The comprehension of ambiguous idioms in aphasic

Neuropsychologia 44 (2006) 1305–1314

The comprehension of ambiguous idioms in aphasic patients Cristina Cacciari a , Fabiola Reati b , Maria Rosa Colombo c , Roberto Padovani a , Silvia Rizzo b , Costanza Papagno b,∗ a

Dipartimento di Scienze Biomediche, Università di Modena e Reggio-Emilia, Italy b Dipartimento di Psicologia, Universit` a di Milano-Bicocca, Italy c Servizio di Riabilitazione, Azienda, Ospedaliera Ospedale di Legnano, Italy

Received 7 April 2005; received in revised form 12 January 2006; accepted 22 January 2006 Available online 28 February 2006

Abstract The ability to understand ambiguous idioms was assessed in 15 aphasic patients with preserved comprehension at a single word level. A stringto-word matching task was used. Patients were requested to choose one among four alternatives: a word associated with the figurative meaning of the idiom string; a word semantically associate with the last constituent of the idiom string; and two unrelated words. The results showed that patients’ performance was impaired with respect to a group of matched controls, with patients showing a frontal and/or temporal lesion being the most impaired. A significant number of semantically associate errors were produced, suggesting an impairment of inhibition mechanisms and/or of recognition/activation of the idiomatic meaning. © 2006 Elsevier Ltd. All rights reserved. Keywords: Idioms; Aphasia; Figurative language

1. Introduction Idioms are among the most common forms of figurative language (Gibbs, 1999). The meaning of idioms is highly conventional in that their meaning generally cannot be predicted from the meaning of their constituent parts. Notwithstanding, the syntactic as well as many of the semantic features characterizing their constituents are still preserved. For instance, one cannot slowly “kick the bucket”, neither idiomatically nor literally, since “to kick” cannot be used to refer to a slow action (Hamblin & Gibbs, 1999). Idioms do not form a unitary class and rather vary along a number of syntactic and semantic dimensions (Nunberg, Sag, & Wasow, 1994; Cacciari & Glucksberg, 1995). First, idioms vary as to their semantic transparency, namely in the extent to which the motivation for their structure can be recovered. Idioms can involve figuration (although there are idioms not involving

Abbreviations: LBD, left brain-damage; rTMS, repetitive transcranial stimulation ∗ Corresponding author at: Dipartimento di Psicologia, Universit` a di Milano Bicocca, Edificio U6, Piazza dell’Ateneo Nuovo 1, 20126 Milano, Italy. Tel.: +39 02 64486738; fax: +39 02 64486706. E-mail address: [email protected] (C. Papagno). 0028-3932/$ – see front matter © 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.neuropsychologia.2006.01.012

figuration at all, such as “by dint of”) and can be originally metaphorical (e.g., “take the bull by the horns”), even if speakers may not perceive the figure originally involved (Nunberg et al., 1994). When an idiom is semantically opaque, the speaker needs to know the stipulated meaning that cannot be derived neither from the image evoked nor from the constituent word meanings. Second, idioms vary as to their decomposability, namely in the extent to which the idiomatic interpretation can be mapped onto single constituents (Gibbs, Nayak, & Cutting, 1989). Third, idioms vary in the extent to which they can be syntactically transformed still retaining their idiomatic meaning (Gibbs & Gonzales, 1985). Finally, some idioms do not have any wellformed literal counterpart while others instead have it and are “ambiguous” (e.g., “break the ice”). Early theories of idiom comprehension assumed a literal meaning priority and a search for a figurative interpretation only when the literal one was defective. For example, Bobrow and Bell (1973) proposed the Idiom List hypothesis according to which idioms were fixed expressions whose meaning was sought in an idiom list whenever the literal reading of the string made no sense in context. Differently, the most influential Lexical Representation hypothesis posited that idioms behaved as long, morphologically complex words stored in the mental lexicon together with the other lexical units (Swinney & Cutler, 1979).

1306

C. Cacciari et al. / Neuropsychologia 44 (2006) 1305–1314

Linguistic processing of the string and retrieval of the idiomatic meaning was supposed to proceed in parallel with the second faster than the first. Gibbs (1984) proposed a more extreme version of this hypothesis arguing that people did not engage in any linguistic analysis at all and could entirely bypass the literal meaning directly accessing the figurative interpretation of the idiom string. A last different model was proposed by Cacciari and Tabossi (1988), the Configuration hypothesis, according to which idioms are configurations of words that undergo a linguistic analysis until enough information had accumulated to prompt the recognition of the idiomatic nature of the string and the subsequent activation of the related figurative meaning. All these hypotheses are based on language unimpaired participants and assume that, in order to understand an idiom, lexical integrity is required. However, this is precisely what cannot be taken for granted with aphasic patients whose lexicalsemantic and/or a syntactic impairments have shown to limit figurative language comprehension (Papagno, Tabossi, Colombo, & Zampetti, 2004). However, a widely accepted view in the neuropsychological literature assumes that damage to the right hemisphere, but not to the left one, has major consequences on the processing of figurative language in general, and of idiomatic expressions in particular (e.g., Van Lancker & Kempler, 1987). Recent evidence questioned this view (e.g., Tompkins, Boada, & McGarry, 1992) showing that idiomatic meaning comprehension is impaired in left-hemisphere damaged patients (Papagno & Genoni, 2004) and in healthy people as well, when the activity of their left hemisphere is selectively disrupted by repetitive transcranial magnetic stimulation (rTMS) (Oliveri, Romero, & Papagno, 2004). The proponents of the right hemisphere hypothesis also posited that idioms were processed like unitary, non-syntactically analysed strings, an assumption contradicted by evidence that language-unimpaired speakers syntactically parsed the idiom string even after its idiomatic meaning was already retrieved (Peterson, Burgess, Dell, & Eberhard, 2001). More specifically, the results collected so far on aphasic patients show that they have a strong bias toward the literal interpretation of the idiom string even when the idiom only has a non-literal interpretation (Papagno et al., 2004). However, this result might at least in part depend on the testing modality that has proved to be extremely relevant in neuropsychological studies, even more than in studies on language-unimpaired participants (Tompkins et al., 1992). The two tasks more often used in idiom studies with patients, the string-to-picture matching task and the oral definition task (but see Hillert, 2004), are both problematic. The string-to-picture matching task can underestimate idiom comprehension, because the picture representing the literal interpretation (often a bizarre image, especially when the idiom string has not a literal meaning) can strongly interfere with the correct response, similarly to what happens in the Stroop effect (MacLeod, 1991; Stroop, 1935). This has proved to be the case for patients with probable Alzheimer’s disease (AD) (Papagno, Lucchelli, Muggia, & Rizzo, 2003), whose idiom comprehension skills were tested using a sentence-topicture matching task. The 15 AD patients involved in this study were asked to select the picture corresponding to the figurative meaning of the idiom string between two alternatives, respec-

