ARTICLE IN PRESS

Brain and Language xxx (2006) xxx–xxx www.elsevier.com/locate/b&l

Electrophysiological and behavioral measures of the inXuence of literal and Wgurative contextual constraints on proverb comprehension 夽 Todd R. Ferretti a,¤, Christopher A. Schwint a, Albert N. Katz b a

Centre for Cognitive Neuroscience, Department of Psychology, Wilfrid Laurier University, 75 University Avenue, Waterloo, Ont., Canada N2L 3C5 b University of Western Ontario, London, Ont., Canada Accepted 3 July 2006

Abstract Proverbs tend to have meanings that are true both literally and Wguratively (i.e., Lightning really doesn’t strike the same place twice). Consequently, discourse contexts that invite a literal reading of a proverb should provide more conceptual overlap with the proverb, resulting in more rapid processing, than will contexts biased towards a non-literal reading. Despite this, previous research has failed to Wnd the predicted processing advantage in reading times for familiar proverbs when presented in a literally biasing context. We investigate this issue further by employing both ERP methodology and a self-paced reading task and, second, by creating an item set that controls for problems with items employed in earlier studies. Our results indicate that although people do not take longer to read proverbs in the literally and proverbially biasing contexts, people have less diYculty integrating the statements in literal than Wgurative contexts, as shown by the ERP data. These diVerences emerge at the third word of the proverbs. © 2006 Elsevier Inc. All rights reserved. Keywords: Proverb comprehension; Slow-cortical waves; ERP; N400; Late positivity; Self-paced reading

1. Introduction Over the last decade researchers have begun to examine how people comprehend Wgurative statements when they are placed in contexts that are consistent with their Wgurative or literal meanings. One of the main goals of this research is to understand the time-course in which people construct Wgurative interpretations for statements that could be interpreted literally or Wguratively. DiVerent theoretical accounts of Wgurative language processing make clear predictions about the time-course in which Wgurative meanings should be constructed. For example, according to the standard pragmatic model (Grice, 1975; Searle, 1979), people Wrst construct the literal meaning of 夽 This research was supported by a Canadian Foundation for Innovation (CFI) grant held by the Wrst author, and by separate NSERC Discovery Grants held by the Wrst and third authors. * Corresponding author. E-mail address: [email protected] (T.R. Ferretti).

the complete phrase and only attempt to construct a Wgurative interpretation when that meaning is perceived as inconsistent with the preceding context. Giora’s graded salience model (Giora, 2003) posits that the most salient or familiar usage of an expression, such as the nonliteral sense of a familiar proverb, will be aroused at the initial moments of processing regardless of the nature of the preceding context. Other models, such as the direct access models (Gibbs, 1995), and the constraint-based model (Katz & Ferretti, 2001), are context-dependent and argue that at the initial moments of processing the Wgurative meaning of as statement is accessed at least as quickly as the literal sense as long as the contexts are suYciently constraining. Results from studies employing on-line reading time measures demonstrate that people do not take longer to read a statement used Wguratively than used literally when placed in contexts that strongly constrain for the Wgurative interpretation, thereby providing support for models that do not posit a delay in accessing Wgurative meanings (e.g.,

0093-934X/$ - see front matter © 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.bandl.2006.07.002

Please cite this article as: Todd R. Ferretti et al., Electrophysiological and behavioral measures of the inXuence of literal and Wgurative contextual constraints on proverb comprehension, Brain and Language (2006), doi:10.1016/j.bandl.2006.07.002.

ARTICLE IN PRESS 2

T.R. Ferretti et al. / Brain and Language xxx (2006) xxx–xxx

Gibbs, Bogdanovich, Sykes, & Barr, 1997; Katz & Ferretti, 2001). Despite these Wndings, recent research involving Event-Related-Brain-Potential methodology (ERP) suggest that people have more diYculty integrating the Wnal word of statements when they are intended to be taken as nonliteral statements rather than as literal statements (e.g., Coulson & Van Petten, 2002; Katz, Blasko, & Kazmerski, 2004; Pynte, Besson, Robichon, & Poli, 1996). The primary brain potential of interest in these studies has been the N400 component (Kutas & Hillyard, 1984). The N400 component is typically interpreted as indexing the ease of semantic integration of words in context; words that are more diYcult to semantically integrate elicit an N400 with larger amplitudes than words that are easier to integrate. The results of the studies cited above show that, Wrst, negative amplitudes that peak about 400 ms after the presentation of the critical word tend to be largest for sentence Wnal words when they are embedded in Wgurative compared to literal contexts and, second, the N400 is often followed by a late positivity, in which there is a larger positive amplitude for words presented in Wgurative than literal contexts. Researchers have interpreted these Wndings as showing that conceptual integration is more diYcult for statements intended to be taken Wguratively than literally (e.g., Coulson & Van Petten, 2002) and, because these diVerences occur during the processing of the statements (Katz et al., 2004), they cannot be accounted for by models that hold that literal meanings must be processed prior to generating a Wgurative meaning (e.g., Grice, 1975). In the following research we build on previous research in several ways. First, we measured ERPs when people read familiar Wgurative statements presented in literally and Wguratively biasing contexts. In our case, in addition to examining N400 and late positivity eVects to single words as above, we also examine slow-cortical wave brain potentials that develop over multiple words in sentences and clauses. Previous research has shown that slow-wave potentials are sensitive to the ease in which people integrate sentences and clauses into a developing mental model of the text (e.g., King & Kutas, 1995; Münte, Heinze, & Kutas, 1998). Ease in integrating sentences is associated with a frontally distributed positivity which is sustained over the sentences, typically taken to indicate decreased working memory load relative to sentential structures that are harder to integrate. Thus, in addition to providing a measure of working memory load, slow-cortical wave potentials are similar to N400 and late positivity components found for single words in that they enable us to examine when a critical statement becomes more diYcult to integrate with preceding contexts. To our knowledge, this research is the Wrst to concurrently investigate both slowwave potentials that develop over sentences and single word averages to provide converging indices on when and how Wgurative statements are integrated in discourse contexts. In this study we employ proverbs to examine the eVects of contextual constraints on Wgurative language processing.

