefore any differences in TOT rates must be caused by
nonmnemonic factors. Furthermore, because the phonological representations of
the target names were studied the same number of times, theories such as
transmission deficit theory would not have predicted differences simply on the
basis of the familiarity of the cue. Thus, a separate inferential mechanism must
be invoked to explain these data.
Accessibility Heuristic
Koriat (1993, 1995) argued that feeling-of-knowing judgments are based on the
amount and intensity of partial information that rememberers retrieve when they
cannot recall a target answer. Koriat’s view is inferential because the judgment
is only made based on accessible information, and it is not based on inaccessible
but available targets. This is an important distinction. In direct-access models, a
present lexical representation in memory is sufficient to cause a TOT. In
Koriat’s model, an inference is made based on accessible information (i.e., that
which can be retrieved). Thus, this view falls in the inferential camp rather than
the direct-access camp. Schwartz and Smith (1997) also hypothesized a similar
basis for TOTs, arguing that the amount and intensity of retrieved information
may play a role in the etiology of TOTs. In our view, TOTs are caused by any
information that is retrieved, including information that may actually be
tangential to the sought-after target.
THEORIES OF- TOT ETIOLOGY 67
To test this hypothesis, Schwartz and Smith (1997) used the TOTimal
methodology. They presented participants with lists of fictional animal names.
Each name was paired with the name of a country, and some of the names were
also accompanied by line drawings of fictional animals. For example,
participants might have seen “Yelkey-Panama,” which indicated that the
“yelkey” is an animal that lives in Panama. Of the animals for which line
drawings were provided, half were also accompanied by information pertaining
to the size and diet of the animal. The three encoding conditions were minimum
information (just the name-country pair), medium information (the name-
country pair and line drawing), and maximum information (the name-country
pair, line drawing, and diet and size information). The conditions were designed
to allow different amounts of information to be retrieved when the participants
were given the country name as a cue for the retrieval of the animal name. If the
participants could not retrieve the name of the animal associated to a given
country cue, they were then asked for a TOT judgment and asked to guess the
first letter of the animal’s name. The participants were also asked to retrieve as
much related information as they could. Related information included both
biographical information and descriptions of the appearance of the animal.
The encoding manipulation (amount of related information) did not affect
recall of animal names and it did not affect the recognition of unrecalled names.
This suggests that three conditions were composed of roughly equivalent items
for which possible TOTs could exist. However, the rates of TOTs differed
among the conditions. Consistent with Koriat’s (1993) accessibility theory, there
were more TOTs reported in both the medium and maximum information
conditions than in the minimum information condition. In addition, the number
of reported TOTs was correlated with the amount of related information
retrieved. Thus, when more information was accessible to participants, they
were more likely to experience TOTs. This basic finding was found in three
separate experiments. Again, direct access theory, based on the strength of the
representation, would not have predicted differences in conditions that were
equated for correct recall and recognition. However, the conditions differed in
the amount of related information that was recalled. Therefore, it is not the
strength of the phonological representation that is driving the differences in
TOTs; it is more likely that it is the accessibility of the related information.
In a recent study, MacLin (2000) also tested Koriat’s (1993) accessibility
theory as it applied to TOTs. MacLin presented famous faces as cues for the
retrieval of the names of these people. For faces for which the name was not
recalled, MacLin asked for TOT judgments, and he asked the participants to
report any related information that they might recall about the person. MacLin
found that, like Schwartz and Smith (1997), the retrieval of related information
was positively correlated with the experience of a TOT, thus supporting Koriat’s
accessibility view.
68 CHAPTER 3
information correlate with knowledge of the TOT target word, and the retrieved
information may cue the TOT word.
The TOT has been referred to as a cognitive heuristic (e.g., Schwartz, 1999b;
Smith, 1994). The heuristic account cogently explains why some TOTs are not
resolved or recognized. These TOTs may be spurious experiences caused by
familiar cues or the retrieval of related information. Indeed, in the case of
illusory TOTs, the person cannot know the target answer (Schwartz, 1998).
Thus, cases of “negative” TOTs (Vigliocco et al., 1997), TOTs for incorrect
items (Rastle & Burke, 1996), and illusory TOTs (Schwartz, 1998) may be best
thought of as an error produced by inferential heuristics. Heuristics can,
therefore, account for both the general accuracy of TOTs and the existence of
unresolved, unrecognized, or illusory TOTs.
Illusory TOTs in particular provide some empirical support to the heuristic
concept. Schwartz (1998) included questions for which there was no correct
answer among normal general-information questions. For example, one of the
questions was, “What is the only living reptile that flies?” The correct answer is
that there is no such animal. However, frequently participants reported TOTs for
these questions. Regardless of whether they were experiencing TOTs via direct
access for an incorrect target (e.g., sugar glider) or were basing their TOTs on
inferential means (lots of related information, e.g., that the Komodo dragon is
the world’s largest lizard), the TOT is an error because the target does not exist.
Thus, metacognitively speaking, some TOTs are errors. Errors in cognition
suggest a mechanism that typically works well, but can be “thrown a curve.”
Indeed, that is the case for TOTs. Most studies show high accuracy, but
Schwartz’ study showed that errors can occur as well.
The heuristic approach to TOTs is also consistent with Tulving’s (1989) view
on the doctrine of concordance. TOTs are accurate at predicting memory
performance because they arise from processes correlated with retrieval, but not
directly from retrieval itself. This correlation does not imply that the cognitive
processes are identical, but only that they are linked. In fact, the linkage may
occur only because of features of the external environment. Concordance does
not preclude significant correlations between cognitive processes and
phenomenology. Rather, it suggests that the processes may be separable, as
indeed they appear to be with some TOTs.
Finally, the heuristic approach converges with the importance of heuristic
processing in a number of other cognitive domains, from decision making to
cognitive mapping. In cognitive mapping, for example, people appear to rely on
heuristics to determine compass directions (Tversky, 1981, 1991). Thus, people
will judge that Miami, FL is east of Pittsburgh, PA, based on the heuristic that
east coast cities are further east than inland cities. In fact, Miami is further west
than Pittsburgh. In the decision-making literature, people’s reliance on cognitive
heuristics has been the focus of intense research for many years (e.g., Kahneman
& Tversky, 1973; 1984; Piattelli-Palmarini, 1994). In the face of uncertainty,
70 CHAPTER 3
SUMMARY
The focus of this chapter was theoretical accounts of TOT etiology. TOT
etiology refers to the processes that cause the phenomenology of the TOT. Two
basic classes of theories were evaluated: direct-access theory and inferential
theory. Direct-access theory states that TOTs occur when a target’s strength is
above some hypothetical knowledge threshold but below a retrieval threshold.
The theory has been instantiated in a number of specific models, most notably
Burke et al.’s (1991) transmission deficit theory, which postulates that TOTs are
caused by access to a semantic representation but a failure to retrieve the
phonological representation. I reviewed the rather extensive evidence that
supported this model. Inferential theory hypothesizes that TOTs are caused by
the accumulation of indirect clues that the rememberer uses to reconstruct the
likelihood of recalling the target. Evidence, which supports inferential theory,
exists in both cue familiarity studies and from studies which manipulate the
retrieval of related information. Inferential theory accounts for accuracy by
suggesting that the TOT serves a heuristic function, usually correctly informing
us which unrecalled targets we can eventually recall. Direct-access and
inferential views are not mutually exclusive, and any complete explanation of
the TOT is likely to require both.
4
TOTs as a Window on Retrieval
In most arenas of our life, it is not our confidence or our potential that matters.
Indeed, it has become almost a truism of modern society that it is results that
count. For a professional athlete, it does not matter how well he or she performs
in practice. It is performance in the big game that counts. Similarly, for a
researcher, it is not the creativity of his or her ideas, but whether these ideas can
be translated into published articles and funded grants. No less for our cognitive
systems. Our TOTs are meaningless unless they really do tell us something
about the retrieval process. In chapter 3, TOTs were discussed as heuristics;
heuristics are only good when they are accurate predictors of the future.
Therefore, in this chapter, I detour from my primary concern with the etiology,
phenomenology, and function of TOTs. Here, I focus on the lexical retrieval
process. In particular, I discuss how TOT research informs theories of lexical
retrieval. Therefore, to paraphrase a famous president, let us ask not what TOTs
are, but what they tell us about the cognitive processes that allows us to speak.
Before we return to the gritty details of theory, recall that precarious pedestal
on which I have placed TOTs. They sit atop a carefully constructed tripod of
research in memory, language, and metacognition. Brown and McNeill (1966)
realized that the TOT was a crucial phenomenon in both memory and language
research. The TOT can be thought of as a memory lapse, but it is a memory
lapse for a linguistic entity. TOTs can also be thought of as speech production
error, one in which the speaker fails to retrieve a particular word. Obviously, this
issue is two sides of the same coin. Elsewhere in this volume I make the case
that metacognition represents a different coin altogether; however in this chapter
I focus what most of the research on TOTs has been directed, namely, theories
of lexical retrieval. Thus in this section, the concern is on the nature of the
memory lapse or speech production error, and less on whether that condition
also includes a TOT experience.
Think about retrieving a word from memory, perhaps the name of someone with
whom you work or the name of one of your first cousins. Introspectively, the
retrieval process is a black box. When you are asked,” What is the name of your
first cousin?” the phonology of the names simply appears, or pops up at the
71
72 CHAPTER 4
form of the word, which can then be outputted by the speech production system
(see Fig. 4.1).
Consider a person asked the question, “What animal is the closest living
relative to humans?” The question initiates a lexical retrieval process. According
to the general model, the first stage involves the retrieval of particular semantic
information. The speaker accesses various semantic attributes of the concept to
be articulated. In the example given, semantic concepts may be accessed such as
animal, primate, smart, uses tools, wild, cute, and so forth. This initial process of
retrieval results in the lemma (Caramazza & Miozzo, 1997; Levelt, 1989). The
lemma maintains an abstract representation of the to-be-spoken word that
combines both the semantic information and various syntactic components (e.g.,
pluralization, grammatical gender, part of speech, etc.). In this case, that would
include information such as noun, regular pluralization, and animate). According
to theory, however, the lemma does not contain any phonological information.
Once the lemma representation has been formed, the second stage of lexical
retrieval is initiated. In this second stage, the rememberer compiles an abstract
description of how the target word chimpanzee is constructed phonologically.
This lexeme is then processed by a motor production unit that actually
articulates the word (see Fig. 4.1).
Normal lexical retrieval is an extremely rapid process, and therefore it is
difficult to break down the lexical retrieval process experimentally into its
component parts, although this is a major thrust of modern psycholinguistic
research (Martin et al. 1994). It is precisely because normal lexical retrieval is so
rapid that the TOT plays such an important role in delineating theories of lexical
retrieval. During TOTs, the retrieval process is slowed. This slowing of an
otherwise rapid retrieval process may allow researchers to examine the lemma-
level retrieval and the lexeme-level retrieval independent of each other. Thus,
for scholars interested in lexical retrieval, the TOT holds the promise of
allowing researchers to examine what the lexical retrieval process looks like in
slow motion. Indeed, Brown (1991) wrote that “examining it [word retrieval] in
a temporary ‘holding pattern’ imposed by the TOT has the potential to reveal
subtleties of normal retrieval functions, similar to how slow-motion photography
clarifies the dimensions of a hummingbird’s flight” (p. 204). During retrieval of
a TOT word, the lexical retrieval process has been slowed significantly.
Despite the agreement on this basic hierarchy of lemmas and lexemes, there
are a great many variants of lexical retrieval models. In some versions of the
models, the two stages are serial, that is, the lemma must be retrieved in advance
of the initiation of the phonological retrieval (see Butterworth, 1989, 1992;
Miozzo & Caramazza, 1997). In interactive models, activation flows back and
forth between different levels (e.g., Burke et al., 1991). In interactive models,
the search for the lexeme is partially dependent on the success of the first stage.
Thus, in both serial and interactive models, the lemma represents an important
74 CHAPTER 4
FIG. 4.1 Part of the lexical system showing the relation between the lemma and
other levels of lexical representation. The lemma and lexeme levels show the
Italian words for the lexical concepts tiger, train, to go, and to drink. Adapted
from Bock and Levelt (1994), Jescheniak and Leveit (1994), and Roelofs
(1992).
preliminary step in accessing word forms (see Harley & Bown, 1998; Miozzo &
Caramazza, 1997).
Recently, Caramazza and Miozzo (1997) challenged the general view that
lexical retrieval involves two distinct stages. They claim that the two-stage view
has too often been assumed and then imposed on existing data. They find no
compelling reason in the literature to support a two-stage view. Instead, they
think of lexical retrieval as not requiring a lemma or lexeme distinction. They
TOTs ON RETRIEVAL 75
base this challenge to the existing theory on TOT data. Their study and rationale
will be reviewed later in this chapter in a more thorough manner.
Most researchers agree that much of the evidence used to support two-stage
models comes from errors in ordinary speech (see Caramazza & Miozzo, 1997).
Garrett (1975) argued that two kinds of errors are observable. The first are called
word-exchange errors, which involve substituting a word of the same
grammatical class but of a different phonology, thus corresponding to lemmas.
For example, one might substitute blanket for pillow (Tweney, Tkacz, &
Zaruba, 1975). The second are exchange errors which involve substitutions of
similar but incorrect phonological structures, with no relation to meaning, thus
corresponding to lexemes (see Caramazza & Miozzo, 1997). For example, one
might substitute mittens for muffins (Tweney et al., 1975). In this research
tradition, TOTs are also seen as speech errors, in which the lemma has been
retrieved, but the complete lexeme has not. Studying what is accessible during
TOTs tells us what kinds of information (other than phonological) are
represented at the level of the lemma. Indeed, James (1890) wrote that “the
rhythm of the lost word may be there without the sound to clothe it; or the
evanescent sense of something which is the initial vowel or consonant may
mock us fitfully, without growing more distinct” (p. 251). James’s quote is
consistent with the view that TOTs are seen as instances in which we can
analyze the lemma, largely uninterrupted by the processes that generate the
lexeme. Thus, according to this view, lemmas are activated and lexemes are
only partially activated in TOTs.
In chapter 3, I presented Burke et al.’s (1991) transmission deficit model in
discussing the etiology of TOTs. In this context, the transmission deficit model
is an example of a two-stage interactive model of word retrieval. In the model,
priming spreads from an initial semantic representation (lemma) to a second
phonological level (lexeme). If sufficient priming spreads from the semantic
level to the phonological level, activation of a word form occurs, and the
rememberer can say the word. In the model, priming is a graded phenomenon,
but activation is all or none; that is, a particular word may be highly primed, but
below the activation threshold and therefore not recalled. It is this state of high
priming, which is still below activation threshold, that triggers a TOT in the
model (refer to Fig. 3.2, this volume). In summary, TOTs occur when lemmas
are retrieved, but there is insufficient retrieval of the lexeme.
According to most theorists, syntactical and semantic information is
represented at the level of the lemma. When a rememberer experiences a TOT,
semantic and syntactical information should be relatively easy to report, despite
the inability to report the whole word, or even any of its phonological features.
The models predict that semantic information will always be accessible in
TOTs, that syntactical information also should be readily accessible during
TOTs, but complete phonological information may not be accessible. On the
other hand, aspects of the phonological lexeme may be accessed, albeit not
76 CHAPTER 4
completely during a TOT. Therefore, the models also predict that partial
phonological information will be accessible during some TOTs. This chapter
first considers phonological access in TOTs and then syntactic access.
The purpose of TOT research within the context of lexical retrieval models is
that TOTs represent instances in which the normal lexical retrieval process has
slowed down or is experiencing a selective failure (Burke et al., 1991). In this
framework, analysis of the kinds of information accessible during a TOT may
shed light on the process whereby words are retrieved. In particular, because the
assumption is that during a TOT the lemma has been fully retrieved, it is
possible to focus on the nature of the phonological retrieval process when a
rememberer is experiencing a TOT. Indeed, Rastle and Burke (1996) wrote that
“TOTs provide a means of isolating phonological processes within a
conceptually-driven task” (p. 594).
Both one-stage (Caramazza & Miozzo, 1997) and two-stage (Burke et al.,
1991) theories predict that during a TOT, participants should be able to retrieve
phonological information about the unrecalled target. In the two-stage views, a
lemma is retrieved, but the lexeme is not completely retrieved. However, the
rememberer may be able to activate some aspects of the lexeme, such as the first
letter or first syllable. Similarly, a related phonological word may be activated.
The one-stage view is also compatible with the retrieval of phonological
information. That is, the retrieval process simultaneously pulls out both semantic
and phonological information, each of which may be complete or incomplete.
The data support the hypothesis that rememberers access aspects of the
phonology during a TOT (see Brown, 1991; Schachter, 1990). Although much
of this data has already been discussed, it is important to review it again in this
context. In chapter 2, the concern was whether the retrieval of any partial
information was more supportive of direct access or inferential theories. In the
current context, it is important to distinguish between different kinds of
linguistic information such as phonological, orthographic, semantic, or syntactic
information. Phonological information includes similar-sounding words, initial
phonemes of the word, the final phoneme of the word, as well as syllabic
information, all of which have been demonstrated in the lab. Note, of course,
that the phonology itself does not distinguish between one- and two-stage views,
a point to which I shall return.
Retrieval of Phonologically-Related Words. In chapter 3, the blocking
hypothesis was introduced. The blocking hypothesis contends that TOTs are
caused by the retrieval of items known to be incorrect. Although the evidence
that blocking causes TOTs or lexical retrieval failures is lacking, there is
considerable data that suggest that TOTs are often accompanied by persistent
TOTs ON RETRIEVAL 77
The crux of the debate between those who argue for a two-stage model of lexical
retrieval and those who argue for a one-stage theory partially hinges on the issue
of syntactical access during a TOT (Caramazza & Miozzo, 1997). In the two-
stage theory, the lemma represents a prephonological stage in which semantic
and syntactical information have been compiled. Syntax here means information
pertaining to the part of speech, article required, grammatical gender (e.g.,
whether nouns are masculine or feminine), and whether the word agrees with the
sentence context in tense and number. After this information is assembled in the
lemma, the retrieval of the lexeme (or phonology) ensues. In the one-stage view,
the syntax is added to the retrieval process in parallel with the retrieval of the
phonological information and compiled into the lexeme.
According to the basic two-stage model (Levelt, 1989), syntactical
information is represented along with semantic information in the lemma.
Therefore, if the lemma is completely activated in a TOT, syntactical
information about the word should be accessible even if phonological
information is not accessed. Recall that phonological information is not accessed
later at the level of the lexeme. In the two-stage model, both semantic and
syntactic information are accessible before phonology. Thus, rememberers
experiencing a TOT should have access to syntactical information, regardless of
whether they have access to phonological information. On the other hand, in a
one-stage model syntactical and phonological information are not retrieved
sequentially, and therefore may be independent of each other. Thus, the relation
between the retrieval of syntactical information and the retrieval of phonological
information can inform us about the underlying processes. Caramazza and
Miozzo (1997) stressed the importance of looking for correlations between the
retrieval of the two kinds of information during TOTs. If there is no correlation
between them, then it is likely that they are retrieved independently and in
parallel, in keeping with a one-stage model.
Grammatical Gender. In some languages, a key element of a word is its
grammatical gender. In some of these languages, a noun may be male, female,
or neuter, whereas in other languages, words are either male or female. For most
nouns, there is no relation to biological gender and grammatical gender. Thus, a
common word such as table may be masculine in one language (e.g., French) but
feminine in another (e.g., Spanish). Because gender is a key aspect of syntax,
many theorists postulate that grammatical gender is stored in the lemma for each
word (Garrett, 1992; Levelt, 1989). If so, it is possible that people experiencing
TOTs in those languages with grammatical gender will often be able to correctly
recall the gender of a missing word (Miozzo & Caramazza, 1997; Vigliocco et
al., 1997; Vinson & Vigliocco, 1999).
80 CHAPTER 4
Three initial studies in this area were conducted with Italian speakers
(Caramazza and Miozzo, 1997; Miozzo and Caramazza, 1997; Vigliocco et al.,
1997). Each study examined grammatical gender and its relation to the TOT. In
Italian, all nouns are either masculine or feminine, and as Vigliocco et al. (1997)
pointed out, the assignment is seemingly arbitrary and not based on any
conceptual distinctions. For example, there is an Italian word stella, which is
feminine, and a word astro, which is masculine. Both translate to the English
word star. Usually, gender is correlated with noun endings. Masculine words
end in “o” and feminine words end in “a,” but there are many exceptions (e.g.,
the word plasma is masculine). Grammatical gender has important syntactical
consequences in sentence formation, as it determines agreement within the
sentence (Vigliocco et al., 1997).
Now consider an Italian speaker in a TOT for a particular word, say,
amniocentesi (amniocentesis), which is feminine and irregular (Caramazza &
Miozzo, 1997). Even if the speaker can retrieve the final phoneme of the target
word, it does not provide that speaker with any information concerning
grammatical gender. According to two-stage theory, the lemma stage compiles
semantic and syntactical information prior to the retrieval of phonological
information. Thus, the speaker should be able to retrieve the grammatical gender
in the absence of retrieving the specific phonological form of the word. The data
from these studies will now be evaluated.
Vigliocco et al. (1997) presented participants with definitions of words and
asked them to retrieve the target word. If the participants could successfully
retrieve the word, they were then given the next definition. If they could not
recall the target for a particular definition, the participants were asked to rate
their feeling of knowing on a 1 to 5 scale, with 5 indicating a TOT. Then they
guessed the grammatical gender, the number of syllables, as many letters of the
word as they could, and any related words that came to mind.
Problematically however, Vigliocco et al. (1997) did not use rememberers’
ratings of how well they knew the word to determine whether the rememberers
were experiencing a TOT. In their published report, they completely ignore
whether the participants felt they were in a TOT Instead, TOTs were assigned to
all unretrieved targets for which the rememberer made partial information
guesses. Positive TOTs were defined as those that were later recognized as
correct, and negative TOTs were those in which the rememberer did not
associate the target word with a hypothetical intended word.
