You are on page 1of 30

Journal of Pragmatics 34 (2002) 457486

A new look at literal meaning in understanding

what is said and implicated
Raymond W. Gibbs Jr.*
Department of Psychology, University of California, Santa Cruz, CA 95064, USA

What role does literal meaning play in language comprehension? This question has been vigorously debated in research on gurative language understanding. The standard pragmatic view
proposes that people must analyze the complete literal meaning of indirect and gurative utterances before pragmatic information is consulted to infer speakers nonliteral messages. Most of
the psycholinguistic research shows, however, that given sucient context people understand
nonliteralmeanings without rst analyzing the complete literal meaning of an expression (i.e., the
direct access view). Several lines of research have recently attempted to demonstrate that people
still analyze aspects of literal meaning when understanding metaphors, irony, idioms, and proverbs. I critically evaluate this new work and suggest that it does not contribute sucient evidence
against the direct access view. Nonetheless, I argue that other research suggests how people
analyze aspects of what speakers say as part of inferring what speakers implicate. This conclusion has several implications for specifying the role of pragmatics in ordinary utterance
interpretation. # 2002 Published by Elsevier Science B.V.
Keywords: Figurative language; Literal meaning; Conversational implicatures; Psycholinguistics

1. Introduction
A signicant issue that continues to draw the attention of linguists, philosophers,
and psychologists concerns the role that literal meaning plays in language interpretation. Many utterances in both conversation and written texts appear to communicate meanings that vary in some way from what is literally said. Consider the
following exchange between two college students:
Steve: Are you going to the big party this weekend?
Beth: Didnt you hear that Bob is going to be there?
* Tel.: +1-408-459-4630.
E-mail address:
0378-2166/02/$ - see front matter # 2002 Published by Elsevier Science B.V.
PII: S0378-2166(01)00046-7


R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486

Beths response does not provide a direct answer to Steves question, but conveys a
possible reply if we assume that Steve knows how Beth feels about Bob being at the
party (i.e., she dislikes her ex-boyfriend Bob and so would not even think about going
to a party he attended). In this instance, Beths literal utterance underdetermines
what she implied in context in the sense that Beth means what she says but also
wants to communicate something more than what she said.
Speakers in other situations say things they do not literally mean in any way, even
if the literal meaning of what they say is sensible. For example, if John says to Peter
My grandmother nally kicked the bucket, he does not wish to communicate that his
grandmother struck her foot against a pail. Instead, John intends to communicate
some nonliteral meaning, such as that his grandmother died, which appears on the
surface to have little to do with what he literally said.
In other cases, speakers literal utterances are anomalous. When Harry says to
Chad that Cigarettes are time bombs, he does not literally imply that cigarettes
explode unexpectedly, but rather he intends to communicate a metaphorical meaning, such as that cigarettes are similar in certain respects to time bombs (i.e., cigarettes are like time bombs in that smoking them will eventually, suddenly cause one
serious harm).
Each of these examples illustrates a problem for pragmatic theories of language
understandinghow does one infer what speakers imply given what they literally
say? There have been many proposals oered in cognitive science in response to this
question (see Gibbs, 1994 for a review). I have been critical of theories that assume
listeners/readers must rst analyze the literal meanings of utterances before applying
pragmatic information to derive what speakers implicate (i.e., the standard pragmatic model). As I have argued elsewhere (Gibbs, 1984, 1989, 1994), people often
appear to directly understand what speakers intend to communicate when using
gurative language without having to process the literal meanings of speakers
utterances (i.e., the direct access model). In recent years, several researchers have
criticized the direct access view and suggested that listeners still process the literal
meanings of gurative utterances.
My aim in this article is to briey respond to these recent empirical ndings. I
criticize how some researchers have operationally dened literal meaning in their
experiments and suggest that the direct access model should still be preferred over
theories that assume literal meanings have priority over gurative interpretations.
An important goal here is to clarify some of the misunderstandings that have arisen
over my previous claims that people do not analyze the literal meanings of gurative
expressions before deriving their indirect or nonliteral meanings. Nonetheless, I also
argue, based on newer ndings from my own lab, that people may possibly analyze
something about what speakers say as part of guring out what speakers implicate.
This conclusion has several implications for both linguistic and psychological theories
of utterance interpretation. Most notably, the empirical evidence rmly points to the
intuitive possibility of a distinction between understanding what speakers say and
implicate, even if there remains no credible evidence that people automatically analyze the non-pragmatic, literal meanings of utterances during ordinary language

R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486


2. The standard pragmatic view

The traditional view of how listeners understand speakers messages that depart
from what speakers literally say comes from H. Paul Grices theory of conversational
implicature (Grice, 1989). Grice argued that the inferences needed to understand
Beths comment Didnt you hear that Bob is going be there? or Johns remark Cigarettes are time bombs are derived from certain general principles or maxims of conversation that participants in talk-exchange are mutually expected to observe (Grice,
1975, 1989). Among these are the expectation that speakers are to be informative,
truthful, relevant, and clear in what they say. When an utterance appears to violate
any of these maxims, as Beth and John appear to do, listeners are expected to derive
an appropriate conversational implicature about what the speaker intended to communicate in context given the assumption that he or she is trying to be cooperative.
For instance, understanding a metaphorical comment, such as Cigarettes are time
bombs, requires that listeners must rst analyze what is said literally, then recognize
that the literal meaning (i.e., that cigarettes are literally bombs that are set to
explode at a specic future time), or what the speaker says, is contextually inappropriate, and only then infer some meaning consistent with the context and the idea
that the speaker must be acting cooperatively and rationally (i.e., smoking cigarettes
can have devastating eects for a person at a later time). Understanding literal
utterances doesnt demand the extra step of guring out how a speakers intended
meaning diers from his or her literal statement. The standard pragmatic view suggests, then, that indirect and gurative language should always be more dicult to
process than roughly equivalent literal speech. More generally, this traditional view
assumes that understanding what speakers literally say requires accessing of semantic information, while understanding what speakers implicate in context demands
pragmatic information that is more dicult to access than semantic knowledge.

3. The direct access view

The results of many psycholinguistic experiments have shown the traditional,
Gricean view to be incorrect as a psychological theory (see Gibbs, 1994; Glucksberg,
1998). Numerous reading-time and phrase classication studies demonstrate that listeners/readers can often understand the gurative interpretations of metaphors,
irony/sarcasm, idioms, proverbs, and indirect speech acts without having to rst
analyze and reject their literal meanings when these expressions are seen in realistic
social contexts. People can read gurative utterances (i.e, Youre a ne friend meaning
Youre a bad friend) as quickly as, sometimes even more quickly, than literal uses
of the same expressions in dierent contexts, or equivalent non-gurative expressions. These experimental ndings demonstrate that the traditional view of indirect
and gurative language as always requiring additional cognitive eort to be understood has little psychological validity.
An alternative view of gurative language use suggests that people can comprehend the intended meanings of many nonliteral utterances directly if these are seen


R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486

in realistic social contexts (Gibbs, 1994). The direct access view simply claims that
listeners need not automatically analyze the complete literal meanings of linguistic
expressions before accessing pragmatic knowledge to gure out what speakers mean
to communicate. The complete literal meaning of a linguistic expression is itself a
dicult concept to dene. Following Frege (1952), the literal meaning of any sentence is traditonally thought to be its compositional meaning (i.e., the combined
meaning of the individual words apart from context). Under the direct access view,
people can, given the right context, understand speakers communicative messages
without having to rst construct a compositional analysis for the utterance. At the
same time, the idea of a complete literal meaning of any utterance implies a highly
non-pragmatic analysis of what is said. For instance, under the standard pragmatic
model, when a speaker utters a statement as simple as The door is open, she only
literally says There is a uniquely identiable door somewhere in the universe and it
is open. Similarly, when a speaker utters Jane has three children, she only literally
says Jane has at least three children, but may have more than three (Grice, 1975).
In this sense, the complete literal meaning of any utterance is its compositional
meaning apart from enriched pragmatic knowledge (i.e., peoples background
knowledge about the world and the local context).
The direct access view does not claim that listeners never access something about
what the individual words mean (perhaps, but not necessarily, these words literal
meanings) during processing of what speakers imply. Nor does the direct access view
claim that people never take longer to process a gurative meaning than to understand a literal one (sometimes referred to as the processing-equivalence hypothesis, see Giora et al., 1998). People may sometimes take a good deal of time to
process novel poetic metaphors, for example. Yet it is not at all clear that the additional time needed to understand some novel expression is necessarily due to a preliminary stage in which the non-pragmatic, literal meaning for an entire utterance is
rst analyzed and then rejected. Listeners may take longer to understand a novel
expression because of the diculty in integrating the gurative meaning with the
context and not because listeners are rst analyzing and then rejecting the expressions literal meaning. Various empirical evidence supports this idea (Schaw, 1995;
Shinjo and Myers, 1987). For this reason, we simply should not infer that the literal
meaning for an entire phrase or expression must have been analyzed simply because
people take longer to read novel instances of gurative language than to process
either familiar gurative expressions or equivalent literal statements.
I oer these cautionary remarks about the direct access model because scholars
often argue that the direct access view can not be right given either dierences in
processing for familiar and unfamiliar instances of gurative language, or because of
longer reading times for gurative language than for equivalent literal expressions
(cf. Giora, 1995, 1997). Moreover, many critics of the direct access view attempt to
empirically demonstrate that the literal meanings of words in, for instance, metaphor and irony, appear to be accessed. These scholars sometimes argue that sensitive, on-line methods are needed to examine if and when literal meanings are
accessed during language interpretation. I agree with this point to a large degree.
But all the direct access model posits is that people do not automatically analyze the

