You are on page 1of 19

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/247504022

ERP Correlates of Individual Differences in the


Comprehension of Nonliteral Language

Article  in  Metaphor and Symbol · October 2006


DOI: 10.1207/s15327868ms2104_4

CITATIONS READS

28 378

2 authors:

Dawn G Blasko Victoria A Kazmerski


Pennsylvania State University Pennsylvania State University
48 PUBLICATIONS   1,301 CITATIONS    32 PUBLICATIONS   777 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

STEM Education View project

Individual Differences in Spatial Cognition View project

All content following this page was uploaded by Dawn G Blasko on 06 March 2014.

The user has requested enhancement of the downloaded file.


METAPHOR AND SYMBOL, 21(4), 267–284
Copyright © 2006, Lawrence Erlbaum Associates, Inc.

ERP Correlates of Individual Differences


in the Comprehension
of Nonliteral Language
Dawn G. Blasko and Victoria A. Kazmerski
Penn State Erie
The Behrend College

In recent years, the field of cognitive neuroscience has flourished as new tools allow
researchers to peer inside the workings of the living brain. At the same time, those in-
terested in the experimental study of language have begun to move beyond their ini-
tial focus on literal language to look at the more complex and potentially more inter-
esting areas of figurative language. The purpose of this article is to highlight ongoing
research into individual differences in nonliteral language processing, and to exam-
ine how the study of the electrophysiology of the brain can help illuminate theories of
figurative language. We inquire into how one can explain individual differences be-
tween subject populations in the time course of comprehension using event-related
brain potentials (ERPs). We review studies on individual differences in metaphor
processing, examine how self-reported differences in relational aggression relate to
the processing of sarcasm, and preview some ongoing studies on expertise in poetry.

Although researchers often make the simplifying assumption in cognitive science


that adults without obvious dysfunction are basically alike, this assumption is
likely to fail in the domain of figurative language. It could be argued that figurative
language requires knowledge by the speaker and the listener of the semantic, prag-
matic, and social context surrounding the utterance to an even greater extent than
most literal language. Take, for example, the situation faced by a group of col-
leagues conversing at a company party. One member of the group points across the
room and says, “Here comes Joe, our company’s database.” For the newest em-
ployee in the group, this statement suggests a great deal of relevant information.
Joe has probably been employed at the company a long time and he has a great deal

Correspondence should be addressed to Dawn G. Blasko, Penn State Erie, The Behrend College
5091 Station Road, Erie, PA 16563. E-mail: dgb6@psu.edu
268 BLASKO AND KAZMERSKI

of knowledge about its inner workings. Therefore, Joe might be a very good person
with whom to establish positive relations. However, for others that had been previ-
ously discussing Joe’s recent mysterious promotion there might be additional sar-
castic connotations to the remark. Perhaps Joe has used his voluminous store of in-
formation against colleagues to advance his own career. Finally, at the most
prosaic level, the noise and divided attention of the typical cocktail party may leave
some people with the mistaken impression that Joe is the company database man-
ager. So in a single everyday conversation, the same phrase may be comprehended
in a number of ways depending on the speaker, the listener, and the communicative
environment. Although the field is only beginning to explore these complexities, in
this article we examine a few examples of studies examining individual differences
in three types of nonliteral language: metaphor, sarcasm, and poetry. We provide
electrophysiological data that provides insights to the temporal processes of com-
prehension and begins to examine the underlying biological processes in the brain.
Metaphor is one of the most widely studied tropes and the model of processing
that has been most examined is sometimes called the three-stage model or the stan-
dard pragmatic model (Searle, 1979). According to this view, when a metaphor is
encountered it is first processed as literal language and, only if that attempt fails, is
a special figurative process activated that uses pragmatic principles to infer mean-
ing. This same model has often been applied to understand other forms of figura-
tive language, including sarcasm and proverbs. The standard pragmatic model was
extremely successful in stimulating research because it made clear and testable
predictions about the time course of processing. It suggested that metaphors
should take longer than literal utterances and that metaphor processing should be
optional if a judgement of literal meaning is all that is required. In contrast to much
earlier work coming out of philosophy and literature, in which extended poetic
metaphors were often studied, tests of this model encouraged the examination of
the very earliest moments of metaphorical processing, typically on very simple
nominal metaphors of the form “X is Y.” It also spurred the development of online
techniques of assessing meaning activation.
Several of these methods have contributed greatly to the understanding of
nonliteral language comprehension. Some of the earliest studies that tested predic-
tions of the standard pragmatic model used systems that could follow each eye
movement as readers moved through text. For example, Inhoff, Lima, and Carroll
(1984) showed that when presented in supportive context, metaphors could be read
as quickly as the same sentences intended literally. Eye-tracking research also re-
futed another interesting prediction of the standard model, that once the processing
of a metaphor was completed via a special figurative mode of processing, any met-
aphor that followed, even if unrelated to the first, would be understood more
quickly. Blasko and Briihl (1997) found that when one metaphor immediately fol-
lowed another, there was no evidence of facilitation unless the second metaphor
was closely related in meaning.
INDIVIDUAL DIFFERENCES IN NONLITERAL LANGUAGE 269

