Big Question 2
(last update 2020-05-12)
What are the characteristics and consequences of internal brain organization for language?
The human brain provides a neurobiological infrastructure that allows us to acquire and process language, and that co-determines the characteristics of spoken (and sign) and written language. The internal organization of the brain and its cognitive architecture both determine and constrain the space of possibilities for human language. This internal organization can be called the Kantian brain for language. It has resulted in a language-readiness of the human brain that is found nowhere else in the animal kingdom. The big question is to characterize the Kantian brain for language.
Currently BQ2 is in the process of building links between the various sub-projects. Each sub-project has had the opportunity to present their most recent work/ideas/questions of interest, and BQ2 is now in a phase of bridging the sub-projects to try to define new research questions based on collaborations between sub-themes. To foster such collaborations, meetings are planned where pairings of sub-projects will present ideas that culminate from joint brainstorm sessions about potential links between one another’s work and expertise. In the long run the hope is that such combinations of expertise and perspectives will lead to innovative and cutting-edge projects that address the overarching goal of how the human brain supports language processing.
Highlight 1: White-matter connectivity of the left posterior middle temporal gyrus (L_pMTG) – differentiation from chimpanzees to humans
Team members: Sierpowska, Bryant, Janssen, Mangnus, Römkens, Roelofs, Kessels, Freches, Mars, and Piai
The main objective of this study was to compare white matter connectivity of the left posterior middle temporal gyrus (hereafter: L_pMTG) structural bottleneck between human and chimpanzee. Additionally, the connectivity of the anterior temporal lobe (L_ATL) bottleneck was explored, as well as the connectivity of both bottlenecks within the right hemisphere.
Regions of interest (ROI) for both bottlenecks were extracted based on the AAL neuroanatomical atlas. Subsequently, FSL Probtrackx probabilistic tractography pipeline was applied to diffusion weighted MRI (DWI) data in order to define the extent of white matter converging towards the ROI. Finally, these maps were compared to the distribution of canonical language-related white matter tracts obtained using the same methodology.
Visual inspection of the results revealed an extensive ventral system of white matter pathways (including inferior frontal-occipital fasciculus - IFOF) originating from the L_ATL seed in both humans and chimpanzees. Importantly, the maps did not substantially differ between the two species. In humans, the probabilistic tracking from L_pMTG showed that the ventral white-matter system extends to both the right hemisphere via the tapetum and to the dorsal pathways for language via the connection between the posterior superior temporal sulcus and the inferior parietal lobe. In chimpanzees, this circuitry was similar with regard to the interhemispheric connections, but connectivity to the dorsal stream was less robust than in humans.
Quantification of these (dis)similarities indicated that the L_pMTG difference in connectivity may be mainly explained by how this circuitry connects towards the canonical, dorsal pathways for language (up to R=.83, p<0.001 for the left posterior branch of the arcuate fasciculus), whereas this pattern is explained by the connectivity towards the ventral pathways for language for the L_ATL maps (up to R=.9, p<0.001 for the left inferior longitudinal fasciculus). Quantification of these differences for the right hemisphere is currently ongoing.
Highlight 2: White-matter bottleneck in small vessel disease: A lesion- symptom mapping study of language-executive functions
Team members: Camerino, Sierpowska, and Piai
Small vessel disease (SVD) is characterized by presence of white matter lesions (WML). Executive functioning and processing speed are the most effected cognitive domains in SVD and are correlated with total WML volume. However, WML location might better explain cognitive symptoms of SVD than total WML volume. This project investigated the relation between WML locations and executive-language functions in SVD.
This study included a cohort of 442 SVD patients without dementia, with varying burden of WML. The Stroop (word reading, color naming, and color-word naming) and the verbal fluency tests were used as measures of language production with varying degrees of executive demands. The digit symbol modality (DSMT) was used as a control task as it does not require verbal abilities. A voxel-based lesion symptom mapping (VLSM) approach was used.
A relationship was found between WML and language- executive functioning in a core fronto-striatal network in SVD, independent of lesion size (figure below). This circuitry formed by the caudate nuclei, forceps minor and thalamic radiations, seems to underlie language-executive functioning.
Figure 2. VLSM Results. The statistically significant cluster of voxels associated with worse performance in each of the neuropsychological tests are shown. These maps show colorized depictions of t-test results evaluating performance of the patients on a voxel-by-voxel basis. High t-scores (red) indicate that lesions to these voxels have a stronger relationship with behavior. Dark purple voxels indicate regions where the presence of a lesion has a weaker relationship with the behavioral measure. The max t value is used for the color scale. Only voxels that were significant at P = 0.05 (controlling for the expected proportion of false positives) are shown. All results are corrected for age, sex, education and lesion volume.