Allow me to add an aside here. Although not really relevant to the interests of
Vigliocco, this study tells us nothing about what kinds of information are
accessible during a TOT, or what is the etiology of the TOT. It does shed light
on the nature of syntactical access in words for which the phonology is lacking,
however, and therefore is relevant for debates concerning the nature of word
retrieval.
TOTs ON RETRIEVAL 81
Vigliocco et al. (1997) found that gender was retrieved correctly 84% of the
time when rememberers were in (experimenter-defined) positive TOT but only
53% of the time when in a negative TOT. This suggests that rememberers do
have access to syntax even when the phonological form is missing. In further
support of this claim, Vigliocco et al. reported that rememberers retrieved
gender correctly 80% for the positive TOTs, when only those TOTs for which
no phonological information was reported, not significantly different from the
overall rate. Moreover, when they considered only those irregular words in
which grammatical gender could not be determined by retrieving the last letter,
the rate was still 80%. Thus, the results support the notion that syntactical and
phonological information are represented at different levels of the word-retrieval
system, and that it is possible to accurately retrieve syntactical information
without retrieving the full phonological form.
Fortunately, the problem of relying on experimenter-defined TOTs in the
Vigliocco et al. (1997) article was corrected in a pair of articles on the
accessibility of gender in Italian during TOTs (Caramazza & Miozzo, 1997;
Miozzo & Caramazza, 1997). I discuss the Miozzo and Caramazza article first,
as it mainly dealt with empirical issues, and then turn to the more theoretical
article of Caramazza & Miozzo.
Miozzo and Caramazza (1997) used a very similar paradigm, asking Italian-
speaking participants to provide the words matching stimulus definitions or
pictures. If they did not recall the target word, they indicated whether they were
in a TOT, and then were required to guess the gender and the initial phoneme.
However, unlike Vigliocco et al. (1997) Miozzo and Caramazza used
rememberers’ subjective judgments of TOTs to classify the unrecalled items.
They compared those items that the participants rated as TOTs with the set of
“don’t know” responses (n-TOTs), which occurred when the participant
indicated that he or she could not retrieve the target and was not in a TOT
Miozzo and Caramazza refer to the n-TOTs as “don’t know” or “DK” responses,
but I will use n-TOTs to remain consistent with other sections in this book.
Miozzo and Caramazza found that grammatical gender was correctly reported in
71% of the TOTs, but only 47% (chance) for the n-TOTs responses. Moreover,
rememberers guessed the initial phoneme 72% for TOTs, but only 58% for n-
TOTs.
Miozzo and Caramazza’s (1997) data are more convincing than Vigliocco et
al.’s (1997) work because the comparison is between subjectively-defined TOTs
and n-TOTs and not between experimenter-defined TOTs and items for which
no partial information was reported. Nonetheless, both studies suggest that
syntactical information, in this case grammatical gender, can be correctly
reported in the absence of the retrieval of the target word. Thus, at a first glance,
the data could be taken to support a two-stage model of lexical access in which
phonology is retrieved only after a lemma, complete with syntactical
information, is retrieved. However, Miozzo and Caramazza urged caution.
82 CHAPTER 4
In fact, Miozzo and Caramazza (1997) argued that the data suggest that
syntax and phonology may indeed dissociable. This suggestion is based on the
observation that, in their Experiment 2, the correct gender was correctly reported
77.8% of the time, and the correct first letter was reported 78.8% of the time.
According to the two-stage theory, this could not happen because for
phonological information to be retrieved, an intact lemma must first be
retrieved, in which the syntax is already known. Therefore, for two-stage models
to be correct, the syntactical retrieval must be higher than the phonological
retrieval. Moreover, if there is control for guessing, the 1% difference would
actually yield a great advantage for the retrieval of phonological information
during TOTs. In other words, guessing the gender correctly will occur 50% of
the time, but guessing the first letter is much less likely, given there are 26
possible first letters.
Thus, Miozzo and Caramazza (1997) rejected serial models of word retrieval
because phonological access during TOTs and n-TOT responses were higher
than syntactical access. If the lemma is always retrieved in TOTs, and it is
phonological access that is deficient, one would not expect better retrieval of
phonology than any aspect of syntax. This led Miozzo and Caramazza to follow
with a second study in which they directly tested the relation between retrieved
syntax and retrieved phonology (Caramazza & Miozzo, 1997).
Caramazza and Miozzo’s (1997) experiments were similar to the earlier
study, except that, in the new article, they were able to correlate gender retrieval
with initial phoneme retrieval. They considered this a crucial test of the role of
syntax in the lemma, and whether it is necessary to consider the lemma stage at
all. If the two-stage model is correct, and a lemma contains syntactic
information, then there should be a strong positive correlation between the
retrieval of gender and the retrieval of phonology during TOTs. That is, if the
syntax precedes the phonology, then phonological retrieval should occur only
after syntactical retrieval. In a one-stage model, the retrieval of syntax and
phonology is potentially independent. Thus, a low or zero correlation would
support a one-stage view.
In their first experiment, Caramazza and Miozzo (1997) found that gender
was retrieved correctly for positive TOTs nearly 74% of the time, whereas the
correct initial letter was retrieved only 28.3% of the time, roughly consistent
with a two-stage model. However, when they examined the correlations they
found that the correlation between the likelihood of retrieving the correct gender
and the likelihood of retrieving the correct phoneme was -.085, not significantly
different from zero. This zero correlation supports the one-stage view which
suggests that the retrieval of syntax and phonology are independent of each
other. In other words, retrieval of phonological information during a TOT does
not depend on the prior retrieval of a syntax-driven lemma.
Caramazza and Miozzo (1997), however were concerned about the low rate
of retrieval of the initial phoneme. In a second more simple experiment, the rate
TOTs ON RETRIEVAL 83
of initial phoneme retrieval rose to 57.5%, only slightly lower than the 67.8%
correct gender retrieval. Nonetheless, the correlation was still not significantly
different from zero (.034), again finding no comfort for the two-stage models of
lexical retrieval. Therefore, Caramazza and Miozzo concluded that two-stage
models are not necessary to account for the retrieval of syntax and phonology
during TOTs. The retrieval of each seems to be independent of each other,
supporting a one-stage model.
Interestingly, the story is not complete. Recently, Silverberg et al. (1999)
attempted to replicate the retrieval of gender during TOTs in another language,
Hebrew. Unlike Italian, Hebrew speakers were unable to access gender during
TOTs. In fact, in their study, gender access was just slightly above chance for
both TOTs and n-TOTs. This suggests that, unlike other aspects of TOTs which
may be universal (e.g., phenomenology), access to gender is variable depending
on the structure of the language. Silverberg et al.’s data can also be interpreted
to support Caramazza and Miozzo’s (1997) view that a lemma stage is not
necessary to explain lexical access. If it were, then syntax would be assembled
there, and Hebrew speakers would have access to gender even when they did not
have full phonological access.
The relation between TOTs and grammatical gender is unclear. In Italian,
rememberers have access to grammatical gender during TOTs. Recently, this
has been replicated in Spanish (Gonzalez & Miralles, as cited in Vigliocco et al.,
1999; Schwartz, Castillo-Andrade, & Gonzalez, 1998). However, Hebrew
speakers do not access gender while experiencing TOTs. Caramazza and
Miozzo (1997) argued that the data on gender suggest that it is not necessary to
invoke a lemma stage in retrieval to explain TOT data. Nonetheless, as we see in
the following section, there is strong support for the notion that some syntactical
information is accessible to rememberers, regardless of language. We turn to
other syntactical features now.
Other Aspects of Syntax. An interesting aspect of language is the way in
which we pluralize certain nouns. Linguists distinguish between mass and count
nouns (Pinker, 1999; Vigliocco et al., 1999). Count nouns are objects that are
countable and are those that get pluralized normally in speech. For example, the
plural of the word cucumber is cucumbers and the plural of bird is birds. Mass
nouns refer to those objects that are not countable and usually exist in groups;
they do not get pluralized. For example, the plural of asparagus is asparagus and
the plural of thunder is thunder. As you can see from these examples, there are
many instances in which the division of words into mass or count nouns seems
arbitrary. Thus, it is likely that this syntactical division must be represented in
memory for each word.
Vigliocco et al. (1999) examined whether participants could determine if a
noun was a mass or a count noun, in the absence of retrieving the actual word in
English. That is, if participants could not retrieve the target word, they were
asked to guess whether the word was pluralized in a mass or count fashion.
84 CHAPTER 4
Much of the research supports the one-stage view of lexical access. I concur
with the Caramazza and Miozzo (1997) position: the TOT data support
Caramazza and Miozzo’s assertion that a one-stage model is more predictive
than two-stage models. In their view, a semantic intention (that is, to produce a
particular linguistic idea) is followed by two parallel processes, one that invokes
the correct syntax, and one that calls the correct phonology. During a TOT,
either or both of these processes may be partially or completely disrupted. That
there is no correlation between the retrieval of partial syntactic information and
the retrieval of partial phonological information suggests that syntax is not
compiled at a lemma stage prior to phonological retrieval.
TOTs have played a role in the development of two other lexical theories. TOTs
recently contributed to our understanding of lexical representation in bilinguals
(e.g. Askari, 1999; Gollan & Silverberg, in press). TOTs also play a role in the
understanding of the relation between hand gestures and spoken language
(Beattie & Coughlan, 1999; Frick-Horbury & Guttentag, 1998). I discuss the
issues of TOT and bilingual representation first and then address gestures.
One of the cognitive issues being studied about bilinguals is the nature of
semantic representation. One the one hand, the single-store view is that
bilinguals represent meaning of words in both languages in a common
representational system (e.g., Altarriba, 1992, Altarriba & Mathis, 1997; Chen &
Ng, 1989). On the other hand, the dual-store view postulates that meaning is
represented separately for each language the person knows (Askari, 1999;
Tulving & Colotla, 1970). Contemporary views favor the single-store view, at
least for speakers who are native or fully fluent in both languages (Altarriba,
1992; Gollan & Silverberg, in press). Whether this is true for those who are less
fluent in a second language is beyond the scope of this discussion. I have no
intention or expertise in delineating the important experiments and ideas in this
area. However, recently two studies have looked at this issue using TOT
methodologies, and each study drew opposite conclusions about the nature of
bilingual semantic representation (Askari, 1999; Gollan & Silverberg, in press).
Askari (1999) examined semantic and phonological priming in Farsi-English
bilinguals. Farsi is an Indo-European language, a descendent of Persian, now
spoken in Iran. Askari’s bilinguals were mostly Farsi-domi-nant, but had been
speaking English since childhood. Her study had two objectives. First, she
wished to test models of bilingual representation. Second, she wished to test
whether Burke et al.’s (1991) transmission deficit model best explained the
pattern of TOTs.
86 CHAPTER 4
During normal human speech, most people combine their spoken language
(articulation) with an array of hand gestures (gesticulations; Frick-Horbury &
Guttentag, 1998). Many researchers have wondered about exactly what is the
interplay between hand gestures and spoken language (e.g., Krauss, Dushay,
Chen, & Rauscher, 1995). Some theorists hypothesize that gestures help us
retrieve words by providing the speaker with retrieval cues (Krauss et al., 1995)
Two new studies have examined the relation between iconic hand and arm
gestures and the likelihood of a TOT (Beattie & Coughlin, 1999; Frick-Horbury
& Guttentag, 1998). Gestures include clasping hands, raising hands above the
head, scratching eyebrows, and drumming fingers. lconic gestures include
indicating size by gesturing outward with the hands and pantomime gestures to
demonstrate the potential use of an object when the name cannot be retrieved.
Frick-Horbury and Guttentag (1998) conducted a standard Brown-McNeill
(1966) type study, in which participants were asked to recall low-frequency
targets when presented with their definitions. Because Frick-Horbury and
Guttentag were primarily interested in the role that gestures play in the retrieval
of words, half of the participants were prevented from making gestures while
doing the task, and the other group was free to move their hands as they pleased.
This experimental manipulation had no effect on overall TOT rates or TOT
resolution. However, when Frick-Horbury and Guttentag looked only at those
participants who were free to gesture, they found that many more TOTs were
accompanied by gestures than were n-TOTs. Thus, a TOT compels us to wave
our hands trying to retrieve the target, although the hand gestures apparently do
not help us resolve TOTs. Frick-Horbury and Guttentag’s (1998) fascinating
study is unique in that it examines the hand movements that accompany
subjective TOTs and frustrating word retrieval. Beattie and Coughlin (1999)
also looked at the relation of hand gestures and lexical retrieval, but they did not
ask the participants to make TOT judgments. TOTs were determined by the
experimenter when they engaged in some unspecified combination of verbal
statements (e.g., “oh God, I know it"), facial expressions (such as winces), body
movements, head movements, and foot and leg movements.
Beattie and Coughlan (1999) found that when rememberers were “free to
gesture,” they recalled no more targets than when they were instructed to “keep
their arms folded.” However, when they were free to gesture, they experienced
substantially more TOTs than when they were to keep their arms folded (112 to
72). Moreover, more TOTs were resolved when they were free to gesture. Thus,
TOTs and TOT resolution increases when people are free to gesticulate.
On the surface, there is again a contradiction between Frick-Horbury and
Guttentag’s (1998) data and that of Beattie and Coughlan (1999). One study
found that restricting hand gestures did not affect TOT rates and resolution, but
90 CHAPTER 4
the other did. I suspect the difference lies in the methodology. Because Beattie
and Coughlan did not rely on subjective TOTs, it is possible that they classified
as TOTs items that Frick-Horbury and Guttentag’s participants would have
labeled n-TOTs. Although there are only two studies from which to draw
inferences, it is the Frick-Horbury and Guttentag study that is more relevant to
those interested in the TOT, as they actually measured TOTs. Nonetheless,
further study is necessary to resolve this discrepancy.
In terms of the standard two-stage model of retrieval, it is possible to
consider that gestures, like semantics and syntax, are represented at an early
level of the word retrieval process (i.e., the lemma). Thus, allowing for gestures
strengthens the representation of the lemma, allowing for stronger activation of
phonological levels, which then increase TOTs and the likelihood of resolution.
However, it is also possible to interpret these results in terms of inferential
models. Gestures may serve as cues that the target item is known and is in
memory, and they increase the probability of resolution of unrecalled items. The
gestures may increase TOTs because rememberers correlate their likelihood of
retrieval with their state of gesturing. If they find themselves gesturing, it is
assumed that they are likely experiencing a TOT. Thus, the very freedom to
gesture increases TOT rates.
The common ground in both Frick-Horbury and Guttentag (1998) and Beattie
and Coughlan (1999) is that gestures increased when people were experiencing
TOTs. It is unclear how to best interpret these data. It is possible that the
gestures are used to help infer that a TOT is present, as just discussed. It is also
possible that people intuit that gestures help them retrieve words. Given that
TOTs are feelings that a word is retrievable, it may be that people choose to
gesture more when experiencing a TOT. If, as Beattie and Coughlan suggested,
gestures do help recall, then it is good strategy to gesture during TOTs.
Reading and writing has been extensively studied in Chinese because Chinese,
unlike most languages, uses a logographic writing (Xu, Pollatsek, &Potter,
1999; Zhang, Perfetti, & Yang, 1999). Most languages, such as English, use a
phonological alphabet, in which each letter represents a particular sound in each
word. In Chinese, however, characters refer to concepts, not to sounds. Sun,
Vinson, and Vigliocco (1998) examined a fascinating phenomenon in Chinese
writing called the tip-of-the-pen phenomenon (TOP). According to Sun et al.,
there is an idiom in Chinese (Ti-Bi-Wang-Zi) that means literally, “Pick up a
pen but forget how to write a character.” Thus, the TOP refers to the “condition
in which a speaker knows a word but cannot write it.” In other words, the
Chinese speaker can say the word aloud, but cannot remember the character
TOTs ON RETRIEVAL 91
used to write the word. Sun et al. found that Chinese participants were able to
retrieve partial orthographic information when in a TOP, similar to the partial
phonological information retrieved during a TOT. Thus, the TOP may mirror the
TOT in both phenomenology and its relation to partial retrieval. TOPs have
never been reported in alphabetic languages, because knowing the pronunciation
always gives you at least a head start on how to write it. However, it would be
interesting to see if TOPs could be induced for exception or irregular words in
English (e.g., mood, blood, good), perhaps in nonnative writers.
SUMMARY
The main goal of this chapter is to describe the ways in which TOT research has
been used to explore issues of word retrieval, in general, and to assess two
versus one-stage theories of word retrieval, in particular. I also touched on how
TOTs have been used to address issues in bilingual representation. In these
research frameworks, TOTs have been used as a “window” on word retrieval,
that is, TOTs allow researchers to address what happens when retrieval is
slowed or blocked. As has been shown in this chapter, research has indicated
that both phonological and syntactical aspects of the to-be-retrieved word are
accessible during TOTs, and that this has been interpreted as supporting both
one- and two-stage views.
In this chapter, I discussed theories of lexical retrieval. Two-stage theory
stipulates that word retrieval occurs in two stages. In the first stage, a lemma is
retrieved. Lemmas contain both semantic and syntactical information specifying
a target word. In the second stage, a lexeme is retrieved, which specifies the
phonological form of the word. In the one-stage view, the lemma is removed,
and syntactical information is not retrieved until the level of the lexeme.
TOTs have been used to address this issue. The studies find that both
phonological and syntactical information is accessible during a TOT. Partial
gestural information is retrievable for American Sign Language signers in
TOTs. Some researchers contend that this pattern of findings supports the two-
stage view. Caramazza and Miozzo (1997), alternately, argued that the statistical
independence of phonological and syntactic retrieval suggests that the one-Stage
model might better account for the data. If syntax were retrieved before
phonology, then the retrieval of phonology would be partially dependent on the
prior access to syntax. Similar logic has been used to compare single-store and
dual-store views of bilingual representation. TOTs are more common in
bilinguals suggesting greater access to a common semantic access when
individual phonological access in each language may be lacking.
5
Theories of Metacognition
An old saying about men is that they never stop to ask for directions, no matter
how lost they actually are. The proverbial “lost man” in this joke is so sure that
he knows where he is that it does not occur to him to stop somewhere and ask
for directions. The man’s failure can be considered an instance of poor
metacognition. To an outside observer, the man is clearly lost, but that is not
subjectively salient to our man. Thus, his metacognitive experience is
inaccurate. Moreover, because he does not know that he is lost, he does nothing
to ameliorate the situation, such as calling and getting new directions. The man
has poor metacognitive monitoring skills, and he cannot adequately control his
performance.
Although we each may know people similar to the one in the joke, research
has shown that fortunately, our metacognitive systems are actually quite good at
informing us about what we know, do not know well, or do not know at all.
Moreover, the research also suggests that based on our metacognitive awareness,
we often make quite sophisticated decisions about how to go about learning,
remembering, or probably finding our way when lost. This chapter provides an
overview of recent theory and research in the domain of metacognition.
In chapter 4, the discussion centered on how TOT has been used as a means
to study the processes of lexical retrieval. In this chapter, I adopt the
metacognitive perspective, which views the TOTs as a metacognitive experience
about word retrieval. The metacognitive approach will be introduced in general
and not directly applied to TOTs until chapter 6, although the metacognitive
perspective has been discussed previously in the context of the importance of
phenomenology (chap. 2) and etiology (chap. 3).
I consider the theory of monitoring and control, mostly drawing on the
theoretical work of Thomas Nelson and Louis Narens (e.g., Barnes, Nelson,
Dunlosky, Mazzoni, & Narens, 1999; Nelson, 1996; Nelson & Narens, 1990;
1994), drawing mostly from the literature on feelings of knowing and judgments
of learning. Nelson and Narens stressed the importance of the function of
metacognition. In chapter 6, I apply it to the work on TOTs and the focus will be
on the possible functions that TOTs may serve to keep the human mind working
efficiently and effectively.
92
THEORIES OF METACOGNITION 93
Nelson (1996) stressed the importance of two potential and integrated functions
of metacognition, which he referred to as monitoring and control Monitoring
refers to peoples’ ability to become aware of how well their cognitive processes
are working. To do this, the monitoring process builds a model of how the basic
process is working. For monitoring to be accurate, information must flow from
the “object” level process to the “meta” level process. Monitoring process, then,
can serve the basis for conscious metacognitive judgments, although in some
cases, the monitoring may be hidden from conscious access (e.g., Reder &
Schunn, 1996). Control refers to the decision that the person makes based on the
output of the monitoring system. If the monitor informs the person that learning
is going poorly, the control processes can redirect attention to the to-be-learned
items. Control is a process whereby the object level process affects changes in
the meta level process. Control, then, changes how the system works. For
control to be effective, it must be able to change the object level processes and
change it in adaptive ways. Control processes may be conscious (e.g., Mazzoni
& Cornoldi, 1993) or nonconscious (Reder & Schunn, 1996). In either case,
their effectiveness is partially dictated by the monitoring effectiveness and
partially by the choice of good strategies.
The key function of metacognition, then, is to monitor what the cognitive
processes are doing and then if needed affect change in them. For a system to be
able to monitor and control itself, it must construct a model of its own ongoing
function (Nelson & Narens, 1994). This monitoring system can represent what
the organism is doing in real time. For this model to work, Nelson hypothesized
that at least two levels of organization must exist—an “object-level” and a
“meta-level.” The object-level is ordinary cognition, retrieving information,
making decisions, and recognizing objects. The meta-level models the activity
of the object-level and therefore can make predictions about its future.
The second feature of Nelson’ s framework is that the object-level and the
meta-level must be connected such that information can continuously flow from
the object-level to the meta-level, and that information can flow back from the
meta-level to the object-level (see Fig. 5.1). This flow of information from the
object to the metalevel is what is called monitoring; that is, the meta-level needs
constant input from the object-level system to construct a model of its function.