Another useful online technique was semantic priming, where meaning activa-
tion was inferred if judgements were faster to a semantically related target (Blasko
& Connine, 1993; Glucksberg & McGlone, 1999; Wolff & Gentner, 2000). For ex-
ample, Blasko and Connine used crossmodal priming, where participants listened
to recorded metaphors and, at the same time, were asked to make a lexical decision
(word/nonword) to a visual target that was unrelated or related to either the figura-
tive or literal meaning of the vehicle. They found that metaphorical meaning was
activated immediately, but only if the metaphors were highly familiar and highly
apt.
One of the most widely studied techniques for examining the earliest stages of
metaphor processing is a Stroop-like task often termed the metaphor interference
task (Gluckberg, Gildea & Bookin, 1982). In this task, readers are asked to judge
the literal truth of a series of short utterances. Participants are presented metaphors
(e.g., “divorce is a nightmare”) and scrambled metaphors (e.g., “divorce is a ta-
ble”) created from the same topic and vehicle terms. The readers are expected to
respond “no” to both of these nonliteral conditions, but if metaphor processing is
automatic and obligatory, than rejecting the real metaphors will be more difficult
than rejecting the scrambled metaphors. A number of studies found that this was
the case as long as the metaphors used were apt (Gildea & Glucksberg, 1983;
Glucksberg et al., 1982).
Although most of the predictions of the standard pragmatic model did not stand
up to rigorous testing, findings that some sorts of figurative language did take lon-
ger than literal language, such as less familiar metaphors (Blasko & Connine,
1993) and proverbs (Honeck, Welge, & Temple, 1998; Turner & Katz, 1997) sug-
gested that in some cases, readers might default to a literal interpretation. In recent
years, the focus toward online processing has continued and time course issues
have been critical to testing more recent models of metaphor as well. According to
Gentner and Wolff’s (1997) structure mapping model, at the earliest stage of pro-
cessing, topic and vehicle terms are aligned and similar attributes are extracted,
then in a second stage, inference processes apply the appropriate characteristics of
the vehicle to the topic. This model predicts that priming with either the topic or
the vehicle should facilitate the first stages of processing.
In contrast, according to categorization or dual reference models, the vehicle re-
fers to both the basic level term and the subordinate level name of an ad hoc cate-
gory representing the ground of the metaphor (Glucksberg, 2001; Glucksberg &
Keysar, 1990). For example, in the metaphor “my job is a jail,” the vehicle “jail”
refers to an actual prison as well as to a category of things that are restrictive and
confining. According to this view, the topic and vehicle always play different roles
and priming with the vehicle should be more effective. These predictions have
been tested in a number of priming studies with conflicting results (Gentner &
Wolff, 1997; Glucksberg, McGlone, & Manfredi, 1997). One problem is that even
online tasks, such as priming, have limitations. In the time it takes to make a mea-
270 BLASKO AND KAZMERSKI

surable response, many overlapping cognitive processes may be occurring and the
typical priming task can only measure one point in time. In addition, most priming
tasks require the participant to complete a rather unnatural secondary task. For ex-
ample, the participant might be asked if words are related or whether a target is a
real word (lexical decision task). Therefore, it would be useful to have a task that
can allow normal listening or reading and yet compare the time course of process-
ing between conditions in a sensitive way.

EVENT-RELATED BRAIN POTENTIALS (ERPS)

ERPs are recordings of the brain’s electrical activity that are time locked to a spe-
cific event or stimulus. ERPs provide a true moment-to-moment online measure
with millisecond accuracy, and so are a powerful tool for investigating time course
issues. They allow the comparison of patterns of brain activation between condi-
tions in real time, from the point of stimulus presentation through processing and
interpretation to response. In some cases, this sensitivity can help clarify the rea-
sons for response time differences and, in other cases, it can detect differences that
are not apparent in response times.
The ERP waveform has positive and negative deflections relative to the base-
line, the amplitude of which is measured in microvolts, and largely interpreted as
indexing the degree of engagement in a task. These deflections or components can
be correlated with sensory, cognitive, and motor processing. Because activity is
shown across a time period rather than at one point in time, as is the case in re-
sponse time studies, subtle individual differences may emerge that might not be
evident in behavioral data. An additional quality of ERPs is that data can be col-
lected simultaneously from multiple scalp sites. Differences in the scalp topogra-
phy across conditions or groups imply that differences exist in the brain areas en-
gaged in these conditions or groups. Source localization techniques allow
researchers to make inferences about the brain structures that may be involved in
producing the waveforms (Slotnick, 2005). This can be important for studies of in-
dividual differences, because it may be that different populations of subjects arrive
at similar behavioral responses using very different strategies that involve different
neural structures.
There are several components of ERPs that are relevant to the study of lan-
guage. ERP components are identified with respect to functional significance and
putative neural sources. They are labeled by their latency with respect to stimulus
onset and polarity (positive or negative). The earliest components, 0–200 ms, are
usually related to early sensory processing of the stimuli and to attentional de-
mands; however, some of these are also sensitive to linguistic differences such as
the emotionality of a word or its lexical frequency. The P3 (P300) is one of the
most widely studied components. As elicited in classical oddball studies, the P300
INDIVIDUAL DIFFERENCES IN NONLITERAL LANGUAGE 271

or P3b may reflect the process of updating the internal representation of the con-
text of the environment (Donchin, 1981). However, Verlerger (1988) argued that
the P300 is related to a closure of processing, and in his recent work suggested that
the P300 reflects processing between the classification and preparation of re-
sponses (Verleger, Jaskowski, & Wascher, 2005).
Two ERP components have been directly tied to the study of literal language
comprehension. The N400, a negative deflection occurring 400–600 ms post-
stimulus onset, is the component that has been most widely studied in the domain
of language. It has been shown to be sensitive to semantic violations (e.g., Kutas &
Hillyard, 1980) as well as to lexical ambiguity (Van Petten & Kutas, 1987). The
P600–900, or late positivity, is elicited when there is a syntactic violation but is
also sensitive to semantics. These components have been found to vary as a func-
tion of a variety of linguistic qualities such as word frequency effects (Rugg,
1990), concreteness effects (Kazmerski, Squires, & Lovrich, 1992; Kounios &
Holcomb, 1994), and variations in the complexity of full sentences (Kutas & Van
Petten, 1994; for a review see Kutas, Federmeier, Coulson, King, & Munte, 2000).