Proverbs are unique relative to metaphors or irony, the most commonly studied forms of non-literal language, because they can be regarded as generally true statements both literally and Wguratively (e.g., lightning really doesn’t strike the same twice). In contrast, a metaphor such as Children are precious gems, is true Wguratively but not literally. One consequence of this property of proverbs is that when a familiar proverb is placed within a context that invites a literal reading, there is inevitably a greater degree of conceptual overlap between the target proverb and words in the context than when the same target is placed in a context that invites a non-literal proverbial reading. One might expect that this diVerence alone would lead to faster reading times for familiar proverbs placed in literal biasing contexts relative to Wgurative biasing contexts in previous research. However, Katz and Ferretti (2001, 2003) using word-by-word self-paced reading methodology found that people read familiar proverbs at the same rate whether or not they are presented in literal or Wgurative contexts, and whether or not explicit markers are employed to signal the reader that an upcoming statement should be interpreted literally or Wguratively. 2. Experiment 1 In Experiment 1 we use ERP methodology to examine how people interpret familiar proverbial statements in literal and Wgurative biasing contexts. There are several advantages to using ERP methodology. First, ERP methodology is well known for having a Wne temporal resolution. Second, individuals can simply read for comprehension without having to push buttons to advance through text. Third, ERP methodology provides insight into the electrophysiological response of the brain to processing familiar proverbs. As discussed above, slow-wave potentials, the N400, and late positivity components (LPC) have all been shown to be a sensitive measure of when diYculties arise in integrating words and sentences in discourse contexts. Moreover, topographic diVerences in brain potentials provide additional insight into how diVerent populations of neurons contribute to diVerences found between conditions. Our main predictions for slow-wave potentials and single word averages are similar. SpeciWcally, if people have less diYculty integrating the proverbs in literal than Wgurative contexts due to greater conceptual overlap in the literal contexts, then we would expect to Wnd more positive slowwave amplitudes at frontal locations for the literal than Wgurative condition. The diVerence in slow wave amplitudes should begin at the point in which the literal condition becomes easier to integrate than the Wgurative condition, and this diVerence should be sustained over the proverbial statements. Based on previous research on Wgurative language processing (e.g., Coulson & Van Petten, 2002; Katz et al., 2004), we also expect to Wnd a smaller N400 and a smaller LPC at the word(s) in the proverbial statement in which people begin to have less diYculty in the

Please cite this article as: Todd R. Ferretti et al., Electrophysiological and behavioral measures of the inXuence of literal and Wgurative contextual constraints on proverb comprehension, Brain and Language (2006), doi:10.1016/j.bandl.2006.07.002.

ARTICLE IN PRESS T.R. Ferretti et al. / Brain and Language xxx (2006) xxx–xxx

literal conditions. Alternatively, if the proverbial statements are interpreted as easily in the two contexts, then we would not expect to Wnd any diVerences in the ERP results.

Table 1 Mean norming ratings (on a 1–7 likert scale) for Wgurative and literal paragraphs used in Experiment 1 and 2 Ease of comprehension

2.1. Participants Twenty-four participants participated in the ERP experiment and 99 students participated in the rating studies. All participants were native English speaking undergraduate students from Wilfrid Laurier University who received course credit for their participation. 2.2. Materials and procedure 2.2.1. Rating studies We Wrst conducted a rating study to ensure that our proverbs were familiar to our population. Twenty-seven participants were presented with 116 proverbs and were asked to rate on a 7 point scale how familiar each proverb was to them (1 D not at all familiar, 7 D very familiar). We then selected 58 familiar proverbs (all seven words long) that served as the basis for constructing our target paragraphs. For each of the 58 proverbs, we constructed a context that invited a literal reading and a context that invited a Wgurative reading. Each of the contexts had 4 sentences preceding each proverb, and the sentence before the proverb was always identical across the two contextual conditions. The contexts were in narrative form and described interactions between people (see examples 1a and 1b). For the literal biasing contexts we ensured that the content words of the familiar proverbs did not appear in the contexts. This ensured that any advantage found for the literal contexts could not be a result of repetition priming between the words in the contexts and words in the proverbs. (1a) Figuratively biasing context “What you need is an investment to shelter your proWt”, said Ann. “But it’s been a volatile market since the crash,” replied George. “Look, I lost a lot of money last year.” “Don’t worry, you’ll be alright,” she assured him. Lightning never strikes the same place twice.” “How can you be certain?”, he asked. “Its true, the market goes in cycles; it won’t crash again for years,” she replied. (1b) Literally biasing context “Let’s take shelter from the rain under this broken tree,” said Ann. “But it’s dangerous to hide under a tree during a storm,” replied George. “Look, the tree has been hit once already.” “Don’t worry, you’ll be alright,” she assured him. “Lightning never strikes the same place twice.” “How can you be certain?,” he asked. “Its true, once the energy dissipates, it takes a while to rebuild,” she replied. We then validated our stimuli with two additional rating studies. In order to ensure that the proverbs were equally comprehensible in the literal and Wgurative contexts, we asked 48 participants to rate each paragraph for how easy they thought the proverbs were to comprehend in the contexts (1 D not all easy to comprehend, 7 D very easy to com-

3

M Figurative paragraphs 5.5 Literal paragraphs 5.3 DiVerence +.2 ¤¤

SE .1 .1

Context Wgurativeness M 5.5 2.3 +3.2¤¤

SE .2 .2

p < .01.

prehend). In the second study, 24 participants provided ratings for how literal or Wgurative they thought the proverbs were in the diVerent contexts (1 D very literal, 7 D very Wgurative). This rating study ensured that our contexts were literal or Wguratively biased as claimed. Based on the results of these rating studies we selected 32 proverbs and their corresponding literal and Wgurative biasing contexts to serve as the experimental stimuli. The mean ratings for these items are presented in Table 1. Overall, the items were rated similar for ease of comprehension (t(31) D 1.15, p D .26), and the Wgurative contexts received higher Wgurative ratings than the corresponding literal contexts (t(31) D 11.38, p < .001). Finally, the 32 proverbs were rated relatively high for familiarity, M D 4.9. The 32 proverbs were placed across two experimental lists with the restriction that each participant read each proverb only once and each proverb appeared in a Wgurative and literal context. Ninety-six Wller trials similar in narrative form and length were included in each list to create an experimental environment in which all of the items would be read in their usual manner, i.e., without inducing a strategy to read for Wgurative meaning. Accordingly, none of these trials included Wgurative statements. The low proportion of Wller trials relative to Wgurative trials is a control that is missing from previous research investigating the eVect of context on interpreting literal and Wgurative language. In total, only 25% of the paragraphs involved a familiar proverb, in half of those trials the proverb was used literally and in half of the trials the proverb was used Wguratively. Participants sat in a chair in front of a computer monitor located in an electrically shielded chamber. They were instructed to read the words one at a time and to answer periodic comprehension questions by pressing buttons labeled “Yes” and “No”. The 32 Experimental paragraphs and 96 Filler paragraphs were presented one word at a time in the center of a computer screen. All words were presented for a duration of 300 ms with an SOA of 500 ms. The Wnal word was always followed by 2000 ms of blank screen. 2.2.2. EEG recording and analysis The electroencephalogram (EEG) was recorded from 64 electrodes distributed evenly over the scalp. See Fig. 1 for a schematic diagram of the layout of the 64 channel cap and the corresponding electrode labels. Eye movements and blinks were monitored via additional electrodes placed on