The flow from the meta-level to the object-level is considered the control
function; that is, the meta-level can now direct the object-level to change or
redirect its processing in line with what the meta-level computes to be a more
desirable goal.
Nelson and Narens (1990, 1994) defined monitoring in the following way:
“The fundamental notion underlying monitoring…is that the meta-level is
informed by the object-level.” (Nelson & Narens, 1990, p. 127; emphasis in the
original). They defined control in the following manner: “The fundamental
94 CHAPTER 5
notion underlying control…is that the meta-level modifies the object-level, but
not vice-versa” (Nelson & Narens, 1994, p. 12, emphasis in the original). Thus,
the meta-level represents and changes the object-level.
Nelson and Narens (1994) compared the metacognition model to the
workings of an ordinary home thermostat. A thermometer, which serves as the
monitoring function, measures the ambient temperature of the room. The
thermometer can be thought of as a model of the outside world. Information
flows from the world (average motion of molecules) to the thermometer,
allowing the thermometer to maintain a model (albeit a simplified one) of one
dimension of the outside world. For the thermostat to work, the system must
have a control element as well. Indeed, when the temperature rises above or falls
below a certain threshold temperature, the thermostat mechanism, which serves
as the control function, turns on the appropriate heating or cooling unit. When
the air conditioner is turned on, the system affects the outside world by lowering
the temperature. By analogy, metacognitive judgments such as judgments of
learning, feelings of knowing, or tip-of-the-tongue states, can serve the purpose
of monitoring object-level cognition.
FIG. 5.1.
In this model, the two levels are continuously interacting with each other.
Once the air conditioner has been turned on, the thermometer continues to
measure the now-falling temperature. When the temperature returns to a
predetermined base level, the thermostat shuts off the air conditioner and allows
the process to begin anew. Similarly, according to the Nelson-Narens
framework, mental monitoring and control are engaged in continuous dance. A
control function may alter behaviors by causing the remember to focus on more
difficult study items. This may, in turn, affect the memory strength of items
being studied, which in turn alters the items selected by the control level as
items to be restudied (e.g., Thiede & Dunlosky, 1999).
THEORIES OF METACOGNITION 95
Nelson and Narens (1990) introduced one more nuance into this theory. They
asserted that the concepts of meta-level and object-level are relative, and that a
particular level that is meta to one level, may be the object of another level. This
hierarchy of levels exists for both monitoring and control. Thus, one level,
called Lx, may be superordinate to one level, Ly, but superordinate to another
level, Lz. Thus, information continually flows from lower levels to higher
levels, in the form of monitoring, and back again, in the form of control. In this
way, Nelson and Narens related their conception of monitoring and control to
earlier notions of control (e.g., Atkinson & Shiffrin, 1968). In practice however,
most research has simply looked at whether metacognitive monitoring is
accurate and the extent to which it directs control.
An example of this system in action comes from the quickness with which
people can say that they do not know something (e.g., Klin, Guzman, & Levine,
1997; Kolers & Palef, 1976; Glucksberg & McCloskey, 1981). In these studies,
participants are asked to decide as quickly as possible whether they know the
answer to a question. Monitoring allows people to quickly decide that some
answers are simply not known, and there is no need to initiate a memory search.
For example, given the question, “What was Benjamin Franklin’s horse’s
name?” most people would quickly respond “don’t know.”
It is also possible, although Nelson and Narens are not explicit about it, to
describe the monitoring/control framework in an evolutionary perspective (e.g.,
Pinker, 1997; Tooby & Cosmides, 1995). This is one of the inherent advantages
of Nelson and Narens’ functional approach, despite the difficulty in
experimentally testing such evolutionary ideas concerning the human mind.
Specifically, the argument is that the ability to monitor and control cognitive
processes may have given our human ancestors an evolu-tionary advantage over
those with less well developed monitoring and control abilities. For example,
consider two Pleistocene hunter-gatherers heading out from the settlement in
search of food. If one expresses more confidence in his knowledge of the likely
way home, he will be followed. This is only adaptive if confidence correlates
with accuracy. If not, following either member will do. A hunter-gatherer with
poorer accuracy risks causing the whole group to go hungry. Of course, many
animals find their way to food and back home without the benefit of
metacognition. But, the organism with metacognition has far more options and
the ability to choose its destiny rather than simply following its strongest
instinct.
At this point, any discussion of evolutionary advantages for a monitoring and
control system are pure speculation. Nonetheless, the functional approach
advocated by Nelson and Narens is interpretable within such evolutionary
thinking. Following the lead of Tooby and Cosmides (1995), it may be possible
to design metacognition experiments to tap into these primordial metacognitive
problems. For example, one might test how well people can monitor their
knowledge of spatial locations, or whether they feel they have learned
96 CHAPTER 5
Monitoring
The study of monitoring in memory dates back to the seminal work of J.T.Hart
in the mid- 1960s. Like Brown and McNeill on the east coast of the United
States, Hart, on the west coast, set out to examine the accuracy of feeling-of-
knowing judgments (FOK). He was interested in whether rememberers could
accurately predict whether they would recognize targets that they could not
recall. Thus, aside from some definitional differences between what a TOT was
and what a FOK was, Hart (1965) and Brown and McNeill (1966) started out
with very similar paradigms. Hart (1965) asked participants to retrieve the
answers to general-information questions (e.g., What is the capital of
Denmark?). If they recalled the answer (Copenhagen), they moved on to the
next item. If they could not recall it, they were asked to make a FOK for
recognition of the item. Hart simply asked people to indicate “Yes” for a
positive FOK or “No” for a negative FOK. He then gave them a recognition test
to assess their accuracy. Hart found that people were better at recognizing
targets when they had experienced a FOK than when they did not.
Unlike Brown and McNeill, who were primarily interested in the information
accessible during a TOT, Hart (1965) was mostly interested in whether the FOK
accurately predicted recognition performance, and thus, his interest was in the
relation between the judgment and performance. In this way, Hart initiated the
study of metacognitive monitoring, that is, how well people know what they
know and what they do not know. Hart, indeed, (1967) coined the phrase “the
memory monitoring system.”
Prospective and Retrospective Monitoring. Nelson and Narens (1990, 1994)
divided monitoring into two basic categories, prospective and retrospective
monitoring. Prospective monitoring has been the classic do main of
metamemory (e.g., Arbuckle & Cuddy, 1969; Hart, 1965, Underwood, 1966).
Prospective monitoring concerns judgments about how likely information will
be remembered in the future. Retrospective monitoring refers to the subjective
THEORIES OF METACOGNITION 97
FIG. 5.2.
98 CHAPTER 5
(e.g., Nelson, 1984, 1988; Schwartz & Metcalfe, 1994). Judgments of learning
also tend to be accurate, with the correlation being often greater than .9 (Nelson
& Dunlosky, 1991; Dunlosky & Nelson, 1994). Ease-of-learning judgments
have not been as well studied, but the studies that have been published suggest
better-than-chance accuracy as well (Leonesio & Nelson, 1990). However,
metacognitive judgments are far from perfect, and there are circumstances in
which judgments may actually mispredict performance (e.g., Benjamin et al.,
1998; Koriat, 1995).
I will interject a brief methodological note. In metacognition research, there
has developed a consensus to use gamma correlations to assess the accuracy of
judgments (see Nelson, 1984; Schwartz & Metcalfe, 1994). Gamma correlations
yield −1 to 0 to 1 values, and are therefore comparable to other correlation
statistics. They are nonparametric, allowing researchers to draw inferences from
ordinal data, such as metacognitive judgments.
Recently, within the field of metacognition, considerable research has been
devoted to the delayed judgment of learning (JOL) effect (e.g., Dunlosky &
Nelson, 1994; Nelson & Dunlosky, 1991; Nelson, Dunlosky, Graf, & Narens,
1994; Spellman & Bjork, 1992). This is a monitoring effect, in which the
gamma correlations between judgments and performance are extremely high.
The contrast between immediate and delayed JOLs is instructive in
understanding both monitoring and control. I have chosen to devote space to
reviewing the delayed JOL effect for two reasons. First, the effect demonstrates
how subtle manipulations can alter the accuracy of metacognitive judgments.
Second, explanations of the effect bring into play the role of mechanistic
theories in accounting for metacognitive monitoring.
In a typical JOL task, participants are given word pairs to learn, such as “dog-
spoon.” Later, they are given the cue word “dog” and asked to make a judgment
as to the likelihood that they will recall the target word at some point later. In an
immediate JOL, the judgment is called for immediately after the participant has
studied the pair. In a delayed JOL, the judgments is called for at some point
much later after the participant has studied the word pair (5 min, 15 min, 24 hrs).
The differences in the accuracy of the judgments is remarkable. Gamma
correlations for the immediate JOLs are respectably better than chance, usually
around +.5. However, the delayed JOLs show an astonishing accuracy, with
gamma correlations usually exceeding +.9 (e.g., Dunlosky & Nelson, 1994). The
only difference is the delay between the time of learning and time of judgment.
Explanations have varied, but they all center on one major component of the
difference between the immediate and delayed judgments (e.g., Bjork, 1994;
1999; Dunlosky & Nelson, 1997). In an immediate JOL, the target is still likely
to be in short-term memory. Thus, when the participant makes the judgment, it
is made with the answer clearly in mind. Bjork (1994) argued that this interferes
with the assessment of how difficult the retrieval will be because the participant
cannot test its difficulty by practicing the retrieval. However, when a delayed
100 CHAPTER 5
JOL is made, the target answer is no longer in short-term memory, and must be
retrieved from long-term memory. Under these circumstances, the judgments
can be made on the basis of an actual retrieval attempt. Successful retrieval of
the target at the time of judgment then makes a much better predictor of later
retrieval than the judgment made without the benefit of the attempt. Given the
similarity between what the rememberer does when the JOL is made and the
later task of retrieving the target, accuracy is extremely good.
I am simplifying the theories here for the sake of brevity, and there is
considerable debate about the specifics of the explanation of the accuracy
differences between the two types of JOLs. However, the common ground
among the theories is the notion that the delayed JOL situation allows re-
memberers to make accurate judgments by making the monitoring maximally
straightforward. Participants can use the retrieval process itself to serve as the
monitoring process in a delayed JOL. However, in an immediate JOL, this
option is not open to the rememberer because the target word is easily accessible
in short-term memory. In the immediate JOL, the rememberer must use
strategies, which are not only less accurate, but more attention demanding of the
individual because the situation does not allow for easy monitoring.
Supporting this idea concerning the accuracy of delayed JOLs and the
relative difficulty of immediate JOLs, Nelson et al. (1998) found that alcohol
intoxication has different effects on immediate and delayed JOLs. In their study,
alcohol impairment did not affect the accuracy of delayed JOLs at all. However,
for immediate JOL, sober controls showed much better accuracy than did their
alcohol-impaired compatriots. This suggests at the reason for the high accuracy
of the delayed JOLs. If one can recall the target at the time of judgment, then
there is a good chance that it will be recalled at the time of test. However, for
immediate JOLs, when one cannot benefit from a dry run at retrieval (no pun
intended), more sophisticated monitoring strategies must be employed, which
are disrupted by the alcohol.
Incidentally, the above reasoning suggests that TOTs may be relatively
difficult judgments to make. In TOTs, the target word is not accessible.
Therefore, the strategies used to infer its presence must tap into sophisticated
strategies, regardless of whether they are direct access or inferential mechanism.
Thus, the delayed JOL findings can be interpreted to predict that TOT accuracy
should be above chance, but less than perfect. Indeed, in studies in which
gamma correlations have been computed, TOT accuracy usually comes out
around +.6 (e.g., Schwartz, 1998; 2001). Thus, when viewed as metacognitive
judgments, TOTs are accurate, but not perfectly so.
Under normal circumstances, both feelings of knowing and judgments of
learning show high correlations between the judgments made and the later
memory performance. Indeed, in the case of delayed JOLs, the accuracy is near
perfect. This rosy picture of metacognitive monitoring stands in contrast to
earlier views of human introspective ability, which showed some major deficits
THEORIES OF METACOGNITION 101
in insight (e.g., Nisbett & Wilson, 1977). However, most researchers agree that
metacognition is an inferential task, and thus is subject to error.
Metacognitive perspectives of monitoring emphasize both the demonstration
of accuracy of judgments and an acknowledgment that accuracy may deviate
from perfection, or that some judgments may not reflect object-level cognition at
all (see Nelson & Narens, 1994; Nisbett & Wilson, 1977). Indeed, this latter
contention leaves open the possibility that some monitoring may not only not
reflect accurate ongoing cognition, but may actually distort the object-level
cognition. Therefore, the monitoring/control perspective fits neatly with
Tulving’s criticism of the doctrine of concordance (Tulving, 1989). Monitoring
research does not assume the judgments made by people will necessarily reflect
the object-level cognition.
When will metacognitive monitoring actually mispredict performance? The
delayed JOL effect suggests that accuracy will be high in circumstances in
which the basis of monitoring is related to the to-be-predicted task in a
straightforward manner. Circumstances that call for complex monitoring
processes will show lower accuracy. If the experimental situations can be
constructed such that rememberers are using the wrong bases for monitoring,
judgments that participants make may, in fact, be negatively related to the task
that they are predicting.
Several studies have now identified situations in which monitoring not only
does not predict performance, but it mispredicts it consistently (Benjamin et al.,
1998; Koriat, 1995). Koriat’s study found that usually accurate feeling-of-know-
ing judgments could be mispredict performance if “trick” questions were used.
Koriat (1995) presented Israeli participants with deliberately chosen “trick”
questions, such as “What is the capital of Australia?” or “In what U.S. state does
one find Yale University?” More Israelis incorrectly respond “Sydney” and
“Massachusetts” than they do with the correct answers (Canberra and Conn-
ecticut, respectively). Moreover, they express high confidence in these incorrect
answers or in their ability to recognize correct answers. Other questions, such as
“What is the process by which plants produce food?” (photosynthesis) may elicit
less confidence in their abilities, but more correct answers. Koriat asked
participants to make feeling-of-knowing judgments (predictions of future
recognizability) for misleading and non-misleading questions. For the mis-
leading questions, the feeling-of-knowing judgments was negatively correlated
with recognition performance, that is, confidence was higher for those questions
that were later answered incorrectly than those answered correctly. Thus, at least
in this specific situation, metacognitive monitoring was inaccurate.
Benjamin et al. (1998) also uncovered a number of situations in which
metacognitive monitoring was inaccurate, but their focus was on JOLs. In one
study, Benjamin et al. asked people to retrieve answers to general-information
questions. However, the JOLs were directed at whether the participant would be
able to free recall the target answer later in the absence of the question. If they
102 CHAPTER 5
did not recall the answer themselves, the answer was provided. The logic of this
experiment was as follows. Some questions (e.g., What is the capital of France)
will be easy, and others (e.g., What is the capital of Argentina?) may be more
difficult. The JOL task, however, is directed at the likelihood of free recalling
Paris and Buenos Aires. Some research (Gardiner et al., 1973) suggested that the
more difficult items in cued recall become the easier items in free recall.
Benjamin et al. suspected that participants would use the ease of retrieving in
cued recall to predict the likelihood of free recall. If this is so, then the JOLs
should mispredict free recall. To operationalize difficulty in cued recall,
Benjamin et al. used the amount of time it took participants to retrieve the
answer or decide that they did not know. More difficult items took longer
whereas easier items required less time. For example, compare the speed of your
response to a question such as “What is the capital of France?” versus a question
like “What is the capital of Argentina?” For most Americans, at any rate, Paris
will be quickly retrieved, but Buenos Aires may take some time.
Under these circumstances, the participants mispredicted their performance.
They gave high probability judgments to the ones that had been easy to retrieve
initially, and low judgments to those that had been retrieved with difficulty. In
the free recall test, they were more likely to recall the originally-difficult ones,
like “Buenos Aires” and less likely to recall the originally-easy ones, like
“Paris.” Therefore, the data supported the original hypothesis of Benjamin et al.
(1998). Under certain misleading circumstances, the processes for monitoring
involve errors, and these errors result in inaccurate monitoring.
To summarize, metacognitive monitoring is generally accurate. This is true
for both judgments of learning and feeling of knowing. Under some circumsta-
nces, it can be highly accurate (e.g., Dunlosky & Nelson, 1994). However, acc-
uracy only occurs when the monitoring process constructs an accurate model of
the object-level process. There are circumstances in which this does not occur,
and accuracy can be at or below chance (Benjamin et al., 1998; Koriat, 1995).
Control
Mazzoni et al. found a strong negative correlation between the JOLs and the
amount of time that the rememberers studied the items. The rememberers chose
to study those sentences given low JOLs (i.e., the difficult items) for longer
durations than those sentences given high JOLs (i.e., the easy ones).
Presumably, the participants thought that they could recall the easy ones without
much study, but that the difficult ones required more study. This negative
correlation between JOLs and allocation of study time is the typical finding in
these studies and has been replicated many times (see Nelson, 1993; Son &
Metcalfe, 2000). Therefore, it appears that rememberers are able and willing to
use their metacognitive monitoring, as measured by the JOLs, to their
advantage, by using monitoring to influence control decisions that they make.
However, just as monitoring is not completely effective, neither is control.
Nelson and Leonesio (1988) identified a phenomenon that they labeled the
“labor-in-vain” effect. This refers to the observation that participants may
allocate more study time to the difficult to-be-learned items, but that does not
ensure that they will, in fact, learn these items. Indeed, in all of the studies
correlating JOLs to study time allocation, the participants fail to completely
compensate for the original difficulty differences. This means that although they
study the more difficult items for longer durations than the easy items, they still
recall more of the easy items. Thus, Nelson and Leonesio argued that control,
like monitoring, is not perfect.
I have found an everyday analog to this labor-in-vain effect when I learn the
names of students in my class. In a typical psychology class at my university,
there are usually about five women for every one man. As a function of that
disparity, I find that I learn the names of the men more quickly than the names
of the women. Knowing this, I emphasize the women’s names in my learning
attempts because my judgments of learning tells me that they are harder.
Nonetheless, I still learn each individual man’s name more quickly than each
individual woman’s name despite my efforts to compensate.
In general, the strategy that learners adopt is to focus on the items that they
judge to be more difficult. They assume that the easy items require less time or
less effort. This strategy is presumably highly effective when the learner has
sufficient time to study and learn the difficult items. However, what happens
when the overall amount of time is more limited? One possibility is that learners
will continue to focus on the difficult items. Another possibility is that people
will realize the hopelessness of learning the difficult ones in the allotted time
and will focus their efforts on the easy ones.
Thiede and Dunlosky (1999) tested these two possible scenarios by observing
how participants selected items for restudy, that is, which items the participants
choose to restudy, not necessarily how long they studied each item. In their
experiments, they varied the amount of total time that participants had to master
a list of word pairs. In some conditions, the total study time was very limited,
whereas in other conditions, participants had almost unlimited time to study the
THEORIES OF METACOGNITION 105
items. When study time was unlimited, Thiede and Dunlosky found the usual
negative correlation between JOLs and selection of items for restudy. However,
when study time was limited, participants opted for a different strategy and
choose the easy items for study, generating a positive correlation between JOLs
and study selection.
Son and Metcalfe (2000) conducted a similar study, in which students were
asked to allocate study time to a number of different short expository essays.
Like Thiede and Dunlosky (1999), Son and Metcalfe found that the allocation of
study time was positively correlated to the judgments of learning when overall
time was limited. The students spent more time studying those items that they
felt would be easier to remember. In simple terms, the participants realized that
they would not have sufficient time to master the difficult items, and therefore,
chose to concentrate their efforts on those items that they thought they had a
better chance of remembering anyway. Interestingly, however, Son and Metcalfe
found that judgments of interest were more predictive of study than judgments
of learning. Thus, although the participants did use some strategies to enhance
learning, they were more likely to follow their interests when expository text
was used. In other words, a grade of “B” might be satisfactory for Physics, but
students want “A’s” in their Psychology classes. These data suggest that
rememberers have complex control mechanisms at their disposal (see also
Nelson et al., 1994). They will switch control strategies depending on the time
constraints and on the intrinsic interest of the materials being studied.
Control at Retrieval. Recent studies also suggest that control functions
influence decisions at the time of retrieval (e.g., Bames et al., 1999; Reder,
1987, 1988; Reder & Schunn, 1996). Monitoring can help people to determine
whether particular information is accessible in memory. If the answer is
positive, participants may spend more time attempting retrieval and may try
more retrieval strategies to recall the item. Although equally important, control
at retrieval has received less attention from researchers than control at encoding,
but there have been a few relevant studies. Rememberers must decide at the time
of retrieval whether it is worth the effort to attempt retrieval, how to recall items,
and whether to admit that one does not know. These decisions may have
important cognitive and social implications. I briefly describe some of the
research designed to test retrieval control.
Reder and her colleagues (1987, 1988; Reder & Schunn, 1996) were
interested in how metacognition affected strategic decisions in the regulation of
recall. In particular, they were interested in whether people relied on memory
search or opted for other strategies such as calculation (in math problems) or
inference (in general-information problems). For example, some questions may
be so difficult (How many minutes have passed since the Big Bang?) that most
people will quickly respond “I don’t know.” Other questions (“How many legs
do antelopes possess?”) may not be stored in memory, but may be easily
inferred once the question is posed. Similarly, a question such as, “How many
106 CHAPTER 5
windows does your current home have?” may not be stored in memory and
hence not immediately retrievable, but may be easily determined by mental
counting. Other questions (“Who was the first president of the United States)
draw forth a quick and definitive response from memory. One of the functions
of metacognition, according to Reder and Schunn, is that it allows us to make
rapid decisions about our state of knowledge. For Reder and Schunn,
metacognition directs the strategies that people use to solve problems or answer
questions. Interestingly, Reder and Schunn argued that most metacognition is
outside of the domain of consciousness. One does not necessarily access one’s
confidence when answering such questions. Nonetheless, the metacognitive
process is directing retrieval strategies.