METAPHOR AND ERPS

Recently, the use of ERPs has been extended to the study of metaphor. In one of the
first studies, familiar metaphors and less familiar metaphors (“those fighters are
lions;” “those apprentices are lions”) were compared with literal sentences
(Pynte, Besson, Robichon, & Poli, 1996). Sentences were presented either without
context or in a relevant or irrelevant context. The results showed enhanced N400
amplitude for the last word of the metaphors in comparison to those in literal sen-
tences, but no difference for the late positive components. The N400 effect was re-
versed when metaphors were presented in context; that is, it was larger for familiar
metaphors in irrelevant context than unfamiliar metaphors presented in relevant
context. These results were interpreted to refute the standard pragmatic model and
to suggest that when metaphors are familiar and contextually relevant, the meta-
phorical meaning is the only one activated. In a more recent study with 11 partici-
pants (Tartter, Gomes, Dubrovsky, Molholm, & Stewart, 2002), sentences were
presented that contained less familiar poetic metaphors. These metaphors were
shown to elicit an N400 effect that was larger in comparison to a literal completion,
suggesting more difficulty with processing of the metaphor condition in compari-
son with the literal condition.
Based on this research, it is clear that ERPs are a valuable technique to examine
time course issues among a variety of stimuli and contexts. However, none of these
studies examined whether some of the inconsistency in the literature might be ex-
plained by individual differences. In one behavioral study (Blasko, 1994) that used
self-paced word-by-word reading, characteristics of the metaphors were shown to
272 BLASKO AND KAZMERSKI

interact with characteristics of the comprehender to determine the time course of


processing. Metaphors were presented in either normal (topic–vehicle) or reversed
(vehicle–topic) order in short adjacent sentences. Some metaphors were high in se-
mantic complexity and others were lower, so that a common feature was more eas-
ily accessible. Readers were pretested for their working memory capacity, and the
results showed that working memory impacted reading times in the most difficult
condition when the topic and vehicles were in different sentences and were pre-
sented in reversed order. The low working memory readers were more likely to be
literally “garden-pathed” than the high working memory readers and also seemed
poorer in comprehension. In a second study, readers with higher working memory
produced richer and more in-depth interpretations of metaphors than those with
lower working memory (Blasko, 1999).
Our recent work (Kazmerski, Blasko, & Dessalegn, 2003) has focused on using
ERPs to investigate individual differences in metaphor processing. To tap early
processes, we used the metaphor interference effect described earlier (Glucksberg,
et al., 1982). Consistent with previous work, we found evidence that even when
people are asked to inhibit metaphorical meaning and judge literal truth, meta-
phors took longer to reject than scrambled sentences, suggesting that the meta-
phorical meaning was processed automatically and directly (i.e., without having to
process and reject a literal meaning first). It is important that we found that literal,
metaphorical, and scrambled sentences showed discernibly different ERP pat-
terns. This pattern can be seen on the left side of Figure 1—young subjects. ERPs
to metaphors showed reduced N400 amplitudes relative to scrambled metaphors
and the literal sentences showed a positive-going wave in the same region. Al-
though the waveforms of the metaphors and the scrambled did not differ at 750 ms,
the time that both were judged literally untrue, the metaphors showed a late
positivity (950 ms) in comparison to the literal and scrambled in what we termed a
metaphor rebound effect. Notably, many subjects reported thinking about the met-
aphorical meaning in the interval between trials. The most interesting finding was
that the data showed clear patterns of individual differences. Participants were pre-
tested on a brief IQ measure that contained verbal (vocabulary and compre-
hension) and performance subscales (patterns and matrixes). We found that when
participants were partitioned based on IQ scores, the behavioral metaphor inter-
ference effect was statistically significant for participants with IQs of 114 and
above, but not for those with IQs below 100. This was clearly reflected in the ERP
recordings.
In a subsequent experiment, we compared young to older adults and found that
older adults showed just as large a metaphor interference effect as young adults
(see also Light, Owens, Mahoney, & La Voie, 1993). However, as shown in Figure
1, the pattern of ERPs across the brain was very different. Whereas younger adults
showed the typical pattern of larger N400 for scrambled than metaphor, the older
adults showed little of this effect. Rather, they showed an enhanced late positivity
INDIVIDUAL DIFFERENCES IN NONLITERAL LANGUAGE 273

FIGURE 1 Grand-averaged waveforms of event-related brain potentials recorded to meta-


phors (solid line), scrambled metaphors (dotted line), and literal sentences (gray dashed line) in
younger (left) and older (right) adults.

