Abstract

We investigated the neural correlates of idiomatic sentence processing using event-related functional magnetic resonance imaging. Twenty-two healthy subjects were presented with 62 literal sentences and 62 idiomatic sentences, each followed by a picture and were required to judge whether the sentence matched the picture or not. A common network of cortical activity was engaged by both conditions, with the nonliteral task eliciting overall greater activation, both in terms of magnitude and spatial extent. The network that was specifically activated by the nonliteral condition involved the left temporal cortex, the left superior medial frontal gyrus (Brodmann area 9), and the left inferior frontal gyrus (IFG). Activations were also present in the right superior and middle temporal gyri and temporal pole and in the right IFG. In contrast, literal sentences selectively activated the left inferior parietal lobule and the right supramarginal gyrus. An analysis of effective connectivity indicated that the medial prefrontal area significantly increased the connection between frontotemporal areas bilaterally during idiomatic processing. Overall, the present findings indicate a crucial role of the prefrontal cortex in idiom comprehension, which could reflect the selection between alternative sentence meanings.

Introduction

Although the crucial role of the left hemisphere (LH) in language function is undisputed, several aspects of linguistic processing have been attributed to the right hemisphere (RH). In particular, a crucial role has been attributed to the RH in figurative language processing. Figurative language allows speakers to communicate meanings that differ in various ways from what they literally say. There are several forms of figurative language. These include metaphors, where ideas from dissimilar knowledge domains are either explicitly (my love is like a red rose) or implicitly (our marriage is a roller-coaster ride) compared; idioms, where a speaker's meaning cannot be derived from an analysis of the words' typical meanings (to kick the bucket); proverbs, where speakers express widely held moral beliefs or social norms; irony, where a speaker's meaning is usually the opposite of what is said; hyperbole, where the speaker exaggerates the reality of some situation (I have ten thousand things to do), etc (for a detailed lists of different forms of figurative language, see Gibbs 1999). Data concerning the role of the RH in figurative language comprehension have been obtained originally from neuropsychological studies, which often did not distinguish among different forms of figurative language, even if it is not known whether the same processes and networks are involved in their interpretation. For example, Winner and Gardner (1977) found that aphasic left brain damaged (LBD) patients were able to perform correctly when they had to associate metaphoric expressions with the corresponding picture, whereas right brain damaged (RBD) patients, despite being able to give correct verbal explanations, chose the picture corresponding to the literal meaning. Although this finding has been taken as evidence that RBD patients are impaired in figurative language, it may only indicate that RBD patients are defective in picture matching (Papagno and Caporali 2007).

More recently, neuropsychological evidence has accumulated against a unique role of the RH in figurative language. The majority of studies has been concentrated on metaphors. No significant difference has been found between RBD patients and healthy participants in metaphor comprehension (Giora et al. 2000) but rather a significant disadvantage of LBD patients relative to both RBD patients and controls. More specifically, significant negative correlations were found between metaphor comprehension and damage to the left middle temporal gyrus and to the junctional area of the superior temporal and supramarginal gyri.

Additional studies with different methodologies (e.g., priming) that initially supported the RH hypothesis have recently questioned a strict dichotomy of the roles of the 2 hemispheres in figurative language processing (Burgess and Chiarello 1996). For instance, an investigation of semantic priming for literal and metaphoric associates presented to either the left or the right visual field showed priming effects for metaphorically related targets for the left visual field, whereas literally related targets were primed solely for the right visual field (Anaki et al. 1998); on the other hand, 2 more recent divided visual field priming experiments (Kacinik and Chiarello 2007) have shown that processes in both hemispheres can support metaphor comprehension.

Contrasting results are also provided by using functional imaging, such as positron emission tomography (PET) (Bottini et al. 1994) and functional magnetic resonance imaging (fMRI) (Rapp et al. 2004). In the former PET study, a RH role is reported, although activation of both RH and LH regions was seen when only 6 normal subjects decided whether or not a sentence was plausible as a metaphor. On the other hand, the latter fMRI experiment using a different task (reading metaphors in contrast to literal sentences) revealed significant activations in the left lateral inferior frontal (Brodmann area [BA] 45/47), inferior temporal (BA 20), and posterior middle/inferior temporal (BA 37) gyri (Rapp et al. 2004). Similar results have been obtained with subjects deciding whether 2 out of 3 auditorily presented adjectives had a similar meaning (Lee and Dapretto 2006). One condition required accessing the literal meaning of the middle word (hot–cold–chilly), whereas the other condition required accessing its metaphorical meaning (hot–cold–unfriendly).

Another form of figurative language, namely idioms, has also been proposed to depend on the RH. Idioms are not a unitary class, as they differ according to a variety of features, such as literality (whether or not they also have a literal meaning), compositionality (whether or not the constituent words can be mapped directly on their idiomatic referents), and opacity/transparency (whether or not the meaning can be derived from the image evoked) (Glucksberg et al. 1993).

In the earliest neuropsychological literature, as already pointed out, metaphors and idioms were considered together, as well as so-called “familiar phrases,” including courtesy phrases or formulas (Van Lancker and Kempler 1987; Kempler et al. 1999). Based on a double dissociation found in RBD and LBD patients (poor performance on idioms and good performance on novel sentences in RBD patients vs. good performance on idioms and poor performance on novel sentences in LBD patients), Kempler et al. (1999) stated that the RH is preferentially involved in processing figurative language in normal healthy adults, suggesting that literal and idiomatic language are mediated by different cerebral structures. These neuropsychological findings indirectly suggest that familiar phrases are processed like unitary, nonsyntactically analyzed elements, a process which is subserved by the RH. According to this view, the meaning of an idiom is directly (and holistically) retrieved from memory. In the authors' opinion, this “familiar-phrase-as-single-word” hypothesis, which considers that familiar phrases behave formally (structurally), but not semantically, like words, is compatible with Swinney and Cutler's “lexical representation hypothesis,” which posits that idioms behave as long, morphologically complex words stored in the mental lexicon together with the other lexical units (Swinney and Cutler 1979). According to this view, the linguistic processing of the string and the retrieval of the idiomatic meaning proceed in parallel, with the second process being faster than the first. An extreme version of this hypothesis (Gibbs 1984) suggests that people do not engage in any linguistic analysis at all, entirely bypassing the literal meaning and directly accessing the figurative interpretation of the idiom string.

As in the case of metaphors, more recent neuropsychological studies, which test separately different aspects of figurative language, have provided data against a specific role of the RH in idiom comprehension (Nenonen et al. 2002; Papagno et al. 2004; Cacciari et al. 2006). Nenonen et al. (2002) examined verb and noun phrase idiom reading in a deep dyslexic and found that verb phrase idioms were read as poorly as comparable free phrases. Other studies on aphasic patients with semantic deficits have found a severe impairment in a sentence-to-picture-matching task using nonambiguous idioms with a significant bias toward the literal interpretation (Papagno and Genoni 2004). Aphasics were impaired even on ambiguous idioms, when a sentence-to-word matching task was used (Cacciari et al. 2006). Converging results come also from functional imaging. Zempleni et al. (2007) found a common activation of the language network when the processing of idioms was compared with literal sentence interpretation; in addition, idiomatic processing engaged the inferior frontal gyri (IFG) bilaterally, as well as the left middle temporal gyrus.

Additional evidence of the LH involvement is provided by repetitive transcranial magnetic stimulation (rTMS) studies in normals. Transcranial magnetic stimulation (TMS) studies of idioms attempted to integrate anatomical results with the current psycholinguistic models of figurative language interpretation (Oliveri et al. 2004). Using an off-line paradigm, rTMS applied over left, but not right, BA 22 increased reaction times (RTs) and reduced accuracy, with no difference between literal and nonliteral sentences. These findings, suggesting that a syntactic analysis takes place during idiom comprehension and that the same anatomical regions are involved in idiom and literal sentence comprehension, support the “configurational hypothesis” (Cacciari and Tabossi 1988), according to which the individual components of the idiom are analyzed as normal linguistic items until a key part is processed that prompts recognition of the idiomatic nature of the string and activation of its figurative meaning. In another rTMS study, which used an online paradigm, Rizzo et al. (2007) found that both the right and the left dorsolateral prefrontal cortices are involved in monitoring the response in the case of idiomatic sentences. An interference with the activity of this region produced a decrease in accuracy, with a significant increase of literal errors.

