Rethinking prerolandic regions role in language processing: insights from a neurolinguistic study of noun-verb dissociation in motor neuron disease.

Background Selective deficit of verb (V) vs. noun (N) processing has been frequently reported in the motor neuron disease–frontotemporal dementia (MND-FTD) complex. Prerolandic regions atrophy has been traditionally thought to account for N-V dissociation in the MND-FTD complex via selective impairment of action semantics (Embodied Cognition Theory, ECT) [1]. Nonetheless, prerolandic regions might be engaged in post-semantic stages of V processing. Indeed, neurofunctional evidence in both the healthy [2] and damaged [3] brain hints at prerolandic regions involvement in morpho-phonological processing of V stimuli regardless of their semantic status. Besides, behavioral evidence shows that Ns and Vs can dissociate in MND patients without semantic impairment [4]. Furthermore, task-difficulty effects – i.e., both executive dysfunctions and global cognitive efficiency decrease - might account to a degree for N-V dissociation in MND patients. Executive functioning (EF) is known to possibly be defective in MND patients [4], and N-V dissociation has been described in neurological disorders that are not characterized by language dysfunctions as pathognomonic behavioral features [5]. This study therefore aimed at assessing the validity of post-semantic and task-difficulty-effect accounts vs. ECT-framed explanations with regards to posited neurocognitive mechanisms underlying N-V dissociation in MND patients. Methods Thirty consecutive MND patients and 29 age- and education-matched healthy control (HC) participants were recruited and underwent global cognitive screening. The two groups were compared on tasks evaluating N and V lexical retrieval (visual confrontation naming), object- and action-semantics (picture-to-picture matching), and V lexical and morphosyntactic implementation (V argument structure production from visual confrontation of both picture and lexical target items), while controlling for an Executive Functioning Index (EFI) computed from phonemic and alternate (phonemic/semantic) verbal fluencies tasks. Effects of motor content (actionality) of Ns and Vs and argument structure complexity (ASC) of Vs (transitive vs. unergative vs. unaccusative) on respective lexical retrieval were assessed at the single-item level via logistic linear mixed models. In order to obviate high inter-individual variability and ceiling effect in performance, between-groups comparisons were implemented via overdispersed generalized linear models by considering the number of errors on each task as the outcome variable (Negative Binomial distribution). Results Global cognitive screening did not classify any of the participants as cognitively impaired. Mean accuracy on every language measures was ≥ 80% in both groups. Both groups performed significantly worse in V than in N naming (p<.001). MND patients performed significantly worse than HC participants in V naming (p=.02) and both object- (p<.001) and action- (p=.015) semantic tasks. Both groups performed comparably on remaining linguistic measures. The EFI did not discriminate MND patients from HC participants but was mostly related to V naming in MND patients (r=.67; p<.001).Single-item-level analyses revealed a significant Group*Grammatical class*Actionality three-way interaction effect (p=.002) showing that MND patients had significantly more difficulty than HC participants in retrieving low- rather than high-actionality Vs. A single-subject-level display of the aforementioned interaction further suggested that the lower the overall naming performance of MND patients, the more both the N-V discrepancy widened and the facilitating effect of actionality on V retrieval was pronounced. With regards to the ASC effect, single-item-level analyses revealed that MND patients found it significantly more difficult to retrieve transitive Vs (p=.004) when compared to HC participants; the two groups were nonetheless equally sensitive to both unaccusative and unergative aurgument structures. Discussion Our results are inconsistent with ECT-framed explanations for N-V dissociation in MND patients. First, patients showed semantic deficits that were not limited to the action domain. Second, actionality affected patients’ V retrieval in sharp contrast to ECT predictions, according to which MND patients were supposed to find it more difficult to process high-motor-content Vs because of their action knowledge impairment. The aforesaid effect of actionality did selectively pertain to patients’ V retrieval, i.e. did not affect patients’ N retrieval or naming performances of HC participants. Hence, it can be speculated that actionality resembled a concreteness effect in affecting V retrieval. N-V dissociation in MND patients is therefore likely to reflect a magnification of a differential processing demand for Vs vs. Ns that is intrinsic to the neurocognitive system. Indeed, N-V discrepancy was also found in HC participants. Consistently, EF contribution was highly relevant to patients’ V processing, and the lower patients’ cognitive efficiency was, the more evident both the N-V discrepancy and the facilitating effect of actionality on V retrieval. Nonetheless, patients’ V deficit was not widespread, as they appeared to be selectively sensitive to V ASC (transitive V deficit). This might imply prerolandic involvement in V post-semantic processing possibly at the lemma level. These findings discourage ECT-framed explanations for language deficit in motor disorders and suggest that prerolandic regions role in language processing should be reconsidered. York C et al. J Neurol. (2014) 261:1073-79. De Zubicaray G et al. J Cogn Neurosci. (2013) 25:1957-74. Branscheidt M et al. J Neurophysiol. (2017) 119:621-30. Papeo L et al. Cortex. (2014) 64:136-47. Kambanaros M et al. Aphasiology (2017) 31:49-66.