tively, representing the figurative and the literal interpretation. Idiom comprehension was very poor with respect to a group of matched controls and was correlated with the performance on tasks assessing executive abilities. When the idiom test was repeated using an unrelated picture, instead of the one depicting the literal meaning of the idiom string, the patients’ performance significantly improved suggesting that the picture representing the literal interpretation strongly interfered with idiomatic meaning retrieval. The second task employed in many idiom comprehension studies, the oral definition task, has important limitations as well: a deficit in speech output, as found in non-fluent aphasic left brain-damaged (LBD) patients, can in fact lead to an underestimation of their ability to understand idioms since they might be unable to provide a verbal explanation of an idiomatic meaning they actually know. Therefore, the aim of this study was to investigate idiomatic meaning comprehension in aphasic patients, without a semantic deficit at a word level, employing a task that might overcome the limitations just outlined. Hence, to avoid the “interference” of the literal interpretation and any verbal fluency effect, we used a string-to-word matching task (see below). We employed idiomatic expressions that have both a figurative and a literal interpretation (i.e., ambiguous idioms) as experimental material. Since these idioms have not yet been tested in aphasic patients, the results of the present study might extend what we currently know on idiomatic processing, allowing a more fine-grained assessment of idiomatic meaning comprehension impairment in aphasia. The figurative interpretation of familiar idiom strings, as those employed in this study, is usually more frequent than the literal one. Consequently, the idiomatic interpretation is almost always the dominant one, and the literal the subordinate one (Cronk, 1990). In any case, the participants were advised that the idiom string was meant idiomatically even though it also had a literal interpretation. Given the heterogeneity of idiomatic expressions in terms of syntactic structure, literal ambiguity, semantic transparency, and so forth (see Nenonen, Niemi, & Laine, 2002), we cannot exclude that different processing mechanisms might be involved in processing different types of idioms. If the right hemisphere hypothesis is correct, then we should expect aphasic patients to comprehend idiomatic expressions as control participants do. A normal performance in aphasic patients with normal word comprehension skills also would support the view that idioms are just long words, as proposed by the Lexical Representation hypothesis. On the contrary, if idiom processing indeed requires morpho-syntactic and lexical-semantic processing (Cacciari & Tabossi, 1988; Peterson et al., 2001), then we should expect aphasic patients to be impaired in understanding idiomatic expressions as previously shown for idioms without a literal meaning (Papagno et al., 2004), and for verb phrase idioms (Nenonen et al., 2002). In sum, the aim of this study was to assess ambiguous idiom comprehension in aphasic patients characterised by a good comprehension at a word level. An impairment in idiom comprehension, but not in word comprehension, would further question the Lexical Representation hypothesis.


2. Materials and method 2.1. Participants Fifteen aphasic patients (12 men and 3 women, age range 36–77, education range 5–18) were selected on the basis of the following criteria: • Only LBD right-handed patients with a single focal vascular lesion were included. The site and extension of the lesion was evaluated by means of a CT-scan. • A minimum of 5 years of education was requested. • Aphasia of a mild to moderate severity. The severity was assessed by means of a standardised language examination (the Italian version of the Aachener Aphasie Test: AAT) and of the Token Test (De Renzi & Faglioni, 1978): only patients with an intact comprehension of single words, as shown by the results in the word comprehension subtest of the language examination, participated in the study. • Absence of non-verbal neuropsychological deficits, such as visuo-perceptual and non-verbal intelligence difficulties, was required. Demographical and clinical data of the patients are reported in Table 1. Fifteen normal participants were matched one by one for age and educational level to the patients, as controls.