R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486


complete context-free, or literal meanings, of entire utterances before deriving their

gurative meanings. People may indeed analyze aspects of what words mean when
understanding indirect and gurative language (Gibbs, 1994).
Consider the research on lexical ambiguity resolution. Psycholinguists have for
decades debated whether the multiple meanings of ambiguous words (e.g., bug in
The computer scientist found a bug in his program) are all momentarily understood or
only the contextually appropriate meaning is activated (see Simpson, 1994 for a
review). Research from various on-line studies provide support for both the contextsensitive model of lexical access (Glucksberg et al., 1986; Vu et al., 1998), in which only
the contextually appropriate meaning is activated, and context-insensitive views
(Swinney, 1979; Burgess et al., 1989), in which all the ambiguous words meanings are
momentarily activated until contexts comes in to disambiguate the words meaning.
Many psychologists now favor the position of highly frequent or dominant meanings
always being immediately activated regardless of context (Tabossi and Zardon, 1993).
Some work shows that processing polysemous words, unlike ambiguous words, results
in the continued activation of inappropriate sense for some time after the critical word
has been heard (Williams, 1992).
But virtually all theories posit that the contextually appropriate meaning of an
ambiguous word is determined within a short period of time (several hundred milliseconds at most). Thus, even theorists espousing a purely modular view of lexical
ambiguity resolution do not argue that people carry the literal meanings of words till
they get to the end of linguistic statements before pragmatic information, including
context, is accessed to infer speakers contextually appropriate messages. In this way,
the research on lexical ambiguity resolution does not directly bear on the specic claims
of the direct access view of gurative language interpretation. Another way of putting
this is that even if individual word meanings are momentarily activated regardless of
context, this should not be understood as implying that the complete literal meanings of
entire sentences are analyzed before speakers intended meanings are recognized.

4. Literal meanings in recent psycholinguistic ndings

Several studies have been published in recent years that attempt to illustrate the
importance of literal meaning in gurative language understanding. This work shows
that some activation of individual word meaning takes place in gurative language
understanding. But scholars then erroneously argue that this evidence indicates
something about literal meaning. As I demonstrated in considerable detail earlier
(Gibbs, 1994, chapter 2), it is not clear what it means to say that a word or expression
has a literal meaning. The concept of literal is quite complex, and people appear to
have at least ve, if not more, versions of literality (Gibbs et al., 1993). As noted
above, most views of literal meaning assume that it refers to something about what
entire sentences mean apart from context and enriched pragmatics. Several of the
recent studies conate aspects of literal and gurative meaning.
For instance, one study examined irony comprehension by asking people to read
statements that could have either literal or ironic meaning depending on the context


R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486

(Giora et al., 1998). Consider the ironic and literal meaning of the statement you are
just in time in the following contexts:
Anna was a great student, but she was absent-minded. One day when I was well
through my lecture, she suddenly showed up in the classroom. I said to her, You are
just in time. (ironic statement)
Anna was a great student and very responsible. One day she called to tell me she did
not know when she would enter the classroom. However, just as I was starting, she
entered the classroom. I said to her You are just in time. (literal statement)
Participants read these stories one line at a time on a computer screen. The results
showed that people took more time to read the nal statements in ironic contexts
than when these same expressions were seen in literal contexts. This nding is consistent with the standard pragmatic model, and contrary to the ndings reported by
Gibbs (1986a) where people actually took less time to read ironies than literal uses
of the same expressions. A closer examination of at least a few of the contexts used
in each study suggests one possible explanation for these discrepant ndings. Consider one short story context from the Gibbs (1986b) studies:
Gus just graduated from high school and he didnt know what to do with his life. One
day he saw an ad about the Navy. It said that the Navy was not just a job, but an
adventure. So Gus joined. Soon he was aboard a ship doing all sorts of boring things. One
day as he was peeling potatoes, he said to his buddy, This sure is an exciting life.
The reason why people might nd the ironic remark This sure is an exciting life as
easy to process as when this same sentence was seen in a literal context (e.g., where
the speaker said something truthful about the exciting life he was leading) is because
the context itself sets up an ironic situation through the contrast between what Gus
expected when he joined the Navy and the reality of it being rather boring. Because
people conceive of many situations ironically (Gibbs, 1994; Lucariello, 1994), they can
subsequently understand someones ironic, or sarcastic, comment without having to
engage in the additional computation that may be required when ironic remarks are
seen in situations that are inherently less ironic. This possibility suggests, not surprisingly, that various contextual factors inuence linguistic processing so it might not be
the case that irony always takes longer to process than literal language (see Katz and
Lee, 1993; Katz and Pexman, 1997 for studies on the inuence of dierent discourse
factors in irony comprehension and Giora, 1995 for a discussion of irony versus literal processing in terms of discourse coherence). People may still need to draw
complex inferences when understanding some ironic statements, but part of these
inferences can occur before one actually encounters an ironic utterance. Once again,
if the context is sucient, people may not necessarily analyze the complete literal
meaning of a sarcastic comment before deriving its intended gurative meaning.
The most vigorous defense of the standard pragmatic model is seen in Temple and
Honecks (1999) work on proverb understanding. Temple and Honeck (1999) dispute, on methodological grounds, the ndings of an earlier study on proverb
understanding showing that people are faster to process proverbs when they are

R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486


used guratively than when they are intended to have literal meaning (Kemper,
1981), which is clearly evidence in favor of the direct access view. They argue that
most studies, like Kempers, investigating the empirical claims of the standard
pragmatic model (which Temple and Honeck refer to as the multistage model of
gurative processing) focus on familiar, conventional expressions such as indirect
requests and idioms. As I have suggested earlier (Gibbs, 1989, 1994), this characterization of the psycholinguistic literature is inaccurate. Several studies have
examined comprehension of novel gurative language, including novel metaphors
and sarcastic indirect requests, and shown that these novel forms can be processed
as quickly as equivalent literal speech. Again, these ndings do not imply that people must always process novel gurative language as quickly as literal language. Yet
the fact that various works along this line have been reported speaks against the
claim that the direct access view only applies to familiar gurative language.
Temple and Honeck criticize the direct access view by arguing In our view, the
traditional problem is that the direct access position places essentially all the basis
for understanding on the context and none whatsoever on the proverb. That is,
people are presumed to already have an understanding of some topic, with the proverb serving as a mere redundant conrmation of it (1999: 47). As I argue above,
this characterization of the direct access position makes little sense. People clearly
do something with the words they hear and the fact that a listener hears a proverb as
opposed to some other statement must matter in the ongoing interaction between
speakers and listeners or writers and readers. Once more, the direct access position
only claims that listeners do not automatically process the complete non-pragmatic,
literal meanings of gurative expressions before bringing in pragmatic information
to infer speakers intended messages. The social context that exists at any one
moment (i.e., the speakers and listeners common ground) is rarely constraining
enough so that listeners know with certainty what speakers will say and intend to
mean before they utter their words.
Temple and Honecks hypothesis for their multistage view of proverb understanding is called the conceptual base theory (Honeck, 1997). This theory, like to
the standard pragmatic model, assumes that a complete construction of a proverbs
literal meaning is an essential step in a theory of how proverbs are understood. Literal meaning under Temple and Honecks view is based on the dierent values of
the words and their syntactic combination, activated background knowledge, lexicalized phrasal constituents, and their conventional usages (1999: 48). It is unclear
what aspects of activated background knowledge come into play in this view of
literal meaning. Yet it seems evident that the literal meaning analyzed by listeners
refers to a proverbs complete, nongurative meaning, determined apart from the
context in which it appears.
To test their hypothesis, Temple and Honeck asked participants to read two twosentence contexts. In the rst condition, one context was relevant to the literal meaning of a proverb and the other was literally irrelevant to the proverb (i.e., the literal
condition). In the second condition, one context was relevant to the gurative meaning of a proverb and the other was guratively irrelevant to the proverb (i.e., the gurative condition). Participants in a rst study either saw all literal or all gurative


R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486

contexts. For both conditions, after reading both contexts, participants read a novel
proverb (e.g., The cow gives good milk but kicks over the pail), and then quickly judged
which context the expression was most meaningfully related to. For both conditions,
then, the literally relevant or guratively relevant contexts were the ones that participants were expected to select. The latencies to make these choices were measured. The
data showed that reaction-times were longer in the gurative condition than in the
literal one, a nding that is taken as supporting the multi-stage view of proverb
comprehension. Follow-up studies essentially replicated this main nding.
The primary diculty with these studies is that the choice reaction-time task
doesnt accurately measure peoples immediate comprehension of proverbs in context. Asking participants to rst read a proverb and then make a judgment between
two contexts conates the process of comprehension with the process of judging the
appropriateness of a proverb against two dierent situations. This metalinguistic
task is dierent from what people ordinarily do in everyday conversations or when
reading. For instance, people hearing a linguistic statement dont have to judge
which of two dierent statements it relates to. Instead, people interpret a statement
by inferring the speakers communicative intention given the context at hand (i.e.,
the speaker and listeners common ground). Psycholinguistic research has been
increasingly critical of comprehension measures, such as Temple and Honecks
choice RT task, that are not sensitive to immediate psychological processes. Temple
and Honeck strongly claim, nonetheless, that the choice reaction-time method is
somehow more sensitive than simple reading time measures.
At the same time, the fact that there is a dierence in choice RT to gurative and
literal contexts does not imply that a proverbs complete literal meaning was analyzed as part of peoples understanding of its nonliteral interpretations. Nor do the
data directly imply that people used the literal meaning to infer the proverbs gurative interpretations (in the gurative condition). Temple and Honeck admit as
much when they state First, like Kempers (1981) study, this study has not directly
shown that literal meaning was in fact used to construct nonliteral meaning. Rather,
it was shown that it takes less time to understand that a proverb, such as The used
key is always bright, is about keys, brightness, and a general key-using schema, than
about frequently used instruments retaining their functional value (1999: 67). This
point is important because psycholinguistic theories of gurative language understanding have not detailed how the analysis of literal meaning may actually contribute to peoples understanding of speakers nonliteral messages. I question the
method employed by Temple and Honeck (1999), and do not agree that their ndings directly imply people analyze a proverbs complete literal meaning before its
nonliteral meaning is understood. Yet I will later suggest that aspects of what
speakers pragmatically say when using novel proverbs, to take one example, may
play some role in inferring what speakers implicate by their use of these expressions.
The majority of recent studies examining the role that literal meaning plays in
gurative language understanding employ on-line methodologies that are sensitive
to rapid activations of meanings. These on-line studies are presumed to be better
indicators of literal meaning activation than are more global measures of utterance
comprehension, such as reading time and phrase classication techniques. Yet, I