Please cite this article as: Todd R. Ferretti et al., Electrophysiological and behavioral measures of the inXuence of literal and Wgurative contextual constraints on proverb comprehension, Brain and Language (2006), doi:10.1016/j.bandl.2006.07.002.

ARTICLE IN PRESS 4

T.R. Ferretti et al. / Brain and Language xxx (2006) xxx–xxx

Fig. 1. Schematic diagram showing the electrode labels and sites for the 64 channel electrode caps used in Experiment 1.

the outer canthus and infraorbital ridge of each eye. Electrode impedances were kept below 5 k. EEG and was processed through a Neuroscan Synamps2 ampliWer set at a bandpass of 0.05–100 Hz, and was digitized at 250 Hz. 3. Results and discussion The data were re-referenced oV-line to the average of the right and left mastoids. High frequency noise was removed by applying a low-pass Wlter set at 30 Hz. ERPs were then computed in epochs that extended from 200 ms before the Wrst word of the sentence to 1000 ms after the Wnal word’s onset (i.e., ¡200 to 4000 ms). Single word epochs were calculated that extended from 100 ms before each word to 1000 ms following the onset of the words. Trials contaminated by blinks, eye-movements, and excessive muscle activity, were rejected oV-line before averaging; 25% of the trials were lost due to artifacts in the slow wave averages, and 20% of the trials were lost in the single word averages. We discuss the slow-wave analyses Wrst, followed by the single word analyses. 3.1. Slow-wave averages (¡200 ms to 4000 ms) Fig. 2 shows the slow-wave potentials for both contextual conditions at one frontal and one central/parietal electrode site (FZ, CPZ) located down the midline, and Fig. 3

shows the topographical distribution of the results for all channels. As can be seen, the slow-cortical waves for both contexts become progressively more positive across the proverbs. However, at about the third word of the proverbs the literal condition starts to become signiWcantly more positive than the Wgurative condition, and this positivity is sustained over the remaining words in the proverbs. The observed diVerences between the context types were largest over frontal sites, smallest over posterior sites, and had a similar distribution across the right and left hemispheres. 3.1.1. Overall analysis We conducted 3-way ANOVAs on the mean amplitude for each condition at 7 separate regions of interest: one for each 500 ms word region in the proverbs. The main factors of interest were Context (literal vs. Wgurative) and Electrode Site, both of which were within participants variables. List was used as a between participant factor to stabilize any variance caused by rotating participants across the diVerent lists (Pollatsek & Well, 1995). Table 2 shows the results for each region of interest. All p-values in this and subsequent analyses are reported after Epsilon correction (Huynh-Felt) for repeated measures with greater than one degree of freedom. As illustrated in Table 2, there was no signiWcant main eVect of Context and no interaction between Context and Electrode Site at the Wrst two words of the proverbs.

Please cite this article as: Todd R. Ferretti et al., Electrophysiological and behavioral measures of the inXuence of literal and Wgurative contextual constraints on proverb comprehension, Brain and Language (2006), doi:10.1016/j.bandl.2006.07.002.

ARTICLE IN PRESS T.R. Ferretti et al. / Brain and Language xxx (2006) xxx–xxx

5

Frontal (FZ) -5.0 -2.5 0.0 2.5 µV 5.0 7.5 10.0 12.5 15.0 17.5 20.0 -200.0 300.0 800.0 1300.0 1800.0 2300.0 2800.0 3300.0 3800.0

-5.0 -2.5 0.0 2.5 5.0 µV 7.5 10.0 12.5 15.0 17.5 20.0 -200.0 300.0 800.0 1300.0 1800.0 2300.0 2800.0 3300.0 3800.0

ms

ms

Central/Parietal (CPZ)

µV

-5.0 -2.5 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 -200.0 300.0 800.0 1300.0 1800.0 2300.0 2800.0 3300.0 3800.0

-5.0 -2.5 0.0 2.5 5.0 µV 7.5 10.0 12.5 15.0 17.5 20.0 -200.0 300.0 800.0 1300.0 1800.0 2300.0 2800.0 3300.0 3800.0

ms

ms

Fig. 2. Experiment 1 mean amplitudes at one frontal site (FZ) and one central/parietal site (CPZ) located on the midline Wltered with a low pass Wlter set at 30 Hz (left columns) and at .7 Hz (right columns) to reveal slow-wave development over proverbs. The Wgurative condition is shown in blue and the literal condition is shown in red. +10.0

Figurative Context

+9.1 +8.1 +7.2 +6.3 +5.3 +4.4 +3.4 +2.5

Literal Context

+1.6 +0.6 -0.3 -1.3 -2.2 -3.1 -4.1

3rd word

4

th

word

th

th

5 word

6 word

th

7 word

-5.0

Proverb Region Fig. 3. Topographical distribution of mean amplitudes in micro-volts at all electrode sites for each 500 ms word region starting at the third word of proverbs through the seventh word for Wgurative contexts and literal contexts.

However, at the third word of the proverbs, the main eVect of Context was signiWcant—greater positivity was found for proverbs located in literal compared to Wgurative contexts. From the fourth word through to the sixth word, the interaction between Context and Electrode Site was signiWcant. At the seventh word, the interaction was marginally signiWcant (p < .08).

3.1.2. Slow-wave distribution analysis In order to examine the observed interactions between Context and Electrode Site, we conducted a 5-way ANOVA on the proverb regions in which the interaction was signiWcant. In addition to the Context and List Factors used in the overall analysis, we added Hemisphere (left vs. right), Laterality (lateral vs. medial), and Anteriority

Please cite this article as: Todd R. Ferretti et al., Electrophysiological and behavioral measures of the inXuence of literal and Wgurative contextual constraints on proverb comprehension, Brain and Language (2006), doi:10.1016/j.bandl.2006.07.002.