2.2. Material 2.2.1. Idiom comprehension Twenty-three familiar verbal idioms were selected (see Appendix A). They were ambiguous in that the string had a literal meaning as well. In order to control for the semantic transparency of the idiom string, a rating study was run in which 30 language-unimpaired participants were presented with the idiom list and were asked to rate how easily could the idiomatic meaning be derived from the meaning of the words forming each idiom string. A score ranging from 0 to 3 was assigned where 0 meant “not at all” and 3 “perfectly” (mean 1.4, range 0.4–2.2). We also controlled for the plausibility of the literal meaning of the idiom string. The same 30 language-unimpaired participants were also asked to rate the extent to which the literal meaning of the string was plausible. A score ranging from 0 (implausible) to 3 (absolutely plausible) was assigned to each string (mean 1.9, range 0.25–3) (see Appendix A). Finally, they were asked to rate the familiarity of each idiom string with a score of 0 (the idiom was unknown), 1 (known but not sure about the meaning), 2 (familiar with a well-known meaning), or 3 (highly familiar) (their mean familiarity ratings ranged from 1.19 to 2.98) (see Appendix A). We used a sentence-to-word matching task in which each idiomatic expression was paired with four target words matched in terms

1307

of length (mean number of characters 6.0, range 4–11) and written frequency (mean 10.4, range 2–26) (see Appendix B): the target word corresponded to the idiomatic interpretation of the string (henceforth idiomatic target; e.g., WINE, for “alzare il gomito”, “to raise the elbow”, i.e., to drink too much); one foil was semantically associated with the last constituent word of the idiom string (in the previous example, LEG); and two words were unrelated foils (TREE, BOX). Specifically, the first type of unrelated target was either an abstract or a concrete word depending on the nature of the idiomatic target: the unrelated target was abstract if the idiomatic target was abstract, and concrete if the idiomatic target was concrete. A target word was considered as concrete based on the availability of the word referent to sensory experience. The second type of unrelated target was a word that could plausibly complete the verb in the verb phrase (BOX). Each idiom was presented in a syntactically simple sentence formed by a subject followed by the idiom string (e.g., “he has raised the elbow”). The participants’ task was to point to the word that matched the idiomatic meaning of the sentence: the examiner read the sentence followed by the four alternatives that were written on cards and remained in front of the patient until a decision was taken. For example, the examiner read the sentence “he has raised the elbow” and then the four alternatives (“wine, leg, box, tree”) were presented. Participants were informed that they would be read sentences commonly used in everyday conversation that had a non-literal interpretation. An example was provided to clarify that the string had to be interpreted figuratively. The rationale whereby we selected these four types of targets is the following: the choice of the idiomatic target should reflect the knowledge and availability of the idiomatic meaning of the idiom string. The choice of the semantically associate foil might reflect an attempt at interpreting the string literally when the patient did not know the idiom meaning, or alternatively that s/he was unable to access the idiomatic interpretation of the idiom string. The semantic associate foil, however, does not reflect the literal meaning of the sentence and its choice is clearly an error. The two unrelated foils should signal an impaired performance of both the idiomatic and the literal processing of the string. We tested the strength of the relationship between: (a) the semantically associate foil and literal meaning of the string, and (b) the idiomatic target and the idiomatic meaning of the idiom string. To this aim, 25 participants were asked to rate on two five point-rating scales ranging from 0 (not at all related) to 5 (perfectly related) the specific relatedness of (a) and (b). Idiomatic targets were rated as significantly more related to the idiom strings than semantically associate targets to the literal meaning of the strings [4.57 (range 3.68–4.92, S.D. 0.28) versus 1.61 (1–3.28, S.D. 0.67)] [t (44) = 19.49, p < 0.00001], a result that might help the choice of an idiomatic target by aphasic patients. At the end of the experiment the patients were asked to specify whether they knew the meaning of each idiom string and to rate how familiar was each idiom, as was done with normal participants: the patients’ idioms mean familiarity ratings were very similar to those provided by language unimpaired participants

Table 1 Demographical and clinical data of the 15 aphasic participants

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Age

Gender

Education

Etiology

Lesion site

Aphasia type

Interval post-stroke

59 68 67 71 66 68 49 36 77 74 60 66 64 59 61

M M M M M M M W W W M M M M M

11 5 5 16 5 8 11 13 8 5 13 5 6 17 18

I I H I I I I I H I I I I H H

F–P F–T, basal g F–T F–P F–T Insula, T–P T–P T–P T–P T–P T–P T–P P T–P P

Broca Broca Amnestic Amnestic Broca Wernicke Broca Broca Wernicke Wernicke Wernicke Amnestic Amnestic Wernicke Broca

6 17 10 14 22 14 24 10 6 3 5 2 3 10 5

Legend: M = man; W = woman; I = ischemic; H = haemorragic; F = frontal; T = temporal; P = parietal.

1308


Table 2 Results of the 15 aphasic patients on the linguistic tasks

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Token test

Word comprehension

Sentence comprehension

Idiom comprehension: choice of the correct target word

Literal associate

19.5 16 14 28 19.5 15 26 18.5 18 13.5 22 19 16.5 15 17

77 78 79 71 77 79 80 77 80 78 80 80 78 79 80

53 49 49 61 53 43 60 51 60 34 53 53 53 37 53

13 5 14 16 7 17 18 16 16 14 15 19 20 20 20

8 12 5 7 7 6 3 3 4 6 7 2 1 2 1

and ranged from 1.22 to 3 (the familiarity ratings for each idiom string of both groups are reported in Appendix A). Word comprehension (Laiacona, Barbarotto, Trivelli, & Capitani, 1993). The word comprehension test included 80 common nouns belonging to eight different categories (fruits, vegetables, animals, furniture, vehicles, tools, body parts, musical instruments). For each stimulus, five pictures, corresponding to the target and to four foils, respectively, were arranged vertically in a column on a card. The foils belonged to the same category, therefore, only semantically associate errors were possible. For example, if the target word was “strawberry”, other four words denoting fruits were presented together with the correct one. A stimulus word was read aloud by the examiner. The task of the patient was to choose the picture corresponding to the target. Sixty controls reached a score of 77/80 (97%) correct answers, with a range of 72–80. 2.2.2. Syntactic comprehension Patients performed a literal sentence comprehension test, in the pointingto-picture modality, including 64 sentences (from a battery designed by Miceli & Capasso, unpublished). The examiner read the sentence to the patient who was asked to choose one picture of two, three or four alternatives. Languageunimpaired participants produce a maximum of two errors in this task.