An attempt to accommodate contrasting data on figurative language lateralization in general is provided by Giora (Giora 1997; Giora et al. 2000). The “graded salience hypothesis” suggests that salient meanings are processed in the LH, regardless of figurativity, whereas nonsalient meanings, involving distant semantic relationships, are processed in the RH. Alternatively, and specifically for idioms, Huber-Okrainec et al. (2005) have proposed, by examining children with callosal agenesia, that interpreting idioms requires interhemispheric integration. The corpus callosum, therefore, would mediate interhemispheric interactions especially during idiom comprehension development.

As a general point, models that are based on tasks or classes of task will have limited explanatory value outside the immediate cognitive domain of those tasks. Indeed, comparison between studies is not straightforward in many cases because of variations in tasks and methods. The effect of methodology has already been underlined in an fMRI study on phonological short-term memory (Logie et al. 2003). A significant interaction between response type and the brain regional activation during semantic memory tasks has also been demonstrated (Jennings et al. 1997). In the case of patients, besides the effects of the type of task and the measures employed (Papagno and Caporali 2007), reorganization processes and compensatory strategies in the chronic phase cannot be totally excluded.

Therefore, converging evidence using different methodologies and tasks would strengthen the results concerning the respective role of the LH and RH in idiomatic language processing. What we assume is that no single experiment is definitive, but a sufficiently rich and coherent body of data will place major constraints to rule out alternative hypotheses.

In the present paper, we aim at identifying the anatomical structures involved in idiom processing using fMRI and at integrating our results in the current models of idiom comprehension in order to add further evidence to support one of the 2 alternative explanations (e.g., lexical hypothesis vs. configurational hypothesis). Cognitive psychology studies as well as neuropsychological findings seem to dismiss the possibility that idioms are treated as long, morphologically complex words (Peterson et al. 2001; Papagno and Genoni 2004). More specifically, we expect that idiom comprehension involves the same areas engaged by literal language processing, as proposed by Oliveri et al. (2004). Moreover, the additional requirements of idiomatic processing can be expected to involve additional areas, in particular in the prefrontal cortex, as suggested by neuropsychological (Papagno et al. 2006) and neurophysiological findings (Rizzo et al. 2007). The involvement of the prefrontal cortex may be due to its role in high-level cognitive operations, such as cognitive control (Koechlin et al. 2003), response monitoring (Ridderinkhof et al. 2004), and coordination between multiple cognitive operations (Ramanani and Owen 2004).

Material and Methods

Materials

Sixty-two idioms in Italian and 62 literal Italian sentences were selected for this study. All idioms were judged as highly familiar in a population of 101 subjects of different age (range 22–65), from different regions of Italy, 30 being ambiguous (the literal interpretation was also plausible, i.e., “to cut the rope” meaning “to escape”) and 32 not (the literal interpretation was not plausible, i.e., “to stay at the ribs” meaning “to follow someone at a very close distance”). However, we reassessed familiarity in the participants' group, after they had completed the fMRI session. The mean index of familiarity on a 3-point scale was 2.68 (range 1.5–2.9, standard deviation [SD] = 0.3). The degree of transparency was checked by asking 30 healthy subjects (mean age = 26, range 20–32) to rate how easily the meaning of the idiom could be deduced from the concrete image evoked by the sentence: a score ranging from 0 to 5 was assigned, where 0 meant “not at all” and 5 “entirely.” This index varied between 1.9 (item 55) and 4.4 (item 16) (mean = 3.1, SD = 0.6) (see Appendix). Literal sentences were matched in length with idioms. For each sentence, be it literal or idiomatic, a line-drawing picture was created. Half of them correctly represented the meaning of the sentence, half did not. In the case of idioms, the correct picture represented the figurative meaning, that is, for the idiom “to have a green thumb,” the picture represented a woman carefully tending some plants (Fig. 1A). The wrong picture represented the opposite of the figurative meaning, that is, for the idiom “tirare la cinghia” (to pull one's belt), meaning “to starve, being very poor” the picture represented a rich man eating voraciously in a restaurant (Fig. 1B). The 30 healthy subjects who rated transparency were also asked to rate how closely the picture corresponding to the idiomatic meaning represented the idiom or its opposite. A score ranging from 0 to 5 was assigned, where 0 meant “not at all” and 5 “perfectly.” The mean score for idiomatic relatedness was 3.6 (range 1.6–4.8, SD = 0.7). In the case of literal sentence, the correct picture exactly matched the sentence (such as in Fig. 1C), whereas the wrong picture differed in one element, that is, for the sentence “ the man opens the window” the wrong picture represented a man opening a door (see Fig. 1D). Idiomatic and literal sentences were matched for length and syntactic complexity (see Appendix).

Figure 1.

Examples of the pictures used in the experimental task. (A) A correct picture for the idiom “avere il pollice verde” (to have a green thumb). (B) A wrong picture representing the opposite of the idiomatic meaning of “tirare la cinghia” (to pull one's belt, meaning “to starve, being very poor”). (C) A correct picture for the literal sentence “the boy is eating an apple.” (D) A wrong picture for the literal sentence “ the man is opening the window.”

Figure 1.

Examples of the pictures used in the experimental task. (A) A correct picture for the idiom “avere il pollice verde” (to have a green thumb). (B) A wrong picture representing the opposite of the idiomatic meaning of “tirare la cinghia” (to pull one's belt, meaning “to starve, being very poor”). (C) A correct picture for the literal sentence “the boy is eating an apple.” (D) A wrong picture for the literal sentence “ the man is opening the window.”

Participants

Twenty-two right-handed students from the University of Milano-Bicocca volunteered to take part in the experiment (13 women and 9 men; mean age 25, range 19–40). They all gave informed consent to participate in the study. They were all Italian native monolingual students. None of them had a history of neurological or psychiatric disorders. Handedness was assessed by using the Handedness Inventory Questionnaire (Oldfield 1971). The study was approved by the local Ethics Committee.

Procedure

The experiment was run in 4 blocks (corresponding to the 4 scanning sessions, see below). Each block consisted of 31 trials. In each trial, subjects were presented with a written sentence, either literal or idiomatic, and a picture. First, the sentence was presented on the upper part of the screen. After 2000 ms, a picture appeared below the sentence. Both the sentence and the picture remained on the screen for additional 3400 ms. Participants were instructed to judge whether the picture represented the meaning of the sentence or not and responded by pressing one of 2 response keys with the right index finger (green button for matching and red button for nonmatching sentence–picture pairs). RTs and accuracy were taken as behavioral data. Stimuli were presented in a randomized order by means of Presentation 0.91 (Neurobehavioral Systems, Albany, CA) and viewed via a back-projection screen located in front of the scanner and a mirror placed on the head coil. Behavioral responses were collected via a fiber-optic response box.

fMRI Acquisition

Magnetic resonance imaging scans were acquired on a 3-T Intera Philips body scanner (Philips Medical Systems, Best, NL) using an 8 channels SENSE head coil (sense reduction factor = 2). Whole-brain functional images were obtained with an echo-planar T2* sequence, using blood oxygenation level–dependent contrast. Each functional image comprised 30 contiguous axial slices (4 mm thick), acquired in interleaved mode, with a repetition time of 2200 ms (echo time: 30 ms; field of view: 240 mm × 240 mm). Each participant underwent 4 functional scanning sessions of 136 brain volumes each. The time interval between functional scanning sessions was between 1 and 2 min, that is, the time necessary to initialize a new scanning sequence.

Data Analyses

Statistical parametric mapping (SPM2, Wellcome Department of Imaging Neuroscience, London, UK) was used for image realignment (Andersson et al. 2001), normalization to the Montreal Neurologic Institute (MNI) standard space, smoothing by a 6-mm full-width half-maximum Gaussian kernel, and General Linear Model statistical analysis (Friston et al. 2002). We adopted a 2-stage random-effects approach to ensure generalizability of the results at the population level (Penny and Holmes 2003).