ARTICLE IN PRESS 6

T.R. Ferretti et al. / Brain and Language xxx (2006) xxx–xxx

Table 2 Experiment 1 grand average (n D 20) slow-wave results for each of the 7 word regions (500 ms epochs) in the proverbs

Word 1st 2nd 3rd 4th 5th 6th 7th a ¤ ¤¤

Context

Electrode £ Context

F<1 F < 1.1 F(1,18) D 4.33¤ F(1,18) D 2.55 F(1,18) D 2.95 F(1,18) D 2.45 F(1,18) D 2.53

F<1 F<1 F < 1.1 F(61,1098) D 2.83¤¤ F(61,1098) D 2.49¤ F(61,1098) D 2.63¤ F(61,1098) D 1.96a

Table 4 Experiment 1 grand average N400 (300–500 ms) results for each of the 7 words in the proverbs

Word 1st 2nd 3rd 4th 5th 6th 7th ¤¤

D p > .05 < 10. p 6 .05. p < .01.

(prefrontal vs. frontal vs. parietal vs. occipital). All of these additional factors were within participants variables. The results of this analysis are presented in Table 3. We only report the results for the interactions involving context because those are the only interactions that are of theoretical interest here. As shown in Table 3, the Context by Anteriority interaction was signiWcant at all regions examined with, in each case, the mean amplitudes being more positive in the literal compared to the Wguratively biased contexts at prefrontal sites, all p’s < .002, and at frontal sites, all p’s < .001. In contrast, there were no diVerences between the two contexts at parietal and occipital sites. 3.2. Single-word averages 3.2.1. N400 analysis (300–500 ms) The results of the N400 analysis are presented in Table 4. Fig. 4 shows the N400 at one frontal electrode (FZ) and one central/parietal electrode (CPZ) located along the midline for the third word of the proverbs. Fig. 4 also shows the topographical distribution of the amplitudes for the two conditions in the N400 region for all electrodes. At the third word of the proverbs, the literal condition was less negative than the Wgurative condition, F(1,18) D 20.80, p < .001, and this eVect of context interacted with electrode site, F(61,1098) D 3.76, p < .01. No other main eVects or interactions for the N400 were found at any of the other words in

Context

Electrode £ Context

F<1 F < 1.25 F(1,18) D 20.80¤¤ F<1 F<1 F(1,18) D 2.47 F<1

F < 1.31 F<1 F(61,1098) D 3.76¤¤ F<1 F<1 F < 1.84 F<1

p < .01.

the proverbs. We examine the interaction between context and electrode site in the topographical distribution analysis reported below. 3.2.2. N400 distribution analysis The same 5-way distribution ANOVA reported above was also performed on the mean N400 amplitudes. The results of this analysis are reported in Table 5. As shown in the table, there was a marginal Context by Anteriority interaction, F(3,54) D 3.56, p < .06. This marginal eVect occurred because the N400 region was less negative for literal contexts than Wgurative contexts at frontal (mean diVerence D 2.65 V; F(1,54) D 42.90, p < .001) and prefrontal sites (mean diVerence D 2.11 V; F(1,54) D 27.14, p < .001) than at parietal (mean diVerence D 1.44 V; F(1,54) D 11.02, p < .05) and occipital sites (mean diVerence D .95 V; F(1,54) D 5.50, p < .05). The interaction between Context and Laterality reached signiWcance, F(1,18) D 4.51, p < .05. This interaction occurred because the diVerence between Wgurative and literal contexts was larger at medial electrode sites (mean diVerence D 2.02 V; F(1,18) D 136.74, p < .001) than at lateral sites (mean diVerence D 1.50 V; F(1,18) D 75.51, p < .001). Finally, the three-way interaction between Context, Anteriority, and Laterality also was signiWcant, F(3,54) D 6.06, p < .01. This interaction occurred because the diVerence between the Wgurative and literal contexts at anterior and posterior electrode sites was larger at medial rather than lateral sites. The observed diVerences in N400

Table 3 Experiment 1 topographic distribution results for the slow-wave averages at each of the 4 regions in which Context and Electrode Site interacted Proverb region 4th word C£A C£L C£H C£A£L C£A£H C£L£H C£A£L£H ¤ ¤¤

5th word ¤¤

F(3,54) D 7.12 F<1 F<1 F<1 F<1 F<1 F < 1.16

F(3,54) D 5.68 F<1 F<1 F < 1.60 F<1 F<1 F < 1.12

6th word ¤¤

F(3,54) D 5.62 F<1 F<1 F<1 F<1 F<1 F<1

7th word ¤¤

F(3,54) D 4.33¤ F<1 F < 1.13 F < 1.67 F<1 F<1 F<1

p < .05. p 6 .01.

Please cite this article as: Todd R. Ferretti et al., Electrophysiological and behavioral measures of the inXuence of literal and Wgurative contextual constraints on proverb comprehension, Brain and Language (2006), doi:10.1016/j.bandl.2006.07.002.

ARTICLE IN PRESS T.R. Ferretti et al. / Brain and Language xxx (2006) xxx–xxx

7

+5.0

A

B +4.4

Frontal (FZ)

Figurative Context (300–500 ms) +3.8

-5.0 -2.5

+3.1

0.0

µV

+2.5

2.5 5.0

+1.9

7.5 10.0 -100.0

+1.3

150.0

400.0

650.0

900.0 +0.6

ms

0

Central/Parietal (CPZ)

Literal Context (300–500 ms)

-0.6

-5.0 -1.3

-2.5 0.0

µV

-1.9

2.5 5.0

-2.5

7.5 -3.1

10.0 -100.0

150.0

400.0

650.0

900.0

-3.8

ms -4.4

-5.0

Fig. 4. (A) Experiment 1 single word grand averages at one frontal site (FZ) and one central/parietal site (CPZ) located on the midline for the third word of the proverbs. The Wgurative condition is shown in blue and the literal condition is shown in red. (B) The topographical distribution of the mean amplitudes for all electrode locations in the N400 region (300–500 ms) at the third word for both contexts.

between medial and lateral locations is frequently found (e.g., Federmeier & Kutas, 1999), and are consistent with research involving intracranial electrode recordings that show that the neurogenerators for the N400 are likely located in the anterior medial temporal lobes (e.g., McCarthy, Nobre, Bentin, & Spencer, 1995). 3.2.3. Late positivity analysis (700–1000 ms) The results of the LPC analysis for all electrode sites are reported in Table 6. As shown in the table, there was a signiWcant main eVect of context at the third word, F(1,18) D 4.90, p < .05. This eVect occurred because the literal condition was more positive than the Wgurative condition. There also was a signiWcant interaction between context and electrode site at the third word, F(61,1098) D 8.53, p < .001, and at the fourth word, F(61,1098) D 2.56, p < .01. The sustained frontal positivity in the slow-wave potentials for the literal condition is the primary cause of the main eVect of context and the interaction with electrode site. At posterior locations the LPC pattern is reversed (especially at the fourth word)—the literal condition is less positive than the Wgurative condition. The distribution analysis reported below provides statistical evidence for these claims. No other main eVects or interactions were found at any of the other words in the proverbs.