2.3. Anatomical templates Lesions were mapped onto standard templates following the procedure described by Damasio and Damasio (1989). First, the angle of incidence in

Unrelated 1

2 4 1 0 4 0 1 1 1 1 0 1 0 0 1

Unrelated 2

0 2 3 0 5 0 1 3 2 2 1 1 2 1 1

Idiom target comprehension 14 11 18 18 11 15 14

17 13 18

which tomographical cuts were obtained was determined by taking into account the relative position of anatomical, cerebral and bone landmarks. The set of best-fitting templates was then chosen and each patient’s lesion charted on the appropriate templates using an x/y-plotting approach.

3. Results On the basis of their performance on the AAT, six patients were diagnosed as having Broca’s aphasia, five were classified as Wernicke’s aphasics, and the remaining four had an amnestic aphasia. The severity of aphasia varied from mild to moderate, without significant difference between non fluent and fluent aphasics, as shown by their performance on the Token Test [U (6, 9) = 17, p = 0.24] (see Tables 2 and 3). All patients showed a good comprehension of single words ranging from 71/80 to 80/80 (mean 78.2, 97.75%) (see Table 2), confirming the results obtained on the AAT with a limited (10) number of stimuli. Syntactic comprehension was mildly impaired, ranging from 34 to 61 (mean 50.8, S.D. 7.76). The mean number of correct responses of aphasic patients in the idiom comprehension task was 15.33 (66.65%) (S.D.

Table 3 Results of the 15 normal participants in the idiom comprehension task

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Age

Education

Idiom comprehension correct

Literal associate

Unrelated 1

Unrelated 2

61 70 65 62 66 79 50 33 66 73 64 68 58 60 68

11 5 8 18 5 6 11 13 8 5 8 13 13 18 8

23 22 22 23 21 23 23 23 23 19 21 21 23 23 23

0 1 0 0 0 0 0 0 0 2 1 1 0 0 0

0 0 1 0 1 0 0 0 0 0 0 0 0 0 0

0 0 0 0 1 0 0 0 0 2 1 1 0 0 0


4.43; range of correct responses: 5–20). In contrast, controls (see Table 3) produced on average 22.13 (96.22%) correct responses (S.D. 1.19, range 19–23). Since the numerical difference between the two types of unrelated errors was small, and statistically not significant (14.78% of unrelated abstract/concrete errors versus 20.86% of verb completion errors, t = −0.96, n.s.), the two error types were pooled together. The ANOVAs on the number or errors were conducted considering both participants (Fp ) and items (Fi ) as random factors. In the by-participant analyses, error type (semantically associate foil versus unrelated foils) was a within-subject factor, and group (patients versus controls) a between-subject factor. In the by-item analyses, error type and group were within-subject factors. A significant effect of the group factor was observed [Fp (1, 28) = 32.91, p < 0.0001; Fi (1, 22) = 43.9, p < 0.0001], being patients’ performance significantly worse than that of controls. Also the error type factor was significant only in the byparticipant analysis [Fp (1, 28) = 6.48, p = 0.01; Fi (1, 22) = 1.7, p = 0.21], being patients’ semantically associate errors (64.34%) more frequent than unrelated errors (35.65%). The interaction between group and error type was also significant in the byparticipant analysis and close to significance in the by-item analysis [Fp (1, 28) = 7.32, p = 0.01; Fi (1, 22) = 3.1, p = 0.09]. Post hoc analyses on the by-participants results (Sheffé test) showed that patients produced significantly more semantically associate errors (p < 0.0001) and more unrelated errors (p < 0.0001) than controls. Moreover, aphasic patients, but not controls (p = 0.99), produced significantly more semantically associate errors than unrelated errors (p < 0.01). The correlation between semantic associate errors and relatedness scores was significant (r = 0.71, p < 0.0001) in that the higher the relatedness of the semantically associated target to the string, the more the patients chose it. The correlation between semantically associate errors and word comprehension scores (r = −0.35) was not significant, although a ceiling effect in the word comprehension test (97.75% correct responses) might account for this result. No significant correlation was also observed between semantically associated errors and sentence comprehension scores (r = 0.02). Unrelated errors were less frequent in Wernicke’s than in Broca’s patients (6.9% versus 18.1% of all responses), but the small sample of patients within each group did not allow us to perform any further analysis. No significant correlation was found between the number of correct responses and the following factors: educational level, semantic transparency of the idiomatic meaning, plausibility of the literal meaning, Token Test scores, sentence comprehension scores (the correlation values are reported in Table 4). Although the semantic well-formedness of the concurrent, but subordinate, literal meaning of the string did not specifically affect the patients’ responses, as suggested by the lack of correlation between the patients accuracy and the mean plau-

1309

Table 5 Number of correct target choices and of semantic and unrelated (UR) foils for each type per idiom in patients Idiom

Number of correct responses Semantic UR

Alzare il gomito Vuotare il sacco Sputare il rospo Vedere le stele Essere al verde Prendere la porta Levare le tende Perdere la testa Avere il pollice verde Prendere un granchio Tirare la cinghia Perdere il filo Cambiare pagina Mettersi le mani nei capelli Tagliare la corda Scendere a rotta di collo Avere le mani bucate Saltare la mosca al naso Mettere la pulce nell’orecchio Avere carta bianca Alzare i tacchi Attaccare un bottone Stringere i denti