This project is the result of the collaboration between partners of the BQ2 with different expertise like neuroimaging and clinical neuropsychology. The results of this study provide valuable information about the effect of vascular lesions on language function.
Highlight 3: Mapping the brain’s feedforward and feedback architecture for language with neural oscillations
Team members: Lewis, Hagoort, and de Lange
This project is designed to develop linguistic paradigms in which the relative contributions of feedforward and feedback information streams can be manipulated in order to study the associated neural architecture. The study utilized MEG to probe high and low frequency neural oscillations as indices of feedforward and feedback information, respectively. This initial phase of the experiment employs a lexical decision task with Dutch words and non-words presented visually under varying levels of visual degradedness.
Visual degradedness is achieved through low-pass spatial filtering, somewhat akin to noise-vocoding with speech stimuli. Based on previous literature investigating the visual system, we treated alpha/beta (8-19 Hz) power as an index of feedback and mid gamma (56-76 Hz) power as an index of feedforward signalling in the MEG. In a visual lexical decision task the primary source of feedforward information would be occipital regions responsible for visual processing of the stimulus, while the primary source of feedback information would be left temporal regions responsible for lexical-semantic processing. Preliminary results (N=10 participants) suggest precisely such a dissociation (Figure 3), where in the low degradedness conditions (when participants can read the words) there appears to be a larger alpha/beta desynchronization for words compared to non-words at temporal sensors but not at occipital sensors.
Figure 3. Mean power values in the Alpha/Beta (8-19 Hz) and Mid Gamma Power (56-76 Hz) range for different levels of visual degradedness.Dots represent mean power values for each participant. The data is collapsed across the two highest and two lowest levels of degradedness.
At occipital sensors gamma synchronization is higher for non-words than for words, suggesting greater prediction error for non-words in regions responsible for visual processing. This difference does not appear (or is greatly reduced) over left temporal sensors. These preliminary findings suggest that for word reading feedforward and feedback signalling may indeed proceed via high and low frequency neural oscillations. Future phases of the project will investigate how these signals may be affected when words (both visually degraded and not) are inserted into sentence contexts with differing degrees of semantic constraint (i.e., differing availability of feedback information).
This project pursues a hot topic in language neuroscience and directly addresses a core question of BQ2: the role of low and high frequency neural oscillations in feedback and feedforward signalling in the brain to support language processing. It also addresses the important question of the extent to which links observed in the visual system between high and low frequency neural oscillations on the one hand, and feedforward and feedback signalling on the other, may generalize to higher cognitive functions supported by alternative brain systems. It brings experts from psycholinguistics and the cognitive neuroscience of language into direct contact with experts from prediction and attention in visual processing as well as experts on cutting edge methods for the analysis of neural oscillations.
Synergy with other Big Questions
Once there is clarity on future directions for collaborative projects within BQ2, the plan is to try to find points of contact with other related BQs in the consortium and potentially foster cross-BQ collaborations on particular aspects of projects where such links present themselves. For instance, neuroimaging data to be collected can be used to evaluate models developed by BQ1. Some of the analysis methods from BQ2 can be further utilized in BQ4 and BQ5.
More concretely, the BQ2 team members Piai (Tenure Track, see page 52) and Janssen (PhD Candidate) have developed a neuropsychological measure in collaboration with De Swart (CLS): the SynTest. The SynTest is a sentence-to-picture matching task that assesses sentence comprehension with increasing grammatical complexity. It is specifically developed to aid in the differentiation between logopenic and non-fluent primary progressive aphasia (PPA) patients and data collection for validation has now started in multiple Dutch hospitals. A computerized version of this test is being implemented in the BQ4 test battery by Menks (Postdoc).
Lewis has joined the BQ5 team and will play a role in projects within that BQ involving MEG. The ideas within the BQ5 about successor representations align well with his investigations into predictive processing and top-down vs bottom-up information flow within the BQ2 and could be mutually informative. Furthermore, ideas within the BQ5 about the role of cognitive maps for language processing may also be of interest to many in the BQ2. Lewis has been involved as a member of search committees for the hiring of 2 postdocs within the BQ5.
Guilherme Blazquez Freches
Floris de Lange