Table 5 Experiment 1 N400 and LPC topographic distribution results at each of the 4 regions in which context and electrode site interacted

C£A C£L C£H C£A£L C£A£H CxLxH CxAxLxH a ¤ ¤¤

3rd word (N400)

3rd word (LPC)

4th word (LPC)

F(3,54) D 3.56a F(1,18) D 4.51¤ F(1,18) D 2.14 F(3,54) D 6.06¤¤ F<1 F<1 F < 1.88

F(3,54) D 13.54¤¤ F(1,18) D 2.69 F(1,18) D 2.43 F(3,54) D 5.18¤¤ F < 1.51 F<1 F<1

F(3,54) D 2.69 F<1 F<1 F < 1.39 F<1 F<1 F(3,54) D 2.41a

D p > .05 < 10. p < .05. p 6 .01.

3.2.4. Late positivity distribution analysis The results of the topographic distribution analysis for the third and fourth words are presented in Table 5. At the third word, context and anteriority interacted, F(3,54) D 13.54, p < .001. This interaction occurred because the literal condition was much more positive than the Wgurative condition at prefrontal (mean diVerence D 2.71 V; F(1,18) D 33.05, p < .001) and frontal sites (mean diVerence D 2.65 V; F(1,18) D 31.57, p < .001), whereas these diVerences were much smaller and in the opposite

Please cite this article as: Todd R. Ferretti et al., Electrophysiological and behavioral measures of the inXuence of literal and Wgurative contextual constraints on proverb comprehension, Brain and Language (2006), doi:10.1016/j.bandl.2006.07.002.

ARTICLE IN PRESS 8

T.R. Ferretti et al. / Brain and Language xxx (2006) xxx–xxx

Table 6 Experiment 1 grand average late positivity results (700–1000 ms) for each of the 7 words in the proverbs

Word 1st 2nd 3rd 4th 5th 6th 7th ¤ ¤¤

Context

Electrode £ Context

F<1 F(1,18) D 2.22 F(1,18) D 4.90¤ F<1 F < 1.45 F<1 F < 1.80

F<1 F<1 F(61,1098) D 8.53¤¤ F(61,1098) D 2.56¤ F<1 F < 1.28 F < 1.14

p < .05. p < .01.

direction at parietal (mean diVerence D .35 V; F < 1) and occipital sites (mean diVerence D .30 V; F < 1). No other interactions reached signiWcance. Both the single-word and the slow-wave analyses reported above demonstrate that people had less diYculty integrating the proverbs when they are embedded in literal than in Wgurative contexts. Both types of averages converge to show that these diVerences began to occur by the third word of the proverbs and, in the slow-wave averages, these diVerences were sustained over the remaining words of the proverbs. The slow-wave results also demonstrated that the largest diVerences were found at prefrontal and frontal electrode sites, a Wnding that is consistent with previous research examining the relationship between slow-cortical waves in sentence processing and working memory load (e.g., King & Kutas, 1995; Münte et al., 1998). We take the slow-wave results as compelling evidence that people had less diYculty integrating the familiar proverbial statements when they were placed in literal rather than Wgurative-biasing contexts. The N400 and LPC Wndings in the single-word averages for the third and fourth word are more diYcult to interpret as a result of the large diVerences in slow-wave amplitudes at anterior electrode locations. For example, the distribution analysis for the N400 analysis demonstrated a two-way interaction between anteriority and context whereby the literal context was much less negative than the Wgurative contexts at anterior positions than at posterior locations. These diVerences are due to amplitudes at anterior locations capturing both the greater slow-wave positivity for the literal contexts and the less negative N400 in the literal contexts. Thus, it could be the case that our N400 results (and LPC results) are primarily driven by the slow-wave diVerences rather than diVerences in the N400 and LPC components. In order to establish whether our N400 and LPC eVects at the third word are signiWcant at posterior locations (where they typically are maximal) without the inXuence of the slow-wave diVerences at anterior locations, we conducted separate N400 and LPC analysis with all electrode locations anterior to the central electrodes removed from the analysis. We also conducted the same analysis for the LPC region at the fourth word.

The results of the N400 analysis without anterior electrodes at the third word revealed a robust main eVect of context, F(1,18) D 13.52, p < .01. As expected, the N400 for the literal condition (M D .16 V) was less negative than for the Wgurative condition (M D 1.67 V). However, the main eVect of context for the LPC region at the third word (F < 1) and fourth word (F(1,18) D 2.70, p D .12) did not reach signiWcance. Therefore, our results suggest that our N400 eVects at the third word are genuine, whereas our LPC eVects at the third and fourth word are primarily driven by the diVerences in the slow-wave amplitudes at anterior electrode locations. Our N400 results are consistent with previous Wgurative language research showing that people have less diYculty integrating the Wnal word of statements when they are intended to be taken as literal statements rather than as metaphoric or sarcastic statements (e.g., Coulson & Van Petten, 2002; Katz et al., 2004; Pynte et al., 1996). Interestingly, the N400 diVerences found between the two contexts were localized to the third word. This pattern is diVerent than the slow-wave analysis that indicated people had less diYculty integrating the familiar statements in the literal contexts from the third word through to the seventh word. These diVerences between single word and slow-wave averages should be interpreted with caution, however, as the 100 ms pre-stimulus baselines for the single word averages following the third word may be inXuenced by N400 eVects resulting from each preceding word. For example, because the SOA for each word was 500 ms, the pre-stimulus baseline for the fourth word will fall within the typical N400 range (i.e., 400–500). As a result, the amplitudes of the wave forms for the subsequent word will be increased or decreased to the degree that they are away from the baseline value in the 400–500 ms following the onset of the preceding word. The multiword averages reported above can be useful in this regard because they were conducted with a baseline that started 200 ms before the onset of the Wrst word of the proverb. As shown in Fig. 2, there is some evidence, albeit indirect, that N400 amplitudes are less negative in literal than Wgurative contexts at most of the other words in the proverbs, including the second, fourth, Wfth, and seventh word. These diVerences in N400 amplitude, however, are much smaller than that found at the third word. Although there was some evidence of a less positive LPC for the literal than Wgurative contexts at the fourth word, the diVerence between the two conditions was not signiWcant when only the central and posterior electrode locations where examined. Thus, the LPC data are similar to previous research (e.g., Coulson & Van Petten, 2002) as there was less positivity associated with the literal condition than the Wgurative condition at scalp locations in which the LPC is typically maximal, although in our case these diVerences did not reach signiWcance. The question remains, however, about why the diVerences between contextual conditions in the slow-cortical waves begin to be signiWcant at the third word and, simi-