11 7 3 3 14 7 11 12 9 13 10 9 4 14 14 13 14 8 11 6 14 10 13

4 5 0 11 1 7 2 3 2 0 2 4 11 0 0 0 1 5 3 6 1 4 2

0 3 12 1 0 1 2 0 4 2 3 2 0 1 1 2 0 2 1 3 0 1 0

sibility ratings of the idiom strings, it may have affected the errors performed by the patients. Indeed this was the case, as shown by two one-way ANOVAs conducted on semantic versus unrelated errors considering only the idiom strings (15 out 23) whose mean plausibility score was higher than 1.5 (in a three-point scale ranging from 1 indicating totally implausible to 3, absolutely plausible) [Fp (1, 14) = 16.8, p = 0.001; Fi (1, 14) = 6.5, p = 0.023]. The results show that semantic errors were significantly more frequent than unrelated errors (74.7% versus 25.3%). Despite the fact that patients affirmed to know all the idioms employed in the study, the correlation between the familiarity of the idiom strings and the choice of the idiomatic targets was significant (r = 0.46, p = 0.028), while in normal participants almost reached significance (r = 0.39, p = 0.06): the more an idiom was perceived as familiar, the more the idiomatic target was chosen. This familiarity effect is consistent with previous evidence on language unimpaired participants who had faster reaction times on familiar idiom strings than on unfamiliar ones (Cronk & Schweigert, 1992; Schweigert, 1986). The errors performed by patients were reanalyzed considering only the idiom strings whose mean familiarity rating was equal or higher than 2.00 (corresponding to a “familiar with a well known meaning” score). Two one-way ANOVAs were conducted on semantic versus unrelated errors that did not significantly differ [Fp (1, 14) = 1.1 p = 0.31; Fi (1, 16) < 1] (Table 5).

Table 4 Correlation matrix for aphasic patients

Idiom comprehension

Idiom familiarity

Semantic transparency

Plausibility

Token test

Sentence comprehension

Educational level

r = 0.19, p = 0.38

r = 0.005, p < 0.98

r = −0.08, p = 0.71

r = 0.06, p = 0.83

r = 0.02, p < 0.95

r = 0.47, p = 0.07

1310


Previous evidence (Franklin, 1989) showed that aphasic patients are more impaired in comprehending abstract/low imageability words than concrete/high imageability words. We controlled whether abstract target words elicited more errors than concrete ones in our patients and indeed this was the case [F (1, 21) = 4.74, p < 0.05]. Therefore, we decided to assess the extent to which the target words employed in the present study were understandable by our patients when used out of context. Due to the abstractness of many of these items, we used a verbal definition paradigm that unfortunately prevented five patients with moderate to severe production deficits from providing an answer. Ten out of 15 patients were able to provide a correct definition of the targets in 64.78% of the responses. Finally, we wondered whether the abstract or concrete nature of the action or state expressed by the idiom string had any influence on the errors performed by patients (e.g., “alzare il gomito”, to lift the elbow, i.e., to drink too much, refers to a concrete action; “perdere la testa”, to loose the head, i.e., to get mad, an abstract action) but no difference emerged [Fp < 1].

possible effect of lesion site, we divided the patients into two groups, depending on whether their score on idiom comprehension was equal or below 19 (which corresponds approximately to two S.D.s below the mean of the control group). The results showed that the patients 12–15, who performed approximately as controls, either had a lesion confined to the temporo-parietal white matter area (12, 14) or to the cortico-subcortical parietal area (13, 15). The remaining patients (who had a lower score) either had a cortico-subcortical temporo-parietal lesion (6–11) or a cortical and/or subcortical frontal lesion (1–5). In particular, the two patients with the most severe deficit in idiom comprehension showed an involvement of both the frontal and the temporal region. We examined the effect of lesion site comparing patients with a frontal versus non-frontal lesion by means of the Mann–Whitney test. A significant effect of the lesion site was found, with frontal patients being more impaired than nonfrontal ones [U (6, 9) = 7.5, p = 0.02]. 4. Discussion

3.1. Lesion site The lesion sites for all patients are schematically represented in Fig. 1. The templates showed that four lesions were purely subcortical (3, 5, 12, 14), while the remaining involved the frontal, temporal or parietal cortex. In order to control for a

We assessed the ability of 15 LBD aphasic patients to understand ambiguous idioms using a string-to-word matching task. Aphasic patients were significantly more impaired in idiom comprehension than matched controls. Previous studies (Papagno & Genoni, 2004; Papagno et al., 2004) showed that

Fig. 1. Lesion site of the 15 aphasic patients.


the comprehension of idioms without a literal interpretation (non ambiguous idioms) was seriously impaired in aphasic patients. In the present study, the understanding of another class of idioms proved to be impaired in aphasics, namely those having a literal interpretation as well. We will separately discuss the main results of this study. 4.1. Idiom comprehension and literal language Our patients were not impaired at a word comprehension level. However, semantically based errors were evident in idiom comprehension: semantically associate errors were indeed significantly more frequent than unrelated errors (64.35% versus 35.65%, respectively). Two explanations might be provided for the high level of selection of semantically associated foils: the first relies on the idea that no access to the corresponding figurative meaning can occur without identification of the idiomatic nature of the string (Cacciari & Tabossi, 1988). What might be deficient in aphasic patients might precisely be the idiom recognition mechanism: they might be delayed or even impaired in identifying the idiomatic nature of the idiom string even when enough perceptual input has accumulated. In a different vein, one can hypothesize that semantically associate errors reflect an impairment in inhibiting the word meaning associated to the final constituent word of the idiom string or a faster activation of that meaning. If this were the case, then the retrieval of the figurative meaning would be blocked by a sort of processing loop in which the patient is unable to get rid of the literal meaning of the string. Unfortunately our data do not allow distinguishing among these alternative explanations. One might wonder whether the choice of a semantically related target indeed implies that some level of processing of the sentential literal meaning occurred or it only reflects a lexical association effect. We cannot rule out the possibility that an associative mechanism might have contributed to our results especially because of the choice of foils that were semantically associated with the last word of the string. However, these results as well as previous evidence on the comprehension of idioms in brain-damaged patients (Papagno et al., 2003; Papagno et al., submitted) suggest that their impairment in comprehending the idiomatic meaning of a string extends far beyond a single-word level. In this study, we presented the idiomatic sentence without any contextual support, even though we are aware that a biasing context may prime the idiomatic interpretation, making it available more easily and quickly (for evidence on context effects in language unimpaired participants, see Cacciari, Corradini & Padovani, 2005). The choice of a minimal neutral context as a carrier for the idiom string was motivated by the need to reduce the complexity of the sentence structure presented to aphasic patients: had the interpretation provided by the patient be incorrect, it would have been difficult to disentangle idiom comprehension problems from contextual information processing impairments. The instructions provided at the beginning of the experiment, however, clearly specified that the participant had to provide the non-literal interpretation of the sentence.