to scrambled metaphors compared to real metaphors, peaking at about 800 ms


poststimuli presentation. This shows that similar behavioral effects may be pro-
duced by what appears to be qualitatively different processes (Kazmerski et al.,
2003).
ERPs can also be an important tool to examine differences between normal and
clinical populations. Recently, Iakivova, Passerieux , Laurent, and Hardy-Bayle
(2005) used the metaphor interference task while recording ERPs in patients with
schizophrenia. They found that the patients were less accurate than controls in
making a semantic judgement about the stimuli. However, for those judged to be
accurate, the schizophrenia patients showed similar patterns to the controls in that
the ERPs showed larger N400s for incongruous sentences than for metaphorical
sentences.
Coulson and Van Petten (2002) also found a larger N400 effect to metaphors
than both literal sentences and intermediate literal mappings. The literal mappings
were designed to encourage readers to map meaning from a different conceptual
domain. In addition to the N400 findings, metaphors also showed a pattern of
larger late posterior positivity than either the literal or mapping condition. It is in-
274 BLASKO AND KAZMERSKI

teresting that the literal mapping condition elicited a large frontal positivity that
was not found in either metaphor or literal conditions. Coulson and Van Petten
noted that “a substantial proportion (74%) of the literal mapping sentences de-
scribed situations of pretense, lying and mistaken identification, so that compre-
hension depends on understanding the mental states of the actors” (p. XX). The de-
velopment of a theory of mind in which one understands that another person’s
understanding of a situation may differ from one’s own has been demonstrated to
be important to the comprehension of nonliteral language, especially in the case of
metaphor and irony (Happe, 1996). This has physiological consequences; for ex-
ample, Fletcher, Dhappe, Frith, and Baker (1995) found that situations requiring
understanding of theory of mind show greater blood flow in the prefrontal cortex.
This may provide an important clue as to the regions of the brain where one might
expect to see different patterns of activation for sarcastic utterances.

SARCASM

Sarcasm is a very interesting form of nonliteral language for several reasons. As


mentioned, it requires an understanding of the speaker’s knowledge and intent,
which makes it an ideal topic to explore in social cognition. It is also one way that a
speaker can harm or wound another, yet save face, because there is also a humor-
ous side to sarcasm. In comparison to metaphor, it is possible to have a high degree
of control over the stimuli because contexts can be developed that influence the
same sentence to be understood in either a literal–sincere or sarcastic way. For ex-
ample, two friends are driving down the road and a car suddenly pulls in front of
them. In one condition, the driver narrowly avoids an accident and the friend ex-
claims sincerely, “You really are a great driver!” And in the other case, the driver
plows into the car in front of them and the same statement is uttered as sarcastic
criticism.
As with metaphor, the time course of processing has been an issue with theoret-
ical implications. There has been some research examining the online processing
of sarcasm using word-by-word reading times, and these findings suggest that the
effects of processing sarcasm can sometimes be observed in spill-over effects to
the following sentences (Pexman, Ferretti, & Katz, 2000). Because ERPs are a
more sensitive measure of processing, we reasoned that they might allow us to see
in more detail the time-course of the processing of sarcasm.
In our first study (Katz, Blasko, & Kazmerski, 2004), we measured brain differ-
ences in the processing of a sentence depending on whether it was presented in a
literal or sarcastic context. Participants read short paragraphs, word-by-word, that
ended with the same sentence intended as either a sincere compliment or sarcastic
criticism. The reading times to the final word did not differ between sarcastic and
sincere contexts. However, the ERPs showed that the waveforms for the sarcastic
INDIVIDUAL DIFFERENCES IN NONLITERAL LANGUAGE 275

FIGURE 2 Grand-averaged waveforms of event-related brain potentials recorded to the last


word of the utterance in the context requiring a literal (solid line) or sarcastic (dashed line) inter-
pretation.

and literal began to diverge as early as 350 ms after the final word was presented
(see Figure 2). The sarcastic condition showed more negativity from 500–700 ms
followed by a larger late positivity for the sarcastic condition (peaking at 900–950
ms) to the sentence when it was to be understood sarcastically than when it was in-
tended sincerely.
In this study, we had a higher density of electrodes (64 channels), which al-
lowed us to conduct a preliminary analysis to estimate the source of this compo-
nent. A principle component analysis was conducted followed by a source analy-
sis. A source analysis of a component identified in the late region (660–900 ms)
revealed that for the literal utterance a single dipole model explained 87% of the
variance with the source of the activity located in the region of the anterior
cingulate gyrus (maximum global field power [GFP] at 836 ms). This region of the
brain, just superior to the corpus callosum, is thought to integrate sensory input
with emotions. In comparison, the maximum GFP for the sarcastic condition was
earlier (GFP at 744 ms) and the model was stronger, explaining 93% of the vari-
ance. It showed a source that was in the same region but more inferior than that
seen in the literal. Unlike the literal condition, in the sarcastic condition there was
also a later component that explained 93% of the variance (GFP at 1,012 ms). The
276 BLASKO AND KAZMERSKI