R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486


argue that these on-line studies only examine the immediate activation of individual
word meaning and do not specically assess the analysis of an utterances complete
literal meaning during immediate language understanding.
For instance, one research project examined comprehension of familiar and less
familiar metaphorical expressions (Blasko and Connine, 1993). Participants in these
experiments heard dierent sentences and made lexical decisions at various times to
visually presented word strings. For example, as participants heard the sentence The
belief that hard work is a ladder is common to this generation, they were visually
presented a letter string immediately after hearing the word ladder. The letter string
visually presented was related to some aspect of the sentences literal meaning (e.g.,
rungs), a letter string related to the sentences metaphoric meaning (e.g., advance), or
a control word unrelated to the sentence (e.g., pastry). The results revealed that
participants were equally fast in responding to the literal and metaphorical targets,
which were both faster than the latencies to the controls. This was true both when
participants made their lexical decisions immediately after hearing the critical word
(e.g., ladder), and when the same decisions were made 300 ms after hearing the critical word. However, when participants made these same types of lexical decisions to
literal and metaphorical targets having heard less familiar expressions, such as The
thought that a good professor is an oasis was clung to by the entire class, only literal
targets were primed immediately after hearing the critical word (e.g., oasis), while
responses to the metaphorical targets were facilitated only 750 ms after the critical
Again, this study examines dierent aspects of meaning (word vs. phrasal) when it
compares response times to literal (word) and metaphoric (phrasal) targets. The literal target rung is a simple semantic associate of the word ladder, while the metaphoric target advance only relates to the general meaning of the entire expression.
This makes it dicult to conclude anything about the time-course under which literal meanings of an entire sentence are activated compared to gurative meanings of
these expressions (i.e., the hypotheses proposed by the standard pragmatic and the
direct access models). Even if one wishes to reconceive of literal meaning as only
relating to individual word meaning, this study does not allow one to compare
activation of literal word meanings with gurative word meanings. Moreover, the
words used as literal and metaphoric targets do not seem to reect very distinctive
literal and gurative meanings. The literal target rung, for instance, is related to the
idea of advancing (i.e., the gurative target) given that climbing ladders, even literally speaking, is one kind advancing along some physical path.
Many of the criticisms raised here apply to several of the other recent studies on
literal meaning in gurative language understanding. For example, one study presented participants with tape-recorded sentences, such as After the excellent performance, the tennis player was... that could end with either a literal expression, such as
in seventh position, or an idiomatic phrase, such as in seventh heaven (Cacciari and
Tabossi, 1988). The idiomatic phrases were either highly familiar and predictable
(e.g., people could predict that the word following in seventh was heaven) or less
predictable. After hearing the last word in each sentence, the participants made a
lexical decision about a visually presented target (i.e., they decided if the letter string


R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486

presented was a word). These target words could be related to the last word in the sentence just heard in three dierent ways: to the literal meaning of the last word (e.g.,
saint), to the idiomatic sense of the entire sentence (e.g., happy), or unrelated (e.g.,
An analysis of the speeded lexical decisions indicated that when participants heard
predictable idioms, they responded faster to the idiomatic targets than to the literal
targets. But when idioms were less predictable, responses to the idiom targets were
facilitated only after a 300 ms delay, suggesting that the idiomatic meanings took
some time to be activated. Responses to literal targets were facilitated both immediately after hearing the last word of the sentence and also 300 ms later. These data
imply that literal meanings seem to have some priority over gurative ones during
comprehension of less predictable idioms and that the literal meanings of words
remain active during idiom processing even if they are not relevant to the gurative
interpretation of the entire idiom phrase.
What light do these empirical ndings shed on the debate between the standard
pragmatic and direct access views? Cacciari and Tabossi (1988) propose a conguration model suggesting that people process an idiom literally until a key word has
been heard (e.g., heaven). After that, the idiom is processed according to its conventional, gurative meaning. This model sits between the standard pragmatic and
direct access views by not insisting that the complete literal meaning of a phrase be
understood before its idiomatic interpretation is derived. Yet this model assumes
that literal processing is the default mode by which idioms, similar to all language,
are understood.
I have two reactions to this proposal. First, the Cacciari and Tabossi study compared speeded responses to targets reecting the meanings of individual words (i.e.,
literal targets) with responses to targets reecting the gurative meaning of an entire
idiomatic phrase (e.g., the idiom targets). But these targets reect very dierent
levels of meaning (i.e., word versus phrase) and so it remains unclear whether there
are really distinct modes of literal and idiomatic processing that are not confabulated with word and sentence processing mechanisms. The fact that some
aspects of word meaning are accessed immediately during idiom understanding is
not surprising. Yet this study has not fairly tested whether compatible literal and
gurative processes are ongoing at both the word and phrasal level.
The second issue with this study is that the activation of a particular meaning (i.e.,
literal or idiomatic) is assumed to reect the output of entirely dierent linguistic
processes. The possibility remains that activation of dierent kinds of meaning (i.e.,
literal or idiomatic) may really reect dierent types of meaning accessed by a single
linguistic process. The fact that scholars label one kind of meaning as literal and
another idiomatic doesnt necessarily indicate that dierent processes operate
(i.e., a literal processing mode and a idiomatic or gurative processing mode) to
access these meanings (either in a serial or parallel manner). Furthermore, I again
emphasize that none of the work discussed here attempts to dene what is meant by
the term literal meaning either at the word or complete sentence level.
Another way of looking at this issue of equating dierent meanings with dierent
processes is to again consider the experimental work on lexical ambiguity resolution.

R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486


Consider the sentence The engine couldnt stand the constant wear, containing the
polysemous word stand. As several studies (Gibbs et al., 1995), and a quick look in
the dictionary reveal, stand has multiple, related meanings, some of which might be
viewed as literal (e.g., He stands at attention and others gurative (e.g., A one night
stand). One could easily argue about whether it is best to consider physical senses of
stand as literal, and nonphysical uses as gurative. I think this is a very dicult issue
to resolve, as it calls into question the very denitions of literal and gurative
(see Gibbs, 1994, chapter 2 for an extended discussion of this point). Yet even if one
claims that a principled distinction exists between literal and gurative meanings,
the activation of the contextually appropriate meaning(s) of stand in The engine
couldnt stand the constant wear could be a result of the same lexical access process
that gives rise to the more literal or physical senses of stand as in He stands at
attention. Thus, one does not have to posit completely dierent linguistic processes
(i.e., literal vs. gurative) to activate dierent meanings.
I suggest this comparison to the work on lexical ambiguity resolution because of
the tendency among researchers to assume that dierent processes must underlie
activation of dierent types of linguistic meaning. My suggestion is that scholars
resist interpreting the ndings of dierent on-line studies of sentence processing,
including those looking at literal meaning in gurative language understanding, as
necessarily demonstrating dierent linguistic processes. An important consequence
of this idea is that dierences in the activation of literal and gurative meanings
should not be viewed as evidence for the primacy of literal processing in utterance
Another reason to question whether dierent linguistic meanings reect dierent
linguistic processes is the fact that there are numerous, perhaps many dozens of,
types of meaning. For instance, there are many types of gurative meaning, including metaphoric, idiomatic, metonymic, ironic, satirical, proverbial, hyperbolic, oxymoronic, and so on (Gibbs, 1994). Scholars often assume within the context of a
single set of studies that there are two processes at work during gurative language
understanding, such as literal vs. idiomatic, literal vs. metaphoric, or literal vs. ironic. Yet, if there are numerous types of meaning, must there be dozens of types of
linguistic processes all at work, or potentially at work, when language is understood? Psycholinguists have not addressed this question primarily because they
focus too narrowly on only one kind of gurative meaning against a simple view of
literal meaning.
Finally, consider the ndings of a similar set of studies looking at irony comprehension (Giora and Fein, 1999). These studies examined peoples understanding of
familiar (e.g., Very funny) and less familiar (e.g., Thanks for your help) ironies in
comparison to literal uses of the same expressions in appropriate contexts. Participants read stories ending with either literal or ironic remarks. After reading the nal
sentence, they were presented with a letter string and had to quickly respond whether that string was a meaningful word. For instance, after reading the statement
Thanks for your help, they were presented with either an ironic test word (e.g., angry)
or a literal test word (e.g., useful). These test words were presented either 150 ms or
1000 ms after participants read the nal statements.