Please cite this article as: Todd R. Ferretti et al., Electrophysiological and behavioral measures of the inXuence of literal and Wgurative contextual constraints on proverb comprehension, Brain and Language (2006), doi:10.1016/j.bandl.2006.07.002.

ARTICLE IN PRESS T.R. Ferretti et al. / Brain and Language xxx (2006) xxx–xxx

larly, why the largest diVerences in the N400 amplitude are found at third word. We believe that it is at this point that people have received enough of the familiar proverbial statements to recognize them as common Wgurative expressions. As a result, it is at this point that people can begin to integrate the well-known meanings of these phrases with the discourse contexts and, therefore, we begin to see diVerences in the ease in which people can integrate the meanings of the proverbs in the literal and Wgurative contexts. This claim is consistent with previous self-paced reading results by Katz and Ferretti (2001) that show by the second word of the proverbs people began to diVerentiate between familiar and unfamiliar proverbs, and this was true for both the literal and Wgurative contexts. However, it is important to note that an additional baseline condition would be required to ensure that the recognition of familiar proverbs employed here occurs at the same point in the proverbs. In summary, the slow-wave averages and the single word averages clearly show that integrating familiar proverbs into literal discourse contexts is less diYcult than when they are placed in Wgurative discourse contexts. However, these Wndings are not consistent with previous self-paced reading experiments that have shown that people read familiar proverbs at the same rate whether presented in Wgurative or literal biasing contexts (Katz & Ferretti, 2001, 2003). The self-paced reading data with proverbs are consistent with a large body of evidence for other instances of highly familiar forms of non-literal language and, as such, the Wndings of Experiment 1, based on an arguably more sensitive measure of temporal processing, have important theoretical implications. Because a diVerent set of passages were employed in the present study than in the earlier self-pace reading studies, it is diYcult to directly compare the present results with the earlier null eVect. For example, it could simply be the case that a self-paced reading study involving the current item set would produce reading times that are more consistent with the present ERP results. Consequently, Experiment 2 addresses whether the pattern observed here with ERP data would be replicated in a self-paced reading task when the same set of items are employed. 4. Experiment 2 Experiment 2 was designed to examine people’s moment-by-moment comprehension of familiar proverbs when they are presented in contexts that are either biased toward their literal or Wgurative meanings. The reading experiment is a replication and extension of Katz and Ferretti (2001, 2003). The main diVerence here is that we only investigate familiar proverbs and we examine proverbs that all have the same length. Using proverbs that are all the same length permits a Wner-grained analysis of word-byword processing than was employed in the previous experiments, in which the variable number of words in the middle region of the proverbs were averaged together as a single region. A second important diVerence is that we employ almost three times the number of items (i.e., 32 vs. 12), thus

9

providing an increased representative dataset and thus a greater likelihood of detecting reading time diVerences if they exist. Based on the results of previous studies examining the inXuence of literal and Wgurative contexts on Wgurative language comprehension (Gibbs et al., 1997; HoVman & Kemper, 1987; InhoV, Lima, & Carroll, 1984; Katz & Ferretti, 2001, 2003), it is expected that overall reading latencies for proverbs placed in literal and Wgurative biasing contexts should be similar across each of the critical regions examined. However, if the lack of a diVerence between the two contextual conditions found in previous research is due to small item sets and problems associated with averaging reading times in the mid-region of the proverbs, then we may Wnd, as we did with the ERP data in Experiment 1, some advantage in reading times for proverbs in the literal biasing contexts due to the greater conceptual overlap between the context and the proverbs than would be found between the proverbs and the words in the corresponding Wgurative biasing contexts. 5. Method 5.1. Participants Twenty-four participants participated in the reading experiment. All participants were native English speaking undergraduate students from Wilfrid Laurier University who received course credit for their participation. 5.2. Materials and procedure The materials were identical to Experiment 1. 5.2.1. Self-paced reading task The paragraphs were displayed on a 17 in. Apple monitor controlled by a Macintosh G4. They were presented using PsyScope (Cohen, MacWhinney, Flatt, & Provost, 1993) in a one-word-at-a-time moving window format. Paragraphs were initially presented on the screen with each non-space character replaced by a dash. Participants pressed a button to reveal the Wrst word of the paragraph. Each subsequent button press revealed the next word and replaced the previous word with dashes. Participants read each paragraph in this manner and then answered a yes–no comprehension question. Testing sessions began with 4 practice items. Participants then read the remaining 128 experimental trials. Each session lasted approximately 1 h. Reading latencies for each word were recorded by the computer and were measured as the time interval between successive button presses. 5.3. Design The analysis was conducted on 10 diVerent regions in and around the proverbs; the word immediately preceding the proverbs, each of the seven words comprising the prov-

Please cite this article as: Todd R. Ferretti et al., Electrophysiological and behavioral measures of the inXuence of literal and Wgurative contextual constraints on proverb comprehension, Brain and Language (2006), doi:10.1016/j.bandl.2006.07.002.