Barnes et al. (1999) designed an experiment to test the strategic regulation of
memory retrieval. They presented participants with general-information
questions for retrieval. They provided penalties for lack of speed, and rewards
for number of correct answers. Thus, participants were penalized for each
second the question was on the screen, but were rewarded whenever they
provided a correct answer. They compared conditions in which the penalty for
slowness was low vs. high, and when the reward for correctness was low vs.
high in two separate experiments. These incentives were predicted affected the
metacognitive control processes.
Barnes et al. found that when the incentive for speed was high, the
rememberers spent less time at retrieval, and the number of correct answers
decreased. However, when the incentive for correct answers was high, the
rememberers spent more time attempting retrieval and generated more correct
answers as a consequence. In a second experiment, Barnes et al. found that when
the incentive was high, participants made fewer commission errors, although
they recalled less overall. These findings suggest that rememberers are capable
of using sophisticated control processes at retrieval as well.
In the Barnes et al. (1999) study, explicit measures of metacognition were not
made. However, other research has demonstrated correlations between the
magnitude of feeling of knowing and retrieval latency. Nelson et al. (1984) and
Nelson et al. (1998) showed that FOK judgments were positively correlated with
retrieval latency for omission errors, that is, the longer the response time to
recall, the higher the FOK judgment had been. These results were replicated by
Costermans et al. (1992). I will present similar evidence for TOTs in the next
chapter.
I discuss one more study concerning control processes and metacognition.
The study (Koriat & Goldsmith, 1996) differs from the others reviewed here
because it concerns retrospective monitoring and its effect on control rather than
prospective monitoring. Retrospective monitoring refers to j udgments made
about the correctness of an already-retrieved target. These judgments are often
called simply confidence judgments and have been studied heavily in the
eyewitness memory area (e.g., Loftus et al., 1989). Confidence in a retrieved
THEORIES OF METACOGNITION 107
answer can affect control decisions. Koriat and Goldsmith were interested in
whether these confidence judgments would affect decisions to output answers
when the situation called for different criteria of correctness.
Consider a question such as “What is the capital of California?”
(Sacramento). There are three possible categories of answers, a correct one, an
incorrect one (commission error), and a “don’t know” response (omission error).
A “don’t know” response is not incorrect, and therefore does not count as an
“inaccurate” memory. On the other hand, if one says “San Francisco,” one has
committed a commission error, which does count against the accuracy score.
Now consider the person who, when asked this question, originally thinks of San
Francisco, and then realizes that the capital is a smaller, less well-known city,
but is not sure of its name. If forced to guess, the person may say San Francisco,
even though the confidence is low for that name. Thus, Koriat and Goldsmith’s
(1996) interest was in how participants withhold low-confidence answers like
San Francisco. The withholding of answers is defined as the control function,
whereas the confidence judgment is the monitoring process that informs the
control process. This monitoring/control function applies to both incorrect and
correct answers.
In their experiment, Koriat and Goldsmith presented participants with
general-information questions. They were required to answer all questions and
then were asked to make a confidence judgment (on a 0 to 100% scale) for each
question. Thus, even if they had no idea what the answer was, they were
required to make a guess. Presumably, such guesses elicited answers
accompanied by low confidence judgments. After they answered all of the
questions, the instructions switched to free-report conditions. The questions
were shown again, and now the participants were free to answer only those
questions that they were sure were correct. Koriat and Goldsmith could then
examine the relation between the original confidence judgments and the
subsequent decision to report or withhold. To make matters more interesting, in
one condition, the participants were offered a financial incentive to answer
correctly, coupled with an equal disincentive for wrong answers (moder-ate-
incentive condition). In a second condition, the disincentive was much larger
than the incentive (high-incentive condition).
The results were quite striking. First, the participants were accurate with their
confidence judgments. Those answers given high confidence ratings were much
more likely to be correct than those given lower confidence ratings. Indeed, the
gamma correlation was .87 for recall. There was also a strong positive
correlation between confidence and whether the answer would be volunteered in
the second phase. Moreover, in the high-incentive condition, the participants
screened out more of their incorrect answers, resulting in an increase in overall
accuracy. This came at a cost of a decrease in quantity, but it does show that the
participants were able to monitor their output (via confidence) and control their
accuracy (via volunteering or withholding answers).
108 CHAPTER 5
Koriat and Goldsmith (1996) argued that these data are consistent with a
metacognitive model. Rememberers monitor the accuracy of their answers,
reflected in their confidence judgments. Then, depending on the incentives, they
can use their confidence judgments to decide which answers they will output
and which they will withhold. Control is represented by the volunteering or
withholding of answers. Therefore, we see the usefulness of the metacognitive
model of monitoring and control in both prospective judgments (feelings of
knowing and judgments of learning) and retrospective judgments (post-answer
confidence).
To summarize, there has been a recent surge of interest in the processes by
which metacognition is used to control the processes of learning and
remembering. The studies suggest that people can monitor their mnemonic
processes accurately, and that people can use the output of monitoring to
effectively control the ongoing learning and remembering. The majority of the
work from the metacognition literature is based on studies that assess FOK
judgments at the time of retrieval or JOLs at the time of encoding.
The question to be addressed in chapter 6 is whether the metacognitive model
of monitoring and control can be successfully applied to TOTs. In my work, I
treat TOTs as a metacognitive state, akin to judgments of learning and feelings
of knowing (e.g., Schwartz, 1994, 1999; Schwartz & Smith, 1997). However,
many researchers prefer to think of TOTs simply as deficits in lexical retrieval
(e.g., Vigliocco et al., 1997). Thus, to demonstrate that TOTs do belong under
the rubric of metacognitive theory, monitoring and control in particular, it is
important to show that TOTs serve both monitoring and control functions for the
cognitive processes of a human rememberer.
SUMMARY
109
110 CHAPTER 6
inferential clues mislead the participants into experiencing TOTs when they do
not know the target answer. I describe these studies next.
From the perspective taken here, a TOT means the feeling that a particular
word will be recalled. The word feeling refers to a subjective state of the
individual and not “cold” judgments that they make based on logic and
conscious inference. The feeling is produced by an unconscious inference based
on clues of memorability because, in general, the inference is accurate and the
TOT feeling serves as a good predictor of future recall. If we manipulate clues
to cause the person to make a TOT inference when they do not have an item in
memory store, illusory TOTs result. Therefore, illusory TOTs are
phenomenological experiences in which the person feels that the target is
recallable, but the target memory is either unavailable, forgotten, or indeed, was
never acquired (Schwartz, 1998; Smith, 1994).
Illusory TOTs expand on the earlier concept of subjective TOTs (Burke et al.,
1991; Jones & Langford, 1987; Perfect & Hanley, 1992). Subjective TOTs refer
to TOTs reported by the participants, which are not followed by successful
memory performance. The target is neither resolved nor recognized, nor is any
partial information retrieved. In contrast, objective TOTs are those in which
there is evidence of target knowledge. The distinction between subjective and
illusory TOTs is a theoretical one. With a subjective TOT it is possible that,
under the right retrieval conditions, rememberers may be able to evidence
knowledge of the target. In an illusory TOT, no knowledge of the target is
possible. Therefore, some subjective TOTs, but not all, may be illusory, but all
illusory TOTs are subjective.
To illustrate this point, consider the following question: “What is the name of
the capital city of Canada?” A person may experience a TOT for this name. An
objective TOT is one in which the person retrieves the partial information that
the first letter is “O” and that the city is located along the Ontario-Quebec
border, or a TOT in which the person later recognizes “Ottawa” from a list of
alternatives. A subjective TOT is one in which the person experiences the TOT
but actually has no supporting objective knowledge and may later misidentify
the capital, perhaps because the person thought the capital was Vancouver. In
contrast, if the question is “What is the capital city of Bormea?” any TOT is
necessarily subjective and necessarily illusory because no such country exists. In
the lab, we can test illusory TOTs with fictional questions such as these.
I conducted two experiments designed to document the existence of illusory
TOTs (Schwartz, 1998). In the first study, I used TOTimals as stimuli (see
Smith, 1994; Smith et al., 1994; Smith, Brown, & Balfour, 1991). Rememberers
were presented with a series of six pictures of fictional animals (drawn to
resemble real animals), combined with some biographical infor-mation about the
animals, such as diet, size, and habitat (see Fig. 3.3). Four animal pictures were
paired with two-syllable fictional names (e.g., Yelkey, Rittle, etc.). Two of the
animals were not paired with names. Later, the rememberers were presented
FUNCTIONAL ASPECTS OF THE TOTs 113
with only the pictures and were asked to recall the animals’ names. If they could
not recall the names, they were asked whether they were experiencing TOTs.
The critical question follows: Would rememberers experience TOTs for the
names of the unnamed animals? The answer was yes—illusory TOTs occurred
and at a rate almost comparable to TOTs for targets with names. Rememberers
experienced more TOTs of the unrecalled, but named, animals (23%) than for
the unnamed animals (16%). Thus, 16% of items for which no name had been
provided induced a TOT for the name of the TOTimal.
In the second study, I used general-information questions to elicit illusory
TOTs. Remembers gave answers to 100 general-information questions
(Schwartz, 1998). Of these questions, 80 were adapted from Nelson and Narens’
(1980) norms and had one-word answers. I added 20 questions to which there
was no correct answer (unanswerable questions). Thus, rememberers could not
possibly retrieve targets because there were no targets to retrieve. For example,
one question was, “What is the name of the Canadian writer who wrote the
novel The Last Bucket? The only correct answers are “nobody” or “no such
novel.” Participants read the questions, both answerable and unanswerable, and
typed in their response on a computer keyboard. Any guess on the unanswerable
questions were treated as errors of commission. Following errors of commission
or of omission, participants made TOT judgments. For the unanswerable
questions, any TOTs were considered illusory. Once again, findings show more
TOTs for the answerable than for the unanswerable, but there were still a great
many illusory TOTs. Illusory TOTs occurred for 18% of the unanswerable
questions, less than half than after answerable questions, but significantly above
zero.
In a follow-up study, Schwartz et al. (2000) examined illusory TOTs for
unanswerable general-information questions, using a paradigm nearly identical
to the work just described. However, in this study, because we were interested in
the phenomenological characteristics of TOTs, participants were asked to judge
emotionality, strength, and imminence for their TOTs. Thus, the
phenomenology of TOTs using both answerable and unanswerable general-
information questions were compared on each of these phenomenological
characteristics. Replicating Schwartz (1998), illusory TOTs were less common
than TOTs following answerable questions. However, illusory TOTs did occur,
and again at comparable rates.
There were differences in the experienced phenomenology of the TOTs of
answerable and unanswerable items. Illusory TOTs were more likely to be
judged less emotional, weaker, and less imminent than were the TOTs for
answerable items. Nonetheless, some illusory TOTs were judged to be strong,
emotional, or imminent. That strong and emotional experiences did occur for
some illusory TOTs suggest that their phenomenology can be quite compelling.
We were curious as to why, however, the illusory TOTs were less subjectively
powerful in general. We suspected that the answerable questions led to more
114 CHAPTER 6
retrieval of partial and related information, thus providing more clues for the
answerable questions than for the unanswerable questions.
It is possible to interpret illusory TOTs within a direct access or lexical
access framework. Perhaps for each illusory TOT, the participant has an
(incorrect) answer in mind that is not accessible. Thus, the participant may be
experiencing a TOT for “loggerhead” for the turtle that lays its eggs underwater.
Although the answer is wrong, the participant has a real target in mind. Thus,
illusory TOTs are not inconsistent with most direct access models. However,
even so, they do support a monitoring perspective on TOTs. Illusory TOTs do
occur, so even if they are based on direct access to an incorrect answer, they do
lead to errors both of prediction and of memory. Therefore, researchers
interested in lexical retrieval must be careful that TOTs are “objective” ones.
Illusory TOTs support a monitoring perspective because they indicate
monitoring failures, regardless of whether one assumes an inferential or direct
access basis for TOT etiology.
Control refers to the decisions that are made based on monitoring cognitive
processes. In the previous chapter, the relation among judgments such as FOKs,
JOLs, and learning behaviors was discussed. Numerous studies have now shown
that metacognitive judgments are closely correlated with various learning and
retrieval strategies (see Son and Metcalfe, 2000, for a recent review). In this
section, I discuss some putative control functions for TOTs and review the
nascent research on TOTs and the control of retrieval.
My assumption is that a major function of TOTs is the control of the retrieval
process. Because TOTs usually indicate successful monitoring recall and
recognition, it is now important to ask the question, How do TOTs control
behavior? With regard to retrieval, the following questions may be addressed:
How do TOTs direct or redirect the retrieval process for it to be successful? Do
TOTs cause us to search our memory for targets longer, or do they drive us to
ask friends or consult dictionaries and encyclopedias to search for the missing
target? Do they cause us to cue ourselves in various ways to aid in effective
retrieval? Most of these question remain unanswered partially because the
monitoring and control approach to TOTs is so new. In this chapter, I propose a
few preliminary answers to some of these questions.
One possible metacognitive function of the TOT is to cause one to search longer
for an item that is currently inaccessible. To be more specific, the TOT tells us
that the unrecalled word is both in memory store and imminent, so we devote
FUNCTIONAL ASPECTS OF THE TOTs 115
more time in our attempt to retrieve the item. Research on feelings of knowing
reviewed in the previous chapter found that feeling-of-knowing judgments are
related to the amount of time people spend attempting retrieval of the target
(Barnes et al., 1999, Costermans et al., 1992; Nelson et al., 1984). Items that are
high in feeling of knowing will elicit longer retrieval attempts because
rememberers think they can retrieve the item if they keep trying. Similar logic
can also be applied to TOTs and retrieval time. TOTs might effect retrieval by
causing people to direct more search time to the TOT targets. Presumably, if
TOTs indicate that the item is stored in memory and could be located, a
plausible strategy would be to spend more time attempting retrieval of those
items for which one is in a TOT than for an item for which one is not in a TOT.
In a recent set of studies (Schwartz, 2001), I examined the relation of TOTs
to the time it took people to retrieve answers or decided that they could not
retrieve the answers. Thus, like the FOK studies, I explored the relation between
a metacognitive judgment (the TOT) and retrieval latency. These latencies were
measured by computer, and they meant something slightly different depending
on condition. For correct recall, retrieval time refers to the amount of time
needed to initiate the response on the computer keyboard. For incorrect recall or
errors of commission, retrieval time was defined in the same way: time to
initiate memory response. However, for errors of omission (“don’t know”
responses), retrieval time was the amount of time necessary to initiate the “don’t
know” response (a question mark in the experiments) on the computer keyboard.
Once again, I used the standard Brown-McNeill (1966) paradigm, in which
participants were asked general-information questions. Following omission and
commission errors, the computer asked the participants whether they were
experiencing a TOT. In the series of four experiments, two called for
emotionality judgments along with the TOTs, whereas the other studies asked
for dichotomous (“yes” or “no”) TOT judgments. Finally, in all experiments, an
eight-alternative forced-choice recognition test was presented. The critical
variable was the amount of time participants spent attempting retrieval during
TOTs and during n-TOTs.
All four studies showed a clear association between retrieval time and the
likelihood of a TOT (see Table 6.1; Schwartz, 2001). TOTs were associated
with longer retrieval times than were n-TOTs. Although this effect appeared to
be moderated by other factors, which I discuss later, there was a positive
correlation between the presence of a TOT and the amount of time spent
attempting retrieval of unrecalled targets, particularly those that would become
omission errors. There was a general trend in this direction for TOTs after
commission errors as well. Although the data are only correlational, they clearly
support the notion that one control function of TOTs is to direct the amount of
time participants spend in retrieval search.
116 CHAPTER 6
Experiment l
Correct Emotional Nonemotional N-TOT
omission 13.3 (1.07) 8.1 (.56) 7.3 (1.00)
commission 12.3 (1.17) 8.2 (.56) 8.8 (.33)
correct 7.7 (.28)
Experiment 2
Correct TOT N-TOT
omission 13.8 (1.13) 7.6 (.42)
commission 12.3 (.89) 11.0 (.72)
correct 8.9 (.48)
Experiment 3
Correct Emotional Nonemotional N-TOT
omission 25.2 (2.60) 16.1 (1.77) 12.7 (1.11)
commission 16.9 (1.26) 15.0 (1.70) 17.3 (1.53)
correct 12.6 (.87)
Experiment 4
Correct TOT N-TOT
omission 14.9 (.92) 7.7 (.34)
commission 16.3 (1.03) 11.3 (.53)
retrospective 16.0 (.84) 10.0 (.54)
correct 11.1 (.46)
Note. Retrieval times in seconds (standard errors in parentheses) as a function of
correct retrieval, TOT substate, and the type of error. From Schwartz (2001). TOT=Tip-
of-the-tongue phenomenon.
n-TOTs (7.3 sec) were the shortest. This basic pattern was found to be
statistically significant for both TOTs induced by general-information questions
and those induced by word definitions.
When I submitted this article to Memory & Cognition for publication, I
received an interesting set of reviews. The reviewers were generally positive
about the study but had many questions about the methodology, in particular,
that I collected TOTs following commission errors, a practice generally not used
in TOT experiments. It led one reviewer (Alan Brown) to reflect on the common
experience of brief TOTs that occur before we retrieve a word. Brown claimed
that many TOTs go unmeasured because resolution occurs so quickly.
Moreover, if my view of metacognitive control was correct, TOTs followed by
rapid resolution should also show longer retrieval times than those successfully
recalled without a TOT experience.
Thus, at the suggestion of Brown, I conducted Experiment 4 (Schwartz,
2001), in which participants were asked to make retrospective TOT judgments
after successful retrieval had occurred. In this study, if people correctly recalled
the target, they were asked if they had temporarily experienced a TOT before
they retrieved it. Thus, retrospective TOTs are those made following correct
retrievals, in which people indicate that they experienced a quick TOT in
advance of correct retrieval. Many people report that TOTs can be quite short-
lived and quickly followed by resolution (e.g., Burke et al., 1991), but no lab
study had examined retrospective TOTs. Experiment 4 showed that the presence
of a TOT can be correlated with retrieval time of correct answers as well as with
the unrecalled items.
As far as I know, this represented a unique contribution to the TOT literature.
Although some questionnaire studies have been done retrospectively (Burke et
al., 1991; Read & Bruce, 1982), this represented the first lab study to rely on
participants’ reports as to whether they had experienced a momentary TOT
before they retrieved the correct answer. The results suggested that the
methodology was tapping into something not all that different than the more
typical TOTs after retrieval failure.
In the study, TOTs were experienced following 37% of failures to recall the
answer, averaging across both omission and commission errors. However, for
the retrospective TOTs, their rates were much lower. Participants rated 22% of
their correct answers as having been preceded by momentary TOTs. Although
significantly lower than the TOT rates for omissions and commissions, this
represented a large enough TOT rate to compare retrieval latencies.
Retrieval latencies, measured in seconds, were longer for TOTs than n-TOTs
for both omission and commission errors. Furthermore, retrospective TOTs also
showed longer retrieval times for correct answers than did retrospective n-TOTs,
which were also for correct answers (see Table 6.1). Thus, the retrospective
TOTs indicated a relation between TOTs and retrieval latencies in a way that
was similar to the more typical TOTs after retrieval failure. These data also
118 CHAPTER 6
support the theory that TOTs are involved with metacognitive control.
Presumably, the temporary experience of the TOT directed the rememberer to
continue searching for the missing target, rather than make a fast “don’t know”
response. The extra time led to a successful retrieval of the target. This is the
essence of metacognitive control.
An important caveat to these findings must be acknowledged, and that is the
experiments from my study (Schwartz, 2001) only demonstrate a correlation
between retrieval times and the likelihood of TOTs. Causality cannot be inferred
from the study. It is likely that being in a TOT causes the person to spend more
time attempting retrieval, but other explanations are possible. For example,
because TOT judgments were made after the retrieval attempt, it is conceivable
that participants may have given TOT judgments to those items that took them a
long time to decide that they did not know the answer. However, this reasoning
begs the question; that is, what processes lead them to take longer to retrieve? It
is, of course, also possible that a third unknown factor, such as familiarity,
drives both the long retrieval latencies and the likelihood of a TOT (Metcalfe et
al., 1993; Nelson et al., 1984). Nonetheless, I prefer the notion that the TOT
drives the long retrieval latency, as this is most consistent with the monitoring
and control framework.
To summarize, at this juncture we have some tentative evidence that TOTs
direct retrieval. My data (Schwartz, 2001) showed a strong association between
TOTs and retrieval times. Although other explanations are possible, it is most
reasonable to interpret this within the metacognitive framework. However, it is
likely that TOTs also serve other control functions. I discuss two lines of
evidence that suggest this. One line comes from studies that were interested in
other aspects of the TOT but can be reinterpreted in light of monitoring and
control theory. The other line includes some specific experiments currently
ongoing (at the time of writing), which directly test the idea that TOTs inform or
control retrieval. I start first with the earlier studies.
A possible function of TOTs is that they may inform the rememberer that a
particular item is prone to being forgotten. Thus, if it is inaccessible at one point
in time, it may be inaccessible at another point in time. To counter the perceived
tendency of a word to be forgotten, a TOT may be produced to make the word
more distinctive. Once the item is made more distinctive, it may become more
memorable and so more easily retrieved in the future, consistent with a great
deal of research on memory and distinctiveness. In essence, the TOT is a self-
FUNCTIONAL ASPECTS OF THE TOTs 119
Ryan et al. (1982) conducted an interesting study of TOTs that is often ignored
by other researchers in the area, but is very relevant from the perspective of
metacognition. Ryan et al. were interested in whether the TOT served a
“motivational” function. What they meant by motivation is that the TOT alerts a
person that an item is potentially accessible, and that the TOT drives the person
to want to remember that item. This is certainly intuitive—resolving a TOT
seems much more satisfactory when one does it by oneself than when someone
else provides the target word. One nonobvious implication, however, is that
TOTs may be distracting when one needs to concentrate on another task. If
attention is diverted or motivation is elsewhere, the TOT may interfere with
performance on nonretrieval tasks. Thus, they predicted that rememberers in
TOTs will be distracted from other tasks because of a conscious or nonconscious
drive to resolve the TOT.