source of the activity was also in the region of cingulate gyrus but was more
lateralized to the right and slightly more superior to the earlier component. We
might speculate that this later component is related to processing of the negative
emotional connotations of the sarcastic utterances.
Recent work by Pexman and colleagues (Ivanko, Pexman, & Olineck, 2004;
Pexman, 2005) identified individual differences in the processing and production
of irony. Ivanko et al. correlated participants’ responses on a sarcasm self-report
scale, with their responses to an irony interpretation and an irony production task.
They found that the self-report measures corresponded to both the interpretation
and production scores. They also found that these differences may have online pro-
cessing implications. In their second experiment, they reported that individuals
with higher self-reported sarcasm scores tended to read ironic statements more
quickly than did those with lower self-reported sarcasm scores.
Because sarcasm is often used as a social weapon, we wondered if individual
differences in relational aggression (RA; a form of indirect aggression intended to
damage the social relations of others) might not influence the processing of sar-
casm. In an ongoing study, college students were asked to complete a self-report
aggression scale (Morales, 1999). Based on the sections that focused on relational
aggression, 14 high and 16 low RA students were selected. As before, participants
read word-by-word context paragraphs that biased a target utterance towards either
a sincere or sarcastic meaning. ERPs were recorded continuously and time locked
to the onset of the final word of the target sentence. In this experiment, we also in-
cluded positive emotion and negative emotion conditions. This made it possible to
examine whether the differences between the literal and sarcastic ERPs were due
primarily or entirely to the differences in emotional valence (Shaffer, Blasko, &
Kazmerski, 2004). The results of the reading time analysis showed that only the
low RA group showed reliably slower reading times to the last word of the sarcas-
tic sentences in comparison to the literal, suggesting that they might be more sensi-
tive to the criticism in the sentences. As seen in Figure 3, the patterns of ERPs re-
flected this observation. The sarcastic and literal waveforms separated even earlier
than in the previous experiment, and analysis of the mean amplitudes showed sig-
nificant group × condition × location effects in the first bin measured (100–300
ms). The low RA group showed a larger P300 than the low group. It is possible that
low RA readers felt more surprise at the sarcastic criticism demonstrated between
friends. It is interesting that the N650 effect seen in sarcastic utterances in the pre-
vious experiment was now seen only in the high RA participants. The most promi-
nent component in the waveform was the late positivity that was larger in the sar-
castic condition than the literal. It is worth noting that when processing sarcasm,
the high RA groups’ late positive component was earlier and smaller in compari-
son to the low RA. This may reflect less effortful processing of the negative conno-
tations of the sarcasm. The suggestion that the size of the late positive component
is, at least in part, related to the perception of negativity, is supported by the fact
INDIVIDUAL DIFFERENCES IN NONLITERAL LANGUAGE 277

FIGURE 3 On the left are midline grand-averaged waveforms of event-related brain poten-
tials recorded in a sarcastic context for individuals rated high (solid line) and low (dashed line)
on relational aggression. On the right are contour maps with an indication of the source of acti-
vation for the High RA (top, GFP = 276) and Low RA (bottom, GFP = 266) groups.

that the stimuli with no sarcasm but a negative outcome showed larger late
positivity than the sarcastic condition.
Source analysis of the early region (maximum GFP = 110 ms) showed that
models for the low RA subjects were slightly left lateralized to the posterior frontal
lobe, again near to the region of the cingulate gyrus. This model explained 93% of
the variance. By comparison, a source model that explained 95% of the variance
for high RA subjects (GFP = 158 ms), was more right lateralized in the temporal
lobe near the junction of the parietal lobe.
The issue of whether sarcasm enhances or mutes emotional negativity is a
controversial one. Dews and Winner (1995) proposed the tinge hypothesis that
irony serves to reduce the strength of the actual statement. In fact, high RA ado-
lescents often dismiss their behavior by saying that they were just teasing or that
it was only a joke (Werner & Nixon, 2005). In contrast, other research has sug-
gested that sarcastic irony might actually enhance the negative emotions felt by
the victim of the barb (Leggitt & Gibbs, 2000). What the current results suggest
is that those who differ in RA may process the same contextual situations very
differently, with the low RA group potentially more sensitive to the negativity
implied in sarcasm.
278 BLASKO AND KAZMERSKI

One interesting question is the extent to which these differences in brain pro-
cessing of sarcasm between those high and low in RA are due to nature or nurture.
The finding that RA is only moderately stable and that some victims of RA eventu-
ally become aggressors themselves (Crick et al., 1999; Werner & Nixon, 2005)
suggests that learning must play an important role, and this is certainly an impor-
tant area for future research. Another question that needs to be answered is
whether interventions designed to reduce relational aggression would be reflected
in changes in the ERPs when processing sarcasm.

POETRY AND EXPERTISE

Individual differences in the use of figurative language may be seen most clearly
by comparing novices to experts in a domain. Take for example, writers and poets
that have spent many hours considering both the relationships between words, as
well as each word’s deeper figurative and emotional connotations. It is possible
that some of these people have innate differences in the way that they comprehend
and produce language, but most research suggests a much larger role for learning.
Comparing novices to experts might tell a great deal about how the brain processes
figurative language before and after such intensive training.
In one study, we compared college students who reported regularly writing po-
etry outside of class for at least 2 years with those that never wrote poetry outside
of a required class. The stimuli consisted of 22 haiku poems written by profes-
sional poets.1 English poets’ haiku presents an interesting case for the study of
nonliteral language because most discussions of haiku often expressly discourage
or forbid the use of metaphor or simile while at the same time encouraging the jux-
taposition of two images or concepts in unusual ways (see Blasko & Merski, 1999,
for a discussion). For example:

All night
the sound of your breathing
the autumn wind

For each poem, we created a matched pseudopoem, matched for length and word
complexity, that ended on the same last word.

