Background
Auditory verbal hallucinations (AVH) have been proposed to result from altered connectivity between frontal speech production regions and temporal speech perception regions. Whilst the dorsal language pathway, serviced by the arcuate fasciculus, has been extensively studied in relation to AVH, the ventral language pathway, serviced by the inferior occipito-frontal fasciculus (IOFF) has been rarely studied in relation to AVH.

Methods
This study used Diffusion Tensor Imaging to investigate whether structural changes in anatomically defined subregions of the IOFF were associated with AVH in patients with schizophrenia. Diffusion tensor imaging scans and clinical data were analyzed for 113 schizophrenia patients, of whom 39 had lifetime experience of AVH (18 had current AVH, 21 had remitted AVH), 74 had no lifetime experience of AVH, and 40 healthy controls.



Results
Schizophrenia patients with a lifetime experience of AVH exhibited reduced fractional anisotropy (FA) in the fronto-temporal fibers of the left IOFF compared to both healthy controls and schizophrenia patients without AVH. In contrast, structural abnormalities in the temporal and occipital regions of the IOFF were observed bilaterally in both patient groups, relative to the healthy controls.

Conclusions
These results suggest that while changes in the structural integrity of the bilateral IOFF are associated with schizophrenia per se, integrity reductions in the fronto-temporal fibers of the left IOFF may be specifically associated with AVH.

Recent attempts to ground models of human language processing in neurobiology have focused on the dorsal and ventral processing streams (Hickok and Poeppel 2004, Friederici 2009, Bornkessel-Schlesewsky and Schlesewsky 2013) based upon neuroanatomical fibre tracts. The models differ, however, in their proposed division of labour between the streams: Friederici (2012) suggests that the dorsal stream processes 'complex' syntax; Bornkessel-Schlesewsky et al (2015) suggest that the dorsal stream processes sequence-dependent, i.e. non-commutative, information, while the ventral stream processes commutative operations.

Here, we examined simple syntactic violations of two forms: sequence ("the red big balloon") and cooccurrence ("the heavy red balloon") in an EEG experiment, comparing the conditions to each other and a control condition ("the big red balloon"). In contrast to Friederici (2012), we predict that these violations are processed in different streams and will thus elicit different responses, both in amplitude and in topography.

Data from 18 subjects (11 female, mean age 24.3±2.5) were examined using linear mixed-effects models (Pinheiro & Bates 2000); subsequent pairwise comparisons were performed using least-square means. Measured at the second adjective, we observed a left-lateralised negativity for both violation conditions, with the cooccurrence condition eliciting a significantly stronger negativity than the sequence condition as well as a more parietal distribution. At the noun, we observed right-lateralised negativity for both conditions, with a more anterior distribution for the the sequence condition.

The division of labour between processing streams along commutativity leads to distinct neural generators for sequence and coocurrence violations and thus distinct ERP signatures.

Language, gesture, and handedness are all indicative of an action-related network that is represented predominantly in the left cerebral hemisphere. To explore the relations among them, we collected fMRI images in a large sample of left- and right-handers while they performed language tasks and watched action sequences. Regions of interest included the frontal and parietal areas previously identified as comprising an action-observation network, and the frontal and temporal areas comprising the primary areas for language production and comprehension. All of the language areas and most of the action-observation areas showed an overall left-hemispheric bias, despite the participation of equal numbers of left- and right-handers. A factor analysis of the laterality indices derived from the different areas during the tasks indicated three independent networks, one associated with language, one associated with handedness, and one representing action observation independent of handedness. These results suggest an evolutionary scenario in which the primate mirror neuron system became increasingly lateralized, and later fissioned onto subsystems mediating language, hand preference, and the execution and observation of manual action.

Eye movements provide an important window into the relationship between brain and behaviour, as they are more closely coupled to neural predictive processes than overt responses (e.g. Friston et al., 2012). This relation has been studied extensively in reading, revealing a correlation between longer fixation times and the amplitude of the N400 ERP (Sereno & Rayner, 2003). Predictions pertaining to perceptual stimulus features (e.g. word form) elicit longer first fixation times, but little is known about the eye-movement correlates of more abstract features (e.g. the animacy of event participants). Here, we tested the prediction – derived from a previous ERP study (Kretzschmar et al., 2015) – that N400 effects for animacy-based prediction mismatches are reflected in later eye-movement measures. Sixteen participants (mean age: 24±1; 13 females) read sentences of the form "A tourist/trawler swam/drifted across the lake ...". Grammatical subjects were either animate ("tourist") or inanimate ("trawler") and verbs were biased towards animate ("swim") or inanimate motion ("drift"). Participants answered comprehension questions on approximately 37% of trials; data were analysed using repeated-measures ANOVAs. As hypothesised, late but not early eye-tracking measures at the position of the verb ("swam/drifted") revealed a cross-over interaction of subject- and verb animacy. Our results demonstrate a decoupling of ERP responses and eye movements in predictive processing during reading. This supports the view that the N400 reflects predictive model updates in temporal receptive windows approximately corresponding to the timescale of words, while eye movements in reading correspond to words themselves and the features associated with them.