R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486

The results showed that when people read less familiar ironies they responded
faster to the literal test words than to the ironic test words in the 150 ms condition,
but there were no dierences in the lexical decision times to the literal and ironic test
words after 1000 ms. In contrast, the literal and ironic test words were responded to
equally fast after both 150 and 1000 ms when people read familiar ironies. This
pattern of data suggests that when people read familiar ironies both literal and ironic meanings are quickly accessed, but only literal meanings are initially activated
when people read less familiar ironic statements. Although Giora and Fein (1999)
favor a salience-rst processing model, as opposed to the standard pragmatic
account, their results support the idea that salient meanings, of perhaps both words
and sentences, are always accessed rst.
The problem with this study is identical to that seen in the experiments on idiom
and metaphor comprehension. Literal test words relate to single words within the ironic statements, while the ironic test words only relate to the gurative meanings of the
entire statements. Once again, the experiment incorrectly compares literal word meaning with gurative utterance meaning. At the same time, the ironic test words do not
even specically relate to what the nal ironic remarks guratively mean. Although the
speaker of Thanks for your help might indeed be angry (in the ironic story context), it is
not obvious that the word angry captures much about what the speaker guratively
means (e.g., that the addressee has not been very helpful). Moreover, the literal test
word useful is to some degree related to the ironic meaning in the sense that within
the story context the addressee was not being useful. It is not surprising, then, to see
quicker responses to these literal test words in the 150 ms condition.
My main point is that each of these on-line studies mistakenly assume, that there
is only one kind of literal meaning (i.e., word meaning). These studies fail to distinguish between the literal meanings of words and the literal meanings of entire
phrases or expressions. The fact that some aspects of word meanings are quickly
activated, and consequently responded to faster than the gurative meanings of
entire phrases is not at all unexpected (see Gibbs, 1984 for discussion of this possibility). These studies do not show that people combine word meanings to form literal meanings for an entire expression as an obligatory part of gurative language
interpretation (i.e., the standard pragmatic model). For this reason, the results of
these on-line studies do not directly bear on either the standard pragmatic or direct
access views.
It is still unclear whether the particular words used in the literal target conditions
in the above studies really reect something about literal meaning as distinct from
gurative meaning. We might for the moment still reasonably adopt the position
that some aspects of word meaning are processed during gurative language processing. Yet it is quite a stretch to conclude that language is processed in a literal
manner until some specic, key word triggers a dierent kind of processing (e.g.,
gurative). It may be more accurate to suppose that dierent kinds of meaning are
activated at dierent points in gurative language processing rather than to suppose
that a completely dierent kind of processing mode kicks in, temporarily taking over
the normal, default literal processing. One need not postulate dierent literal and
gurative processing modes to account for any of the data obtained in these studies.

R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486


Debates over literal meaning in understanding what speakers say and mean extend
into empirical research in neuropsychology. Many studies have shown, for example,
that patients with frontal-lobe damage have impaired understanding for ambiguous
language (Pearce et al., 1995), conventional indirect speech acts (e.g., Can you pass
the salt?) (McDonald and van Sommers, 1993), and sarcastic remarks in which the
intended meaning is the opposite of the literal meaning (e.g., What a great football
game) (McDonald, 1992; McDonald and Pearce, 1996). Other studies show that
patients with right hemisphere damage have diculty understanding metaphoric word
meanings (Brownell et al., 1990), proverbs (Van Lancker, 1990), and humor (Bihrle et
al., 1983). These studies generally demonstrate that patients interpret indirect, or gurative, language in terms of literal meanings and do not correctly infer speakers
implicated, communicative messages.
Some neuropsychologists interpret their ndings as providing support for the standard pragmatic view. Thus, patients failure to correctly interpret nonliteral language
suggests a decit in the application of pragmatic knowledge to infer speakers intended
meanings. For instance, one study asked frontal-lobe patients to interpret whether
conversational remarks were sarcastic without any explicit information regarding the
attitude of the speaker (McDonald and Pierce, 1996). The sarcasm was mainly apparent from the counterfactual nature of the remarks. Thus, participants were presented
with the following exchange:
Mark: What a great football game.
Wayne: So you are glad I asked you.
Each remark here could be meaningfully understood as a literal interpretation of
what was said. This is the literally consistent condition, as the second statement was
a literal response to the rst.
In the literally inconsistent condition, however, the second statement was taken
from the alternative sentence pair to represent the antithesis of the expected response
to the rst statement. Participants in this condition saw the following example:
Mark: What a great football game.
Wayne: Sorry I made you come.
Because the two statements are pragmatically contradictory, the only way in
which they can be seen as meaningful is if one or another of the statements is interpreted to be the opposite of what it literally asserts (and thus a sarcastic comment).
After hearing either of these conversational exchanges, participants were asked
four questions:
(a) Did Mark think the game was good?
(b) Did Mark think the game was bad?
(c) Is Wayne pleased that he asked Mark to the game?
(d) Is Wayne sorry that he asked Mark to the game?


R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486

For literally consistent items, answers to the literal meaning of each question was
deemed correct. For the inconsistent items, there were several possible combinations
of answers, depending on which of the statements was regarded as insincere. Thus,
for the second exchange above, if the subject discerned sarcasm because he believed
the rst statement to be insincere, he would respond NYNY (i.e., No, Yes, No, Yes).
If the second statement was processed as the insincere one he would answer YNYN.
If the subject could not reconcile the contradictory nature of the interchange, the
responses would reect the literal interpretations of the meaning (YNNY), or some
other combination representing a confused interpretation of the exchange.
The results showed that frontal lobe patients were able to process literal meanings
(e.g., their reasonable performance on the consistent items), but a signicant proportion of them could not reconcile the inconsistent items in order to detect sarcasm. McDonald and Pierce (1996) argue that the fact that some of the frontal lobe
patients failed the sarcasm task is consistent with the Gricean view that comprehension of sarcasm requires application of inferential processes in order to reinterpret
literal meanings.
I have no doubt that brain-damaged patients often misinterpret speakers communicative messages. The question is whether their apparent focus on the literal
meanings of expressions reects the operation of the standard pragmatic process
where literal meaning is automatically processed rst during nonliteral language
understanding. For instance, the task in the McDonald and Pierce study actually
required people to interpret the expression What a great football game literally as
there was no particular reason to interpret it any other way. Only when participants read the second utterance in the literally inconsistent condition were they
forced to reinterpret, perhaps, their understanding of the rst expression. This kind
of situation where people need to reinterpret some utterance certainly occurs in
some cases. But this task does not really capture what ordinarily happens when
people see sarcastic comments in contexts that may provide crucial clues to the
speakers nonliteral meanings. If anything, then, McDonald and Pierces task looks
at frontal-lobe patients xedness of meaning rather than the priority of literal
meaning in the on-line processing of sarcastic messages. Furthermore, trying to
infer something about peoples rapid, unconscious processing of language from
their conscious interpretations of meaning makes the mistake of conating the
products of understanding with the processes of understanding (Gibbs, 1994). The
fact that people give dierent interpretations to dierent types of language does
not necessarily imply that these meanings are understood via entirely dierent cognitive mechanisms.
Dierent studies looking at brain processes in normal individuals suggests that literal meanings do not have priority when people interpret gurative language. For
instance, Pynte et al. (1996) examined the time-course of literal and metaphorical
processes in metaphor comprehension by measuring event-related potentials (ERPs).
One advantage of measuring ERPs is that recording brain electrical activity when
participants silently read sentences for comprehension minimizes any decision stage
required in reading time experiments (i.e., where participants must push a button
indicating their comprehension of each sentence read). One endogenous component

R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486


of the ERP is the N400 wave, which is known to be a sensitive index of language
processing, especially for word level processes. The ERPs were recorded when participants read short familiar metaphors (e.g., The ghters are lions), unfamiliar
metaphors (e.g., These apprentices are lions), and literal control sentences (e.g.,
Those animals are lions). Participants read these sentences in isolation, and in relevant or irrelevant contexts.
When people read the sentences in isolation, the terminal words of the metaphors
elicited larger N400 components than did terminal words of literal sentences. Pynte
et al. (1996) suggest that the N400 amplitude between metaphors and control sentences indicates that the incongruous, literal meaning of the metaphor has been
accessed during metaphor comprehension. Analysis of the later ERP components
between 6001000 and 10001400 latency bands, which are sensitive to elaboration
and integration processes, revealed no dierence between metaphoric and literal
sentences. However, in later experiments, when participants read the less and more
familiar metaphors in relevant and irrelevants contexts, there were no dierence in
the N400 for the two types of metaphors in relevant contexts, suggesting, according
to Pynte et al. (1996), that in relevant context literal meanings were short-cut and no
longer accessed.
Overall, the ndings from these studies support the context-dependent view of
metaphor comprehension in which direct access to the metaphoric meanings occur
when the meaning is relevant to the preceding context (although the lack of a literal
control condition is problematic in interpreting these results). This conclusion is
essentially consistent with the direct access view in that people do not analyze the
complete literal meanings of expressions as part of their understanding of what
speakers intended to communicate. Nevertheless, these studies, like many of those in
psycholinguistics, assume from the beginning without justication that there must
be two linguistic processes at work: literal and metaphoric. The data collected here
with ERPs are just as easily explained by the view that there is a single interpretive
process that gives rise to various meanings during utterance interpretation.
Although some of these meanings appear to be literal and others metaphoric, this
does not mean that each arises from completely dierent cognitive mechanisms.

5. Falsifying the direct access view

My main argument has been that recent research in psycholinguistics and neurolinguistics provides little support for the standard pragmatic view and is most consistent with the direct access view of gurative language understanding. The
occasional ndings that some gurative utterances may take longer to process than
literal expressions is by itself insucient to support the standard pragmatic model,
or reject the direct access view. Similarly, showing that people access aspects of what
might be referred to as the literal meanings of words in gurative statements also
does not imply that the direct access view is wrong. Falsifying the direct access view
requires some demonstration that the analysis of a sentences complete literal
meaning is responsible for the extra time needed to comprehend gurative language.