Outside
by the ripling lake
there was a strong wind

1The 17 syllable rule that is often cited was based on the Japanese syllable (kanji), which carries less

information. Therefore, in recent years, most Haiku poets that write in English have abandoned the syl-
lable rule and instead focus on brevity and clarity of image.
INDIVIDUAL DIFFERENCES IN NONLITERAL LANGUAGE 279

The utterances were recorded by a male speaker and read with a poetic
three-part prosodic contour. Participants were then asked to judge whether each
stimulus was a real poem. ERPs were time-locked to the final word of each utter-
ance. Behavioral results showed that both poets and nonpoets could distinguish the
real poems at over 80% accuracy and with similar speed. The only difference be-
tween the groups was that the nonpoets were somewhat more accurate in judging
the literal stimuli in comparison to the poems, whereas the poets were equally ac-
curate in both conditions.
Although the behavioral results showed few differences between groups, the
ERP results showed robust differences. Analysis of the mean amplitudes showed
that there was a three-way interaction between left–right hemispheres, utterance
type, and poet group starting 300 ms after the beginning of the last word. ERPs to
the literal sentences were rather similar for the poets and nonpoets, however they
showed large differences for the poems. When the nonpoets listened to the poems,
their ERPs were more negative, especially at 375 ms and again at 900 ms. Poets, on
the other hand, showed more positivity overall then nonpoets. When looking only
at the poetry condition, the poets showed a large positive peak at 500 ms and a
broad late positivity from 900 to 1400 ms post stimulus onset. Figure 4 shows that
the difference between poems and pseudopoems was largest for the frontal elec-
trodes and that the late effect was broadly distributed over temporal and parietal re-
gions and slightly larger in the right hemisphere.
Source localization analysis (SOURCE 2.0, Neuroscan) on these data generally
confirm the conclusions discussed. The source analyses showed that in the earliest
regions (GFP at 250 ms), as the poets were listening to the last word of the poems,
they showed focal activity in the right parietal area. A single dipole source in this
region explained 93% of the variance. The nonpoets also showed activity in the pa-
rietal lobe, but the best model included a second dipole in the prefrontal cortex.
This model (max. GFP at 274 ms) explained 88% of the variance. Activity in the
region associated with the N2–P3 ERP components may invoke aspects of work-
ing memory, and thus reflect that poets may rely less on this region as expertise has
developed in making the semantic leaps necessary to appreciate poetry.

CONCLUSIONS

The results of the experiments summarized here suggest that the processing of
nonliteral language involves a complex social and emotional interaction between
speaker and listener that is made even more complicated by individual differ-
ences. Although the majority of research has focused on metaphor, other sorts of
figurative language, such as sarcasm and poetry, are amenable to empirical
study, and may help to inform researchers’ original theoretical questions. For ex-
ample, is figurative language understood directly or only after a failed attempt to
find a literal interpretation? As we have shown, this question has spawned a
280 BLASKO AND KAZMERSKI

FIGURE 4 The top shows the grand-averaged waveforms of the event-related brain poten-
tials recorded to the last word of poems for nonpoets (solid line) and poets (dashed line). The
bottom shows the dipole location for the source derived at 250 ms.

great deal of important research that has lead to even more interesting questions,
but the answer, as with many questions about complex human behavior, seems to
be, “it depends.”
When a figurative interpretation has a great deal of contextual support or when
the leaps between semantic domains are clear, then processing appears to occur di-
rectly and may even supercede literal processing. But when the links between con-
cepts are more distant, or when they allow for ambiguity, then a judgment must be
made based on a host of factors that includes the social context, and the personal-
ity, knowledge, and expertise of the comprehender. It remains to be seen whether
one model can be developed to handle all of these factors or whether different theo-
ries will be required for different tropes and for different factors. One promising
idea is that the most salient stored meaning will be activated immediately, regard-
less of context, but strong context can later increase the activation of less salient or
more peripheral interpretations (Giora, 2003). In contrast, constraint satisfaction
models work on the assumption that different forms of information will compete
INDIVIDUAL DIFFERENCES IN NONLITERAL LANGUAGE 281

for activation over time, with these constraints providing probabilistic evidence in
support of various alternative interpretations (Katz & Ferretti, 2001; Kintsch,
2000).
The main benefit of this second class of models is that there is ample room for
the type of individual differences that this article discusses. One type of individual
difference results from the speed and efficiency of basic cognitive and neuronal
processes. For example, working memory capacity is thought by many to be re-
lated to the efficiency of the allocation of limited attentional resources (Engel,
2002). In all complex language processing, the comprehender must rapidly acti-
vate and develop one set of relevant meanings and inferences while ignoring or
suppressing others that also may be semantically related, but are irrelevant to the
circumstances. A teenager hanging out with a group of her friends after school
may simultaneously be interpreting multiple conversations for their literal and
nonliteral connotations using contextual knowledge, body language, gestures, fa-
cial expression, and prosody. At the same time, she must plan her response, but
only after assessing the social hierarchies of the speakers, and considering the so-
cial consequences of potential retorts. Those with faster and more efficient brain
processes will be at an advantage, but others may develop specific strategies for
coping with less resources.
Many older people seem to be masters of coping in the face of slower and
less efficient neuronal processes. In our studies, college students with lower ver-
bal IQ, which is correlated with working memory, showed no automatic activa-
tion of the metaphorical meaning, but older people, regardless of IQ, did show
metaphor activation. They answered more slowly and accurately and showed a
very different pattern of brain waves than young adults. One possibility is that
the older people compensated by using the richer schemas and semantic net-
works that they developed across the years. This illustrates a second theme that
crosses all three types of nonliteral language discussed here. The familiarity that
comes with expertise, whether it is metaphor, sarcasm, or poetry, has conse-
quences for brain processing. When the experienced poet ponders the connota-
tions created by the juxtaposition of different concepts, they show different pat-
terns of brain activation than those without such experiences. As we have shown,
frontal regions that are active when nonpoets listen to poetry are not needed by
poets. It appears that processing becomes more efficient and recruits fewer re-
sources with experience.
One last point is that in several cases reported herein, including the poetry
study, response times and accuracies did not show individual differences; how-
ever, the pattern of ERPs showed large and reliable differences. Why is this the
case? One possibility is that the ratings and response times used in many studies
show only the composite results of many summed cognitive and motor pro-
cesses. ERPs and other brain imaging techniques, although clearly more costly
and time consuming than behavioral studies, provide a more sensitive measure
282 BLASKO AND KAZMERSKI

of online processing. ERPs used in concert with other tools might also help re-
searchers learn more about the way that the brains of different people change
with experience as they develop strategies to interpret and produce nonliteral
language.