First-Level Statistical Models

At the first stage, the time series of each participant were high-pass filtered at 67 s and prewhitened by means of an autoregressive model AR(1) (Andersson et al. 2001). No global scaling was applied. Hemodynamic evoked responses for all experimental conditions were modeled as a set of hemodynamic response function, including the canonical response function, its first derivative (modeling for temporal latency effects), and its second derivative (modeling for dispersion effects). The first-level individual design matrices included the data of all 4 scanning sessions. For each session, we modeled the 4 experimental conditions (idiomatic sentence, true response [IT]; idiomatic sentence, false response [IF]; literal sentence, true response [LT]; literal sentence, false response [LF], with the onset of the hemodynamic response functions time locked to the appearance of the sentence on the screen. A set of Student's t-test contrasts were defined for use at the second statistical level: these consisted in one contrast per experimental condition (IT, IF, LT, LF) per hemodynamic response function (canonical, first derivative, second derivative).

Second-Level Statistical Models

At the second stage of analysis, the effects obtained at the single-subject level were used to compute an analysis of variance (ANOVA) assessing their significance at the group level (n = 22 participants). The ANOVA included the set of all first-level t-test contrasts of all participants. We used F-contrasts to assess the following effects of interest: 1) Conjunction null (Nichols et al. 2005) between all experimental conditions (IT, IF, LT, LF); 2) Main effect of idiomatic versus literal sentences ([IT + IF] vs. [LT + LF]); 3) Direct comparisons (IT vs. IF; LT vs. LF; IT vs. LT; IF vs. LF). The direction (tail distribution) of the effects (idiomatic vs. literal) was assessed by inspecting the plots of the contrast estimates with 90% confidence intervals in SPM2. All reported effects relate to voxel-level statistics (P < 0.05 with family-wise error type correction for multiple comparisons).

We assessed the cytoarchitectonic probability of activations in Broca's area, using the probability maps (Amunts et al. 1999) for BA 44 and BA 45 available with the SPM Anatomy toolbox (http://www.fz-juelich.de/ime/spm_anatomy_toolbox).

Additional Analyses: Excluding for Potential Differences between Ambiguous and Nonambiguous Idiomatic Sentences

Ambiguous idioms can be also assigned a literal interpretation in an appropriate context (e.g., kick the bucket). Given that aphasic patients produce different patterns of errors with ambiguous and nonambiguous idioms, and double dissociations have been reported (Papagno and Caporali 2007), one could expect that different processes recruiting distinctive cortical substrates are involved. In addition, in a recent fMRI study on idiom comprehension (Zempleni et al. 2007), unambiguous sentences elicited significantly increased activation in comparison to ambiguous sentences. To exclude that ambiguous and nonambiguous idiomatic sentences recruited different brain regions or the same brain regions to a different extent, we performed a control statistical analysis, in which the 2 types of idiomatic sentences were modeled separately. To this aim, we specified an additional first-level statistical model for each participant. The same parameters as in the statistical analysis described above were used. However, in this control analysis, we modeled, instead of 4, 6 experimental conditions (ambiguous, idiomatic sentence, true response [AIT]; ambiguous, idiomatic sentence, false response [AIF]; nonambiguous, idiomatic sentence, true response [NIT]; nonambiguous, idiomatic sentence, false response [NIF]; literal sentence, true response [LT]; literal sentence, false response [LF]). In addition, to correct for the potential bias produced by the different number of stimuli across conditions, this model also included a parametric modulation, consisting in a convolution with a constant value for each condition equal to the number of stimuli in that particular condition (AIT: 22; AIF: 8; NIT: 9; NIF: 23; LT: 31; LF: 31). In contrast to the previous analysis, the 4 functional scanning sessions were modeled as one single session. This choice was necessary in order to circumvent the fact that for some experimental conditions there was a very small number of stimuli in some of the functional scanning sessions. Nevertheless, to account for the 4 separate functional scanning sessions, we added 4 constant regressors: each regressor included values of 1 for the scans of one particular session and values of 0 for the scans of the other 3 sessions.

At the second statistical level, we used F-contrasts to assess the following effects of interest: 1) Main effect of idiomatic versus literal sentences ([AIT + AIF + NIT + NIF] vs. [LT + LF]); 2) Effect of ambiguous versus nonambiguous idiomatic sentences ([AIT + AIF] − [NIT + NIF]). The main effect of idiomatic versus literal sentences (contrast 1) was only included in order to check for the consistency between this control analysis and the main analysis described in the paper: importantly, there were no relevant qualitative differences between the 2 analyses for this particular contrast.

Dynamic Causal Modeling

A complementary approach to the investigation of brain functional specialization related to the individual task is the analysis of functional integration, by means of methods which estimate the coupling of brain areas and the influence of experimental manipulations on that coupling (Lee et al. 2006). One of these methods is dynamic causal modeling (DCM), which derives connectivity parameters for the influence of the experimental conditions on brain states, for the intrinsic connections between the brain areas included in the system, and for the changes in connectivity between areas induced by the experimental conditions (bilinear terms) (Friston et al. 2003). We used DCM to measure how the significant main effect “IDIOMATIC > LITERAL” is conveyed from the input area, via the specified connections, to other areas. To this purpose, a set of 5 anatomical regions of interest was selected based on the result of the main effect IDIOMATIC > LITERAL (i.e., the anterior medial frontal gyrus [input area], the left and right IFG, and the left and right anterior middle temporal gyri). We then defined a set of connections between the 5 regions of interest to test whether the activity of the anterior medial frontal gyrus had an influence on the activity in perisylvian areas.

As in the case the General Linear Model analysis, also for the DCM analysis, we adopted a 2-stage random-effects approach. At the single-subject level, we implemented a DCM including the following 5 brain regions: the anterior medial frontal gyrus (x = −4, y = 54, z = 32), the left IFG (x = −50, y = 26, z = 0), the left anterior middle temporal gyrus (x = −58, y = −6, z = −20), the right IFG (x = 52, y = 32, z = −8), and the right anterior middle temporal gyrus (x = 58, y = −8, z = −20).

Spherical volume of interest of 6 mm radius centered on these coordinates were extracted for all participants, using the F-contrast over all the effects of interest. A separate input for all 4 experimental conditions was allowed to enter the DCM at the medial prefrontal area. The resulting perturbation was then allowed to propagate throughout the network via interconnections between the medial prefrontal area and the remaining regions (see Fig. 2). In such a model, in which separate inputs for all experimental conditions are provided, the DCM analysis measures how the significant main effect (in this case IDIOMATIC > LITERAL) that is induced in the input area is conveyed, via the specified connections, to other areas (effective connectivity). In this case, the focus is on the intrinsic connections and not on the bilinear terms. Therefore, we only report the data concerning the intrinsic connections, measuring the effective connectivity of one region over another. The ensuing subject-specific parameters of the intrinsic connections were then taken to a second level for population inference (Penny and Holmes 2003). One sample t-tests on the intrinsic connections were performed. As this was repeated for several connections, we adopted a conservative statistical threshold of P < 0.001.

Figure 2.

DCM used to test the hypothesis of a driving role of the anterior prefrontal area in inhibiting the alternative interpretation in the case of idiomatic sentences, by coordinating the activity in perisylvian areas. The model includes 5 brain regions: the left anterior middle temporal gyrus (LMTG), the right anterior middle temporal gyrus (RMTG), the left inferior frontal gyrus (LIFG), the right inferior frontal gyrus (RIFG), and the anterior medial frontal gyrus (mFG). The MNI coordinates are displayed on the right. The black arrows indicate forward/backward intrinsic connections between 2 anatomical nodes.

Figure 2.

DCM used to test the hypothesis of a driving role of the anterior prefrontal area in inhibiting the alternative interpretation in the case of idiomatic sentences, by coordinating the activity in perisylvian areas. The model includes 5 brain regions: the left anterior middle temporal gyrus (LMTG), the right anterior middle temporal gyrus (RMTG), the left inferior frontal gyrus (LIFG), the right inferior frontal gyrus (RIFG), and the anterior medial frontal gyrus (mFG). The MNI coordinates are displayed on the right. The black arrows indicate forward/backward intrinsic connections between 2 anatomical nodes.