1311

4.2. Neural correlates The fact that all patients were examined in a chronic stage rules out the possibility of areas of dysfunction not detected by CT-scan or MRI. Even though our mapping procedure has some important limitations, a general analysis of the anatomical localization of the lesions shows that two sites are relevant for the patients’ performance in idiom comprehension: a frontal, even subcortical area, and a cortical temporal region. These results are consistent with previous findings obtained with non-ambiguous idioms in an rTMS study (Oliveri et al., 2004). However, the role of frontal structures in idiom comprehension could be twofold. Specifically, after the linguistic analysis of the string is performed, the choice of the correct response among four alternatives requires a selection process coupled with monitoring of the response. This selection and monitoring process is likely to be performed by the central executive whose neural correlates are located in the frontal lobe (Stuss, Eskes, & Foster, 1994). The more severe impairment of patients with a frontal involvement confirms the role of inhibition, since these patients produced a mean of 7.5 (S.D. 2.43) literal errors compared to a mean of 3.22 (S.D. 2.11) literal errors produced by non-frontal patients. rTMS and fMRI studies will be necessary to confirm this hypothesis. A similar result was obtained by Metzler (2001) with ambiguous words: patients with left frontal lesions did not show semantic facilitation for contextually appropriate meaning, relative to controls, in a semantic priming task on ambiguous words (i.e., homograph words). When the ambiguous words were replaced by unambiguous ones, patients with left frontal lesions showed normal levels of semantic priming, suggesting a deficit in selecting contextually appropriate meanings when competing meanings are available. The left prefrontal region also is involved in text comprehension (Ferstl, Guthke, & Von Cramon, 2002) that might share some processing resource and mechanism with idiom processing (e.g., inference making): patients with left frontal lesions were able to detect a coherence gap but were instead unable to draw an inference. Deficits in auditory comprehension associated with lesions of the frontal lobe have been described by Berthier (2001) in transcortical sensory aphasia: such an impairment resulted from a functional damage to posterior neocortical association areas, presumably due to interruption of reciprocal cortico-cortical connections. Moreover, fMRI studies demonstrated that semantic processing tasks also activated the dorsolateral prefrontal cortex (Zahn et al., 1999). Alternative interpretations could be suggested, as for example a tendency to give concrete responses to individual words resulting from premature choice of an interpretation, as found with proverbs by Burgess and Chiarello (1996). However, Ulatowska, Sadowska Kazdielawa Kordys, and Rymarczyk (2000) observed that when patients were required to select the meaning of a proverb among several alternatives, there were fewer concrete responses. Moreover, this interpretation does not exclude that an inhibition process is needed to prevent the patient from giving concrete responses.

1312


4.3. Type of errors The patients’ errors deserve a further comment: the high number of semantically associate errors produced by aphasic patients (64.34%) suggests that, when the linguistic processing of the idiom string is impaired as in aphasia, the target word semantically associate with the last constituent of the idiom string might still be available and selected. This choice might be the outcome of several levels of impairment: first, an impairment of inhibition mechanisms in that the patients were unable to get rid of the constituent word literal meanings; second, an impairment in recognition and activation mechanisms in that the patients’ ability to access and retrieve the idiomatic meaning was damaged and they did not recognize the figurative nature of the string; third, a simultaneous impairment of both inhibition and activation mechanisms. The current data do not allow us to distinguish among these three alternatives. However, it should be noted that aphasic patients were impaired, and even more so, also with idioms without a literal interpretation (see Papagno et al., 2004) suggesting a simultaneous damage to linguistic processing and to idiomatic meaning recognition and activation. Alternative explanations can account for unrelated errors as well: if someone chooses an unrelated target, as controls more often do when they are mistaken, most likely this can suggest that s/he did not know the idiomatic meaning. This can also explain unrelated errors in aphasic patients: they refuse the literal interpretation of the idiom string (that is less frequent than the idiomatic one) and look for a possible alternative, being unable to retrieve the correct idiomatic meaning. Acknowledgments The authors are grateful to Alessandra Caporali for testing two patients. The research has been supported by a FAR 2003 and by two COFIN 2003 grants, respectively, assigned to CP and to CC. Appendix A Experimental idioms with literal and idiomatic English translations. The columns respectively indicate the mean Familiarity ratings for each idiom for language unimpaired participants (UP) and patients (P) and the mean plausibility rating obtained by each idiom string in the Norming phase