They conducted a clever experiment to test this notion. In the experiment, the
participants studied lists of paired associates, such as “trumpet-green.” In an
initial learning phase, the participants studied 52 word pairs for either 7, 5, or 3
sec each. Not surprisingly, it was difficult to learn the words in the set time, and
overall recall was rather poor.
In the recall phase of the experiment, the participants were given the stimulus
(or cue) word and asked to retrieve the target word. If they could not recall the
target word, they were asked if they were experiencing a TOT. Then, for both
TOTs and n-TOTs, the participants were given a number-probe task. A set of
digits was presented on and then disappeared from a computer screen. The
participant had to decide as quickly as possible whether the set that they had just
seen was a member of six test sets now which had replaced it. They responded
“yes” or “no” as quickly as possible. This test was repeated for 20 sec, and then
the participants were given feedback as to their performance on the number-
probe task. The number-probe task was repeated until each participant made
eight correct decisions during three consecutive 20 sec intervals. When the
participants passed the number-probe task, they were given the next stimulus
term in the paired-associate task.
Now here is the interesting part. Ryan et al. (1982) hypothesized that if the
TOT was motivating participants to continue searching, either consciously or
nonconsciously for the unretrieved target, then more cognitive resources would
be drawn away from the probe task following TOTs than following n-TOTs. If a
participant experienced a TOT, then it would be more difficult to concentrate on
the number-probe task than if the participant was not experiencing a TOT. Thus,
Ryan et al. predicted that participants experiencing a TOT for the previous word
would make more errors on the number-probe task during TOT trials than
during n-TOT trials.
FUNCTIONAL ASPECTS OF THE TOTs 121
This is exactly what Ryan et al. found. The mean number correct following n-
TOTs was 7.23 number-probes, significantly higher than the mean number
correct following TOTs, which was 7.00 number-probes. Ryan et al. concluded
that TOTs are “associated with changes in capacity-allocation” (p. 144).
Because some cognitive capacity is focusing on the resolution of the TOT there
is less than full capacity left for the number-probe task. Thus, Ryan et al.
concluded that TOTs “control memory processing” (p. 144).
I think Ryan et al. (1982) were ahead of their time. I suspect their research
was neglected for several reasons. First, most studies use real-world materials
like word definitions and general-information questions. They used word
associates, and therefore, their data lack some of the ecological validity that
other studies have. Second, at the time, most of the TOT research was directed
at the retrieval process per se, or the kinds of information accessible during a
TOT. Ryan et al. were already thinking about control processes. Ryan et al.’s
work is an important and convincing study demonstrating one function of the
TOT, providing the motivation to continue searching memory store.
Strategy Selection. The findings of Schwartz (2001), Gardiner et al. (1973),
and Ryan et al. (1982) support the notion that TOTs act to control the retrieval
process. TOTs cause us to search longer for an elusive target, they alert us to
those items that we know but cannot recall, and they distract us from other tasks
as we continue to ponder our fragmented memories. In the case described by
Ryan et al., TOTs actually hinder us, by drawing attervtion away from the tasks
at hand. Each of these studies hinted at a hypothesized control function of TOTs,
not yet explored: the control of retrieval strategies. Brown (1991) also
speculated that TOTs serve the function of directing our retrieval strategies,
thereby helping us remember the inaccessible target.
For example, a person may just have experienced a TOT for a piece of
information, say a trivia question asked during a time-out during sports coverage
on TV (e.g., “Who was the only basketball player to average 50 points a game
for a single season?”). The person is instantly seized with a TOT. The TOT
alerts the person that this information is available, and starts that person thinking
about how to render it accessible. He or she might then assess what kinds of
partial information is accessible. The partial information may be visual,
phonological, semantic, or some combination of these. For example, the person
might wonder if he or she has a visual image of the person. Or perhaps she may
be able to retrieve partial phonological information. If semantic blockers are
occurring (e.g., Michael Jordan, Kareem Adul- Jabbar), he or she might wonder
whether it is better to continue searching now, or return to it later, after thinking
about something else. The occurrence and nature of the TOT would therefore
drive the nature of the retrieval process. Hopefully, these retrieval processes will
act fast enough that the person will remember “the late great Wilt Chamberlain”
before the answer is given by the TV announcers. Thus, the question that can be
asked is whether TOT experiences cause us to search our memories in different
122 CHAPTER 6
The results of the first experiment were intriguing (Schwartz, 2000). The
TOT rate was 27% of the initially unrecalled items, consistent with earlier data
from my lab. When participants had control over their strategy choice, the most
chosen strategy was a return strategy. Fifty-seven of the TOTs were followed by
the return strategy, whereas only 21% of the TOTs were followed by the search
strategy, leaving 22% in the see strategy category. In contrast, for n-TOTs, 57%
of the time the participants chose the see strategy, 32% chose the return strategy,
and 11% chose the search strategy. This supported the idea that TOTs serve as
motivational states. The participants were far more likely to attempt to self-
retrieve the answer than opt to have it provided for them when they were
experiencing a TOT.
The most startling finding came when the comparison was made between
resolution of TOTs for which the participants had control over strategy and
those for which they did not. When the participants had control over their
retrieval decisions they resolved 60% of the items for which they chose the
return strategy, and they resolved 67% of the items for which they choose the
search strategy. For the group for which strategies were randomly assigned,
participants resolved only 42% of the items in the return strategy and only 44%
of the items in the search strategy. The difference between the participant
control and randomly assigned conditions was statistically significant. Thus,
giving people control over the strategies that they use to resolve TOTs allows
them to resolve those TOTs much better, in this study more than 20% better.
However, for n-TOTs, exercising control over retrieval strategies did not help
resolve more TOTs. In the participant-controlled condition, 20% of n-TOTs
were resolved, whereas 16% of n-TOTs were resolved in the randomly assigned
condition, a nonsignificant difference. Therefore, it is the TOT experience itself
that appears to be directing rememberers toward the correct retrieval strategy.
An interesting aspect of the strategy selection data is that neither pop-up or
search were better at driving resolution; they both did equally well. Nonetheless,
exercising control over the strategy improved resolution. Thus, the advantage for
the control condition must reflect that TOTs direct us toward different retrieval
strategies for different items. Therefore, it is likely that differences in TOT
phenomenology may cause us to choose different retrieval strategies. Currently,
we are investigating what variables (discussed in chap. 2) are correlated with the
decision to choose different retrieval strategies.
Another hypothesis concerning the control functions of TOTs is that TOTs may
serve to facilitate certain social interactions. Indeed, TOTs may be the grease
that smoothes the way for us to be friendly with people that we barely know. For
example, a person whom you have met cannot remember your name. You
remember his, but he cannot remember yours; not a good start to a positive
124 CHAPTER 6
state of research, how one would distinguish between metacognitive control that
was a function of design features or a consequence of other adaptive systems,
particularly given that most of the research is correlational. Nonetheless, the
thought is provocative and worth addressing.
In this chapter, I discussed the idea that TOTs are not just monitors of our
memory failures. Rather, TOTs actively direct our retrieval processes in such a
way as to resolve TOTs, maximize our recall performance, and perhaps even
ease awkward social situations. The first function of TOTs that was discussed
was that TOTs may cause us to spend more time trying to retrieve an unrecalled
target (Schwartz, 2001). This was supported by longer retrieval attempts during
TOTs than n-TOTs. The second function discussed was that TOTs may ensure
greater processing of TOT targets, promoting better retrievability of those
targets in the future (Gardiner et al., 1973). A third function is that TOTs may
motivate us to work harder to retrieve those items (Ryan et al., 1982), and a
fourth function may be that they help us direct retrieval strategies (Schwartz,
2000). Finally, they may also serve as social motivators.
Now, it may be possible to reinterpret each of the individual findings
reviewed in this chapter. However, I think a convincing case has been made that
TOTs do serve a control function for the rememberer. If this is so, it strongly
supports the Nelson-Narens (1990) model, which claims that metacognition
evolved so that effective monitoring could inform an effective control system to
maximize performance of the rememberer.
SUMMARY
Throughout the first six chapters of this book, the discussion has mainly focused
on TOTs in normal adults. Most of the data discussed so far has been drawn
from younger adults, mostly students in introductory psychology courses. It is
now time to consider developmental and neuropsychological perspectives on
TOTs, concentrating on TOTs in “special” populations. TOTs among older
adults, children, and neuropsychological patients are instructive for several
reasons. Naturally, it is important to understand cognitive processes in each of
these populations. It is also important to understand the development of TOT
etiology and phenomenology. Beyond that, each group may offer insights into
both the nature of lexical theory and the nature of monitoring and control
processes. In this chapter, I first consider the literature on TOTs and aging,
primarily because it is the most extensive and well-articulated. Following
discussion of TOTs and aging, I discuss the limited research on TOTs in
children and the limited literature on TOTs and neuropsychological patients.
We have seen throughout this volume the role that TOTs play in two bodies
of thinking: theories of lexical retrieval and theories of metacognition. In
theories of lexical retrieval, TOTs are seen as a “window” into the process of
retrieval, an opportunity to study retrieval when the process is slowed or halted
(e.g., A.S.Brown, 1991; Kohn et al., 1987). In contrast, in metacognitive theory,
the TOT is an experiential feeling that allows rememberers to monitor and
control their retrieval processes. Each of these perspectives has led to great
many interesting facts, theories, and unexplored questions. Research on older
adults, children, and neuropsychological patients may inform and refine each of
these theoretical perspectives.
Unfortunately, the TOT has not been much of a focus in neuropsychological
circles. There a few scattered articles looking at TOTs in clinical populations,
and no published neuroimaging studies. Thus, given the paucity of studies and
the paucity of replications, it is hard to conclude anything from the
neuropsychological literature. However, TOTs and normal aging have been well
studied with some oft-replicated findings. Thus, some firm conclusions can be
drawn based on the aging data. Nonetheless, none of the existing studies on
TOTs and aging have been conducted with metacognitive theory in mind. Thus,
one of my goals in this chapter is to add a metacognitive spin to some of the
research on TOTs and aging, suggesting how such research could be accounted
for by a monitoring and control perspective. However, it should be kept in mind
that all of the studies on TOTs and aging were originally designed to address
issues of word memory or lexi- cal retrieval theory.
126
DEVELOPMENT ISSUES 127
TOT Frequency
Do people experience more TOTs as they age? Most research suggests that
TOTs increase with age from young adulthood to the oldest adults (see Table
7.1). Although there has not been a longitudinal study, of 11 cross-sectional
studies that have been conducted, 10 show clear evidence of age-related
increases in TOT frequency (see Brown, 2000). The exception (Maylor, 1990a)
128 CHAPTER 7
compared people in their 50s, 60s, and 70s. It is conceivable that the Maylor
(1990a) study did not have sufficient age range to document age-related
increases in TOTs. The bottom line, however, is that across all of these studies,
it is apparent that TOT frequency increases as people age from young adulthood
through middle age to older adults to oldest adults.
Despite the null results from Maylor (1990a), the trend appears to be a steady
increase in TOTs over time. For example, Burke et al. (1991) found more TOTs
in their middle-aged group (mean age=38.7) than for the young adult group
(mean age=19.4). Heine et al. (1999) found more TOTs in their oldest adult
group (range 80 to 92) than in the young-old group (60–74). Thus, increases in
TOT rates occur with advancing age and well into late life.
TOT Resolution
Do older and younger adults differ in the rate at which they resolve TOTs? It is
known that TOT rates increase with aging. However, the issue of whether TOT
resolution differs with age does not naturally follow. From the decrement view,
it might be expected that, given the general breakdown in associative networks,
older adults would show a deficit in resolution. For example, if older parti-
cipants are experiencing more TOTs because of breakdowns in phonological
activation, as suggested by Burke et al. (1991), one might hypothesize that fail-
ure to activate sound representations would persist and TOT resolution would be
low. Alternatively, from the incremental knowledge view, older adults may have
more and better cueing strategies with which to resolve TOTs. Unfortunately,
the data are ambiguous with respect to differences in TOT resolution and aging.
Three diary studies (Burke et al., 1991; Cohen & Faulkner, 1986, Heine et al.,
1999) found increases in TOT resolution with aging, although in none of the
studies was the difference found to be statistically significant. Cohen and
Faulkner (1986) found an 81 % resolution rate for young participants, a 66%
resolution rate for middle-aged participants, and an 87% rate for older adults.
Burke et al.’s (1991) rates for young, middle-aged and older participants were
92%, 95%, and 97%, respectively. Heine et al. (1999) found a 91% rate for
younger adults, a 95% for young-old participants, and a 98% rate for the oldest
adults. Thus, the diary findings show trends that older adults are actually better
than younger adults at resolving TOTs. This is consistent with incremental
knowledge views of TOTs and aging, which suggest that there are aspects of
TOT performance in which older adults will do better than younger adults.
129
130 CHAPTER 7
need to be directed at the processes by which older and younger adults retrieve
TOT words, as well as whether there are age-related differences in the control
processes used to resolve TOTs.
Partial Information Retrieval. The retrieval of partial information has played
an important role in the development of theories concerning the etiology of
TOTs. Lexical retrieval theory postulates that partial information represents the
access to the lemma with only partial access or no access to the lexeme.
Inferential theories postulate that TOTs are based on the amount of information
that is retrieved. Similarly, theories of age differences in TOTs also hinge on
access to partial information. According to the Burke et al. (1991) model, based
on decrement theory, aging weakens connections between the semantic level
and the phonological level. TOTs arise when semantic activation is not
accompanied by phonological activation. Thus, according to the model, older
participants should access less phonological information during a TOT,
including fewer retrieved blockers that are phonologically related to the TOT.
Incremental knowledge theory does not make predictions that are substantially
different from the decrement theory here.
The data here strongly support the view that during a TOT less phonological
information is accessible for older participants than younger participants (see
Table 7.2a). Also, more persistent alternates or blockers are accessible for
younger than for older participants. The one exception is the study by Brown
and Nix (1996), who found the opposite pattern for alternates, with the older
participants experiencing more blockers. They also found that younger
participants reported more first-letter information, but this information was not
more accurate than that of the older participants.
Why does the Brown and Nix (1996) study find a different pattern than the
other studies? There is no immediately obvious answer. However, there are
several procedural differences that may point to why this pattern occurred. In
particular, all of the older participants in the Brown and Nix study were alumni
of Southern Methodist University (SMU), and therefore, they all had more years
of education than the SMU students, whereas in the other studies addressing this
question, the older participants were more widely selected. It is unclear exactly
how this would change the rate of persistent alternates, but it seems to have done
so. Second, the Brown and Nix procedure was, by and large, simpler for the
participant. It is possible that the older participants could therefore devote more
attention to the retrieval of each item. Finally, Cohen and Faulkner (1986),
Burke et al. (1991), and Heine et al. (1999) had specific hypotheses that there
would be fewer persistent alternates for the older participants. Brown and Nix’s
instructions were decidedly more neutral. Therefore, it is possible that
participants may have been picking up on the experimenter’s biases in the
former studies.
132
DEVELOPMENT ISSUES 133
To review the theories from the beginning of the chapter, decrement theory
postulates that age-related increases in TOTs are a function of normal biological
decline that takes place in aging. For example, Burke et al.’s (1991) theory
suggests that with aging, breakdowns occur in the associations between
semantic representations and phonological representations, leading to increased
TOTs. Incremental knowledge theory postulates that increases in TOTs with
advancing age occur because older adults have more knowledge of words, and
thus a greater set of materials for which to experience TOTs. Indeed, findings
show that older adults have greater vocabulary and greater general world
knowledge than do younger adults (Dahlgren, 1998; Maylor, 1994, Salthouse,
1991).
I consider the evidence for the incremental knowledge theory first. The theory is
newer, and there are less data that support it. The data we do have are
convincing. However, the main empirical evidence comes from a study by
Dahlgren (1998), which I review later. First, I share my experience in trying to
induce TOTs in older adults, during an unsuccessful experiment, which now
leads me to favor the incremental knowledge view.
The difference in knowledge base between older and younger adults was
recently vividly demonstrated to me. In my work on illusory TOTs (see
Schwartz, 1998; Schwartz et al., 2000) I developed a set of trick questions, for
which there was no correct answer (e.g., “What is the name of Mercury’s
moon?”). These questions were carefully designed so that college students
would not suspect the questions to be unanswerable, and they might be
susceptible to illusory TOTs. In the four experiments reported in Schwartz
(1998) and Schwartz et al. (2000), only one college student out of nearly 200
detected a discrepancy (the participant knew that Mercury had no moons).
However, when I tried to conduct the study at a local senior citizen center, I was
overwhelmed with the participants approaching me to tell me I had made a
mistake on the question sheet. Others asked if they could answer “no possible
answer.” The study was a washout because virtually every older adult detected
the falsity of the questions. Daunted by the task of having to reconstruct my
illusory questions, I abandoned the project. The relevance here is that the older
adults showed both a better vocabulary and greater knowledge, albeit of trivia,
than did the younger college-age adults. Although the data from this experiment
were long ago discarded, I suspect that both the correct recall rates and the TOT
rates were higher for these older adults, probably in part because of their greater
knowledge of the world. I did not analyze normative data on these older adults,
134 CHAPTER 7
Decrement Theory
Decrement theory states that TOTs increase with advancing age because of
failures and breakdowns in the retrieval systems of older adults. Most of the data
supporting this view come from studies that start from a lexical theory per-
spective, based on direct access models. In this section, I evaluate direct-access
models in addition to decrement theory. Recall that direct-access theory
postulates that the TOT is based on access to the unretrieved but present target
word. Thus, TOTs are influenced by variables affecting retrieval. Direct-access
views are divided into blocking, one-stage, and two-stage views. The one-stage
models mirror the activation models and the two-stage models refer to variants
of the transmission deficit theory (e.g., chap. 3).
With respect to aging, the dominant theory has been transmission deficit
theory (Burke et al., 1991; Rastle & Burke, 1996), and all of the studies on
TOTs and aging since 1991, except Dahlgren’s (1998), have been linked to it.
Therefore, the folowing section will be an evaluation of the transmission deficit
model as the prototypical direct-access model.
Aging in the Transmission Deficit Model. Recall that the transmission deficit
model is considered a two-stage model because it postulates two distinct levels
of representation, a semantic level and a phonological level (Harley & Bown,
1998). TOTs occur when semantic activation is insufficient to activate a
phonological representation (Burke et al., 1991). MacKay and Burke (1990)
argued that aging weakens the connections between nodes in the model,
reducing the amount of priming that can spread from semantic to phonological
levels. Because of this, older adults with weaker connections will experience
more TOTs, consistent with the decrement model. Another prediction of the
transmission deficit model, according to Burke et al. (1991), is that older adults
will experience fewer persistent alternates (or blockers) during a TOT than will
younger adults. This hypothesis is made because the deficit that causes TOTs
136 CHAPTER 7
may also “reduce bottom-up priming to the lexical nodes for alternates, making
activation of an alternate less possible” (Burke et al., 1991, p. 549). As we have
seen, this too, is supported by the empirical evidence. Four of five studies
demonstrated significantly more persistent alternates for younger adults than for
older adults. Similarly, older adults should retrieve less partial information than
younger adults, according to the model. This is also supported by the data (see
Table 7.2b). Burke et al.’s data are among the strongest that support the
decrement model of TOTs in aging.
The strong association between age and decreased retrieval of partial
information supports the transmission deficit models, and it also argues against
the incremental knowledge view. If older adults experience more TOTs because
they have a greater vocabulary, as incremental knowledge theory suggests, then
it is likely that they will also experience more semantic blockers or partial
semantic retrieval. Thus, it is likely that at least some of the increase in TOTs
with age is a function of decrements in associative networks, as suggested by
Burke et al. (1991).
With respect to TOT resolution, Burke et al. (1991) claimed that the
transmission deficit model did not make relevant predictions. However, if aging
causes weakened connections between semantic and phonological
representations, it is likely that older adults should resolve fewer TOTs than do
younger adults. The data, however ambiguous, actually suggest the opposite—
that older adults are better at resolving TOTs, or at the very least, TOT
resolution is about the same in younger adults than it is in older adults. None of
the studies found significant age advantages for TOT resolution, so this issue
also requires future study. If these nonsignificant trends hold up under future
testing, the better resolution of older adults would be damaging to decrement
theory and would support incremental knowledge theory.
Inferential Models. Inferential models claim that TOTs are the result of
unconscious inferences that a target is known, based on the recognition of cues
or the retrieval of clues (e.g., Koriat, 1993; MacLin, 2000; Schwartz & Smith,
1997). Inferential theory is compatible with either decrement theory or
incremental knowledge theory. If decrement theory is correct, older adults
experience more TOTs because, despite the breakdown of connections, the clues
that an item is in memory are still there. In inferential models, it is the clues, not
the breakdown, that induces the TOTs. If the incremental knowledge theory is
correct, the increase in knowledge that occurs with aging will be accompanied
by an increase in the number of clues that people can use to infer TOTs. Or,
simply put, more knowledge means more clues and more cue familiarity, which
in turn means more TOTs.
To date, however, inferential theory has not been specifically tested in studies
looking at age effects on TOTs. How well can inferential models explain the
137
138 CHAPTER 7
Because the work on TOTs among aging adults has centered on the lexical
retrieval issues, the role of monitoring and control functions of TOTs among
older adults has not received any attention. It is possible to speculate how some
of the literature can be parsimoniously explained by metacognitive models of
monitoring and control.