Results

Behavioural Data

The literal and nonliteral sentences were matched for length and syntactic complexity. However, subjects' mean RTs significantly differed between conditions (F1,21 = 674.43, P < 0.0001), the subjects being faster with literal (M = 1373.4, SD = 224.1) than idiomatic sentences (M = 1563.3, SD = 299.4), while there was no difference between true/false responses (F1,21 = 23.92, P = 0.72). Indeed, the mean RT for true responses was 1550.9 for idioms and 1364 for literal sentences, whereas for false responses, it was 1551 for idioms and 1386 for literal sentences. The interaction was not significant (F1,21 = 0.22, P = 0.65).

Accuracy was very high in both conditions (M = 56.1, SD = 4.6 out of a total of 62 idioms and M = 58.9, SD = 2.1 out of a total of 62 literal sentences) despite a lower accuracy for nonliteral sentences with respect to literal sentences, which was significant (F1,21 = 11.35, P < 0.005), while there was no difference between true and false sentences (F1,21 = 0.77, P = 0.39), nor the interaction true/false × literal/nonliteral was significant (F1,21 = 0.39, P = 0.85).

We also compared RTs and accuracy for ambiguous and nonambiguous idioms, and no significant difference was found (P = 0.06 and P = 0.2, for accuracy and RTs, respectively). Finally, accuracy was not correlated neither with familiarity (r = −0.11) nor with transparency (r = 0.33, P = 0.13).

Imaging Data

Conjunction Null

A network of bilateral brain activation covering frontal, parietal, temporal, and occipital areas was found. More specifically, in the LH, this network involved the IFG (BA 44/45), the precentral gyrus (BA 6), the postcentral gyrus, the inferior parietal lobule (BA 40), the insula, the middle temporal gyrus, the rolandic operculum, and the fusiform gyrus. The experimental task also engaged a pattern of visual areas including the left middle and left inferior occipital gyri (BA 19 and BA 18). Bilateral subcortical activation was found in the left thalamus and left cerebellum (see Table 1).

Table 1

Conjunction among the 4 experimental conditions

Area Z-score x, y, z CP% 
L inferior frontal gyrus Inf −46, 28, 16 BA 45 (30%) 
L inferior frontal gyrus Inf −52, 16, 28 BA 44 (40%) 
L precentral gyrus Inf −44, −12, 60  
L precentral gyrus Inf −56, 2, 32  
L insula Inf −32, 20, 0  
L postcentral gyrus Inf −34, −30, 52  
L inferior parietal lobule Inf −44, −42, 48  
L rolandic operculum 7.58 −48, −24, 16  
L middle temporal gyrus 7.16 −34, −52, 0  
L fusiform gyrus Inf −38, −52, −24  
L middle occipital gyrus Inf −34, −84, 16  
L inferior occipital gyrus Inf −46, −78, −4  
L thalamus Inf −14, −22, 8  
L cerebellum Inf −32, −68, −20  
R/L medial frontal gyrus Inf −4, 56, 4  
R/L supplementary motor area Inf −2, −2, 56  
R/L anterior cingulate gyrus Inf −2, 42, −4  
R superior frontal gyrus 7.21 38, −6, 64  
R inferior frontal gyrus Inf 50, 28, 16 BA 45 (70%) 
R inferior frontal gyrus Inf 50, 8, 28 BA 44 (40%) 
R insula Inf 32, 22, −4  
R middle cingulate cortex 7.27 10, −50, 32  
R superior temporal gyrus 7.32 62, −4, 0  
R middle temporal gyrus Inf 54, −68, 0  
R fusiform gyrus Inf 44, −52, −20  
R superior occipital gyrus Inf 22, −76, 36  
R middle occipital gyrus Inf 44, −70, 8  
R inferior occipital gyrus Inf 46, −76, −12  
R caudate nucleus Inf 10, 8, 8  
R thalamus Inf 12, −16, 4  
R cerebellum Inf 34, −46, −28  
Area Z-score x, y, z CP% 
L inferior frontal gyrus Inf −46, 28, 16 BA 45 (30%) 
L inferior frontal gyrus Inf −52, 16, 28 BA 44 (40%) 
L precentral gyrus Inf −44, −12, 60  
L precentral gyrus Inf −56, 2, 32  
L insula Inf −32, 20, 0  
L postcentral gyrus Inf −34, −30, 52  
L inferior parietal lobule Inf −44, −42, 48  
L rolandic operculum 7.58 −48, −24, 16  
L middle temporal gyrus 7.16 −34, −52, 0  
L fusiform gyrus Inf −38, −52, −24  
L middle occipital gyrus Inf −34, −84, 16  
L inferior occipital gyrus Inf −46, −78, −4  
L thalamus Inf −14, −22, 8  
L cerebellum Inf −32, −68, −20  
R/L medial frontal gyrus Inf −4, 56, 4  
R/L supplementary motor area Inf −2, −2, 56  
R/L anterior cingulate gyrus Inf −2, 42, −4  
R superior frontal gyrus 7.21 38, −6, 64  
R inferior frontal gyrus Inf 50, 28, 16 BA 45 (70%) 
R inferior frontal gyrus Inf 50, 8, 28 BA 44 (40%) 
R insula Inf 32, 22, −4  
R middle cingulate cortex 7.27 10, −50, 32  
R superior temporal gyrus 7.32 62, −4, 0  
R middle temporal gyrus Inf 54, −68, 0  
R fusiform gyrus Inf 44, −52, −20  
R superior occipital gyrus Inf 22, −76, 36  
R middle occipital gyrus Inf 44, −70, 8  
R inferior occipital gyrus Inf 46, −76, −12  
R caudate nucleus Inf 10, 8, 8  
R thalamus Inf 12, −16, 4  
R cerebellum Inf 34, −46, −28  

Note: Stereotactic coordinates (x, y, z) for significant clusters (random effects, family wise error P < 10−6 corrected) are given in millimeters together with effect sizes (Z-score). We also report the cytoarchitectonic probability (CP%) of activations in Broca's area, using the probability maps (Amunts et al. 1999) for BA 44 and BA 45 available with the SPM Anatomy toolbox. Inf = infinite; R = right; L = left.

In the RH, activations were found in the inferior and superior frontal gyri, the insula, the middle cingulate cortex, the superior and middle temporal gyri, and the fusiform gyrus, as well as in the canonical areas engaged by visual tasks, such as the superior, middle, and inferior occipital gyri. Subcortical activation involved the caudate nucleus and thalamus. Activation was also found in the cerebellum. Bilateral clusters of activation were found in the medial frontal gyrus, supplementary motor area, and anterior cingulum.

Idiomatic versus Literal Sentences

A bilateral frontotemporal network of increased cortical activity was found for idiomatic compared with literal sentences. The idiomatic condition activated bilaterally the IFG and the anterior middle temporal gyrus (BA 21), while in the LH, activation of the superior frontal gyrus in its anterior, medial part, of the inferior temporal gyrus and of the angular gyrus (BA 39) was observed. Finally, in the RH, activation was seen in the middle temporal gyrus and in the temporal pole.

Literal versus idiomatic sentences activated the supramarginal gyrus (BA 40) bilaterally and the superior (BA 7) and inferior (BA 40) parietal lobules in the RH. (see Table 2 and Fig. 3)

Figure 3.

Idiomatic versus literal sentences. (A) Activations for the 2 experimental conditions (white, idiomatic > literal; black, literal > idiomatic) are superimposed on the lateral surfaces of a cortical rendering, representing the mean anatomical T1 image of the 22 participants. Contrasts of parameter estimates, with 90% confidence intervals, for 2 representative activations are shown below: (B) idiomatic > literal: left IFG (x = −50, y = 26, z = 0); (C) literal > idiomatic: left supramarginal gyrus (x = −58, y = −32, z = 40). In these 2 regions, and in all the other significantly activated regions, a common activation pattern was found, with the amplitude of the canonical hemodynamic response function being significantly larger for one of the 2 conditions, and the effects for the first and second derivatives having the reverse sign. A larger effect for the first and second derivative means that the temporal and the dispersion profile of the hemodynamic response, respectively, presented a higher variance. In other words, the activations with a higher amplitude, for example the idiomatic sentences, also presented a consistently more uniform temporal profile and consistently less dispersion for idiomatic versus literal sentences.