Alzare il gomito (to lift the elbow, to drink too much) Vuotare il sacco (to empty the sack, to confess something) Sputare il rospo (to spit the toad, to reveal a secret) Vedere le stele (to see the stars, to experience a pain) Essere al verde (to be at the green, to be broke) Prendere la porta (to take the door, to go away) Levare le tende (to remove the curtains, to leave) Perdere la testa (to loose the head, to get mad) Avere il pollice verde (to have a green thumb) Prendere un granchio (to take a crab, to make an error) Tirare la cinghia (to pull the belt, to be poor) Perdere il filo (to loose the thread, to get lost) Cambiare pagina (to change page, to change your life) Mettersi le mani nei capelli (to put the hands in the hair, to be in despair) Tagliare la corda (to cut the rope, to escape) Scendere a rotta di collo (to go down breaking the neck, to behave riskily) Avere le mani bucate (to have holes in the hands, to spend too much) Saltare la mosca al naso (to jump the fly on the nose, to get angry) Mettere la pulce nell’orecchio (to put the flea in the ear, to suggest something) Avere carta bianca (to have white paper, to be free in taking decisions) Alzare i tacchi (to lift the heels, to go away abruptly) Attaccare un bottone (to attach a button, to speak too much) Stringere i denti (to tighten the teeth, to strongly pursue a goal)

UP

P

Plausibility of literal meaning

2.67

3.00

2.75

2.47

2.67

2.78

2.67

2.22

0.25

2.56

1.56

3.00

2.92

2.56

0.13

1.19

2.11

1.72

2.17

1.56

2.44

2.78

2.22

0.38

2.61

1.56

0.69

2.17

2.33

2.69

2.47

2.89

2.53

2.64

1.89

2.34

2.14

1.22

2.72

2.72

2.22

2.84

2.78

2.67

2.84

1.47

2.22

0.34

2.67

2.89

0.34

1.50

1.22

0.56

2.50

2.56

0.75

2.56

2.56

2.09

2.44

2.33

2.44

2.44

2.33

2.91

2.58

2.33

2.53


1313

Appendix B Word targets for each idiom with English translation Idiom

Idiomatic target

Semantically associate target

Unrelated target 1

Unrelated target 2

Alzare il gomito Vuotare il sacco Sputare il rospo Vedere le stele Essere al verde Prendere la porta Levare le tende Perdere la testa Avere il pollice verde Prendere un granchio Tirare la cinghia Perdere il filo Cambiare pagina Mettersi le mani nei capelli Tagliare la corda Scendere a rotta di collo Avere le mani bucate Saltare la mosca al naso Mettere la pulce nell’orecchio Avere carta bianca Alzare I tacchi Attaccare un bottone Stringere I denti

Vino (wine) Segreto (secret) Confessione (confession) Dolore (pain) Denaro (money) Distacco (separation) Partenza (leave) Matto (crazy) Fiori (flowers) Errore (error) Povertà (poverty) Racconto (story) Novità (news) Guaio (trouble) Fuga (escape) Fretta (hurry) Acquisti (shopping) Ira (anger) Sospetto (suspect) Libertà (freedom) Fuga (escape) Conversazione (conversation) Sacrificio (sacrifice)

Gamba (leg) Zaino (rucksack) Farfalla (butterfly) Cielo (sky) Blu (blu) Finestra (window) Tappeto (carpet) Collo (neck) Giallo (yellow) Gatto (cat) Camicia (shirt) Cotone (cotton) Foglio (sheet) Spalla (shoulder) Nastro (ribbon) Braccio (arm) Taglio (cut) Zanzara (mosquito) Vespa (wasp) Biglietto (ticket) Scarpa (shoe) Giacca (jacket) Bocca (mouth)

Albero (tree) Mistero (mystery) Disprezzo (contempt) Giustizia (justice) Pelliccia (fur) Onestà (honesty) Tranquillità (calm) Turismo (tourism) Montagna (mountain) Economia (economy) Ragionamento (reasoning) Processo (process) Religione (religion) Curiosità (curiosità) Bugia (lie) Affetto (affect) Legno (wood) Emozione (emotion) Allegria (happiness) Fortuna (fortune) Scopo (aim) Bottiglia (bottle) Cortesia (courtesy)

Cassa box) Lavatrice (washing machine) Uovo (egg) Chiav e (key) Medico (doctor) Autobus (bus) Costume (costume) Cucchiaino (spoon) Figlia (daughter) Romanzo (novel) Penna (pen) Premio (prize) Piatto (dish) Tasca (pocket) Busta (envelope) Temperatura (temperature) Lenzuolo (linen) Ostacolo (obstacle) Tavolino (table) Palla (ball) Aquilone (kite) Etichetta (label) Vestito (dress)

References Berthier, M. L. (2001). Unexpected brain–language relationships in aphasia: Evidence from transcortical sensory aphasia associated with frontal lobe lesions. Aphasiology, 15, 99–130. Bobrow, S., & Bell, S. (1973). On catching on to idiomatic expressions. Memory and Cognition, 19, 295–308. Burgess, C., & Chiarello, C. (1996). Neurocognitive mechanisms underlying metaphor comprehension and other figurative language. Metaphor and Symbolic Activity, 11, 67–84. Cacciari, C., Corradini, P., & Padovani, R. (2005). Speed of processing effects on spoken idioms comprehension. In Proceedings of the XXVII Annual Meeting of the Cognitive Science Society. Cacciari, C., & Glucksberg, S. (1995). Understanding idioms: Do visual images reflect figurative meanings? European Journal of Cognitive Psychology, 7, 283–305. Cacciari, C., & Tabossi, P. (1988). The comprehension of idioms. Journal of Memory and Language, 27, 668–683. Colombo, L. (1993). The comprehension of ambiguous idioms in context. In C. Cacciari & P. Tabossi (Eds.), Idioms: Processing, Structure, and Interpretation (pp. 163–200). Hillsdale: LEA. Colombo, L. (1998). Role of context in the comprehension of ambiguous Italian idioms. In D. Hillert (Ed.), Syntax and Semantics: vol. 31 (pp. 379–404). New York: Academic Press. Cronk, B. (1990). Familiarity and likeliness for figurative and literal meanings of idioms. In Paper presented at the Second Annual Convention of the American Psychological Association. Cronk, B., & Schweigert, W. A. (1992). The comprehension of idioms: The effects of familiarity, literalness and usage. Applied Psycholinguistics, 13, 131–146. Damasio, H., & Damasio, A. R. (1989). Lesion Analysis in Neuropsychology. Oxford: Oxford University Press. De Renzi, E., & Faglioni, P. (1978). Normative data and screening power of a shortened version of the token test. Cortex, 14, 41–49.