Research in other metacognitive domains has shown that older adults monitor
their memory at least as well as younger adults, particularly for JOLs (e.g.,
Connor, Dunlosky, & Hertzog, 1997; Dunlosky & Connor, 1997; Dunlosky &
DEVELOPMENT ISSUES 139
TOTs IN CHILDREN
When do children experience their first TOT? At what age, and why? Do
children experience more TOTs because they have not strengthened the
associations between semantic and phonological aspects of a word? Or do they
experience fewer TOTs because their vocabularies are smaller? Are accuracy
rates the same because TOTs are a function of direct-access to an unretrieved
target, or are accuracy rates lower because children have not learned which clues
to pay attention to? Again, when we talk about TOTs in children, we encounter
yet another unexplored area in the TOT research. Although theoretically, the
issue of TOTs in children is important for both metacognitive and
140 CHAPTER 7
psycholinguistic theory, there has been little attempt to address these issues in
developmental studies. The issues of partial knowledge, access to knowledge,
and awareness of that knowledge clearly have important educational
implications. However, to my knowledge, only three studies of limited scope
have asked if and how children experience TOTs (Elbers, 1985; Faust et al.,
1997; Wellman, 1977). Two of those studies (Faust et al., 1997; Wellman, 1977)
were conducted in an experimental fashion, but neither study specifically asked
children to report on TOT states. I also discuss a fourth study, by Butterfield,
Nelson, and Peck (1988), designed as a follow-up to the Wellman (1977) study,
but which used FOKs.
Elbers (1985) recorded several conversations with her 2½-year-old son
concerning a single difficult retrieval in their native language, Dutch. On the
previous day, the child and his mother had visited the aquarium, where the child
saw dolphins for the first time. The next day, when seemingly trying to retrieve
the word for dolphins, the child could not. On several occasions, blocking words
intruded (“soldier,” which sounds similar to the word for “dolphin” in Dutch),
but the child could not retrieve the word “dolphin” even though it was clear that
he was talking about them. On the basis of this anecdote, Elbers concluded that
TOTs occur in children younger than 3 years old.
In an experimental study, Wellman (1977) looked at the accuracy of FOKs in
young children. In addition to collecting FOKs, Wellman also looked for
evidence of actual TOT states by observing when the children experienced
frustration because they could not recall the target word for which they were
searching. Although explicit TOT judgments by the children were not included,
Wellman clearly wished to address the issue of TOT phenomenology.
Wellman (1977) was one of the earliest researchers interested in
metacognition, and his 1977 article was specifically designed to look at
metacognition in young children. One of his concerns was the accuracy of the
childrens’ FOKs, which was to serve as a measure of the children’s monitoring
abilities. Wellman, like later metacognitive researchers, was interested in the
nature of the judgments and how well they predicted performance. In his study,
he asked kindergarten, first-, and third-grade children to name pictures taken
from the Peabody picture vocabulary test. He prescreened the pictures into easy,
medium, and difficult categories to see which would be more likely to elicit
feelings of knowing when the name could not be recalled. If the children could
not name the object, the experimenter asked them for a FOK by asking the
following question: “What if I told you a lot of names, could you pick out the
right name for the picture?” Most of the children understood this question and
did not answer either all “yes” or all “no.” In a final recognition test, the
children were given the names and had to pick which of nine pictures was the
correct one.
The children’s FOKs predicted correct recognition at greater than chance
levels. Furthermore, the third-graders were significantly more accurate than the
DEVELOPMENT ISSUES 141
kindergarten and first-grade children. The younger and older children showed
equivalent recognition and familiarity, suggesting that the older children s
increase in accuracy was due to better metacognitive monitoring. Wellman
(1977) also looked at verbalizations, which included clear statements of
metacognitive intent by the children. For example, some children, particularly
the third-graders, said when indicating a feeling of knowing, “I know I know
that,” or “I know, I just can’t remember.” Wellman also reported that some of
the third-graders “became agitated and frustrated with their inability to recall the
name.” (p. 20). Although, Wellman specifically looked at FOKs rather than
TOTs, the anecdotal reports suggest that some of the children were experiencing
TOTs.
In a study by Butterfield et al. (1988), that only partially replicated
Wellman’s (1977) findings, accuracy differences were not significantly dif-
ferent between older and younger children. The 6-year-olds showed equivalent
accuracy to the 10-year-olds and to the 18-year-olds using gamma correlations.
Butterfield et al., therefore, suggested that metacognitive accuracy was invariant
across these developmental periods. Because of the failure to replicate from one
study to the next, and because neither study really asked children to report what
was on the “tip of the tongue,” the nature of TOTs in tots remains a mystery.
Faust et al. (1997) were also interested in TOTs in young children. They
examined TOT rates and TOT accuracy in normal and language-impaired
children. They suspected that the language-impaired children might show more
TOTs than did the normal children because word retrieval is often a problem for
children with impairment.
Faust et al. (1997) looked at comparable samples of normal and impaired
children who were either in the second or third grade (the children were all
approximately 8 years old). Like Wellman (1977), they showed children pictures
of objects or animals and asked them for the names of the items. If the children
could not recall the name of the item, they were asked whether they knew the
item and might be able to remember it later. If they said they could remember it
later, they were asked to make a FOK (with instructions that were very close to
those normally given for TOTs). The children were subsequently given a
recognition test.
Both language-impaired and normal children quite frequently indicated that
they knew an item that they could not recall right away. Indeed, the language-
impaired children gave the TOT response on 30% of all the items, and the
normal children indicated a TOT response 11% of the time, indicating that by 8
years of age, children are routinely experiencing TOTs.
Faust et al. (1997) examined the accuracy of TOT judgments in both the
normal and the language-impaired children. There were some stunning
differences between the two groups with respect to accuracy. The normal
children resolved 36% of their TOTs, comparable to adults. However, the
language-impaired children resolved only 18%. The recognition results were
142 CHAPTER 7
even more striking. Normal children correctly recognized 45% of their TOTs,
and 0% of their n-TOTs, thus demonstrating good metacognitive accuracy. The
language-impaired children actually recognized more n-TOTs (39%) than they
did TOTs (27%), thereby demonstrating very poor, if any, monitoring capability.
Also, the language-impaired children recalled more false phonological
information than did the normal children.
This study demonstrates that children as young as 8 years old experience
TOTs, and that these TOTs can predict both recall and recognition. The Faust et
al. (1997) study also suggests that there is a relation between TOT accuracy and
language impairment, although the nature of that relation is unclear. One
important implication, discussed by Faust et al., is that clini-cians may be able to
overcome language impairment by teaching good metacognitive skills.
From any perspective, TOTs in children could be seen as an important line of
research. Consider lexical retrieval theory. Research in children could
demonstrate how connections between semantic nodes and phonological nodes
form and develop. If TOTs arise from failures between semantic and
phonological nodes, then as yet unformed connections in children may give rise
to similar phenomena as with the very old. The pattern of TOTs in younger
children may mirror that of older adults.
From the point of view of metacognition, TOTs in children are also of
interest. If metacognition is a complex and late-developing skill, there should be
little evidence of TOTs in young children. However, if TOTs are in place at
quite a young age in fully-formed format, it may suggest an early development
of metacognitive skills and perhaps emphasize the role of nonconscious
processes in metacognition (e.g., Reder & Schunn, 1996). Furthermore, it would
be important to determine if TOTs have the same etiology in children and adults
or if childhood TOTs are qualitatively different from those of adults.
Nonetheless, this research remains in the future.
With respect to TOT phenomenology, it is also interesting to know when
children start using the expression “its on the tip-of-my-tongue,” given that this
expression exists in most languages (Schwartz, 1999b, chap. 2). Although it may
have just not occurred to Wellman (1977) to quote any child as saying it, he
does not report any anecdotes in which children used the expression “on the
tongue.” If TOTs really are universal, it would be expected that the phrase
would be used by children. However, once again, I must say “alas” because the
research is not done.
Perhaps the most promising avenue of research on TOTs, the one with the least
amount of research actually accomplished, is in the domain of Neuro-
psychology. Neuropsychology should give us some tremendous insight into the
DEVELOPMENT ISSUES 143
TOT, if we can bring its tools bear. First, patient studies need to address the
nature of the TOT in both aphasic and amnesic patients. Second, neuroimaging
studies would also enhance our understanding of TOTs considerably. Indeed, the
prominent neuropsychologist, Schacter wrote that, “Although no neuroimaging
studies of TOT states have yet been reported (and may be difficult to carry out
because of the relative infrequency of the TOT phenomenon), such studies could
provide novel insights into the neural correlates of retrieval blocking.”
(Schacter, 1999, p. 188). I could not agree more and would only add that novel
insights would also occur concerning the nature of TOT etiology and
phenomenology as well. D.L.Schacter assured me that such studies are on the
agenda in his lab, but have not yet been completed at the time of this writing
(personal communication, June 9, 2000).
A few studies have discussed the TOT in neuropsychological patients. Many
of them, however, do not explicitly ask the patients to report if they are
experiencing a TOT. Rather, the TOT is inferred from their behavior, as we saw
with Vigliocco et al.’s (1997) study of syntactic access during Italian TOTs. I
urge caution in interpreting these studies because what may seem like a TOT to
an outside observer may not feel like one for the neuropsychological patient. I
discuss these studies here, but remind the reader that the TOTs are third-person
defined, and not first-person subjective TOTs.
Before reviewing the studies, let me strengthen the argument for their
importance. Patient groups could allow for some careful testing of the different
theories outlined in this volume. For example, consider the comparison between
direct-access theory and inferential theory. Direct-access theory postulates that
TOTs are a function of access to an unretrieved word, but with a failure to
access its phonological component. Direct-access theory might predict that
Broca’s aphasics and anomic aphasics, therefore, should experience TOTs very
frequently, as their chief problem is one of access to phonological form. Because
of their intact semantic knowledge, their TOTs should show high accuracy when
measured by recognition tests. Inferential theory suggests that TOTs are based
on piecing together clues. Inferential theory, therefore, makes a different
prediction, at least for the Broca’s aphasics. Broca’s aphasics do not retrieve
much partial information because of the extreme deficit in their retrieval
systems. According to inferential theory, two of the clues we use to infer a TOT
are the amount and the intensity of retrieved partial and related information.
Because Broca’s aphasics do not retrieve much partial information, they may
experience very few TOTs. In contrast, inferential theory might predict that
patients with prefrontal damage may have a more difficult time with accurate
TOT judgments. Prefrontal damage is often associated with monitoring
difficulties (Schacter, Harbluk, & McLachlan, 1984), so it is likely that TOT
accuracy will decline as well.
144 CHAPTER 7
TOTs in Aphasia
the name of a line drawing of an object, they were either given a phonemic cue
or not given a cue. Bruce and Howard found that the Broca’s aphasics benefitted
from the phonemic cues. Like Goodglass et al. (1976), Bruce and Howard
primarily used the label TOT in terms of the ac-cessibility of partial information
combined with the inability to retrieve the full target word.
The two most illuminating studies concerning the nature of TOTs during
aphasia are recent ones involving single-Subject designs with individual anomic
aphasics. The first, Funnell et al. (1996), examined the nature of FOKs in a
strongly anomic patient. Funnell et al. were interested in linking the
experimental work on FOKs with the word-retrieval difficulties in anomia. The
second study, conducted by Vigliocco et al. (1999), examined the nature of
syntactic information that can be retrieved by an anomic patient during word-
finding difficulties.
Funnell et al. (1996) studied the patient HW, who is a highly intelligent man
but became densely anomic after suffering a left-hemisphere stroke in 1989. A
Harvard MBA graduate, HW’s poststroke WAIS-R IQ was still measured at 121
when it was given to him untimed. However, he has a very limited production
vocabulary, and his speech is characterized by endless word-retrieval failures.
Funnell et al. conducted a FOK experiment, in which HW was asked general-
information questions such as “When something is chronologically out of place,
it is called an _____” (anachronism). HW was almost completely unable to
recall any of the words that fitted the sentence. He usually gave very high
feeling-of-knowing judgments and declared that he knew the word absolutely,
but he just could not say it. Indeed, his recognition performance was extremely
good. Thus, it seemed highly likely that HW did indeed have an image of the
word and perhaps related semantic information, but that his deficit was in
activating the phonological form of the word. Funnell et al. argued that their
data supported Burke et al.’s (1991) position on TOT etiology, that is, that TOTs
occur when there is semantic activation without phonological activation—this
seemed to be the pattern in HW. However, Funnell et al. used FOKs, not TOTs,
to assess metacognitive ability.
In a second single-subject study of TOTs and anomia, Vigliocco et al. (1999)
was interested in whether anomics would access syntactic information for words
that they could not recall. They studied the patient MS, a 31-year-old man
rendered anomic after a bout with herpes encephalitis. MS is of normal
intelligence aside from his verbal deficit. In Vigliocco et al.’s paradigm, MS was
presented with pictures of items representing fruits and vegetables, or buildings
and furniture. Half of the presented items were mass nouns (“asparagus”),
whereas half of the presented items were count nouns (“artichokes”). Mass
nouns refer to items that are not grammatically pluralized when speaking of
many of the items, whereas count nouns are. If MS could not retrieve the name
of the object, he provided whatever phonological information he could, guessed
146 CHAPTER 7
at the number of syllables, and indicated whether the unretrieved word was a
mass or count noun.
Vigliocco et al. (1999) defined a TOT as occurring when the patient could
not retrieve the target name but did retrieve partial information about the target.
MS was not asked whether he thought that he was experiencing a TOT. Positive
TOTs occurred when MS later recognized the correct word, whereas negative
TOTs were those items for which recognition was incorrect. Vigliocco et al.
found that MS was able to correctly identify whether a word was mass or count
even when he could not actually retrieve the word. Because of the limited size of
the item pool, Vigliocco et al. were not able to compare correct response rates
for positive TOTs, negative TOTs, and n-TOTs. Nonetheless, the study does
show that syntactical information is accessible to anomic patients. Given the
general similarity between deficits in word retrieval in anomics and TOTs in
normal people, Vigliocco et al. inferred that this study supports a model that
includes syntactical information as important in the production and resolution of
TOTs.
Both studies on anomic aphasia were directed at testing word-access theories.
In both studies, the predominant problem of the patient was accessing the
phonological form of target words. Both patients retained access to semantics,
and Vigliocco et al. (1999) showed that the anomic had access to syntactical
information as well. Therefore, both studies support the general notion that a
complete lemma is accessed before the phonological representation, or lexeme,
is retrieved. It is not clear what each study tells us about the TOT, as the first
study substituted FOKs, and the second study did not measure subjective TOTs.
I suspect it is highly likely that anomic patients will experience TOTs quite
frequently. This is a straightforward prediction from direct access theory, but
also a prediction of inferential theory. According to inferential theory, TOTs are
based on the number and intensity of clues that are retrieved by the rememberer.
Anomic aphasics have no shortage of retrieved clues, although they are usually
semantic, not phonological in nature. Therefore, both models predict high
frequency of TOTs, and good predictive accuracy when recognition is used as
the criterion measure. On the other hand, it is possible that direct and inferential
theory will make different predictions for other population groups, particularly
different aphasic groups. I find it very unfortunate that no labs are currently
addressing these issues.
We now turn our attention to two studies that have been conducted on
patients with neurological diseases. Parkinson’s is a disease that attacks the
substantia nigra of the basal ganglia, causing problems with the control of
voluntary movement. Because the basal ganglia are connected to the prefrontal
cortex, some Parkinson’s patients also show dementia. Alzheimer’s is a
degenerative and eventually fatal neurological disorder, which first affects
memory, but then degrades all cognitive functions.
DEVELOPMENT ISSUES 147
Matison, Mayeux, Rosen, and Fahn (1982) postulated that TOTs might be more
common in Parkinson’s disease, given that Parkinson’s patients often have
difficulties in articulating words. However, Parkinson’s deficits are generally
considered to be dysarthric deficits, that is, difficulties in motor control, and not
aphasic deficits. Matison et al. tested 22 medically stable Parkinson’s patients
with standard neuropsychological testing, including the Boston naming test.
They found that, like aphasics, the Parkinson’s patients benefitted from
phonological cueing of unrecalled words. Therefore, Matison et al. interpreted
Parkinson’s deficits in articulation as being similar to the problems of speech
production in aphasia and in normal individual with TOT states. Matison et al.
did not explicitly ask their patients to provide subjective TOT judgments.
Because of possible frontal cortex damage in Parkinson’s patients, one might
suspect that the TOTs would show lower correlations to resolution and
recognition, but this aspect was not tested by Matison et al.
Astell and Harley (1996) looked at explicit TOT judgments in patients suffering
from Alzheimer’s disease (AD). Astell and Harley gave simple word definitions
to AD patients and asked them to supply the word. For example, one item used
was “timepiece that is worn on the wrist” (watch). If the patients could not
retrieve the target word, they were asked to provide any semantically or
phonologically similar words. TOT states were only recorded if “they indicated
that they knew the word and were unable to retrieve it, and if they carried out an
active search in an attempt to locate the target word. These were often marked
by phrases such as ‘it’s on my tongue’ and ‘I can’t get my tongue around it.’” (p.
200). The patients were then given the target word and asked if that matched the
word that they had in mind.
Compared to the age-matched controls, the AD patients experienced more
TOTs (both resolved and unresolved). This result is partially because the control
patients retrieved more words correctly, leaving a smaller pool available for
TOT states. The AD patients retrieved few phonological blockers relative to the
controls, although they were roughly equivalent for semantic blockers. Thus,
like the other patient groups, the AD patients seemed to be revealing a
phonological deficit while they were experiencing TOTs. Astell and Harley
(1996) interpreted these results as consistent with a two-stage account of
retrieval with an initial semantic phase followed by a subsequent phonological
phase. Like Burke et al. (1991), Astell and Harley think of TOTs occurring
when that phonological stage is impaired.
148 CHAPTER 7
Unfortunately, Astell and Harley (1996) did not measure the accuracy of the
TOTs. Although the low phonological retrieval in the AD patients supports the
lexical access model, it does not address inferential model. Assuming that AD
patients have frontal damage relatively early in the course of AD, we would then
expect to see accuracy deficits in the TOT judgments, according to inferential
theory. Astell and Harley did not consider this because their goal was to test the
two-stage model of lexical retrieval in a patient group.
information will not differ from TOTs to n-TOTs because the frontal patients
can no longer distinguish between the two states.
Temporal-lobe amnesics, on the other hand, show no deficits in
metacognitive monitoring (Shimamura & Squire, 1986). Therefore, their
production of TOTs should be unimpaired. Given their memory deficits, it is
likely that amnesics will generate more TOTs than comparison groups, and that
the TOTs will predict recognition performance equally well (although given
their general recall deficits, not resolution). Certainly, the amnesic group should
show better TOT accuracy than the frontal-lobe patients. Although amnesics
may also show deficits in the retrieval of related phonological and semantic
information, it should be no worse for TOTs than for n-TOTs.
Relevant to these hypotheses is a recent study by Widner, Otani, and
Falconier (1999). They classified normal participantsts as high frontal or low
frontal as a function of their performance on the Wisconsin Card Sort.
Participants then retrieved word definitions, and gave both FOKs and TOTs.
Participants who scored low on the frontal tasks showed lowered FOK accuracy
relative to those with better accuracy, consistent with the hypotheses of frontal
lobe’s role in metacognition. However, there were no differences in either TOT
rates or TOT accuracy. Thus, their data did not support a role of frontal lobes in
TOT judgments. On the other hand, Widner et al. (1999) tested only unimpaired
people who all scored higher than neuropsychological patients with frontal
damage. They did, however, find that FOKs correlated with frontal functioning,
similar to the findings of Souchay et al. (2000). Thus, it is possible that there
simply was not enough variance in the participants to reveal deficits in TOT
accuracy. This hypothesis needs to be tested with patients whose frontal lobes
are truly damaged, not simply inefficient.
Another manner in which to test these views would be to conduct
neuroimaging studies of normal adults. As the quote by Schacter (1999) at the
start of this section indicated, there have been no published studies looking at
TOT formation with neuroimaging equipment, although such studies would be
terribly interesting. It is likely that neuroimaging data would be extremely
important in elucidating issues in TOT etiology and lexical retrieval. Given that
the left prefrontal cortex appears to be implicated in retrieval from semantic
memory, one hypothesis is that TOTs will selectively activate this area relative
to n-TOTs (see Nyberg, Cabeza, & Tulving, 1996). Such a finding would also
support the role of frontal lobes in metacognitive monitoring. If the deficit
causing a TOT is the lack of retrieval of phonological information given the
retrieval of semantic information, one might expect TOTs to activate areas
posterior in the frontal lobe, perhaps Broca’s area. Broca’s area activation
during TOTs would support direct-access models of TOTs. However, all such
notions remain speculation.
Some very preliminary data from D.L.Schacter’s (personal communication,
June 9, 2000) lab points to a role for the anterior cingulate gyrus of the frontal
150 CHAPTER 7
SUMMARY
In this chapter, issues related to TOTs and aging, children’s TOTs, and TOTs in
neuropsychological patients have been addressed. I contrasted the decrement
theory of TOTs and aging with the incremental knowledge theory of TOTs and
aging, and discussed the support for each in the literature. I then discussed how
both direct-access and inferential models of TOT etiology account for the aging
literature. The literature on TOTs in childhood and neuropsychological patients,
unfortunately, is still in its infancy, but some interesting avenues of research in
those domains were suggested.
8
Conclusions and Directions
for the Future
In the first seven chapters, I present a synopsis of what is known about TOTs
and also illustrated how much is not known. I suspect that my readers were
frequently disappointed that issues could not be resolved and that so many
interesting questions remain unsolved. Despite the gaps, loose ends, and open
holes in our knowledge about TOTs, there are solid data pointing to the
importance of the phenomena for a range of psychological issues. More to the
point, relative to many other subjective phenomena of human psychology, there
has been a lot of research conducted on TOTs, which may serve as a model for
other phenomenology-based investigations. In this closing chapter, I present
some general thoughts, link TOTs to other phenomenological experiences, and
offer a summary on TOTs.