Figure 3.

Idiomatic versus literal sentences. (A) Activations for the 2 experimental conditions (white, idiomatic > literal; black, literal > idiomatic) are superimposed on the lateral surfaces of a cortical rendering, representing the mean anatomical T1 image of the 22 participants. Contrasts of parameter estimates, with 90% confidence intervals, for 2 representative activations are shown below: (B) idiomatic > literal: left IFG (x = −50, y = 26, z = 0); (C) literal > idiomatic: left supramarginal gyrus (x = −58, y = −32, z = 40). In these 2 regions, and in all the other significantly activated regions, a common activation pattern was found, with the amplitude of the canonical hemodynamic response function being significantly larger for one of the 2 conditions, and the effects for the first and second derivatives having the reverse sign. A larger effect for the first and second derivative means that the temporal and the dispersion profile of the hemodynamic response, respectively, presented a higher variance. In other words, the activations with a higher amplitude, for example the idiomatic sentences, also presented a consistently more uniform temporal profile and consistently less dispersion for idiomatic versus literal sentences.

Table 2

Direct comparison: idiomatic versus literal sentences

Area Z-score x, y, z CP% 
IDIOMATIC > LITERAL 
    L medial frontal gyrus Inf −4, 54, 32  
    L superior frontal gyrus 6.79 −10, 48, 48  
    L inferior frontal gyrus Inf −50, 26, 0 BA 45 (30%) 
    L angular gyrus 6.16 −50, −58, 24  
    L anterior middle temporal gyrus 7.42 −52, 4, −32  
    L inferior temporal gyrus Inf −52, 6, −40  
    R inferior frontal gyrus 6.02 52, 32, −8 BA 45 (20%) 
    R middle temporal pole 6.87 46, 12, −36  
    R anterior middle temporal gyrus 6.15 54, 0, −24  
    R posterior middle temporal gyrus 5.61 52, −48, 12  
LITERAL > IDIOMATIC 
    L supramarginal gyrus 5.56 −58, −32, 40  
    R superior parietal lobule 5.18 42, −46, 60  
    R inferior parietal lobule 5.45 58, −36, 52  
    R supramarginal gyrus 5.51 62, −28, 36  
Area Z-score x, y, z CP% 
IDIOMATIC > LITERAL 
    L medial frontal gyrus Inf −4, 54, 32  
    L superior frontal gyrus 6.79 −10, 48, 48  
    L inferior frontal gyrus Inf −50, 26, 0 BA 45 (30%) 
    L angular gyrus 6.16 −50, −58, 24  
    L anterior middle temporal gyrus 7.42 −52, 4, −32  
    L inferior temporal gyrus Inf −52, 6, −40  
    R inferior frontal gyrus 6.02 52, 32, −8 BA 45 (20%) 
    R middle temporal pole 6.87 46, 12, −36  
    R anterior middle temporal gyrus 6.15 54, 0, −24  
    R posterior middle temporal gyrus 5.61 52, −48, 12  
LITERAL > IDIOMATIC 
    L supramarginal gyrus 5.56 −58, −32, 40  
    R superior parietal lobule 5.18 42, −46, 60  
    R inferior parietal lobule 5.45 58, −36, 52  
    R supramarginal gyrus 5.51 62, −28, 36  

Note: Family wise error (P < 0.05) corrected. CP% = cytoarchitectonic probability; Inf = infinite; R = right; L = left.

Direct Comparison between the 4 Conditions

No suprathreshold clusters of activation were found, suggesting that, within idiomatic and literal condition, the pattern of activation was unaffected by true/false condition.

Control Analysis: Ambiguous versus Nonambiguous Idiomatic Sentences

No significant differences were found between ambiguous and nonambiguous idiomatic sentences, even at an uncorrected significance threshold of P < 0.001. This indicates that the results reported above were not biased by having ambiguous and nonambiguous idiomatic sentences conflated within the idiomatic experimental conditions.

DCM Analysis

The results are shown in Table 3. One sample t-test analyses showed that, at the group level, all the intrinsic connections included in our model were significantly different from zero for the main effect of idiomatic minus literal sentences. In other words, the idiomatic-specific activation of the medial prefrontal area elicited a significantly greater effective connectivity between the brain regions included in the DCM for Idiomatic versus literal sentences.

Table 3

DCM: estimated parameters for the intrinsic connections, which refer to the coupling between 2 regions (effective connectivity over intrinsic connections, P < 0.001)

Intrinsic connections Mean SD P 
mFG → LIFG 0.086 0.143 <0.0001 
mFG → LMTG 0.074 0.128 <0.0001 
mFG → RIFG 0.068 0.132 <0.0001 
mFG → RMTG 0.054 0.105 <0.0001 
LIFG → LMTG 0.011 0.029 <0.001 
LMTG → LIFG 0.014 0.033 <0.001 
RIFG → RMTG 0.005 0.013 <0.001 
RMTG → RIFG 0.009 0.024 <0.001 
Intrinsic connections Mean SD P 
mFG → LIFG 0.086 0.143 <0.0001 
mFG → LMTG 0.074 0.128 <0.0001 
mFG → RIFG 0.068 0.132 <0.0001 
mFG → RMTG 0.054 0.105 <0.0001 
LIFG → LMTG 0.011 0.029 <0.001 
LMTG → LIFG 0.014 0.033 <0.001 
RIFG → RMTG 0.005 0.013 <0.001 
RMTG → RIFG 0.009 0.024 <0.001 

Note: Mean parameters (mean), SD, and 2-tailed statistical significance (P) are reported. (See Fig. 2 for a description of the areas and of the intrinsic connections of the model). mFG, medial frontal gyrus; LIFG, left inferior frontal gyrus; LMTG, left anterior middle temporal gyrus; RIFG, right inferior frontal gyrus; RMTG, right anterior middle temporal gyrus.

Discussion

In the current study, fMRI was used to examine the brain areas involved in the processing of idiomatic sentences. We used a judgment task, in which the participants had to decide whether or not the meaning of a sentence matched a picture. The results reveal that making yes/no judgments about literal and nonliteral sentences yielded a similar network of activity, involving the language areas of the LH. However, accessing the nonliteral meaning led to overall higher levels of activation. We found an activation of the left temporal cortex, as predicted by neuropsychological and rTMS studies. In addition, in line with previous studies using different methodologies, the left superior frontal (approximately covering BA 9), as well as the left IFG, were specifically involved in processing idiomatic sentences. Activations were also seen in the right superior and middle temporal gyri, in the temporal pole, and in the right IFG. In contrast, literal sentences activated the left inferior parietal lobule and the right supramarginal gyrus.

The findings of the present study, therefore, are in line with recent fMRI studies performed with metaphors, using different paradigms, such as reading and judging metaphor plausibility (Rapp et al. 2004) or choosing the 2 adjectives, out of 3, which had the same metaphorical meaning (Lee and Dapretto 2006). A bihemispheric network, involving the IFG bilaterally and the left middle temporal gyrus, has been also recently reported (Zempleni et al. 2007).

The activation of the left inferior parietal lobule with literal sentences may be related to differential requirements of phonological memory resources in order to assess the matching between the sentence and the picture. Because in the case of idiomatic sentences a picture can only depict a concrete instantiation of the figurative meaning that is generally abstract, the participant needs to decide whether this is appropriate or not. This decision requires a central executive involvement to maintain the sentence and then match it to the picture. In contrast, the detection of a possible mismatch for the literal sentence may require only a verbatim retention of the sentence, for which the phonological loop is necessary. Overall, subjects took longer and were less accurate in the idiomatic condition, suggesting that the greater activity and the prefrontal involvement could in part reflect the additional task demands of holding in memory different interpretations and matching them with the picture, while switching from one meaning to the other. This activation result parallels the one obtained on metaphor comprehension (Lee and Dapretto 2006). A number of investigations have linked activity in the left IFG to the selection of semantic knowledge among competing alternatives or, more in general, to the selection among alternatives without necessarily involving semantics (Zhang et al. 2004). Accordingly, this region was strongly activated in the present study during the idiomatic condition, in which 2 different meanings (the literal and the idiomatic) were presumably retrieved and evaluated in order to accomplish the task.