ARTICLE IN PRESS 10

T.R. Ferretti et al. / Brain and Language xxx (2006) xxx–xxx

erbs, and the two words immediately following the proverbs. The analysis for the two Wnal regions was a gauge of sentence “wrap-up” eVects. We conducted 2-way analysis of variance on each region. The main factor of interest was Context (literal vs. Wgurative), which was within participants (F1) and within items (F2). List (F1) and Item Rotation Group (F2) were used as a between factor to stabilize any variance caused by rotating participants and items across the diVerent lists (Pollatsek & Well, 1995). Any reading latency that was greater than three standard deviations from the mean was removed from the analysis. Items in which the participants incorrectly answered the comprehension question were removed from the analysis. 6. Results and discussion The mean reading times for each of the 10 critical regions are shown in Fig. 5. As can be seen, participants read each word in the proverbial statements at a similar rate, regardless of contextual bias. Moreover, there was no evidence of any diVerences in sentence wrap-up eVects at the two regions following the proverbs. In all regions, the diVerence in reading times was not signiWcant, all F’s < 1.3. The results are consistent with previous research by Katz and Ferretti (2001, 2003) showing that people read familiar proverbs at the same rate when they are embedded in literal and Wgurative biasing contexts. These results are also consistent with an extensive body of evidence that show that people do not need additional time to read other forms of Wgurative language, such as metaphors, idioms, and indirect speech acts, compared to their non-literal counterparts (e.g., Gibbs, 1989; Gibbs et al., 1997; HoVman & Kemper, 1987; InhoV et al., 1984). The results of Experiment 2 are inconsistent with all models that obligate the reader to Wrst process the literal sense of a proverb, such as the standard pragmatic approach of Searle (1979) or a literal-Wrst model speciWc to proverb processing (Honeck, 1997). We recognize that it is diYcult to draw conclusions from experiments that produce a null eVect, but we suggest that there are a number of reasons why the self-paced moving 400 Figurative Context 375

Literal Context

350 325 300 275 250 Before First Second Third Fourth Fifth

Sixth Seventh After 1 After 2

Proverb Region

Fig. 5. Experiment 2 mean reading times (ms). Error bars with larger caps capture the literal condition and bars with smaller caps capture the Wgurative condition.

window procedure used here would have detected diVerences between reading times had they been present. First, in our previous research examining the inXuence of literal and Wgurative contexts on proverb interpretation (e.g., Katz & Ferretti, 2001) we investigate both familiar and unfamiliar proverbs. As mentioned above, in that study we also did not Wnd diVerences in reading times for familiar proverbs but we did Wnd robust contextual inXuences on reading times for unfamiliar proverbs. Thus, in that study we replicate the same null eVect seen here but also show that moving window methodology is sensitive to diVerences in how easily people can construct Wgurative interpretations in the literal and Wgurative contexts for unfamiliar proverbs. Moreover, in the present study we used almost three times the number of items per condition and controlled for the overall length of the proverbs and still did not Wnd any reading time diVerences. Second, as mentioned above, the null eVects observed here are consistent with a large body of evidence that suggests that people often do not need more time to read Wgurative statements in Wgurative than literal contexts. Third, other researchers have shown that moving window methodology is sensitive to the earliest moments of processing in other domains of language comprehension, such as when examining the interaction between syntax and semantics during syntactic ambiguity resolution (e.g., McRae, Spivey-Knowlton, & Tanenhaus, 1998). Fourth, examination of Fig. 5 and our statistical analysis show that there were no diVerences in reading times that were even close to being signiWcant. In short, we believe that our present experiment would have been sensitive to reading times had they been present. A diVerent, but related issue, is how the diVerent presentation rates for each word across the two experiments may have led to the diVerences observed between the experiments. Although words were presented one at a time in both experiments, in Experiment 1 the words were presented for a 300 ms duration with an SOA of 500 ms (note that this is a standard SOA used in ERP investigations of language comprehension), whereas in Experiment 2 people were able to read each word at their own rate. It is interesting to note that people in Experiment 2 read each word for approximately 325 ms on average, only slightly slower than the 300 ms word presentation duration in Experiment 1. However, given the SOA of 500 ms in Experiment 1, this suggests that people were forced to read at a rate of 175 ms per word slower on average in Experiment 1 than Experiment 2 (keep in mind that we comparing two diVerent groups of people). In our opinion, it is diYcult to see how slowing people down while they read by such a small amount of time would make it harder rather than easier to integrate the exact same phrases in Wgurative contexts than for literal contexts. For example, given that the phrases were familiar statements we would expect, as suggested above, that people would probably retrieve the meaning of the phrases long before they Wnished reading those phrases. Presumably giving people more time to retrieve the meaning of those familiar phrases by slowing down presentation

Please cite this article as: Todd R. Ferretti et al., Electrophysiological and behavioral measures of the inXuence of literal and Wgurative contextual constraints on proverb comprehension, Brain and Language (2006), doi:10.1016/j.bandl.2006.07.002.

ARTICLE IN PRESS T.R. Ferretti et al. / Brain and Language xxx (2006) xxx–xxx

rates would give them more time to integrate the meaning of those familiar statements with the preceding discourse contexts. Thus, it is diYcult to see how the slight diVerences in word presentation rate across the two experiments would lead to more diYculty in the Wgurative than literal condition in Experiment 1, but not lead to any diVerences in Experiment 2. 7. General discussion The present research extends previous research on Wgurative language processing in a number of ways. Perhaps the most intriguing Wnding is that our ERP results show that people have less diYculty when processing familiar proverbial phrases placed in contexts that invite a literal interpretation than when placed in Wguratively biasing contexts, whereas no such diVerences were found for the same items in a self-paced reading task. In our case, we directly compared self-paced reading times and ERP potentials to identical stimulus sets and found clear diVerences in the pattern of results across the two measures. The diVerences between the two methodologies speak to the importance of using converging measures to investigate Wgurative language processing. In particular, our results add to a small but growing body of literature that shows how ERP methodology provides a measure that is more sensitive than reading time measures in indexing the diYculty readers have when comprehending Wgurative statements embedded in diVerent contexts (e.g., Coulson & Van Petten, 2002; Katz et al., 2004). In the present case, we examined slow-wave potentials that are known to be sensitive to working memory load during sentence processing and also N400 and LPC amplitudes. As such, the diVerences in ERP amplitudes we observe here between the literal and Wgurative-inviting contexts is consistent with the interpretation that the diVerences reXect a greater ease in integrating the proverb with the discourse context when it is being used in its literal sense than in its familiar non-literal sense, an interpretation also consistent with that given by Coulson and Van Petten (2002) in their ERP study of metaphor processing. Moreover, here we Wnd that this diVerence emerges by the third word of the proverb. Thus problems in integrating the proverbial sense of the sentence into the emerging discourse structure is found well before the reading of the proverb has been completed. Recall that the proverbs we employed were familiar and, therefore, the non-literal, proverbial sense should have been highly salient (see Giora, 2003). Consequently, problems in integration found with the Wgurative context are not likely to be due to a problem in accessing a viable non-literal meaning of the proverb per se, but to diYculty in creating coherent discourse. We argue here that this diVerence is due to the characteristic of proverbs that, when used literally, they tend to be true both literally and Wguratively. Thus a proverb used literally shares greater conceptual overlap with the context than when the same statement is preceded by non-literal context.