I reiterate some of the important themes in the volume: On one hand, TOTs
can be viewed as instances in which known words are not retrieved. In this view,
TOTs serve as “windows” onto word retrieval (Kohn et al., 1987) and inform us
about retrieval processes. On the other hand, TOTs can be thought of as
phenomenological experiences, in which one has a feeling about the likelihood
of retrieval. In this view, it may be better to categorize TOTs as experiences that
provide a “lens” onto human phenomenology (Schwartz, 1999b). That is,
because the phenomenology is so salient and occurs so often, the TOT can serve
as a case study in the investigation of human phenomenology. In this volume, I
do not argue against the former view, but argue strongly for the second view.
Thus, to the extent that some researchers have exclusively endorsed the
“window” view, my arguments have been critical of this singular approach.
Partially as a function of the two different views, different mechanistic
explanations have been advanced to explain TOTs. On the one hand, direct-
access models theorize that TOTs are based on access to the unrecalled target,
usually semantic aspects of it. On the other hand, inferential models theorize that
TOTs are based on clues and cues that a target might be in memory store.
Psycholinguists who argued for a “window on retrieval” view, usually endorse
the direct-access models, whereas metacognition researchers, such as myself,
have stressed the inferential model. I think the jury is far from reaching its
verdict on which, if either, theory should prevail, but all of us benefit from
151
152 CHAPTER 8
the subjective experience of the TOT could become damaged, causing failure to
experience the TOT, even when unrecalled words are potentially retrievable.
Another brain region, responsible for linking those subjective experiences to the
word retrieval process, could become damaged, causing indiscriminate TOTs
without any link to successful resolution. In chapter 7, I argued that it is possible
that damage to the prefrontal cortex may create difficulties in appropriately
producing TOTs or linking TOTs to retrieval. Finally, TOT production,
combined with a deficit in word retrieval, may result in patients who constantly
experience TOTs for known words, but can only recognize but not recall those
TOT targets. This scenario is probably the case in anomic aphasic patients.
Future neuropsychological studies should be geared in this direction.
A WORD ON CONSCIOUSNESS
RELATED PHENOMENA
TOTs are advanced as a case study in phenomenology because the TOT research
can serve as a model for thinking about a number of different psychological
phenomena, such as déjà vû, the tip-of-the-nose, absent-mindedness, and
perhaps, the remembering of dreams. These phenomena, not well studied or well
understand, mirror the TOT in a number of important ways.
Like the TOT, déjà vû is characterized by a strong phenomenological
experience. We sense that we have previously experienced something, although
knowing objectively that it is not possible. Like the TOT, the important aspects
of the déjà vû experience are twofold. On the one hand, it is important to
determine what kinds of memory processes are activated during déjà vû
experiences. On the other hand, déjà vû is essentially a phenomenological
154 CHAPTER 8
experience, which tells us we may know something but not how we know it. It is
monitoring certain processes, which we cannot monitor consciously, and like the
TOT, it serves a control function: It tells us we may have knowledge that is not
currently accessible. It urges us to search memory store to bring the information
into conscious accessibility.
Unlike TOTs, I suspect a diary study would reveal that déjà vû experiences
are very rare, and probably not possible to induce in the lab. Thus, déjà vû
experiences are more difficult to study than TOTs. Nonetheless, I think they
probably shares basic features with the TOT in its phenomenological salience
and in its probable function.
Engen and colleagues (Engen, 1987; Lawless & Engen, 1977; see Brown,
1991) described what they called the “tip-of-the-nose” phenomenon. This refers
to a common experience in which one smells an odor that is very familiar, but
cannot name the odor. Engen claimed that people have almost no lexical or
phonological information for the odor, but are overwhelmed by its familiarity. In
their studies, Engen and his colleagues showed that participants have knowledge
about the odor, as they could identify objects associated with it. In the terms I
have advanced, this is another example in which a strong phenomenological
experience gives rise to an inference that one may have knowledge that is not
currently accessible. In a functional sense, then, the “tip-of-the-nose”
phenomenon can direct search and knowledge acquisition strategies.
“I don’t know his name, but his face sure rings a bell” is the punch line to a
children’s joke about a paraplegic bell-ringer in the cathedral of Notre Dame. It
also rings true as another phenomenon that may share features with the TOT. As
discussed earlier, people generally have good memories for faces, but poor
memories for names. Thus, we are often confronted with situations in which
recall of a name is impossible, although the face is recognized. Indeed, Brennan
et al. (1990) showed that presenting faces does not resolve TOTs for names.
Nonetheless, a familiar face may inform us that we know something about the
person and perhaps should mentally search for that information.
Earlier, the Chinese language phenomenon called the “tip-of-the-pen”
phenomenon (TOP) was discussed (see Brown, 1991, for a discussion of slips of
the pen). In the TOP, a person knows the phonological form of the word, that is,
he or she can say it, but the written character is not accessible. Thus, the TOP
informs the person to direct a mental search for the missing orthographic entry.
There is also the common feeling on awakening of having just had a dream,
which slips away as your mind begins to change from its relaxed dream state to
the awareness that you are already late for work. There is often a feeling that the
whole dream is right on the tip of one’s tongue. We usually feel that we can or
should be able to remember the dream, but are seldom able to do so.
I argue that the TOT can serve as a model for these other phenomenological
experiences. Unlike these other experiences, the TOT is quite common in
everyday life and is easily amenable to laboratory methodology for study. It has
CONCLUSIONS FOR THE FUTURE 155
been relatively well studied over the past 35 years since Brown and McNeill
(1966) first asked a group of Harvard students to let them know if they were
experiencing TOTs in response to obscure word definitions. I suspect that much
of what we know about the TOT is generalizable, and at least in principle, to
many other human phenomenological experiences.
Over the last 10 years, my colleagues, students, and I have been on a quest to
understand an odd but interesting aspect of cognition, the tip-of-the-tongue
phenomenon. To understand just one small aspect of human cognition, it was
necessary to synthesize theories in four domains: psycholinguistics, memory
theory, cognitive aging, and metacognition. Unfortunately, a fifth domain,
neuropsychology, has yet to contribute much. The research has carried me from
interviewing Amharic-speaking cab drivers, to e-mailing Native American Web
sites, to asking students to carry around diaries, to the more traditional work of
running experiments in the lab.
If I can draw any one conclusion about TOTs, it is their strong subjective
presence when we experience them. They can be frustrating, or exciting, or not
at all emotional, but they cause us to rack our minds for unremembered words
and names. And it is this strong subjective experience that deserves our attention
as scientists. Thus, regardless of their etiology, and regardless of their role in
word retrieval, monitoring, and control, more study is needed to understand why
TOTs feel the way they do. ]
It is traditional to end a volume such as this with a quote from a famous
person, perhaps a famous playwright or novelist, who may have casually made
reference to the phenomena under discussion, or failing that, a quote from an old
master such as William James, or failing that, at least a quote from a respected
colleague. Instead, I end with a quote from the aphasic patient HW (Funnell et
al., 1996), a man caught in a perpetual TOT state, always trying to remember
words that he knows but whose retreival prove elusive. When asked to retrieve
the furry animal that attacks snakes (mongoose), HW said: “The problem is I
can’t remember in my own head. It’s not that I can’t say it. I know what it is and
I can’t know how to say the name. I don’t know what the name is of what they
are” (Funnell et al., 1996, p. 190). And that, perhaps, is what TOTs feel like to
us all: the wave of a magician’s wand—now we know it, and now we don’t.
References
157
158 REFERENCES
Dunlosky, J., & Nelson, T.O. (1997). Similarity between cue for judgments of learning
(JOLs) and the cue for test is not the primary determinant of JOL accuracy. Journal of
Memory and Language, 36, 34–49.
Ecke, R (1997). Tip of the tongue states in first and foreign languages: Similarities and
differences of lexicad retrieval failures. Proceedings of the EUROSLA 7 Conference (pp.
505–514). Barcelona, Spain.
Ecke, P., & Garrett, M.F. (1998). Lexical retrieval stages of momentarily inaccessible
foreign language words. Ilha do Desterro: Special issue on the cognitive aspects of
foreign/second language acquisition/learning, 35, 157–183.
Elbers, L. (1985). A tip-of-the-tongue experience at age two? Journal of Child Language,
12, 353–365.
Engen, T. (1987). Remembering odors and their names. American Scientist, 75, 497–503.
Faust, M., Dimitrovsky, L., & Davidi, S. (1997). Naming difficulties in language-disabled
children: Preliminary findings with application of the tip-of-the-tongue paradigm.
Journal of Speech, Language, and Hearing Research, 40, 1026–1036.
Finley, G.E., & Sharp, T. (1989). Name retrieval by the elderly in the tip-of-the-tongue
paradigm: Demonstrable success in overcoming initial failure. Educational Gerontology,
15, 259–265.
Freedman, J.L., & Landauer, T.K. (1966). Retrieval of long-term memory: “Tip-of-
thetongue” phenomenon. Psychonomic Science, 4, 309–310.
Frick-Horbury, D., & Guttentag, R.E. (1998). The effects of restricting hand gesture
production on lexical retrieval and free recall. American Joumal of Psychology, 111, 43–
62.
Funnell, M., Metcalfe, J., & Tsapkani, K. (1996). In the mind but not on the tongue: Feeling
of knowing in an anomic patient. In L.M.Reder (Ed.), Implicit memory and
metacognition (pp. 171–194). Mahwah, NJ: Lawrence Erlbaum Associates.
Gardiner, J.M. (1988). Functional aspects of the recollective experience. Memory &
Cognition, 16, 303–313.
Gardiner, J.M., Craik, F.I.M., & Bleasdale, F.A. (1973). Retrieval difficulty and subsequent
recall. Memory & Cognition, 1, 213–216.
Gardiner, J.M., & Java, R.I. (1993). Recognizing and Remembering. In A.F.Collins, S.E.
Gathercole, M.A., Conway, & P.E.Morris (Eds.), Theories of memory (pp. 163–188).
Hillsdale, NJ: Lawrence Erlbaum Associates.
Garrett, M. (1975). The analysis of sentence production. In G.H.Bower (Ed.), The
psychology of learning and motivation (pp. 133–177). New York: Academic.
Garrett, M. (1992). Disorders of lexical selection. Cognition, 42, 143–180.
Glucksberg, S., & McCloskey, M. (1981). Decisions about ignorance: Knowing that you
don’t know. Journal of Experimental Psychology: Human Leaming and Memory, 7, 311–
325.
Gollan, T., & Silverberg, N. (in press). Tip-of-the-tongue states in Hebrew-English
bilinguals. Bilingualism. Language and Cognition.
Goodglass, H., Kaplan, E., Weintraub, S., & Ackerman, N. (1976). The “tip-of-the-tongue”
phenomenon in aphasia. Cortex, 12, 145–153.
Guralnik, D.B. (1984). Webster’s new world dictionary (2nd College Ed). New York: Simon
and Schuster.
Harley, T.A., & Bown, H.E. (1998). What causes a tip-of-the-tongue state? Evidence for
lexical neighbourhood effects in speech production. British Journal of Psychology, 89,
151–174.
160 REFERENCES
Hart, J.T. (1965). Memory and the feeling-of-knowing experience. Journal of Educational
Psychology, 56, 208–216.
Hart, J.T. (1967). Memory and the memory-monitoring process. Journal of Verbal Learning
and Verbal Behavior, 6, 685–691.
Heine, M.K., Ober, B.A., & Shenaut, G.K. (1999). Naturally occurring and experimentally
induced tip-of-the-tongue experiences in three adult age groups. Psychology and Aging,
14, 445–457.
Hulms, K. (1983). The bone people. New York: Penguin.
Iwasaki, K, Vigliocco, G., & Garrett, M.F. (1998). Adjectives and adjectival nouns in
japanese: Psychological processes in sentence production (pp. 555–568).
Japanese/Korean Linguistics. Stanford, CA: CSLI.
Jacoby, L.L. (1991). A process dissociation framework: Separating automatic from
intertional uses of memory. Journal of Memory and Language, 30, 513–541.
Jacoby, L.L., & Dallas, M. (1981). On the relationship between autobiographical memory
and perceptual learning. Journal of Experimental Psychology: General, 110, 306–340.
James, L.E., & Burke, D.M. (2000). Phonological priming effects on word retrieval and tip-
of-tongue experiences. Journal of Experimental Psychology, LMC, 26, 1378–1391.
James, W. (1890). The principles of psychology: Vol 1. New York: Holt.
Janowsky, J.S., Shimamura, A.P., & Squire, L.R. (1989). Memory and metamemory:
Comparisons between frontal lobe patients and amnesic patients. Psychobiology, 17, 3–
11.
Jones, G.V. (1988). Analyzing memory blocks. In M.M. Gruneberg, P.E.Morris, & R.N.
Sykes (Eds.), Practical aspects of memory: Current research and issues (Vol. 1; pp. 215–
220). New York: Wiley.
Jones, G.V. (1989). Back to Woodworth: Role of interlopers in the tip-of-the-tongue
phenomenon. Memory & Cognition, 17, 69–76.
Jones, G.V., & Langford, S. (1987). Phonological blocking in the tip of the tongue state.
Cognition, 26, 115–122.
Kahneman, D., & Tversky, A. (1973). On the psychology of prediction. Psychological
Review, 80, 237–251.
Kahneman, D., & Tversky, A. (1984). Choices, values, and frames. American Psychologist,
39, 341–350.
Klin, C.M., Guzman, A.E., & Levine, W.H. (1997). Knowing that you don’t know:
Metamemory and discourse processing. Journal of Experimental Psychology: Learning,
Memory, & Cognition, 23, 1378–1393.
Kohn, S.E., Wingfield, A., Menn, L., Goodglass, H., Berko Gleason, J., & Hyde, M. (1987).
Lexical retrieval: The tip-of-the-tongue phenomenon. Applied Psycholinguistics, 8, 245–
266.
Kolb, B., & Whishaw, I.Q. (1996). Fundamentals of human neuropsychology, 4th edition.
New York: Freeman.
Kolers, P.A., & Palef, S.R. (1976). Knowing not. Memory & Cognition, 4, 553–558.
Koriat, A. (1993). How do we know that we know? The accessibility account of the feeling
of knowing. Psychological Review, 100, 609–639.
Koriat, A. (1995). Dissociating knowing and the feeling of knowing: Further evidence for
the accessibility model. Journal of Experimental Psychology: General, 124, 311–333.
Koriat, A. (1997). Monitoring one’s own knowledge during study: A cue-utilization
approach to judgments of learning. Journal of Experimental Psychology: General, 126,
349–370.
REFERENCES 161
Koriat, A., & Goldsmith, M. (1996). Monitoring and control processes in the strategic
regulation of memory accuracy. Psychological Review, 103, 490–517.
Koriat, A., & Lieblich, I. (1974). What does a person in a “TOT” state know that a person in
a “don’t know” state doesn’t know. Memory & Cognition, 2, 647–655.
Koriat, A., & Lieblich, I. (1977). A study of memory pointers. Acta Psychologica, 41, 151–
164.
Kozlowski, L. (1977). Effects of distorted auditory and of rhyming cues on retrieval of tip-
of-the-tongue words by poets and nonpoets. Memory & Cognition, 5, 477–481.
Krauss, R.M., Dushay, R.A., Chen, Y., & Rauscher, F. (1995). The communicative value of
conversational hand gestures. Journal of Experimental Social Psychology, 31, 533–552.
Larsen, S.F. (1998). What is it like to remember? On phenoemal qualities of memory. In C.
P.Thompson, D.J.Hermann, D.Bruce, J.D.Read, D.G.Payne, & M.P.Toglia (Eds.),
Autobiographical memory: Theoretical and applied perspectives (pp. 163–190).
Mahwah, NJ: Lawrence Erlbaum Associates.
Lawless, H., & Engen, T. (1977). Associations to odors: Interference, mnemonics, and
verbal labeling. Journal of Experimental Psychology: Human Learning and Memory, 3,
52–59.
Leonesio, R.J., & Nelson, T.O. (1990). Do different metamemory judgements tap the same
underlying aspects of memory. Journal of Experimental Psychology, LMC, 16, 464–470.
Levelt, W.J.M. (1989). Speaking: From intention to articulation. Cambridge, MA: MIT
Press.
Loftus, E.E, Donders, K., Hoffman, H.G., & Schooler, J.W. (1989). Creating new memories
that are quickly accessed and confidently held. Memory & Cognition, 17, 606–616.
MacKay, D.G., & Burke, D.M. (1990). Cognition and aging: A theory of new learning and
the use of old connections. In T.Hess (Ed.), Aging and cognition: Knowledge,
organization, and utilization (pp. 213–263). Amsterdam: North-Holland.
MacLin, O.H. (2000). The relationship between feeling of knowing and tip of the tongue:
Effects of perceived difficulty, cue familiarity, and partial information. Manuscript
submitted for publication.
Maki, R.H. (1998). Test predictions over text material. In D.J.Hacker, J.Dunlosky, & A.C.
Graesser (Eds.), Metacognition in educational theory and practice (pp. 117–144).
Mahwah, NJ: Lawrence Erlbaum Associates.
Marsh, S., & Emmorey, K. (2000, July). Tip-of-the-finger experiences in deaf signers. Paper
presented at the Theoretical Issues in Sign Language Research Conference, Amsterdam,
the Netherlands.
Martin, N., Dell, G.S., Saffran, E.M., & Schwartz, M.E (1994). Origins of paraphasias in
deep dysphasia: Testing the consequences of a decay impairment to an interactive
spreading activation model of lexical retrieval. Brain and Language, 47, 609–660.
Matison, R., Mayeux, R., Rosen, J., & Fahn, S. (1982). “Tip-of-the-tongue” phenomenon in
Parkinson disease. Neurology, 32, 567–570.
Matlin, M.W. (1998). Cognition (4th ed.). Orlando, FL: Harcourt Brace.
Maylor, E.A. (1990a). Recognizing and naming faces: Aging, memory retrieval, and the tip
of the tongue state. Journal of Gerontology, 45, 215–226.
Maylor, E.A. (1990b). Age, blocking and the tip of the tongue state. British Journal of
Psychology, 81, 123–134.
Maylor, E.A. (1994). Ageing and the retrieval of specialized and general knowledge:
Performance of master minds. British Journal of Psychology, 85, 105–114.
162 REFERENCES
Mazzoni, G., & Cornoldi, C. (1993). Strategies in study time allocation: Why is study time
sometimes not effective? Journal of Experimental Psychology: General, 122, 47–60.
Mazzoni, G., Cornoldi, C., & Marchetilli, G. (1990). Do memorability ratings affect study-
time allocation? Memory & Cognition, 18, 196–204.
Mazzoni, G., & Nelson, T.O. (1995). Judgments of learning are affected by the kind of
encoding in ways that cannot be attributed to the level of recall. Journal of Experimental
Psychology: Learning, Memory, & Cognition, 21, 1263–1274.
Metcalfe, J. (1993). Novelty monitoring, metacognition, and control in a composite
holographic associative recall model: Interpretations for Korsakoff amnesia.
Psychological Review, 100, 3–22.
Metcalfe, J. (1996). Metacognitive processes. In E.L.Bjork & R.A.Bjork (Eds.), Memory (pp.
383–411). San Diego, CA: Academic.
Metcalfe, J. (2000). Metamemory: Theory and data. In E.Tulving & E I.M.Craik (Eds.), The
Oxford handbook of memory (pp. 197–214). New York: Oxford University Press.
Metcalfe, J., Schwartz, B.L., & Joaquim, S.G. (1993). The cue familiarity heuristic in
metacognition. Journal of Experimental Psychology: Learning, Memory, and Cognition,
19, 851–861.
Metcalfe, J., & Wiebe, D. (1987). Intuition in insight and noninsight problem solving.
Memory & Cognition, 15, 238–246.
Miozzo, M., & Caramazza, A. (1997). Retrieval of lexical-syntactic features in tip-of-
thetongue states. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 23, 1410–1423.
Meyer, A.S., & Bock, K. (1992). The tip-of-the-tongue phenomenon: Blocking or partial
activation? Memory & Cognition, 20, 715–726.
Mulligan, N.W., & Hirshman, E. (1997). Measuring the bases of recognition memory: An
investigation of the process-dissociation framework. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 23, 280–304.
Murakami, Y. (1980). On the memory unit within kana-letter and kanji-letter words in the tip
of the tongue phenomenon. Japanese Journal of Psychology, 51, 41–44.
Neisser, U. (1978). What are the important questions? In M.M.Gruneberg, P.E.Morris, &
R.N.Sykes (Eds.), Practical aspects of memory: Current research and issues (Vol. 1; pp.
3–24). New York: Wiley.
Nelson, T.O. (1984). A comparison of current measures of the accuracy of feeling-of-
knowing predictions. Psychological Bulletin, 95, 109–133.
Nelson, T.O. (1988). Predictive accuracy of feeling of knowing across different criterion
tasks and across different subject populations and individuals. In M.Gruneberg, P.Morris,
& R.Sykes (Eds.), Practical aspects of memory: Current research and issues (Vol. 1; pp.
190–196). New York: Wiley.
Nelson, T.O. (1993). Judgments of learning and the allocation of study time. Journal of Ex-
perimental Psychology: General, 122, 269–273.
Nelson, T.O. (1996). Consciousness and metacognition. American Psychologist, 51, 102–
116.
Nelson, T.O., & Dunlosky, J. (1991). The delayed-JOL effect: When delaying your
judgments of learning can improve the accuracy of your metacognitive monitoring.
Psychological Science, 2, 267–270.
Nelson, T.O., Dunlosky, J., Graf, A., & Narens, L. (1994). Utilization of metacognitive
judgments in the allocation of study during multitrial learning. Psychological Science, 5,
207–213.
REFERENCES 163
Nelson, T.O., Gerler, D., & Narens, L. (1984). Accuracy of feeling of knowing judgments
for predicting perceptual identification and relearning. journal of Experimental
Psychology: General, 113, 282–300.