Ferstl, E. C., Guthke, T., & von Cramon, D. Y. (2002). Text comprehension after brain injury: Left prefrontal lesions affect inference processes. Neuropsychology, 16, 292–308. Franklin, S. (1989). Dissociations in auditory word comprehension: Evidence from nine “fluent” aphasic patients. Aphasiology, 3, 189–207. Gibbs, R. (1999). Figurative language. In R. Wilson & F. Keil (Eds.), The MIT Encyclopedia of the Cognitive Sciences (pp. 314–315). Cambridge: MIT Press. Gibbs, R. W. (1984). Literal meaning and psychological theory. Cognitive Science, 8, 275–304. Gibbs, R. W., & Gonzales, G. P. (1985). Syntactic frozeness in processing and remembering idioms. Cognition, 20, 243–259. Gibbs, R. W., Nayak, N. P., & Cutting, C. (1989). How to kick the bucket and not decompose: Analyzability and idiom processing. Journal of Memory and Language, 28, 576–593. Hamblin, J. L., & Gibbs, R. W. (1999). Why you can’t kick the bucket as you slowly die: Verbs in idiom comprehension. Journal of Psycholinguistic Research, 28, 25–39. Hillert, D. (2004). Spared access to idiomatic and literal meanings: A singlecase approach. Brain and Language, 89, 207–215. Laiacona, M., Barbarotto, R., Trivelli, C., & Capitani, E. (1993). Dissociazioni semantiche intercategoriali: Descrizione di una batteria standardizzata e dati normativi. Archivio di Neurologia, Psichiatria e Psicologia, 54, 209–248. MacLeod, C. M. (1991). Half a century of research on the Stroop effect: An integrative review. Psychological Bulletin, 109, 163–203. Metzler, C. (2001). Effects of left frontal lesions on the selection of contextappropriate meanings. Neuropsychology, 15, 315–328. Nenonen, M., Niemi, J., & Laine, M. (2002). Representation and processing of idioms: Evidence from aphasia. Journal of Neurolinguistics, 15, 43–58. Nunberg, G., Sag, I. A., & Wasow, T. (1994). Idioms. Language, 70, 491–538. Oliveri, M., Romero, L., & Papagno, C. (2004). Left but not right temporal lobe involvement in opaque idiom comprehension: A repetitive transcranial stimulation study. Journal of Cognitive Neuroscience, 16, 848–855.

1314


Papagno, C., & Genoni, A. (2004). The role of syntactic competence in idiom comprehension: A study on aphasic patients. Journal of Neurolinguistics, 17, 371–382. Papagno, C., Lucchelli, F., Muggia, S., & Rizzo, S. (2003). Idiom comprehension in Alzheimer’s disease: The role of the central executive. Brain, 126, 2419–2430. Papagno, C., Tabossi, P., Colombo, M., & Zampetti, P. (2004). Idiom comprehension in aphasic patients. Brain and Language, 89, 226–234. Peterson, R. R., Burgess, C., Dell, G. S., & Eberhard, K. (2001). Dissociation between syntactic and semantic processing during idiom comprehension. Journal of Experimental Psychology: Learning, Memory, and Cognition, 27, 1223–1237. Schweigert, W. A. (1986). The comprehension of familiar and less familiar idioms. Journal of Psycholinguistic Research, 15, 33–45. Stroop, J. R. (1935). Studies of interference in serial verbal reactions. Journal of Experimental Psychology, 18, 643–662. Stuss, D. T., Eskes, G. A., & Foster, J. K. (1994). Experimental neuropsychological studies of frontal lobe functions. In F. Boller & H. Spinnler

(Eds.), Handbook of Neuropsychology: vol. 9 (pp. 149–185). Amsterdam: Elsevier (Section 12: The frontal lobes). Swinney, D. A., & Cutler, A. (1979). The access and processing of idiomatic expression. Journal of Verbal Learning Verbal Behaviour, 18, 523– 534. Tompkins, C. A., Boada, R., & McGarry, K. (1992). The access and processing of familiar idioms by brain-damaged and normally aging adults. Journal of Speech and Hearing Research, 35, 626–637. Ulatowska, H. K., Sadowska, M., Kazdielawa, D., Kordys, J., & Rymarczyk, K. (2000). Linguistic and cognitive aspects of proverb processing in aphasia. Aphasiology, 14, 227–250. Van Lancker, D., & Kempler, D. (1987). Comprehension of familiar phrases by left but not by right hemisphere damaged patients. Brain and Language, 32, 265–277. Zahn, R., Huber, W., Erberich, S., Kemeny, S., Specht, K., Krings, T., et al. (1999). Semantic processing of auditory words in acute recovery from transcortical sensory aphasia and in normal subjects. Brain and Language, 69, 301–303.