11

7.1. Implications for models of Wgurative language processing Our results are most consistent with models that posit that people construct a Wgurative interpretation very early in the processing of a Wgurative statement when the contexts are suYciently constraining (Gibbs, 1994; Gibbs et al., 1997; Katz & Ferretti, 2001). The fact that we Wnd diVerences in slow-wave potentials for the two contexts at the third word of the proverb is strong evidence against models of Wgurative language processing which posit that the literal meaning of an entire statement must be processed prior to constructing a Wgurative meaning (e.g., Grice, 1975). The Wndings also appear problematic for another theory that assumes obligatory access of sentence meaning, namely, the graded salience hypothesis (Giora, 2003). Proponents of this approach would need to explain why the familiar (and hence salient) proverbs used in the ERP experiment were more diYcult to integrate into the Wgurative context than into the literal context, because that theory proposes that it should be at least as easy, and maybe even easier, to integrate into the Wgurative context than into the literal context. Taken together, our data is consistent with models that assume we actively construct interpretations during discourse processing, rather than retrieve entrenched meanings from semantic memory. In conclusion, our results indicate the utility of using ERP methodology to investigate slow-wave potentials that develop over sentences and clauses for the on-line investigation of Wgurative language processing. The present research is the Wrst to show that people have less diYculty integrating familiar proverbs presented in literal contexts than Wgurative contexts, data consistent with theories that emphasize the active construction of meaning (literal and non-literal alike) during discourse processing. References Cohen, J. D., MacWhinney, B., Flatt, M., & Provost, J. (1993). PsyScope: A new graphic interactive environment for designing psychology experiments. Behavioral Research Methods, Instruments & Computers, 25, 257–271. Coulson, S., & Van Petten, C. (2002). Conceptual integration and metaphor: an event-related potential study. Memory & Cognition, 30, 958–968. Federmeier, K. D., & Kutas, M. (1999). A rose by any other name: longterm memory structure and sentence processing. Journal of Memory and Language, 41, 469–495. Gibbs, R. W. (1989). Understanding and literal meaning. Cognitive Science, 13, 243–251. Gibbs, R. W. (1994). The poetics of mind. Cambridge, England: Cambridge University Press. Gibbs, R. W. (1995). What proverb understanding reveals about how people think. Psychological Bulletin, 118, 133–154. Gibbs, R. W., Bogdanovich, J. M., Sykes, J. R., & Barr, D. J. (1997). Metaphor in idiom comprehension. Journal of Memory and Language, 37, 141–154. Giora, R. (2003). On our mind: Salience, context and Wgurative language. Oxford, England: Oxford University Press. Grice, H. P. (1975). Logic and conversation. In P. Cole & J. Morgan (Eds.), Syntax and Semantics (Vol. 3, pp. 41–58). New York: Academic Press.

Please cite this article as: Todd R. Ferretti et al., Electrophysiological and behavioral measures of the inXuence of literal and Wgurative contextual constraints on proverb comprehension, Brain and Language (2006), doi:10.1016/j.bandl.2006.07.002.

ARTICLE IN PRESS 12

T.R. Ferretti et al. / Brain and Language xxx (2006) xxx–xxx

HoVman, R., & Kemper, S. (1987). What could reaction time studies be telling us about metaphor comprehension? Metaphor and Symbolic Activity, 2, 149–186. Honeck, R. (1997). A proverb in mind. Mahwah NJ: Erlbaum. InhoV, A. W., Lima, S. D., & Carroll, P. J. (1984). Contextual eVects on metaphor comprehension in reading. Memory & Cognition, 12, 558–567. Katz, A. N., Blasko, D. G., & Kazmerski, V. A. (2004). Saying what you don’t mean: social inXuences on sarcastic language processing. Current Directions in Psychological Science, 13, 186–189. Katz, A. N., & Ferretti, T. R. (2001). Moment-by-moment comprehension of proverbs in discourse. Metaphor and Symbol, 16, 193–221. Katz, A. N., & Ferretti, T. R. (2003). Reading Proverbs in context, the role of explicit markers. Discourse Processes, 36, 19–46. King, J., & Kutas, M. (1995). Who did what and when? Using word- and clause-level ERPs to monitor working memory usage in reading. Journal of Cognitive Neuroscience, 7, 376–395. Kutas, M., & Hillyard, S. A. (1984). Reading senseless sentences: brain potentials reXect semantic incongruity. Science, 207, 203–205.

McCarthy, G., Nobre, A. C., Bentin, S., & Spencer, D. D. (1995). Language-related Weld potentials in the anterior-medial temporal lobe: I. Intracranial distribution and neural generators. Journal of Neuroscience, 15, 1080–1089. McRae, K., Spivey-Knowlton, M. J., & Tanenhaus, M. K. (1998). Modeling the inXuence of thematic Wt (and other constraints) in on-line sentence comprehension. Journal of Memory and Language, 38, 283–312. Münte, T. F., Heinze, H. J., & Kutas, M. (1998). When temporal terms belie conceptual order: an electrophysiological analysis. Nature, 396, 71–73. Pollatsek, A., & Well, A. D. (1995). On the use of counterbalanced designs in cognitive research: a suggestion for a better and more powerful analysis. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21, 785–794. Pynte, J., Besson, M., Robichon, F., & Poli, J. (1996). The time course of metaphor comprehension: an event related potentials study. Brain and Language, 55, 293–316. Searle, J. R. (1979). Expression and meaning: Studies in the theory of speech acts. Cambridge: Cambridge University Press.

Please cite this article as: Todd R. Ferretti et al., Electrophysiological and behavioral measures of the inXuence of literal and Wgurative contextual constraints on proverb comprehension, Brain and Language (2006), doi:10.1016/j.bandl.2006.07.002.