ACKNOWLEDGMENTS

The authors thank those students who helped in the data collection and analysis:
Holly Blasko Drabik, Christine Anderson, Dotty Shaffer, Kristen Shaffer, Matt
Stevenson, Joshua Rowe, Alejandra Marroquin, and Banchiamlack Dessalegn.
This research was funded in part by grants through the Office of Undergraduate
Research at Penn State Erie, The Behrend College.

REFERENCES

Blasko, D. G. (1994). The constraints of working memory capacity on the resolution of semantic inde-
terminacy. Dissertation Abstracts International: Section B: The Sciences and Engineering. Vol
55(2-B), pp. 616.
Blasko, D. G. (1999). Only the tip of the iceberg: Who understands what about metaphor? Journal of
Pragmatics, 31, 1675–1683.
Blasko, D. G., & Briihl, D. S. (1997). Reading and recall of metaphorical sentences: Effects of familiar-
ity and context. Metaphor & Symbol, 12, 261–285.
Blasko, D. G., & Connine, C. M. (1993). Effects of familiarity and aptness on metaphor processing.
Journal of Experimental Psychology: Learning, Memory, & Cognition, 19, 295–308.
Blasko, D. G., & Merski, D. (1998). Haiku poetry and metaphorical thought: An invitation to interdisci-
plinary study. Creativity Research Journal, 11, 9–46.
Coulson, S., & Van Petten, C. (2002). Conceptual integration and metaphor: An event related potential
study. Memory & Cognition, 30, 958–968.
Crick, N. R., Werner, N. E., Casas, J. F., O’Brien, K. M., Nelson, D. A., Grotpeter, J. K., et al. (1999).
Childhood aggression and gender: A new look at an old problem. In D. Bernstein (Ed.), Gender and
Motivation, Vol. 45: Nebraska Symposium on Motivation (pp. 74–141). Lincoln: University of Ne-
braska Press.
Dews, S. & Winner, E. (1995). Muting the meaning: A social function of irony. Discourse Processes,
19, 347–367.
Donchin, E. (1981). Surprise!… Surprise? Psychophysiology, 18, 493–513.
Engle, R. W. (2002). Working memory capacity as executive attention. Current Directions in Psycho-
logical Science, 11(1) 19–23.
Fletcher, P. C, Happe, F., & Frith, U., & Baker, S. C. (1995). Other minds in the brain: A functional im-
aging study of “theory of mind” in story comprehension. Cognition, 57, 109–128.
Gentner, D., & Wolff, P. (1997). Alignment in the processing of metaphor. Journal of Memory & Lan-
guage, 37, 331–355.
Gildea, P., & Glucksberg, S. (1983). On understanding metaphor: The role of context. Journal of Verbal
Learning & Verbal Behavior, 22, 577–590.
Giora, R. (2003). On our mind: Salience, context and figurative language. New York: Oxford Univer-
sity Press.
INDIVIDUAL DIFFERENCES IN NONLITERAL LANGUAGE 283