R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486

There are only two sets of studies that include both of these elements in experimental tests of the direct access view (Gibbs, 1983, 1986c). Participants in one set of
these studies read stories that ended with sentences intended either as indirect
requests or as literal questions (Gibbs, 1983). After reading each story, the participants were presented with a word string. Their task was to decide whether the word
string constituted a meaningful English sentence. Presented below is an example of
two stories followed by each of its possible target strings.
Literal context
Martin was talking with his psychiatrist.
He was having many problems with relationships.
He always seemed hostile to other people.
Martin commented to the psychiatrist,
Everyone I meet I seem to alienate.
The shrink said,
Cant you be friendly?
Literal: Are you unable to be friendly?
Indirect: Please be friendly to other people.
Indirect request context
Mrs. Connor was watching her kids play in the backyard.
One of the neighbors children had come over to play.
But Mrs. Connors son refused to share his toys.
This made Mrs. Connor upset.
She angrily walked outside and said in a stern voice to her son,
Cant you be friendly?
Indirect: Please be friendly to other people.
Literal: Are you unable to be friendly?
The indirect meanings of these stories last lines can be viewed as their conventional
interpretations. When understanding the literal, nonconventional uses of these sentences (i.e., the literal question), participants may analyze the conventional request
meaning of these expressions before deciding that the nonconventional, literal meanings are appropriate. When reading literal questions, participants responses to both the
literal and conventional targets should be fast relative to the time it takes to respond to
unrelated targets. However, participants response times for literal targets should be
slow if they do not ordinarily analyze the literal interpretations of indirect requests.
The results rst showed that people take less time to read indirect requests than
literal uses of the same expressions. This nding replicates Gibbs (1979). The results
of this study showed that when people read indirect requests they were much faster
to make the sentence classication responses for indirect targets than for literal ones.
Moreover, there was no signicant dierence in response times for the literal and
unrelated targets when participants read indirect requests. These data indicate that
there was no residual left from the participants processing of the literal meanings of

R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486


the indirect requests that subsequently facilitated their responses to the literal targets. This suggests that people do not necessarily analyze the literal interpretation of
an indirect speech act during their immediate comprehension. Similar ndings have
been reported for idioms (Gibbs, 1986c).
These studies on idioms and indirect requests include both the speeded responsetimes and complete literal meanings as test phrases that are essential factors in
examining the predictions of the standard pragmatic and direct access views. Moreover, as suggested in Gibbs (1983, 1986c), these data are also inconsistent with
claims that the literal and gurative, or indirect, meanings of idioms or indirect
request are processed in parallel, given that no activation of the idioms or indirect
requests literal meanings was found in these studies (although note above my criticism of the dierent meanings equals dierent linguistic processes idea).
Despite my belief that the Gibbs (1983, 1986c) studies generally represent the right
way of trying to falsify the direct access view, I am not convinced that the specic
method and stimuli used there are completely satisfactory. First, these studies only
examined conventional aspects of gurative language. This is not by itself problematic as the standard pragmatic model applies to all kinds of gurative and indirect
speech. Yet more studies should indeed look at whether the complete literal meanings of novel gurative utterances play some role in their comprehension. One
immediate diculty with studying novel aspects of gurative language is that many
novel metaphors, for instance, are literally anomalous (i.e., My criticism is a branding iron), making it impossible to construct meaningful literal targets for the phrase
classication task. But it might be possible to conduct similar experiments with
novel proverbs which are literally meaningful.
A second problem with the above studies is that the task of asking people to judge
the meaningfulness of an entire sentence, after reading an idiom or indirect request,
may involve post hoc cognitive processes that have little to do with what occurs
when people read the previous statements (i.e., the idioms or indirect requests).
Finally, the design of these studies didnt suciently include conditions to examine
whether the priming eects were due to facilitation of gurative meaning as opposed
to inhibition of literal meaning.
Clearly, there is much more work to be done. Yet, my point still stands that the
direct access view can only be falsied if both processing times are compared for
literal and gurative utterances and it can be shown that the extra time used to
process gurative language is specically due to activation, and then rejection, of a
sentences complete literal meaning. Even if some data are obtained in the appropriate experimental situation, we still need to better understand if literal meaning is
actually rejected, or if a sentences literal meaning sometimes plays a positive role in
interpreting a speakers gurative meaning.

6. Figurative foundation of literal meaning

My claim is that without a better idea of what constitutes literal meaning at both
the word and sentence level, it will be dicult to interpret the ndings of these newer


R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486

studies looking at literal meaning in gurative language interpretation, especially in

regard to the debate over the standard pragmatic and direct access views.
Another important point to remember, or realize for the rst time, is that many
aspects of what might be termed literal speech have their roots in gurative thought
and language. As cognitive linguists have demonstrated, for example, dierent gurative schemes of thought, most notably metaphor and metonymy, underlie systematic
patterns of conventional linguistic expressions, historical development of word and
phrase meanings, and novel linguistic expressions (Lako and Johnson, 1980; Johnson,
1987; Lako, 1987; Lako and Turner, 1989; Sweetser, 1990). A great amount of
experimental research in psycholinguistics shows that people have tacit knowledge of
conceptual metaphors, for instance, that partly motivates how they learn, make
sense of, and comprehend dierent linguistic expressions (Gibbs, 1994). These studies collectively call into question any simple distinction between literal and gurative
meanings and emphasize the gurative nature of much language that people, at
times, may refer to as literal. Simply referring to some pieces of language as literal and others as gurative, or whatever other tropes may be of interest, does
not empirically establish that literal meanings are somehow dierent than gurative
meanings, or are produced and understood by dierent cognitive mechanisms. My
continued plea is that scholars not refer to any linguistic expression as literal
unless theoretical reasons can be clearly stated as to what makes this type of meaning dierent from all other kinds of meaning (e.g., metaphoric, metonymic, poetic,
and so on). I raise this issue here as an important challenge for researchers looking
at the role of so-called literal meanings in understanding what speakers say and

7. The problem of what speakers say

The original proposals on literal meaning in indirect and gurative language
understanding assumed that an analysis of what speakers literally say is in many
cases an important part of interpreting what speakers implicate in context. Consider
again the exchange between two college students:
Steve: Are you going to the big party this weekend?
Beth: Didnt you hear that Bob is going to be there?
Under the traditional Gricean view, what Beth says here only conveys part of the
meaning she wishes to communicate. Thus, what speakers say is indeed part of what
listeners must understand, yet understanding what is said alone is insucient. Listeners presumably take what speakers literally say and expand upon it by applying
enriched pragmatic knowledge to draw a conversational implicature about what the
speaker really meant. This view assumes that what a speaker says is equivalent to an
utterances semantic or literal meaning. Most linguists and psycholinguists maintain
a similar belief about the close, if not isomorphic, relationship between literal
meaning and what speakers say. Many scholars, following Grice, maintain that

R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486


pragmatics play at best only a small part in determining what speakers literally say.
My earlier empirical work on indirect speech acts (Gibbs, 1979, 1983), idioms (Gibbs,
1980, 1986a), and sarcasm (Gibbs, 1986b, 1986c) was consistent with this view in characterizing literal meaning as equivalent to context-free, semantic meaning. To some
degree, the most recent psycholinguistic studies described above also implicitly adhered
to the distinction between context-free literal meaning and pragmatically determined
gurative meaning.
I now argue that this view of literal meaning as equivalent to said or semantic
meaning is incorrect. Signicant aspects of what speakers say, and not just what they
totally communicate, are deeply dependent upon enriched pragmatic knowledge.
Under this revised view, people may indeed analyze aspects of what speakers say as
part of understanding what speakers conversationally implicate.

8. A revised view of what speakers say

In recent years, several linguists and philosophers have persuasively argued that
the traditional, Gricean view ignores the fact that essentially the same sorts of
inferential processes used to determine conversational implicatures also enter into
determining what is speakers say (Sperber and Wilson, 1986; Carston, 1988, 1993;
Recanati, 1989, 1993; Wilson and Sperber, 1993;). Consider a case where a speaker
says to you I havent eaten. In this case, at least once the indexical references and the
time of the utterance are xed, the literal meaning of the sentence determines a
denite proposition, with a denite truth condition, which can be expressed as The
speaker has not eaten prior to the time of the utterance. This paraphrase reects
the minimal proposition expressed by I havent eaten (Recanati, 1989). However, a
speaker of I havent eaten is likely to be communicating not a minimal proposition,
but some pragmatic expansion of it, such as I havent eaten today. This possibility
suggests that signicant pragmatic knowledge plays a role in enabling listeners to
expand upon the minimal proposition expressed to recover an enriched pragmatic
understanding of what a speaker says.
Gibbs and Moise (1997) demonstrated in several experimental studies that pragmatics plays a major role in peoples intuitions of what speakers say (for further
discussion of these ndings, see Gibbs, 1999; Nicole and Clark, 1999). Consider the
expression Jane has three children. According the Gricean view, the interpretation that
Jane has exactly three children comes from applying specic pragmatic information
to the minimally-pragmatic proposition of what is said (e.g., Jane has at least three
children), a process that results in what Grice referred to as a generalized conversational implicature (i.e., implicatures that are normally drawn regardless of the context). But we showed, in a series of experiments looking at students intuitions about
what speakers say, that people do not equate the minimal meaning with what a
speaker says. A rst study asked participants to choose between two paraphrases of
what a speaker says by dierent indicative utterances (e.g., cardinal sentences like
Jane has three children, possession sentences like Robert broke a nger yesterday,
scalar sentences like Everybody went to San Francisco, time-distance sentences like It