Furthermore, no difference was found between ambiguous and nonambiguous idioms, suggesting that, at least at a gross level, processes and neural correlates are the same for both types of idiomatic sentences. However, this is in contrast with recent neuropsychological data on aphasic patients and with the fMRI study by Zempleni et al. (2007), in which an additional right temporal activation was found in the case of ambiguous items and deserves further investigation.

An additional finding in our study was the involvement of the anterior–medial part of the superior frontal gyrus, with a left-sided preponderance. The anterior prefrontal cortex has been recently considered to be specifically involved in integrating the outcomes of 2 or more separate cognitive operations in the pursuit of a higher behavioral goal. Christoff and Gabrieli (2000) suggested that the anterior prefrontal cortex might be specialized for the explicit processing of internal mental states and events. In particular, tasks that require the generation and monitoring of internally generated responses should activate this region, an example being the Tower of London Test, in which possible solutions must be internally generated and then accepted or rejected, before making a move. Alternatively, Ramnani and Owen (2004) proposed that the anterior prefrontal cortex is engaged when problems involve more than one cognitive process: that is when the application of one cognitive operation (such as a rule) on its own is not sufficient to solve the problem as a whole, and the integration of the results of 2 or more separate cognitive operations is required to fulfill the higher behavioral goal. Multiple, related cognitive operations can only be successfully performed if they are coordinated. One can speculate that the coordination of information processing and information transfer between multiple operations across supramodal cortex is an important aspect of anterior prefrontal cortex function. Applying this theory to idiom comprehension, it can be argued that the sentence is linguistically analyzed, producing 2 possible interpretations: the literal and the figurative. Both are matched with the context and eventually one or the other is accepted (or, in some other cases, both may be rejected). In the present experiment, the coordination of the different interpretations with the meaning of the picture is required only in the case of idiom comprehension. In the case of literal sentence, the meaning of the picture needs to be matched with one interpretation only.

An additional support to this view comes from the results of the DCM analysis. In defining our model, we hypothesized that the specific effect of sentence type leading to an activation of the anterior medial frontal gyrus would propagate to perisylvian regions, bilaterally, in terms of an increased effective connectivity for idiomatic versus literal sentences. Our results do indeed favor this view. This may suggest that the anterior medial frontal gyrus has a prominent role in regulating the cross talk between perisylvian language areas and that the coherence of activation in these areas is higher when a choice between 2 competing hypotheses is required, such as in the case of idiomatic sentences. The anterior medial frontal gyrus might mediate the inhibition of the alternative interpretation in favor of the correct one. In this context, the salience of the anterior medial frontal gyrus as an input region, as opposed to other brain regions, was also confirmed by comparing the DCM described here with a second model that included the same brain areas, but in which the psychological input was delivered to the middle temporal gyrus, bilaterally, rather than to the medial frontal gyrus (details not shown). In all the 22 subjects, the model with input to the medial frontal gyrus leads to a better fit than the model with input to the middle temporal gyri.

The present data undermine the hypothesis of a specific role of the RH, also in this type of figurative language, as it has been already demonstrated in the case of metaphors (Lee and Dapretto 2006). Overall, they support the view that idiomatic language follows the same procedure as literal language, plus a selection process between alternative meanings, which is likely to be supported by the IFG, and a supplementary high-level cognitive process, supported by the anterior prefrontal cortex, possibly monitoring and integrating the results of the linguistic analysis and of the selection between competing meanings.

Therefore, the results of the present study provide strong evidence against the lexical hypothesis (Swinney and Cutler 1979) because it is unlikely that such a complex network would be activated in the case that idioms were treated as single complex words. Instead, they are consistent with the configurational hypothesis suggested by Cacciari and Tabossi (1988), which proposes that the individual components of the idiom are analyzed as normal linguistic items, until a key part is processed that prompts the recognition of the idiomatic nature of the string and the activation of its figurative meaning. In functional terms, the linguistic analysis engages an extensive bihemispheric network, with left-sided prevalence. The additional processing required by idiomatic sentences is reflected by the recruitment of other frontotemporal regions in the 2 hemispheres, with a key role of the left anterior and anteromedial prefrontal cortices. The necessary role of the LH within the network is supported by TMS data (Rizzo et al. 2007), showing that rTMS applied over the left BA 9 reduces accuracy in the case of idiomatic but not literal sentences. This is particularly evident when rTMS is applied 80 ms after picture presentation.

Additional converging evidence is provided by the results of neuropsychological studies (Papagno et al. 2006). An impaired performance has been found in both LBD and RBD patients, whose lesions were mapped following Damasio and Damasio's procedure (1989). However, although all LBD patients were significantly impaired, no matter if the lesion was frontal or temporal (although the most severely impaired patients exhibited an overlapping damaged area in the temporal lobe), in the case of RBD, only those with a right prefrontal lesion were impaired, suggesting that this region could be involved in inhibiting the literal interpretation or, alternatively, in retrieving/activating the figurative one. The absence of a right prefrontal lesion effect on literal sentence comprehension is in line with the present experiment, showing that this area is selectively involved in idiom comprehension and not in any task requiring a selective choice. Neuropsychological studies often use sentence-to-picture matching tasks, in which the foils often include the literal which may force a literal interpretation. The absence of literal alternatives in the present experiment should have minimized this effect.

Finally, the graded salience hypothesis (Giora 1997) suggests that accessing the meaning of salient nonliteral meanings, such as the ones used in the present study, would not be contingent upon the availability of contextual information and, as such, retrieval of these meanings need not require RH involvement, relying instead on LH mechanisms. However, we found an activation of the RH even with conventional, familiar idioms. A possible explanation is suggested by Kacinik and Chiarello (2007), who argued that RH can maintain various possible interpretations and this sustained activation of inconsistent meanings may act as a backup system in the event that the interpretation of an expression needs to be revised. Moreover, several neuroimaging studies indicate that the RH is recruited whenever contextual and semantic information needs to be integrated in order to interpret ambiguous language (e.g., Beeman and Chiarello 1998; Rodd et al. 2005).

In conclusion, converging evidence (neuropsychological studies, neurophysiological studies using rTMS, and fMRI studies) suggests that the prefrontal cortex is involved in idiom interpretation. Further studies avoiding the use of pictorial material need, however, to be run in order to confirm this result. In addition, a higher number of literal sentences as compared with idiomatic ones need to be included in the experimental set to avoid specific the deployment of strategies. Finally, a comparison between different types of idioms (ambiguous vs. nonambiguous, opaque vs. transparent and so on) is the next step required to check if the intrinsic linguistic properties of idioms result in a different pattern of brain activity.

This research was supported by a PRIN grant 2003 to SFC and CP. CP was also supported by a FAR grant 2003. Conflict of Interest: None declared.