Nelson, T.O., Graf, A., Dunlosky, J., Marlatt, A., Walker, D., & Luce, K. (1998). Effect of
acute alcohol intoxication on recall and on judgments of learning during the acquisition
of new information. In G.Mazzoni & T.O.Nelson (Eds.), Metacognition and cognitive
neuropsychology: Monitoring and control processes (pp. 161–180). Mahwah, NJ:
Lawrence Erlbaum Associates.
Nelson, T.O., & Leonesio, R.J. (1988). Allocation of self-paced study time and the “labor-in-
vain” effect. Journal of Experimental Psychology: Learning, Memory, & Cognition, 14,
676–686.
Nelson, T.O., & Narens, L. (1980). Norms of 300 general-information questions: Accuracy
of recall, latency of recall, and feeling-of-knowing ratings. Journal of Verbal Learning &
Verbal Behavior, 19, 338–368.
Nelson, T.O., & Narens, L. (1990). Metamemory: A theoretical framework and new
fmdings. In G.Bower (Ed.), The psychology of learning and motivation (Vol. 26; pp.
125–141). San Diego, CA: Academic.
Nelson, T.O., & Narens, L. (1994). Why investigate metacognition? In J.Metcalf &
A.Shimamura (Eds.), Metacognition: Knowing about knowing (pp. 1–26). Cambridge,
MA: MIT Press.
Nisbett, R.E., & Wilson, T.D. (1977). Telling more than we can know: Verbal reports on
mental processes. Psychological Review, 84, 231–259.
Nyberg, L., Cabeza, R., & Tulving, E. (1996). PET studies of encoding and retrieval: The
HERA model. Psychonomic Bulletin & Review, 3, 135–148.
Perfect, T.J., & Hanley, J.R. (1992). The tip-of-the-tongue phenomenon: Do experimenter-
presented interlopers have any effect? Cognition, 45, 55–75.
Piattelli-Palmarini, M. (1994). Inevitable illusions: How the mistakes of reason rule our
minds. New York: Wiley.
Pinker, S. (1997). How the mind works. New York: Norton.
Pinker, S. (1999). Words and rules: The ingredients of language. New York: Basic.
Pitcher, H. (1999). Chekhov: The comic stories. Chicago: Ivan Dee.
Priller, J., & Mittenecker, E (1988). Experimente zum Unterrschied von “Wort auf der
Zunge” und “Gefuhl des Wissens.” Zeitshrift fur Experimentelle und Angewandte
Psychologie, 35, 129–146.
Rastle, K.G., & Burke, D.M. (1996). Priming the tip of the tongue: Effects of prior
processing on word retrieval in young and older adults. Journal of Memory and
Language, 35, 586–605.
Read, J.D., & Bruce, D. (1982). Longitudinal tracking of difficult memory retrievals.
Cognitive Psychology, 14, 280–300.
Reason, J.T., & Lucas, D. (1984). Using cognitive diaries to investigate naturally occurring
memory blocks. In J.Harris & P.E.Morris (Eds.), Everyday memory, actions, and absent
mindedness (pp. 53–70). London: Academic.
Reder, L.M. (1987). Selection strategies in question answering. Cognitive Psychology, 19,
90–138.
Reder, L.M. (1988). Strategic control of retrieval strategies. In G.Bower (Ed.), The
psychology of learning and motivation (Vol. 22). San Diego, CA: Academic.
164 REFERENCES
Reder, L.M., & Ritter, F.E. (1992). What determines initial feeling of knowing? Familiarity
with question terms, not with the answer. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 18, 435–451.
Reder, L.M., & Schunn, C.D. (1996). Metacognition does not imply awareness: Strategy
choice is governed by implicit learning and memory. In L.M.Reder (Ed.), Implicit
memory and metacognition (pp. 45–78). Mahwah, NJ: Lawrence Erlbaum Associates.
Riefer, D.M., Kevari, M.K., & Kramer, D.L. (1995). Name that tune: Eliciting tip-of-
thetongue experience using auditory stimuli. Psychological Reports, 77, 1379–1390.
Roediger, H.L. (1996). Memory illusions. Journal of Memory and Language, 35, 76–100.
Rubin, D.C. (1975). Within word structure in the tip-of-the-tongue phenomenon. Journal of
Verbal Learning and Verbal Behavior, 14, 392–397.
Rubin, D.C. (1998). Beginnings of a theory of autobiographical remembering. In C. E
Thompson, D.J.Hermann, D.Bruce, J.D.Read, D.G.Payne, & M.P.Toglia (Eds.),
Autobiographical memory: Theoretical and applied perspectives (pp. 47–68). Mahwah,
NJ: Lawrence Erlbaum Associates.
Ryan, M.P., Petty, C.R., & Wenzlaff, R.M. (1982). Motivated remembering efforts during
tip-of-the-tongue states. Acta Psychologica, 51, 137–147.
Salthouse, T.A. (1991). Theoretical perspectives on cognitive aging. Hillsdale, NJ: Lawrence
Erlbaum Associates.
Salthouse, T.A. (1993). Speed mediation of adult age differences in cognition.
Developmental Psychology, 29, 722–738.
Salthouse, T.A. (1996). The processing-speed theory of adult age differences in cognition.
Psychological Review, 103, 403–428.
Schachter, P. (1990). What’s in a name? Inferences from tip-of-the-tongue phenomena. In
L.M.Hyman & C.Li (Eds.), Language, speech, and mind: Studies in honour of Victoria A.
Fromkin (pp. 295–321). New York: Routledge & Kegan Paul.
Schacter, D.L. (1996). Searching for memory: The brain, the mind, and the past. New York:
Basic.
Schacter, D.L. (1999). The seven sins of memory: Insights from psychology and cognitive
neuroscience. American Psychologist, 54, 182–203.
Schacter, D.L. (2000). Commentary on perspectives on the tip-of-the-tongue phenomenon.
Paper presented at the American Psychological Society Convention, Miami Beach, FL.
Schacter, D.L, Harbluk, J.L., & McLachlan, D.R. (1984). Retrieval without recollection: An
experimental analysis of source amnesia. Journal of Verbal Learning and Verbal
Behavior, 23, 593–611.
Schwartz, B.L. (1994). Sources of information in metamemory: Judgments of learning and
feelings of knowing. Psychonomic Bulletin & Review, 1, 357–375.
Schwartz, B.L. (1998). Illusory tip-of-the-tongue states. Memory, 6, 623–642.
Schwartz, B.L. (1999a). The phenomenology of naturally-occurring tip-of-the-tongue states:
A diary study. Abstracts of the Psychonomic Society, 4, 24.
Schwartz, B.L. (1999b). Sparkling at the end of the tongue: The etiology of tip-of-thetongue
phenomenology. Psychonomic Bulletin & Review, 6, 379–393.
Schwartz, B.L. (2001). The relation of tip-of-the-tongue states and retrieval time. Memory &
Cognition, 29, 117–126.
Schwartz, B.L. (2000). The strategic selection of retrieval strategies during tip-of-thetongue
states. Unpublished data.
Schwartz, B.L., & Castillo-Andrade, Y. (1999). Access to plural forms in subjectively-
defined tip-of-the-tongue states. Unpublished data.
REFERENCES 165
Schwartz, B.L., Castillo-Andrade, Y., & Gonzalez, J. (1998). Access to grammatical gender
in Spanish tip-of-the-tongue states. Unpublished data.
Schwartz, B.L, & Metcalfe, J. (1992). Cue familiarity but not target retrievability enhances
feeling-of-knowing judgments. Journal of Experimental Psychology: Learning, Memory,
and Cognition, 18, 1074–1083.
Schwartz, B.L, & Metcalfe, J. (1994). Methodological problems and pitfalls in the study of
human metacognition. In J.Metcalfe & A.Shimamura (Eds.), Metacognition: Knowing
about knowing (pp. 93–114). Cambridge, MA: MIT Press.
Schwartz, B.L., & Smith, S.M. (1997). The retrieval of related information influences tip-of-
the-tongue states. Journal of Memory and Language, 36, 68–86.
Schwartz, B.L., & Smith, S.M. (1998). Fluctuation and stability in tip-of-the-tongue states.
Unpublished manuscript.
Schwartz, B.L., Travis, D.M., Castro, A.M., & Smith, S.M. (2000). The phenomenology of
real and illusory tip-of-the-tongue states. Memory & Cognition, 28, 18–27.
Shimamura, A.P., & Squire, L.R. (1986). Memory and metamemory: A study of the feeling-
of-knowing phenomenon in amnesic patients. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 12, 452–460.
Silverberg, N., Gollan, T., & Garrett, M.F. (1999). What you can get from a TOT and what
NOT. Abstracts of the Psychonomic Society, 4, 11.
Smith, S.M. (1994). Frustrated feelings of imminent recall: On the tip-of-the tongue. In
J.Metcalfe & A.P.Shimamura (Eds.), Metacognition: Knowing about knowing (pp. 27–
46). Cambridge, MA: MIT Press.
Smith, S.M. (2000, June). Blocking, tip-of-the-tongue states, and incubation in word
retrieval. Paper presented at American Psychological Society Convention, Miami Beach,
FL.
Smith, S.M., Balfour, S.P., & Brown, J.M. (1994). Effects of practice on tip-of-the-tongue
states. Memory, 2, 31–49.
Smith, S.M., Brown, J.M., & Balfour, S.P. (1991). TOTimals: A controlled experimental
method for studying tip-of-the-tongue states. Bulletin of the Psychonomic Society, 29,
445–447.
Smith, S.M., Ward, T.B., & Finke, R.A. (1995). The creative cognition approach.
Cambridge, MA: MIT Press.
Spellman, B.A., & Bjork, R.A. (1992). People’s judgments of learning are extremely
accurate at predicting subsequent recall when retrieval practice mediates both tasks.
Psychological Sciences, 3, 315–316.
Son, L.K., & Metcalfe, J. (2000). Metacognitive and control strategies in study-time
allocation. Journal of Experimental Psychology: Learning, Memory, & Cognition, 26,
204–221.
Souchay, C., Isingi, M., & Espagnet, L. (2000). Aging, episodic memory feeling-of-
knowing, and frontal functioning. Neuropsychology, 14, 299–309.
Sun, Y., Vinson, D.P., & Vigliocco, G. (1998). Tip-of-the-tongue and tip-of-the-pen in
Chinese. Abstracts of the Psychonomic Society, 3, 32.
Sunderland, A., Watts, K., Baddeley, A.D., & Harris, J.E. (1986). Subjective memory as
sessment and test performance in elderly adults. Journal of Gerontology, 31, 376–384.
Thiede, K.W., & Dunlosky, J. (1999). Toward a general model of self-regulated study: An
analysis of selection of items for study and self-paced study time. Journal of
Experimental Psychology: Learning, Memory, & Cognition, 25, 1024–1037.
166 REFERENCES
Tooby, J., & Cosmides, L. (1995). Mapping the evolved functional organization of mind and
brain. In M.S.Gazzaniga (Ed.), The cognitive neurosciences (pp. 1185–1197).
Cambridge, MA: MIT Press.
Tulving, E. (1983). Elements of episodic memory. New York: Oxford University Press.
Tulving, E. (1985). Memory and consciousness. Canadian Psychologist, 26, 1–12.
Tulving, E. (1989). Memory: Performance, knowledge, and experience. European Journal of
Cognitive Psychology, 1, 3–26.
Tulving, E. (1994). Foreword. In J.Metcalfe & A.Shimamura (Eds.), Metacognition:
Knowing about knowing (pp. vii–x).
Tulving, E., & Colotla, V.A. (1970). Free recall of trilingual lists. Cognitive Psychology, 1,
86–98.
Tulving, E., & Pearlstone, Z. (1966). Availability versus accessibility of information in
memory for words. Journal of Verbal Learning and Verbal Behavior, 5, 381–391.
Tversky, B. (1981). Distortions in memory for maps. Cognitive Psychology, 13, 407–433.
Tversky, B. (1991). Spatial mental models. The Psychology of Learning and Motivation, 27,
109–145.
Tweney, R.D., Tracz, S., & Zaruba, S. (1975). Slips of the tongue and lexical storage.
Language and Speech, 18, 388–396.
Underwood, B.J. (1966). Individual and group predictions of item difficulty for free
learning. Journal of Experimental Psychology, 71, 673–679.
Vigliocco, G., Antonini, T., & Garrett, M.F. (1997). Grammatical gender is on the tip of
Italian tongues. Psychological Science, 8, 314–317.
Vigliocco, G., Vinson, D.P., Martin, R.C., & Garrett, M.F. (1999). Is “count” and “mass”
information available when the noun is not? An investigation of tip of the tongue states
and anomia. Journal of Memory and Language, 40, 534–558.
Vinson, D.P., & Vigliocco, G. (1999). Can independence be observed in a dependent
system? The case of tip-of-the-tongue states. Brain and Language, 68, 118–126.
Wellman, H.M. (1977). Tip of the tongue and feeling of knowing experiences: A
developmental study of memory monitoring. Child Development, 48, 13–21.
White, K.K., & Abrams, L. (1999). The role of syllable phonology and aging in priming tip-
of-the-tongue resolution. Abstracts of the Psychonomic Society, 4, 68.
Widner, R.L., Otani, H., & Falconier, C. (1999). The contribution of the prefrontal cortex to
feeling-of-knowing and tip-of-the-tongue reports. Abstracts of the Psychonomic Society,
4, 13.
Widner, R.L., Smith, S.M., & Graziano, W.G. (1996). The effects of demand characteristics
on the reporting of tip-of-the-tongue states and feeling-of-knowing states. American
Journal of Psychology, 109, 525–538.
Xu, Y., Pollatsek, A., & Potter, M.C. (1999). The activation of phonology during silent
Chinese word reading. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 25, 838–357.
Yaniv, I, & Meyer, D.E. (1987). Activation and metacognition of inaccessible stored
information: Potential bases for incubation effects in problem-solving. Journal of
Experimental Psychology: Learning, Memory, and Cognition, 13, 187–205.
Yarmey, A.D. (1973). I recognize your face, but I can’t remember your name: Further
evidence on the tip-of-the-tongue phenomenon. Memory & Cognition, 1, 287–290.
Zhang, S., Perfetti, C.A., & Yang, H. (1999). Whole word, frequency-general phonology in
semantic processing of Chinese characters. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 25, 858–876.
Author Index
167
168 AUTHOR INDEX
M P
MacKay, D.G., 3, 10, 29–37, 51–63, Palef, S.R., 97
72, 74, 77, 78, 88, 89, 99, 116, 117, Pearlstone, Z., 6
123, 127, 133, 134, 136, 137, 139, 141– Peck, V., 146
144, 152, 154 Perfect, T.J., 8, 9, 17, 37, 57, 66, 91,
MacLin, O.H., 69, 70, 144 117
McLachlan, D.R., 150 Perfetti, C.A., 93
McNeill, D., 4–15, 18, 21, 22, 28, 36, Petty, C.R., 52
43, 47, 53,59, 70, 73, 79, 80, 92, 99, Piattelli-Palmarini, M. 72
100, 116, 121,122,162 Pinker, S. 72, 74, 86, 98, 130
Maki, R.H., 100 Pitcher, H., 1
Marchetilli, G., 107 Pollatsek, A., 93
Marsh, S.V., 22, 28, 80 Posner, M.I., 157
Martin, N., 75 Potter, M.C., 93
Mathis, K.M., 88 Priller, J., 22
Matison, R., 154
Matlin, M.W., 74 R
Mayeux, R., 154 Rajaram, S., 20
Maylor, E.A., 134, 136, 137 Rastle, K.G., 5, 9, 17, 51, 60–63, 68,
Mazzoni, G., 96, 107, 198 71, 89, 90, 91, 133, 142
McCloskey, M., 98 Rauscher, F., 91
Melnick, R., 102 Read. J.D., 123
Metcalfe, J., 45, 47, 50, 65–68, 99, 102, Reason, J.T., 8, 29, 31, 32, 56, 79
123, 144, 151 Reder, L.M.. 66, 67, 96, 109, 110, 149
Miozzo, M., 10, 21, 50, 52, 54, 75, 77– Riefer, D.M., 39
87, 94, 116 Ritter, F.E., 66, 67
Mittenecker, S., 22 Roediger, H.L., 17, 20
Meyer, A.S., 57 Rosen, J., 154
Mulligan, N.W., 29 Rubin, D.C.. 16, 70, 80
Murakami, Y., 22, 54 Ryan, M.P., 52, 125–127, 131
S
N Saffran, E.M., 75
Neely, J.H., 67,
Salthouse, T.A., 32, 136, 139
Narens, L., 96, 96–104, 107, 113, 115,
Sanvito, J., 102
131
Schachter, P., 75, 79, 81
Neisser, U., 18, 21
Schacter, D.L., 130, 149, 150, 156, 157
Nelson, T.O., 50, 53, 96–104, 106–108,
Schooler, J.W., 16
110, 111, 113, 120, 123, 131, 146
Schunn, C.D., 96, 109, 110, 149
Ng, M.L., 88
Schwartz, B.L., 3, 5–9, 13, 14, 22, 41,
Nisbett, R.E., 104
43–48, 50, 52–54, 65–71, 86, 87, 99,
Nix, L A., 137–139, 145
100, 102, 104, 112, 116–124, 127–129,
Nyberg, L., 156
131, 140, 144, 145, 147, 158
Schwartz, M.F., 75
O Sharp, T., 135
Ober, B.A., 29 Shenaut, G.K., 29
170 AUTHOR INDEX
Shiffrin, R.M., 98 V
Shimamura, A.P., 155, 156 Vigliocco, G., 6, 7, 22, 37, 54, 55, 70,
Silverberg, 22, 54, 61, 80, 85 88, 90, 91 71, 82–84, 86, 93, 112, 116, 150–153
Smith, S.M., 5, 13, 19, 29, 41, 44, 45, Vinson, D.P., 93
47, 50, 53, 58, 59, 65, 68–71, 112, 116,
117, 118, 131, 144, 145 W
Spellman, B.A., 102, 103 Wade, E., 3
Son, L.K., 100, 107–109 Ward, T.B., 19
Squire, L.R., 155, 156 Watts, K., 145
Souchay, C., 155 Wellman, H.M., 21, 146–149
Sun, Y., 93 Wenzlaff, R.M., 52
Sunderland, A., 145 Wiebe, D., 45, 47
Whishaw, I.Q., 150
T White, K.K., 60, 62, 63, 65, 89, 90
Thiede, K.W., 98, 107–109 Widner, R.L., 41–43, 53, 130, 156, 160
Tkackz, S., 97 Wilson, T.D., 104
Tooby, J., 98, 99 Worthley, J.S., 3
Tsapkani, K., 151
Travis, D.M., 5 X
Tucker, D.M., 157 Xu, Y., 93
Tulving, E, 6, 15–21, 66, 71, 88, 105,
156, 159, 160
Tunnicliff, J., 16 Y
Tversky, A., 72 Yang, H., 93
Tversky, B., 71 Yarmey, A.D., 7
Tweney, R.D., 97
Z
U Zaruba, S., 97
Underwood, B.J., 100, 101 Zhang, S., 93
Subject Index
171
172 SUBJECT INDEX
Inferential Theory, 50, 55, 65- 72, 116, 79, 83, 93, 115–117, 127, 137, 138,
117, 144, 145, 150, 159 142, 145, 146, 150, 151, 155, 159
Imminence, 5, 6, 29, 30, 37, 44, 45, Persistent alternates, see interlopers
119, 128 Phenomenology, 15–18, 20, 21, 36–48,
Incomplete activation, 51, 59, 60 65, 66, 93, 95, 119, 132, 147, 149, 150,
Interlopers, 8, 31, 32, 56, 66, 79, 136, 157, 160, 162
138, 142 Phonological access, 74, 75, 78, 79, 81,
84, 154
J Phonological information, 10, 29, 78–
Judgments of knowing, 66, 100–106, 82, 93, 94, 115, 116, 137, 153, 156
108, 109, 112, 145 Phonological priming, 57, 58, 63, 80–
85, 88–91, 142
Positive TOTs, see negative TOTs
K
Problem-solving, 45
Korsakoff s amnesia, 155
Prospective montoring, 100
L
Labor-in-vain effect, 108 R
Recall, 6, 105, 106, 108, 110, 114, 124,
Lemma, 75–78, 81, 82, 84, 93, 94, 153
126
Lexeme, 75–78, 81, 82, 84, 93, 94, 153
Relief, 45
Lexical retrieval, 73–77, 85, 95, 112,
Remember/know judgments, 20
132, 155, 157
Retrieval, 6, 17, 18, 51, 52, 74, 83, 91,
95, 107–109, 120, 124, 127
M Retrieval time and TOTs, 120–124, 131
Memory research, 2, 3, 73, 124 Retrospective monitoring, 100, 111,
Metacognition, 18, 95, 99, 102–109, 112
113, 120, 132, 144, 149, 152, 156, 157,
159, 162 S
Metacognitive model, 96–98, 106, 112, Semantic information, 10, 29, 81, 94,
113, 131 138, 155, 157, 159
Metamemory, 50, 102 Semantic priming, 88, 94
Monitoring, 95, 96–108, 111–117, 119, Speech errors, 77
123, 145, 146, 155, 157, 162 Speed-accuracy tradeoff, 136
Motivational states, 125, 126, 128, 131 Stability, see fluctuation
Strategy selection, 96, 104, 107, 109,
N 110, 127–129, 135, 136
Negative TOTs, 9, 83, 86, 87 Strength of TOTs, 29, 30, 37, 119
Neuroimaging, 133, 149, 156, 157 Stress, 40, 43
Neuropsychology, 132, 133, 149, 150, Syntactical access, 81, 83–85, 87, 94,
154–157, 160, 162 115, 116, 153, 155
O T
Omission errors 39, 40, 110, 120, 121, Temporal lobes, 155, 156, 160
123 Third-person TOTs, 6, 83, 150
Orthography, 81, 93 Tip-of-the-finger, 28
Tip-of-the-nose, 161
P Tip-of-the-pen, 93, 94, 162
Partial information, 8, 9, 11, 53–55, 65, TOT