R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486

will take us some time to get there, and temporal relation sentences like Amy bought a
new dress and she went out dancing). Each of these sentences are traditionally viewed
as conveying generalized conversational implicatures. Participants chose signicantly more enriched pragmatic paraphrases of what speakers say (e.g., Jane
has exactly three children), than they did paraphrases that were minimally pragmatic (e.g., Jane has at least three children and may have more than three). A
second study revealed that even when alerted to the Gricean position (i.e., what is
said is equivalent to the minimal proposition expressed), people still reply that enriched pragmatics is part of their interpretation of what a speaker says and not just
what the speaker implicates in context.
The fact that people prefer enriched pragmatic paraphrases for what speakers say
does not mean that they are unable to distinguish between what speakers say and
what they implicate. The ndings of another study reported in Gibbs and Moise
(1997) demonstrated that people recognize a distinction between what speakers say,
or what is said, and what speakers implicate in particular contexts. For instance,
consider the following story:
Bill wanted to date his co-worker Jane.
Being rather shy and not knowing Jane very well,
Bill asked his friend, Steve, about Jane.
Bill didnt even know if Jane was married or not.
When Bill asked Steve about this, Steve replied
Jane has three children.
What does Steve say and what does he implicate by his utterance? Steve implicates
by his statement Jane has three children in this context that Jane is already married. To the extent that people can understand what Steve says, but not implicates,
by Jane has three children, they should be able to distinguish between the enriched
and implicated paraphrases of the nal expressions. The results of one study showed
this to be true. When participants were asked to choose the best paraphrase of what a
speaker says in a context like the above one, they chose one that reected the enriched
pragmatic meaning (i.e., Jane has exactly three children) and not implicature
paraphrases (i.e., Jane is married). These ndings show that pragmatics strongly
inuences peoples understanding of what speakers both say and communicate. It
appears that Grices examples of generalized conversational implicatures are not
implicatures at all but understood as part of what speakers say. More generally, the
Gibbs and Moise (1997) ndings suggest that the distinction between saying and
implicating is orthogonal to the division between semantics and pragmatics.
These data suggest that dierent aspects of pragmatics may be accessed when
people understand what speakers say and implicate. One possibility is that there are
two kinds of pragmatic information or knowledge, primary and secondary, that
become activated during normal language understanding (Recanati, 1993; Gibbs
and Moise, 1997). Primary pragmatic knowledge applies deep, default background
knowledge to provide an interpretation of what speakers say. Under this view, primary pragmatic knowledge relates to deeply held, perhaps non-representational

R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486


(Searle, 1983) knowledge that is so widely shared as to seem invisible. To take a

classic example (Searle, 1978), our interpretation of the expression The cat is on the
mat presupposes an enumerable set of assumptions, such as that the cat chose for
some reason to sit on the mat, and that the cat and mat are on the ground operating
under the constraints of physical laws like gravity and are not oating in space in
such a way that the cat is on the mat by virtue of touching the underneath part of
the mat as in The y is on the ceiling. Our ability to infer what speakers say when
uttering any word or expression rests, to a large part, on deeply held background
knowledge that is very much part of our pragmatic understanding of the world.
Secondary pragmatic knowledge, on the other hand, refers to information from
context that provides an interpretation of what speakers implicate in discourse. For
instance, a speaker who utters The cat is on the mat might implicate that the addressee
should get up and let the cat outside. Listeners draw the appropriate inferences about
what speakers intend by recognizing specic features of the local context based on the
common ground between themselves and speakers (i.e., their mutual beliefs, attitudes,
knowledge). Thus, a speaker and listener may have as part of their common ground
that the cat usually desires to go outside when it sits on the mat by the front door.
Overall, though, listeners stereotypical background knowledge dominates the
application of secondary pragmatic information to reveal what is said by a speakers
utterance as distinct from what the speaker implicates (See Recanati, 1993).
Note that I have not assumed primary and secondary pragmatic knowledge
necessarily reects dierent cognitive processes (Recanati, 1993, describes these as
processes). Moreover, it is dicult, if not impossible, to predict in advance which
pragmatic knowledge is best viewed as primary and prominent in understanding what
speakers say and what pragmatic knowledge is secondary, that is, essential for inferring what speakers conversationally implicate. One possibility is that primary pragmatic information is more salient and accessed more quickly than more elaborate,
local, secondary pragmatic information (cf. Recanati, 1993). Very recent experimental evidence lends credence to this idea.
Two studies by Hamblin and Gibbs (2000) examined the speed with which people
understand expressions in which speakers communicative intentions were either
identical to what they pragmatically said, or varied in some way, thus requiring listeners/readers to derive a conversational implicature. Consider the following stories,
each of which end with the same sentence:
Said/implied identical
Ted and Michele ran into each other at the mall.
Ted asked Michele what she had been doing lately.
Michele said that she had been busy car shopping.
Looking for ideas, Michele decided to consult Ted.
Michele asked Ted about his own car.
Ted mentioned:
I drive a sports utility vehicle. (enriched pragmatic meaning)
Said/implied dierent


R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486

Ted and Michele are planning a trip to Lake Tahoe.

Michele had heard that there was a terrible storm there.
She wondered if it was going to be safe for them to go.
Michele was concerned about the vehicle they would drive.
She asked Ted if he thought they would be okay.
Ted replied:
I drive a sports utility vehicle. (implicature)
In the rst context, what the speaker pragmatically says by I drive a sports utility
vehicle is identical to what he implies in that there is no further pragmatic meaning
he wishes for listeners to infer beyond that he drives a particular kind of car. But in
the second context, the speaker not only says one thing (i.e., about the kind of car he
drives), but also implies something beyond that meaning, namely that his particular
car is safe to drive in a storm.
If people access primary pragmatic information sooner than they do secondary
pragmatic knowledge, readers should take less time to comprehend utterances in which
what speakers mean is identical to what they pragmatically say than to understand
messages in which what speakers say underdetermines what they mean. In the rst
study, twenty-four undergraduate students read stories like the above (there were a total
of 24), line-by-line on a computer screen, pushing a button as they read and understood
each line. We measured the amount of time participants took to read each line (starting
from when the sentence rst appeared on the screen and ending when the participants
pushed the comprehension button). We were specically interested in how long it took
readers to understand the last line of each story. An individual participant saw only one
of the two stories presented above, but across all the conditions, an equal number of
participants saw the said/implied identical and the said/implied dierent stories.
The results showed that people took longer on average to understand what
speakers meant when their utterances varied from what they simply said (1751 ms)
than when speakers meant only what they pragmatically said (an explicature) (1604
ms). These ndings demonstrate that drawing conversational implicatures increases
processing eort beyond the eort needed to understand what speakers say. Moreover, the data are consistent with the idea that people analyze what speakers say as
part of their determination of what speakers imply.
A second experiment investigated processing of what speakers say and imply in a
dierent way. Consider the following story, and two dierent nal expressions:
Bill is a new tenant in an apartment building.
His neighbor Jack has lived there for four years.
Bill was concerned that the building might be too loud.
Bill decided to ask a neighbor about it.
Bill asked Jack since he was the only neighbor Bill had met.
Jack replied,
This is a very noisy building. (said/implied identical)
I usually sleep with earplugs. (said/implied dierent)

R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486


Understanding I usually sleep with earplugs demands that listeners draw a pragmatic inference beyond that needed to understand what this same expression pragmatically says. However, understanding This is a very noisy building in this context
only requires listeners/readers to comprehend what the speaker pragmatically said.
For this reason, participants should take less time to read This is a very noisy building than I usually sleep with earplugs in this context. Twenty-four undergraduate
students participated in this study.
Across all the stimuli in this study, people took signicantly more time to read
sentences necessitating the implicatures (1661 ms) than they did the sentences
requiring only enriched pragmatic said meanings (1511 ms). Once again, it appears
that people more easily understand speakers messages when these are identical to
what they pragmatically say than when what is said undeterdetermines what the
speakers intend to communicate (i.e., conversational implicatures). The data from
both studies suggest that people may actually process what speakers pragmatically
say as an automatic part of their interpretation of what speakers communicate.
My main conclusion from this set of reading-time experiments is that pragmatics
is not simply used in understanding speakers intended meaning, but plays a role in
utterance interpretation from the earliest stages of linguistic processing. In this
sense, Grice was right in suggesting that people may analyze what speakers say
before inferring what they implicate. But Grice and others are incorrect in assuming
that understanding what speakers say refers to minimally-pragmatic meaning and
that enriched pragmatics only has a role in deriving conversational implicatures.

9. What is said and gurative language understanding

How do we reconcile the new view on the pragmatics of what is said with the extensive data from psycholinguistics that people can quickly understand many instances of
gurative language without having to rst analyze the entire literal meanings of these
statements? Does not the fact that people can understand the meanings of metaphors
and ironies, for instance, more quickly than when these same statements are used
literally, argue against the idea that inferring what speakers imply takes longer than
understanding what they simply say?
At rst glance, the ndings of Gibbs and Moise (1997) and Hamblin and Gibbs
(2000) appear to support the standard pragmatic model in which the literal-said
meaning is processed rst before enriched pragmatic knowledge is used to infer what
speakers implicate. My argument, however, is that these empirical ndings support a
revised view of speaker meaning in which listeners analyze what speakers pragmatically say as a critical part of the understanding process. Nevertheless, understanding
what speakers pragmatically say is not all identical to the putative, non-pragmatic
literal meanings of speakers utterances.
Under what conditions will listeners analyses of what speakers say demand more
cognitive eort to understand what speakers mean to communicate? There are several possibilities worth considering. The rst point to note is that the conventionality
of what speakers say will have a facilitatory inuence in understanding what they