Appendix

List of idioms used in the experiment

 Familiarity index (0–3) Transparency (0–5) Picture relatedness (0–5) 
Andare al fresco 2.7 2.5 3.8 
Andare in bestia 2.9 3.8 4.0 
Avere i minuti contati 2.9 4.0 3.4 
Avere il cuore in mano 2.2 2.9 3.0 
Avere il pollice verde 2.9 3.0 4.5 
Avere le mani in pasta 2.5 2.8 3.6 
Avere la puzza sotto il naso 2.9 3.0 3.9 
Avere poco sale in zucca 2.7 2.7 3.2 
Dare del filo da torcere 2.9 2.9 4.4 
Dare i numeri 2.9 2.4 1.6 
Essere a piede libero 2.4 3.4 4.5 
Essere al settimo cielo 2.9 3.7 4.2 
Essere di facili costumi 2.9 3.1 4.1 
Essere in forma 2.9 4.1 3.5 
Essere sul viale del tramonto 2.3 3.4 3.1 
Essere sulla cattiva strada 2.8 4.4 4.0 
Essere uno stinco di santo 2.7 3.0 4.5 
Far venire il latte alle ginocchia 2.6 2.3 3.1 
Fare fiasco 2.7 2.4 4.2 
Farsene un baffo 2.6 2.1 3.5 
Mandare a monte 2.9 2.0 3.6 
Mettersi le gambe in spalla 2.1 2.5 3.0 
Montarsi la testa 2.9 3.1 2.1 
Mordere il freno 1.5 2.3 2.0 
Non vedere l'ora 2.9 3.3 2.4 
Perdere la faccia 2.7 3.1 2.1 
Prendere fischi per fiaschi 2.6 3.3 3.4 
Prendere in castagna 2.7 2.6 4.7 
Restare con un palmo di naso 2.6 2.3 4.2 
Rimetterci le penne 2.8 3.6 4.8 
Spremere le meningi 2.8 3.9 2.0 
Stare alle costole 2.6 3.3 2.6 
Stare in campana 2.3 2.3 2.7 
Tendere le orecchie 2.5 3.9 4.1 
Tenere banco 2.4 2.8 2.4 
Vendere cara la pelle 2.3 3.6 4.1 
Venire alle mani 2.7 3.6 4.0 
Alzare i tacchi 2.7 3.4 3.4 
Alzare il gomito 2.8 3.0 2.3 
Attaccare bottone 2.9 2.7 4.2 
Avere le mani bucate 2.8 3.2 3.7 
Chiudere un occhio 2.9 3.7 3.8 
Essere al verde 2.9 2.6 3.9 
Gettare la spugna 2.9 2.9 4.4 
Levare le tende 2.5 3.8 4.0 
Mangiarsi il fegato 2.7 2.7 4.0 
Mettere la pulce nell'orecchio 2.8 3.0 3.9 
Mettersi le mani nei capelli 2.8 3.6 3.5 
Perdere il filo 2.8 2.9 4.0 
Perdere la bussola 2.5 3.8 3.8 
Perdere la testa 2.9 3.9 2.2 
Prendere la porta 2.3 3.1 3.9 
Prendere un granchio 2.6 2.6 4.4 
Rompere il ghiaccio 2.9 2.7 3.7 
Saltare la mosca al naso 2.0 1.9 2.4 
Scendere a rotta di collo 2.2 3.0 4.1 
Sputare il rospo 2.9 2.8 3.4 
Stringere i denti 2.9 3.6 3.7 
Tagliare la corda 2.8 2.9 4.8 
Tirare la cinghia 2.8 3.0 3.8 
Vedere le stelle 2.8 2.9 4.4 
Vuotare il sacco 2.8 3.5 4.0 
 Familiarity index (0–3) Transparency (0–5) Picture relatedness (0–5) 
Andare al fresco 2.7 2.5 3.8 
Andare in bestia 2.9 3.8 4.0 
Avere i minuti contati 2.9 4.0 3.4 
Avere il cuore in mano 2.2 2.9 3.0 
Avere il pollice verde 2.9 3.0 4.5 
Avere le mani in pasta 2.5 2.8 3.6 
Avere la puzza sotto il naso 2.9 3.0 3.9 
Avere poco sale in zucca 2.7 2.7 3.2 
Dare del filo da torcere 2.9 2.9 4.4 
Dare i numeri 2.9 2.4 1.6 
Essere a piede libero 2.4 3.4 4.5 
Essere al settimo cielo 2.9 3.7 4.2 
Essere di facili costumi 2.9 3.1 4.1 
Essere in forma 2.9 4.1 3.5 
Essere sul viale del tramonto 2.3 3.4 3.1 
Essere sulla cattiva strada 2.8 4.4 4.0 
Essere uno stinco di santo 2.7 3.0 4.5 
Far venire il latte alle ginocchia 2.6 2.3 3.1 
Fare fiasco 2.7 2.4 4.2 
Farsene un baffo 2.6 2.1 3.5 
Mandare a monte 2.9 2.0 3.6 
Mettersi le gambe in spalla 2.1 2.5 3.0 
Montarsi la testa 2.9 3.1 2.1 
Mordere il freno 1.5 2.3 2.0 
Non vedere l'ora 2.9 3.3 2.4 
Perdere la faccia 2.7 3.1 2.1 
Prendere fischi per fiaschi 2.6 3.3 3.4 
Prendere in castagna 2.7 2.6 4.7 
Restare con un palmo di naso 2.6 2.3 4.2 
Rimetterci le penne 2.8 3.6 4.8 
Spremere le meningi 2.8 3.9 2.0 
Stare alle costole 2.6 3.3 2.6 
Stare in campana 2.3 2.3 2.7 
Tendere le orecchie 2.5 3.9 4.1 
Tenere banco 2.4 2.8 2.4 
Vendere cara la pelle 2.3 3.6 4.1 
Venire alle mani 2.7 3.6 4.0 
Alzare i tacchi 2.7 3.4 3.4 
Alzare il gomito 2.8 3.0 2.3 
Attaccare bottone 2.9 2.7 4.2 
Avere le mani bucate 2.8 3.2 3.7 
Chiudere un occhio 2.9 3.7 3.8 
Essere al verde 2.9 2.6 3.9 
Gettare la spugna 2.9 2.9 4.4 
Levare le tende 2.5 3.8 4.0 
Mangiarsi il fegato 2.7 2.7 4.0 
Mettere la pulce nell'orecchio 2.8 3.0 3.9 
Mettersi le mani nei capelli 2.8 3.6 3.5 
Perdere il filo 2.8 2.9 4.0 
Perdere la bussola 2.5 3.8 3.8 
Perdere la testa 2.9 3.9 2.2 
Prendere la porta 2.3 3.1 3.9 
Prendere un granchio 2.6 2.6 4.4 
Rompere il ghiaccio 2.9 2.7 3.7 
Saltare la mosca al naso 2.0 1.9 2.4 
Scendere a rotta di collo 2.2 3.0 4.1 
Sputare il rospo 2.9 2.8 3.4 
Stringere i denti 2.9 3.6 3.7 
Tagliare la corda 2.8 2.9 4.8 
Tirare la cinghia 2.8 3.0 3.8 
Vedere le stelle 2.8 2.9 4.4 
Vuotare il sacco 2.8 3.5 4.0 

References

Amunts
K
Schleicher
A
Buergel
U
Mohlberg
H
Uylings
HBM
Zilles
K
Broca's region revisited: cytoarchitecture and intersubject variability
J Comp Neurol
 , 
1999
, vol. 
412
 (pg. 
319
-
341
)
Anaki
D
Faust
M
Kravetz
S
Cerebral hemispheric asymmetries in processing lexical metaphors
Neuropsychologia
 , 
1998
, vol. 
36
 (pg. 
353
-
362
)
Andersson
JLR
Hutton
C
Ashburner
J
Turner
R
Friston
K
Modeling geometric deformations in EPI time series
Neuroimage
 , 
2001
, vol. 
13
 (pg. 
903
-
919
)
Beeman
M
Chiarello
C
Right hemisphere language comprehension. Perspectives from cognitive neuroscience
1998
Mahwah (NJ)
Lawrence Erlbaum
Bottini
G
Sterzi
R
Paulesu
E
Vallar
G
Cappa
SF
Erminio
F
Passingham
RE
Frith
CD
Frackowiak
RS
Identification of the central vestibular projections in man: a positron emission tomography activation study
Exp Brain Res
 , 
1994
, vol. 
99
 
