[show abstract][hide abstract] ABSTRACT: In sentence processing, storage and ordering of the verb and its arguments (subject and object) are core tasks. Their cortical representation is a matter of ongoing debate, and it is unclear whether prefrontal activations in neuroimaging studies on sentence processing reflect the storage of arguments or their ordering. Moreover, it is unclear how storage during sentence processing relates to the neuroanatomy of storage outside the sentence processing domain. To tackle these questions, we crossed the factor "ordering" (subject-first vs. object-first German sentences) with the factor "storage" (one vs. four phrases intervene between the critical argument and the verb) in an auditory fMRI study. Ordering focally activated the left pars opercularis in Broca's area, while storage activated deep left temporo-parietal (TP) regions. Notably, left TP activation correlated with listener's digit span, while Broca's area activation did not. Furthermore, fractional anisotropy of listeners' left arcuate fasciculus/superior longitudinal fasciculus (AF/SLF) is shown to covary with the functional effect of increased storage demands at sites along the tract. Functionally, the results suggest that storage during sentence processing relies on TP regions, likely shared between sentence processing and other working memory-related tasks, while Broca's area appears as a distinct neural correlate of ordering. We conclude that the abstract notion of sentence processing can be captured by the interplay of concrete cognitive concepts such as ordering and storage.
[show abstract][hide abstract] ABSTRACT: Both functional magnetic resonance imaging (fMRI) and event-related brain potential (ERP) studies have shown that verbal working memory plays an important role during sentence processing. There is growing evidence from outside of sentence processing that human alpha oscillations (7-13 Hz) play a critical role in working memory. This study aims to link this to the sentence processing domain. Time-frequency analyses and source localization were performed on electroencephalography (EEG) data that were recorded during the processing of auditorily presented sentences involving either a short or a long distance between an argument (subject or object) and the respective sentence-final verb. We reasoned that oscillatory activity in the alpha band should increase during sentences with longer argument-verb distances, since decreased temporal proximity should result in increased memory demands. When verbal working memory-intensive long-dependency sentences were compared to short-dependency sentences, a sustained oscillatory enhancement at 10 Hz was found during storage prior to the sentence-final verb, turning into a transient power increase in the beta band (13-20 Hz) at the sentence-final verb. The sources of the alpha oscillations were localized to bilaterally occipital and left parietal cortices. Only the source activity in the left parietal cortex was negatively correlated with verbal working memory abilities. These findings indicate that the parsimonious role of alpha oscillations in domain-general working memory can be extended to language, that is, sentence processing. We suggest that the function of left parietal cortex underlying verbal working memory storage during sentence processing is to inhibit the premature release of verbal information that will subsequently be integrated.
[show abstract][hide abstract] ABSTRACT: In sentence processing, it is still unclear how the neural language network successfully establishes argument-verb dependencies in its spatiotemporal neuronal dynamics. Previous work has suggested that the establishment of subject-verb and object-verb dependencies requires argument retrieval from working memory, and that dependency establishment in object-first sentences additionally necessitates argument reordering. We examine the spatiotemporal neuronal dynamics of the brain regions that subserve these sub-processes by crossing an argument reordering factor (i.e., subject-first versus object-first sentences) with an argument retrieval factor (i.e., short versus long argument-verb dependencies) in German. Using functional magnetic resonance imaging (fMRI), we found that reordering demands focally activate the left pars opercularis (Broca's area), while storage and retrieval demands activated left temporo-parietal (TP) regions. In addition, when analyzing the time course of fMRI-informed equivalent current dipole sources in the EEG at the subcategorizing verb, we found that activity in the TP-region occurs relatively early (40-180 ms), followed by activity in Broca's area (300-500 ms). These findings were matched by topographical correlation analyses of fMRI activations in EEG sensor space, showing that, in the scalp potential, TP-region activity surfaces as an early positivity and IFG activity as a later positivity in the scalp potential. These results provide fine-grained evidence for spatiotemporally separable sub-processes of argument retrieval and reordering in sentence processing.
[show abstract][hide abstract] ABSTRACT: Under real-life adverse listening conditions, the interdependence of the brain's analysis of language structure (syntax) and its analysis of the acoustic signal is unclear. In two fMRI experiments, we first tested the functional neural organization when listening to increasingly complex syntax in fMRI. We then tested parametric combinations of syntactic complexity (argument scrambling in three degrees) with speech signal degradation (noise-band vocoding in three different numbers of bands), to shed light on the mutual dependency of sound and syntax analysis along the neural processing pathways. The left anterior and the posterior superior temporal sulcus (STS) as well as the left inferior frontal cortex (IFG) were linearly more activated as syntactic complexity increased (Experiment 1). In Experiment 2, when syntactic complexity was combined with improving signal quality, this pattern was replicated. However, when syntactic complexity was additive to degrading signal quality, the syntactic complexity effect in the IFG shifted dorsally and medially, and the activation effect in the left posterior STS shifted from posterior toward more middle sections of the sulcus. A distribution analysis of supra- as well as subthreshold data was indicative of this pattern of shifts in the anterior and posterior STS and within the IFG. Results suggest a signal quality gradient within the fronto-temporal language network. More signal-bound processing areas, lower in the processing hierarchy, become relatively more recruited for the analysis of complex language input under more challenging acoustic conditions ("upstream delegation"). This finding provides evidence for dynamic resource assignments along the neural pathways in auditory language comprehension.