R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486

imply. Most studies on gurative language understanding, for instance, only examine
interpretation of conventional language. People nd highly conventionalized uses of
metaphors, idioms, indirect speech acts, etc. very easy to understand. On the other
hand, understanding that I usually wear earplugs communicates The building is noisy
requires listeners or readers to draw a novel inference about the relation between a
statement about the speakers use of earplugs and the topic of the conversation, namely,
whether or not the building is noisy. Inferring this kind of conversational implicature
might, indeed, take people longer to do than to comprehend conveyed meanings of
conventional indirect and gurative language. Although it is certainly true that ones
familiarity with well-known gurative expressions facilitates ones processing of
these statements, several studies also demonstrate that people can process novel
instances of metaphor and irony in context as quickly as they do literal uses of the
same statements (or when what is said and meant are identical) (Gibbs, 1994).
A second explanation for the possible discrepancy between the ndings of the
Hamblin and Gibbs (2000) reading time studies and the previous research on gurative language understanding is that some aspects of gurative meaning are
understood as part of what speakers say and others as part of what speakers implicate. For instance, several linguists have argued that the nonliteral meanings of
certain indirect speech acts (e.g., Can you pass the salt?), metonymies (e.g., The buses
are on strike), and ironies (e.g., Youre a ne friend) are understood as part of our
interpretation of what a speaker says, called explicatures (Groefsema, 1992; Papafragou, 1996), and not derived as conversational implicatures. Under this view, the
context-appropriate meaning of, say, an ironic remark is completely captured by
understanding what the speaker pragmatically said. The listener only needs to recognize how a speakers utterance reects another thought attributed to somebody else
(Papafragou, 1996). Recognizing that what a speaker says echoes some other
thought, or previous utterance, conveys the speakers attitute of amused rejection of
this thought. All of the assumptions needed to infer this aspect of speaker meaning
are understood as part of what is said (or as explicatures).
There are certainly cases of irony that demand further elaboration of what
speakers say to infer their intended communicative messages. For instance, if I say
to a passenger in my car love drivers who signal before turning right after some other
driver has cut in front of me without signaling, the listener will likely need to expand
on what I have said to correctly infer my ironic meaning. There may not be a hardand-fast rule that determines which kind of irony is understood as an explicature,
and which as an implicature. Yet there is sucient pragmatic information, perhaps
part of peoples deep background knowledge, that allows them in some cases to
quickly infer some gurative meanings without having to apply very local, contextually-specic, pragmatic information. This possibility certainly ts in with the
empirical results showing that people can easily comprehend many kinds of gurative language (Gibbs, 1994). But no research, thus far, has explicitly attempted to
link quick processing of gurative language with the new view on the pragmatics of
what is said.
A nal point regarding the various empirical ndings focuses on the dierent roles
that what is said plays in understanding gurative language as opposed to the

R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486


indicative utterances studied in the Gibbs and Moise (1997) and the Hamblin and
Gibbs (2000) experiments. Consider again the utterance Jane has three children when
used to communicate that Jane is married. Understanding what is implicated by
this utterance is accomplished by virtue of our recognition of the pragmatically said
interpretation that Jane has only three children. Many of the novel proverbs studied
by Temple and Honeck (1998) may also be understood in this manner. On the other
hand, understanding the intended meanings of many metaphors and idioms is
accomplished in spite of what these expressions specically say. In many cases of
gurative language understanding, processing what a speaker says is short-circuited
in favor of processing what that utterance is intended to communicate in context.
For example, our understanding of the metaphorical expression The old rock was
brittle with age, stated by one student to another in reference to an elderly professor,
might not require that we rst determine what the speaker specically says. Instead,
the normal process of referential assignment when reading the phrase the old rock
prompts people to quickly seek an alternative gurative meaning that makes sense in
the discourse situation. In some specialized and highly available contexts, the metaphoric interpretation is accessed rst. This quick search for nonliteral meanings in
context provides one main reason why metaphorical utterances can be understood
as fast as, if not faster, than literal uses of the same expressions. In other cases,
understanding what a speaker says will lead us to draw further gurative inferences
as implicatures. Just as a speaker might say Jane has three children to imply that
Jane is married, a speaker might say I love drivers who signal before changing lanes
to ironically implicate that I hate the driver who just switched lanes without signaling. What a speaker says in both of these instances underdetermines what he or
she wants to communicate.
Understanding what speakers actually intend requires that we elaborate on the
pragmatic interpretation of what is said by applying secondary pragmatic information to infer what he or she really implicates. Drawing inferences about what
speakers guratively communicate beyond what they pragmatically say may, under
some circumstances, take additional processing eort. Interestingly, there are occasions when understanding what someone says automatically leads one to infer a
gurative meaning even if the speaker didnt necessarily intend that gurative
meaning to be communicated. For instance, when someone is literally skating on thin
ice, he also guratively is in a precarious situation. Thus, listeners can draw
inferences from something a speaker says to derive a gurative meaning, and this
process takes longer than if people simply understood the phrase skating on thin
ice as having only gurative, idiomatic meaning (Gibbs, 1986a).
There is clearly much further empirical work needed to look more closely at the
role that pragmatics has in understanding what speakers say and implicate by their
use of both expressions such as Jane has three children and dierent aspects of gurative speech. Understanding how dierent aspects of pragmatics interact with
dierent types of linguistic information may provide essential clues to characterizing
peoples on-line comprehension of pragmatic meaning. It may be that some aspects
of indirect language are understood as part of what speakers say, while others are
understood as part of what speakers implicate in specic discourse contexts. People


R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486

may sometimes construct representations of what speakers say as part of, or even
before, what they understand speakers implicate, even if these said meanings are
not related to traditional views of literal meaning.

10. Conclusion
For several decades now, language theorists and researchers from many disciplines have argued about the possible role that literal meaning plays in understanding what speakers intend to communicate. My comprehensive reviews of the
psycholinguistic literature back in 1984 and 1994 led me to conclude that little evidence
supported the idea that people analyzed the literal meanings of nonliteral utterances
during their ordinary interpretations of these expressions (Gibbs, 1994). There have
been various responses to this conclusion from both philosophers (Dascal, 1987) and
psycholinguists. Many psycholinguists, in particular, have quite recently attempted to
show that literal meaning is indeed automatically analyzed when people process gurative language. As I have argued in this article, many of these experimental studies
conate aspects of literal and nonliteral meanings and often confuse what occurs during processing of lexical meaning with what occurs when entire utterances are
interpreted. For these reasons, I claim, once again, that little empirical evidence
exists to support the idea that people process the complete literal meanings of
utterances either before or in parallel to understanding what speakers communicate
by indirect, conveyed, or gurative language. Certain philosophical arguments are
consistent with this perspective on linguistic understanding (Recanati, 1995).
At the same time, it is still entirely unclear even whether people automatically
analyze the literal meanings of individual words during on-line utterance interpretation. The fact that people clearly analyze something about what words mean in
immediate utterance comprehension does not imply that the meanings of the words
activated are necessarily their literal meanings. Without a better idea of what
constitutes literalness at both the word and sentence level, it seems safer not to
assume that processing of literal meanings constitutes the default mode of linguistic understanding. I suggest that the pursuit of literal meaning in theories of
linguistic meaning and understanding is a fruitless exercise, especially when one is
interested in exploring how people ordinarily produce and understand language.
Despite these criticisms of literal meaning in understanding what speakers communicate, there is now good empirical evidence to suggest that people (a) can distinguish between what speakers say and implicate, (b) that understanding what
speakers say and implicate both involve enriched pragmatic knowledge, and (c) that
people may indeed ordinarily process what speakers pragmatically say as part of
their understanding what speakers implicate. Under this new view of speaker
meaning, what are traditionally viewed as generalized implicatures are really part of
the retrieval of what is said. The Gricean distinction between generalized implicatures and what is said is, therefore, unnecessary. Moreover, the distinction
between what is said and what is implicated is orthogonal to the putative distinction
between semantics and pragmatics. According to Kaplan (1989), a semantic theory

R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486


must be grounded in speakers intuitions about what is said (see Cappelan and
Lepore, 1997 for critical discussion of this claim). Yet, if peoples intuitions about
what speakers say involve enriched pragmatic knowledge, then the link between
semantics and what is said has little empirical validity. Most generally, my thesis is
that pragmatic information pervades all aspects of utterance interpretation.
Some scholars may now argue that literal meaning provides the foundation for
determining what is pragmatically said even if what is said and literal meaning can
no longer be equated. I see no reason at this point to agree with this claim. No
empirical evidence from psycholinguistics exists to show that there is some canonical, non-pragmatic meaning that is automatically analyzed at both the word and
sentence level, which in turn feeds into a higher-level pragmatic processor used to
interpret what speakers pragmatically say and implicate. At the very least, before
any claims can be made in favor of literal meaning in normal utterance interpretation, scholars will need to be much more explicit about what constitutes literal
meaning than they have been up till now.
My earlier writings on literal meaning in utterance interpretation were criticized
by some scholars on the grounds that people do not infer speakers intended messages by context alone, and must do something with the actual words heard or read.
Moreover, there must be occasions when some aspect of what speakers say species
part of what they actually implicate in context. Dascal (1987), for instance, argued
for the thesis of moderate literalism to capture some of the pragmatic aspects of what
people say as part of what constitutes literal meaning. Although I see no reason to
posit a level of literal analysis in a theory of utterance interpretation, I now agree
that people distinguish between what speakers say and implicate, that both aspects
of speaker meaning involve substantial pragmatic knowledge, and that people may
analyze what speakers pragmatically say as part of their understanding of what
speakers imply. There is much theoretical and empirical work left to be done on how
dierent aspects of pragmatics shape understanding of what speakers say and
implicate, as well as a need for experimental research examining understanding of
speakers utterances in which what is pragmatically said and implied vary from one
another in dierent ways.

I thank Mira Ariel and Rachel Giora for their helpful comments on earlier drafts
of this paper.

Blasko, Dawn and Cynthia Connine, 1993. Eects of familiarity and aptness on metaphor processing.
Journal of Experimental Psychology: Learning, Memory, and Cognition 19: 295308.
Bihrle, Amy, Hiram Brownell, John Powelson and Howard Gardner, 1986. Comprehension of humorous
and non-humorous materials by left and right brain-damaged patients. Brain and Cognition 5: 399411.