1
(pg. 
164
-
169
)
Burgess
C
Chiarello
C
Neurocognitive mechanisms underlying metaphor comprehension and other figurative language
Metaphor Symb Act
 , 
1996
, vol. 
11
 (pg. 
67
-
84
)
Cacciari
C
Tabossi
P
The comprehension of idioms
J Mem Lang
 , 
1988
, vol. 
27
 (pg. 
668
-
683
)
Cacciari
C
Reati
F
Colombo
M-R
Padovani
R
Rizzo
S
Papagno
C
The comprehension of ambiguous idioms in aphasic patients
Neuropsychologia
 , 
2006
, vol. 
44
 (pg. 
1305
-
1314
)
Christoff
K
Gabrieli
JDE
The frontopolar cortex and human cognition: evidence for a rostrocaudal hierarchical organization within the human prefrontal cortex
Psychobiology
 , 
2000
, vol. 
28
 (pg. 
168
-
186
)
Damasio
H
Damasio
AR
Lesion analysis in neuropsychology
1989
Oxford
Oxford University Press
Friston
KJ
Glaser
DE
Henson
RN
Kiebel
S
Phillips
C
Ashburner
J
Classical and Bayesian inference in neuroimaging: applications
Neuroimage
 , 
2002
, vol. 
16
 (pg. 
484
-
512
)
Friston
KJ
Harrison
L
Penny
W
Dynamic causal modelling
Neuroimage
 , 
2003
, vol. 
19
 (pg. 
1273
-
1302
)
Gibbs
RW
Literal meaning and psychological theory
Cognit Sci
 , 
1984
, vol. 
8
 (pg. 
275
-
304
)
Gibbs
RW
Wilson
R
Keil
F
Figurative language. The MIT encyclopedia of the cognitive science
1999
Cambridge (MA)
MIT Press
(pg. 
314
-
315
)
Giora
R
Understanding figurative and literal language: the graded salience hypothesis
Cogn Linguist
 , 
1997
, vol. 
7
 (pg. 
183
-
206
)
Giora
R
Zaidel
E
Soroker
N
Batori
G
Kasher
A
Differential effects of right- and left-hemisphere damage on understanding sarcasm and metaphor
Metaphor Symb
 , 
2000
, vol. 
15
 (pg. 
63
-
83
)
Glucksberg
S
Brown
M
McGlone
MS
Conceptual metaphors are not automatically accessed during idiom comprehension
Mem Cognit
 , 
1993
, vol. 
21
 
5
(pg. 
711
-
719
)
Huber-Okrainec
J
Blaser
SE
Dennis
M
Idiom comprehension deficits in relation to corpus callosum agenesis and hypoplasia in children with spina bifida meningomyelocele
Brain Lang
 , 
2005
, vol. 
93
 (pg. 
349
-
368
)
Jennings
JM
McIntosh
AR
Kapur
S
Tulving
E
Houle
S
Cognitive subtractions may not add up: the interaction between semantic processing and response mode
Neuroimage
 , 
1997
, vol. 
5
 (pg. 
229
-
239
)
Kacinik
NA
Chiarello
C
Understanding metaphors: is the right hemisphere uniquely involved?
Brain Lang
 , 
2007
, vol. 
100
 (pg. 
188
-
207
)
Kempler
D
Van Lancker
D
Marchman
V
Bates
E
Idiom comprehension in children and adults with unilateral brain damage
Dev Neuropsychol
 , 
1999
, vol. 
15
 (pg. 
327
-
349
)
Koechlin
E
Ody
C
Kouneiher
F
The architecture of cognitive control in the human prefrontal cortex
Science
 , 
2003
, vol. 
302
 
5648
(pg. 
1181
-
1185
)
Lee
S
Dapretto
M
Metaphorical vs. literal word meanings: fMRI evidence against a selective role of the right hemisphere
Neuroimage
 , 
2006
, vol. 
29
 (pg. 
536
-
544
)
Lee
L
Friston
K
Horwitz
B
Large-scale neural models and dynamic causal modelling
Neuroimage
 , 
2006
, vol. 
30
 (pg. 
1243
-
1254
)
Logie
RH
Venneri
A
Della Sala
S
Redpath
TW
Marshall
I
Brain activation and the phonological loop: the impact of rehearsal
Brain Cogn
 , 
2003
, vol. 
53
 (pg. 
293
-
296
)
Nenonen
M
Niemi
J
Laine
M
Representation and processing of idioms: evidence from aphasia
J Neurolinguistics
 , 
2002
, vol. 
15
 (pg. 
43
-
58
)
Nichols
T
Brett
M
Andersson
J
Wager
T
Poline
JB
Valid conjunction inference with the minimum statistic
Neuroimage
 , 
2005
, vol. 
25
 (pg. 
653
-
660
)
Oldfield
RC
The assessment and analysis of handedness: the Edinburgh inventory
Neuropsychologia
 , 
1971
, vol. 
9
 (pg. 
97
-
113
)
Oliveri
M
Papagno
C
Romero
L
Left but not right temporal lobe involvement in opaque idiom comprehension: a repetitive transcranial stimulation study
J Cogn Neurosci
 , 
2004
, vol. 
16
 (pg. 
848
-
855
)
Papagno
C
Caporali
A
Testing idiom comprehension in aphasic patients: the modality and the type of idiom effects
Brain Lang
 , 
2007
, vol. 
100
 (pg. 
208
-
220
)
Papagno
C
Curti
R
Rizzo
S
Crippa
F
Colombo
MR
Is the right hemisphere involved in idiom comprehension? A neuropsychological study
Neuropsychology
 , 
2006
, vol. 
20
 (pg. 
598
-
606
)
Papagno
C
Genoni
A
The role of syntactic competence in idiom comprehension: a study on aphasic patients
J Neurolinguistics
 , 
2004
, vol. 
17
 (pg. 
371
-
382
)
Papagno
C
Tabossi
P
Colombo
M
Zampetti
P
Idiom comprehension in aphasic patients
Brain Lang
 , 
2004
, vol. 
89
 (pg. 
226
-
234
)
Penny
WD
Holmes
AP
Frackowiak
RSJ
Friston
KJ
Frith
CD
Dolan
R
Price
CJ
Frackowiak
RSJ
Friston
KJ
Frith
CD
Dolan
R
Price
CJ
Zeki
S
Ashburner
J
Penny
W
Random effects analysis. Human brain function
2003
San Diego (CA)
Academic Press
(pg. 
843
-
850
)
Peterson
RR
Burgess
C
Dell
GS
Eberhard
K
Dissociation between syntactic and semantic processing during idiom comprehension
J Exp Psychol Learn Mem Cogn
 , 
2001
, vol. 
27
 (pg. 
1223
-
1237
)
Ramnani
N
Owen
AM
Anterior prefrontal cortex: insights into function from anatomy and neuroimaging
Nat Rev Neurosci
 , 
2004
, vol. 
5
 (pg. 
184
-
194
)
Rapp
AM
Leube
DT
Erb
M
Grodd
W
Kircher
TTJ
Neural correlates of metaphor processing
Cognit Brain Res
 , 
2004
, vol. 
20
 (pg. 
395
-
402
)
Ridderinkhof
KR
Ullsperger
M
Crone
EA
Nieuwenhuis
S
The role of the medial frontal cortex in cognitive control
Science
 , 
2004
, vol. 
306
 
5695
(pg. 
443
-
447
)
Rizzo
S
Sandrini
M
Papagno
C
The dorsolateral prefrontal cortex in idiom comprehension: a repetitive TMS study
Brain Res Bull
 , 
2007
, vol. 
71
 (pg. 
523
-
528
)
Rodd
JM
Davis
MH
Johnsrude
IS
The neural mechanism of speech comprehension: fMRI studies of semantic ambiguity
Cereb Cortex
 , 
2005
, vol. 
15
 
8
(pg. 
1261
-
1269
)
Swinney
DA
Cutler
A
The access and processing of idiomatic expression
J Verb Learn Verb Behav
 , 
1979
, vol. 
18
 (pg. 
523
-
534
)
Van Lancker
D
Kempler
D
Comprehension of familiar phrases by left but not by right hemisphere damaged patients
Brain Lang
 , 
1987
, vol. 
32
 (pg. 
265
-
277
)
Winner
E
Gardner
H
The comprehension of metaphor in brain-damaged patients
Brain
 , 
1977
, vol. 
100
 (pg. 
717
-
729
)
Zempleni
MZ
Haverkort
M
Renken
R
Sotwe
LA
Evidence for bilateral involvement in idiom comprehension: an fMRI study
Neuroimage
 , 
2007
, vol. 
34
 (pg. 
1280
-
1291
)
Zhang
JX
Feng
CM
Fox
PT
Gao
JH
Tan
LH
Is left inferior frontal gyrus a general mechanism for selection?
Neuroimage
 , 
2004
, vol. 
23
 (pg. 
596
-
603
)