Inflecting regular and irregular verbs: neuroimaging and behavioural data from the three Italian conjugations

Introduction The present study investigates the production of inflected forms from word-stems with a regular vs. an irregular inflection. This issue has been mostly focused on the theoretical debate between Dual Mechanism theories (rule-based procedures vs. recovery of whole forms stored in an associative network, Pinker & Ullman, 2002) and Connectionist models (single procedure for both regular and irregular forms, McClelland & Patterson, 2002). The two accounts make different predictions about the neural underpinnings of lexical processing for regular and irregular verbs. The declarative/procedural model is a modular theory positing that irregular forms are recovered from the mental lexicon which depends on declarative memory and is subserved by the temporal lobe, whereas regular forms are assembled by means of a mental grammar which involves procedural memory and is subserved by the frontal cortex and the basal ganglia (Ullman, 2001). Concurrent models do not postulate a categorical difference between regular and irregular forms and suggest that different pattern of brain activations elicited by regular and irregular verbs are predicted by phonological, semantic and other probabilistic features of verbs (Joanisse & Seidenberg, 2005). However, an alternative hypothesis, mainly grounded on linguistic, behavioral and simulation data (Burzio, 1998; Bordag & Pechmann, 2009; Colombo et al., 2004; 2006; Laudanna, 2007; Laudanna et al., 2004; Pirrelli & Herreros, 2005; Verissimo & Clahsen, 2009), suggests that all inflected forms in the mental lexicon, regardless of regularity, are organized in clusters on the basis of their paradigmatic relations among word-forms other than the regular/irregular distinction and that inflectional processes are likely to be governed by membership of words to specific clusters, namely the inflectional classes. This third positions predicts the possibility to detect experimental effects driven by the membership of words to inflectional classes rather than to the coarse distinction between regular and irregular. In this study we try to shed light on the issue by exploiting a suitable test case, the Italian verb inflectional system. Here, verbs fall into one of three conjugations each exhibiting different degrees of regularity, numerosity and consistency. By using both behavioral and neuroimaging techniques, we aim at demonstrating that mechanisms underlying speaker’s ability to perform inflectional operations rely on grammatical properties of verbs depending on their membership to different inflectional classes. Methods A past-participle generation task was admnistered in a behavioral (42 participants) and an eventrelated fMRI setting (23 participants). A set of 168 Italian verbs was selected and arranged into 6 subsets of experimental stimuli on the basis of Regularity (regulars vs. irregulars) and Conjugation membership (1st vs. 2nd vs. 3rd). The six subsets were matched on frequency, length, N-count, locus and manner of articulation of the initial phoneme, syllabic structure of the initial syllable, syllabic structure of both the infinitive and past-participle form. Different sets of statistical analyses were carried out on response latencies, accuracy rates and BOLD responses. Analyses testing putative effects of Conjugation were restricetd to response Poster session 1 34 patterns only elicited by regular verbs. Analyses on response measures only elicited by 2nd and 3 rd conjugation verbs tested the interaction between Regularity and Conjugation. Results Main effects of the variable Conjugation were obtained both in the behavioural and in the fMRI experiments: 2nd conjugation regular verbs were responded significantly slower and poorer than 1 st and 3rd conjugation regular verbs. The left middle frontal gyrus and the pre-supplementary motor area exhibited a significantly higher BOLD activation response for 2nd and 3rd conjugation regular verbs than for 1st conjugation regular verbs. In addition, the analyses performed on both 2nd and 3rd conjugation regular and irregular verbs revealed that 2nd conjugation verbs were slower and less accurate than 3rd conjugation verbs and that the perigenual part of the left anterior cingulate cortex was significantly more activated for 2nd than for 3rd conjugation verbs. A main effect of Regularity was also observed: irregular verbs were slower and less accurate than regular verbs. The left prefrontal cortex and the anterior part of the left inferior frontal gyrus were found significantly more activated for irregular when compared to regular verbs. Interestingly, a significant Conjugation by Regularity interaction was observed: behaviorally, the Regularity effect was stronger for 3rd than for 2nd conjugation verbs. In the left and right postcentral gyri, the 2nd conjugation irregular verbs elicited significant weaker activation than regular ones; the 3rd conjugation verbs elicited the opposite pattern. Discussion The present study demonstrates that inflecting Italian verbs from different conjugations, irrespective of regularity, requires different cognitive operations since it elicits differences in behavioral response latencies, accuracy rates and cortical activations. Rather than a segregation between different brain structures devoted to the processing of regular and irregular verbs, different patterns of activation for both the regular and irregular verbs in the same structures were found. Differences among regular verbs belonging to different inflectional classes were also obtained both on behavioral and neural grounds. This pattern of data cannot be easily accommodated within either a dual or a single mechanism account. Dual mechanism theories postulate that all regular verbs follow symbolic inflectional rules, thereby no differences among regular verbs are expected. The observed difference cannot be ascribed neither to any formal feature as predicted by single mechanism accounts: actually, in our work we matched many formal features across the experimental categories. Our data corroborate the idea that language-specific properties of inflectional classes might modulate cognitive operations and biological underpinnings of inflectional behavior.