R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486

Brownell, Hiram, Dee Michel, John Powelson and Howard Gardner, 1983. Surprise but not coherence:
Sensitivity to verbal humor in right-hemisphere patients. Brain and Language 18: 2027.
Brownell, Hiram, Tracy Simpson, Amy Bihrle, Heather Potter and Howard Gardner, 1990. Appreciation
of metaphoric alternative word meanings by left and right brain-damaged patients. Neuropsychologia
28: 375384.
Burgess, Curt, Michael Tannenhaus and Mark Seidenberg, 1989. Context and lexical access: Implications
of nonword interference for lexical ambiguity resolution. Journal of Experimental Psychology: Learning, Memory, and Cognition 15: 620632.
Cacciari, Cristina and Patrizia Tabossi, 1988. The comprehension of idioms. Journal of Memory and
Language 27: 668683.
Cappelan, Herman and Ernest Lepore, 1997. On the alleged connection between indirect speech and the
theory of meaning. Mind and Language 12: 278296.
Carston, Robyn, 1988. Implicature, explicature, and truth-theoretic semantics. In: Ruth Kempson, ed.,
Mental representations: The interface between language and reality, 155181. Cambridge: Cambridge
University Press.
Carston, Robyn, 1993. Conjunction, explanation, and relevance. Lingua 90: 2748.
Dascal, Marcelo, 1987. Defending literal meaning. Cognitive Science 11: 259281.
Frege, Gottlob, 1952. On sense and reference. In: Max Black, Peter Greach, eds., Philosophical writings
of Gottlob Frege, 5678. Oxford: Blackwell.
Gibbs, Raymond, 1979. Contextual eects in understanding indirect requests. Discourse Processes 2: 110.
Gibbs, Raymond, 1980. Spilling the beans on understanding and memory for idioms in conversation.
Memory and Cognition 8: 449456.
Gibbs, Raymond, 1983. Do people always process the literal meanings of indirect requests?. Journal of
Experimental Psychology: Learning, Memory and Cognition 9: 524533.
Gibbs, Raymond, 1984. Literal meaning and psychological theory. Cognitive Science 8: 275304.
Gibbs, Raymond, 1986a. Skating on thin ice: Literal meaning and understanding idioms in conversation.
Discourse Processes 9: 1730.
Gibbs, Raymond, 1986b. On the psycholinguistics of sarcasm. Journal of Experimental Psychology:
General 115: 315.
Gibbs, Raymond, 1986c. Comprehension and memory for nonliteral utterances: The problem of sarcastic
indirect requests. Acta Psychologica 62: 4157.
Gibbs, Raymond, 1989. Understanding and literal meaning. Cognitive Science 13: 243251.
Gibbs, Raymond, 1994. The poetics of mind: Figurative thought, language, and understanding. Cambridge: Cambridge University Press.
Gibbs, Raymond, 1999. Speakers intuitions and pragmatic theory. Cognition 69: 355359.
Gibbs, Raymond, in press. Metarepresentations in staged communicative acts. In: D. Sperber, ed.,
Metarepresentations. New York: Oxford University Press.
Gibbs, Raymond and Jessica Moise, 1997. Pragmatics in understanding what is said. Cognition 62: 5174.
Gibbs, Raymond, Dinara Beitel, Michael Harrington and Paul Sanders, 1995. Taking a stand on the
meanings of stand: Bodily experience as motivation for polysemy. Journal of Semantics 11: 231251.
Gibbs, Raymond, Darin Buchalter, Jessica Moise and William Farrar, 1993. Literal meaning and gurative language. Discourse Processes 16: 387403.
Giora, Rachel, 1995. On irony and negation. Discourse Processes 19: 239264.
Giora, Rachel, 1997. Understanding gurative and literal language: The graded salience hypothesis.
Cognitive Linguistics 7: 183206.
Giora, Rachel and Ofer Fein, 1999. Irony: Context and salience. Metaphor and Symbol 14: 241257.
Giora, Rachel, Ofer Fein and Tamir Schwartz, 1998. Irony: Graded salience and indirect negation.
Metaphor and Symbol 13: 83101.
Glucksberg, Sam., 1998. Metaphor. Current Directions in Psychological Science 7: 3943.
Glucksberg, Sam, Roger Kreuz and Susan Rho, 1986. Context constrains lexical access: Implications for
models of linguistic analysis. Journal of Experimental Psychology: Learning, Memory and Cognition
12: 323335.
Grice, H. Paul, 1975. Logic and conversation. In: Jerry Morgan and Peter Cole, eds., Syntax and
semantics 3: Speech acts, 4158. New York: Academic Press.

R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486


Grice, H. Paul, 1989. Studies in the way of words. Cambridge: Harvard University Press.
Groefsema, Marjoelen, 1992. Can you pass the salta short-circuited implicature. Lingua 87: 103135.
Hamblin, Jennifer, and Raymond Gibbs. in press. Processing the meanings of what speakers say and
implicate. Manuscript submitted for publication.
Honeck, Robert, 1997. A proverb in mind: The cognitive science of proverbial wit and wisdom. Mahwah,
NJ: Erlbaum.
Johnson, Mark, 1987. The body in the mind. Chicargo: University of Chicago Press.
Kaplan, David, 1989. Afterthoughts. In: Howard Wettstein, John Perry, eds., Themes from Kaplan, 565
614. Oxford: Oxford University Press.
Katz, Albert and Christopher Lee, 1993. The role of authorial intent in determining verbal irony and
metaphor. Metaphor and Symbolic Activity 8: 257279.
Katz, Albert and Penny Pexman, 1997. Interpreting gurative statements: Speaker occupation changes
metaphor to irony. Metaphor and Symbol 12: 1941.
Katz, Albert, Cristina Cacciari, Raymond Gibbs and Mark Turner, 1998. Figurative language and
thought. New York: Oxford University Press.
Kemper, Susan, 1981. Comprehension and the interpretation of proverbs. Journal of Psycholinguistic
Research 10: 179198.
Lako, George, 1987. Women, re, and dangerous things: What categories reveal about the mind. University of Chicago Press Chicago.
Lako, George and Mark Johnson, 1980. Metaphors we live by. Chicago: University of Chicago Press.
Lako, George and Mark Turner, 1989. More than cool reason: A eld guide to poetic metaphor. Chicago: The University of Chicago Press.
Lucariello, Joan, 1994. Situational irony: A concept of events gone awry. Journal of Experimental Psychology 123: 129145.
McDonald, Skye, 1992. Dierential pragmatic loss following closed head injury: Ability to comprehend
conversational implicature. Applied Psycholinguistics 13: 295312.
McDonald, Skye and Samantha Pearce, 1996. Clinical insights into pragmatic language theory: The case
of sarcasm. Brain and Language 53: 81104.
McDonald, Skye and Peter van Sommers, 1993. Dierential pragmatic language loss following closed
head injury: Ability to negotiate requests. Cognitive Neuropsychology 10: 297315.
Pearce, Samantha, Skye McDonald and Max Coltheart, 1995. Interpreting ambiguous advertisements:
The eects of frontal lobe damage. Abstracts of the 2nd INS Pacic Rim Conference, Cairns, Australia.
Journal of the International Neuropsychological Society 1: 342344.
Papafragou, Anna, 1996. On metonymy. Lingua 94: 169195.
Pynte, Joel, Mireille Besson, Fabrice-Henri Robinchon and Jezabel Poli, 1996. The time-course of metaphor comprehension: An event-related potential study. Brain and Language 35: 293316.
Nicole, Steve and Billy Clark, 1999. Experimental pragmatics and what is said: A reply to Gibbs and
Moise. Cognition 69: 337354.
Recanati, Francois, 1989. The pragmatics of what is said. Mind and Language 4: 295329.
Recanati, Francois, 993. Direct reference: From language to thought. Cambridge: Blackwell.
Recanati, Francois, 1995. On the alleged priority of literal interpretation. Cognitive Science 19: 207222.
Schraw, Gregory, 1995. Components of metaphoric processes. Journal of Psycholinguistic Research 24:
Searle, John, 1978. Literal meaning. Erkenntnis 13: 207224.
Searle, John, 1979. Metaphor. In: A. Ortony, ed., Metaphor and thought, 92123. New York: Cambridge
University Press.
Searle, John, 1983. Intentionality. New York: Cambridge University Press.
Shinjo, Makiko and Jerome Myers, 1987. The role of context in metaphor comprehension. Journal of
Memory and Language 26: 226241.
Simpson, Greg, 1994. Context and the processing of ambiguous words. In: Morton Ann Gernsbacher, ed.,
Handbook of psycholinguistics, 359374. San Diego: Academic Press.
Sperber, Dan and Deidre Wilson, 1986. Relevance: Communication and cognition. Oxford: Blackwell.
Sweetser, Eve, 1990. From etymology to pragmatics: The mind-body metaphor in semantic structure and
semantic change. Cambridge: Cambridge University Press.


R.W. Gibbs Jr. / Journal of Pragmatics 34 (2002) 457486

Swinney, David, 1979. Lexical access during sentence comprehension: (Re)consideration of context
eects. Journal of Verbal Learning and Verbal Behavior 18: 545569.
Tabossi, Patrizia and Francesco Zardon, 1993a. Context and lexical access. Journal of Memory and
Language 32: 359372.
Tabossi, Patrizia and Francesco Zardon, 1993b. The activation of idiomatic meaning in spoken language
comprehension. In: Cristina Cacciari and Patrizia Tabossi, eds., Idioms: Processing, structure and
interpretation, 145162. Hillsdale, NJ: Erlbaum.
Temple, John and Robert Honeck, 1999. Proverb comprehension: The primacy of literal meaning. Journal of Psycholinguistic Research 28: 4170.
Van Lancker, Diane, 1990. The neurology of proverbs. Behavorial Neurology 3: 69187.
Vu, Hoang, George Kellas and Stephen Paul, 1998. Sources of semantic constraint on lexical ambiguity
resolution. Memory and Cognition 26: 9791001.
Williams, John, 1992. Processing polysemous words in context: Evidence for interrelated meanings.
Journal of Psycholinguistic Research 21: 193218.
Wilson, Deidre and Dan Sperber, 1993. Linguistic form and relevance. Lingua 90: 125.