Glucksberg, S. (2001). Understanding figurative language. New York: Oxford University Press.
Glucksberg, S., Gildea, P., & Bookin, H. B. (1982). On understanding nonliteral speech: Can people ig-
nore metaphors? Journal of Verbal Learning & Verbal Behavior, 21, 85–98.
Glucksberg, S., & Keysar, B. (1990). Understanding metaphorical comparisons: Beyond similarity.
Psychological Review, 97, 3–18.
Glucksberg, S., & McGlone, M. (1999). When love is not a journey: What metaphors mean. Journal of
Pragmatics, 31, 1541–1558.
Glucksberg, S., McGlone, M. S., & Manfredi, D. (1997). Property attribution in metaphor comprehen-
sion. Journal of Memory and Language, 36, 50–67.
Happe, F. G. (1996). Understanding minds and metaphors: Insights from the study of figurative lan-
guage in autism. Metaphor & Symbol, 10, 275–295.
Honeck, R. P., Welge, J., & Temple, J. G. (1998). The symmetry control in tests of the standard prag-
matic models: The case of proverb comprehension. Metaphor & Symbol, 134, 257–273.
Iakivova, G., Passerieux, C., Laurent, J., & Hardy-Bayle, M. (2005). ERPs of metaphoric, literal, and
incongrous semantic processing in schizophrenia. Psychophysiology, 42, 380–390.
Inhoff, A. W., Lima, S., & Carroll, P. (1984). Contextual effects on metaphor comprehension in reading.
Memory & Cognition, 12, 558–567.
Ivanko, S., Pexman, P., & Olineck, K. (2004). How sarcastic are you? Individual differences and verbal
irony. Journal of Language and Social Psychology, 23, 244–271.
Katz, A., Blasko, D., & Kazmerski, V. (2004). Saying what you don’t mean: Social influences on sar-
castic language processing. Current Directions in Psychological Science, 13(5), 186–189.
Katz, A. & Ferretti, T. (2001). Moment-by-moment reading of proverbs in literal and nonliteral con-
texts. Metaphor and Symbol, 16, 193–221.
Kazmerski, V., Blasko, D., & Dessalegn, B. (2003). ERP and behavioral evidence of individual differ-
ences in metaphor comprehension. Memory & Cognition 31(5), 673–689.
Kazmerski, V., Blasko, D., & Marroquin, A. (2003, May). ERP evidence for differences in automaticity
of metaphor processing in younger and older adults [abstract]. Psychophysiology, 40, S52.
Kazmerski, V., Squires, N., & Lovrich, D. (1992). ERP indices of the concreteness effect [abstract].
Psychophysiology, 29, S43.
Kintsch, W. (2000). Metaphor comprehension: A computational theory. Psychonomic Bulletin & Re-
view, 7, 257–266.
Kounios, J., & Holcomb, P. J. (1994). Concreteness effects in semantic processing: ERP evidence sup-
porting dual-coding theory. Journal of Experimental Psychology: Learning, Memory, and Cogni-
tion, 20, 804–823.
Kutas, M., Federmeier, K., Coulson, S., King, J., & Munte, T. (2000). Language. In J. Caccioppo, J.
Tassinary, & G. Berntson (Eds.) Handbook of psychophysiology (2nd ed., pp. 576–601). New York:
Cambridge Press.
Kutas, M., & Hillyard, M. (1980). Reading senseless sentences: Brain potentials reflect semantic in-
congruity. Science, 20, 203–205.
Kutas, M., & Van Petten, C. (1994). Psycholinguistics electrified: Event-related brain potential investi-
gations. In M. Gernsbacher (Ed.), Handbook of psycholinguistics (pp. 83–144). San Diego, CA: Aca-
demic.
Leggitt. J., & Gibbs, R. W., Jr. (2000). Emotional reactions to ironic language. Discourse Processes, 29,
1–24.
Light, L. L., Owens, A. S., Mahoney, P. G., & La Voie, D., (1993). Comprehension of metaphors by
young and older adults. In J. Cerella, J. Rybash, W. Hoyer, & M. Commons (Eds.), Adult information
processing: Limits on loss (pp. 459–488). San Diego, CA, US: Academic.
Morales, J. (1999, April). Hostile attribution and aggression in adolescent peer and romantic relation-
ships. Paper presented at the Biennial Meeting of the Society for Research in Child Development,
Albuquerque, NM.
284 BLASKO AND KAZMERSKI

Pexman, P. (2005). Social factors in the interpretation of verbal irony: The roles of speaker and listener
characteristics. In H. Colston & A. Katz (Eds.), Figurative language comprehension: Social and cul-
tural influences (pp. 209–232). Mahwah, NJ: Lawrence Erlbaum Associates, Inc.
Pexman, P., Ferretti, T. R., & Katz, A. (2000). Discourse factors that influence online reading of meta-
phor and irony. Discourse Processes, 29, 201–222.
Pynte, J., Besson, M., Robichon, F.H., & Poli, J. (1996). The time course of metaphor comprehension:
An event-related potential study. Brain and Language, 55, 293–316.
Rugg, M. (1990). Event-related brain potentials dissociate repetition effects of high- and low-frequency
words. Memory & Cognition, 18, 367–379.
Searle, J. (1979). Metaphor. In A. Ortony (Ed.) Metaphor and thought (pp. XX–XX). Cambridge, Eng-
land: Cambridge University Press.
Shaffer, K., Blasko, D., & Kazmerski, V. (2004, May). Individual differences in relational aggression
influence brain activity for sarcastic language. Presented at the annual meeting of the American Psy-
chological Society, Chicago, IL.
Slotnick, S. (2005). Source localization of ERP generators. In T. Handy (Ed.) Event-related potentials:
A methods handbook (pp. 149–166). Cambridge, MA: MIT Press.
Tartter, V. C., Gomes, H., Dubrovsky, B., Molholm, S., & Stewart, R. V. (2002). Novel metaphors ap-
pear anomalous at least momentarily: Evidence from N400. Brain and Language, 80, 488–509.
Turner, N. E., & Katz, A. N. (1997) The availability of conventional and of literal meaning during the
comprehension of proverbs. Pragmatics & Cognition, 5(2), 199–233.
Van Petten, C. K., & Kutas, M. (1987). Ambiguous words in context: An event-related potential analy-
sis of the time course of meaning activation. Journal of Memory and Language, 26, 188–208.
Verleger, R. (1988). Event-related potentials and cognition: A critique of the context updating hypothe-
sis and an alternative interpretation of P3. Behavioral and Brain Sciences, 11, 343–356.
Verleger, R., Jaskowski, P., & Wascher, E. (2005). Evidence for an integrative role of P3b in linking re-
action to perception. Journal of Psychophysiology, 19, 165–181.
Werner, N. E., & Nixon, C. L. (2005). Normative beliefs and relational aggression: An investigation of
the cognitive bases of adolescent aggression behavior. Journal of Youth and Adolescence, 35(3),
229–243.
Wolff, P., & Gentner, D. (2000). Evidence for role-neutral initial processing of metaphors. Journal of
Experimental Psychology: Learning, Memory, and Cognition, 26, 529–541.

View publication stats

You might also like