Final Report Summary - NEUROLEX (Neurocognitive systems for morpho-lexical analysis: The cross-linguistic foundations for language comprehension)
Language comprehension is a fundamentally dynamic process, where the incoming speech stream is continuously segmented into a sequence of words and morphemes. This speech information interfaces with two markedly different neurocognitive processing systems – a left hemisphere system that is critical for key linguistic processes of grammatical analysis, and a more distributed bi-hemispheric system that supports the broader interpretation of the meaning of the utterance. This view of the neurocognitive language system has emerged from converging interdisciplinary research in psycholinguistics, neuropsychology and cognitive neuroscience, primarily based on English. The NEUROLEX project has taken our understanding of these systems to a new level of generality by conducting parallel investigations in several contrasting language systems (including English, Italian, Polish and Russian, Arabic, , and Mandarin Chinese), while achieving a new level of specificity in terms of the spatio-temporal patterns of different language processing procedures across the brain.
At the start of this project, analysis methods for magneto- and electro-encephalography (MEG and EEG) were very limited in their ability to characterise the dynamic spatiotemporal changes in the brain that underlie language comprehension. To address this crucial limitation, we developed a novel multivariate data modelling and analysis method based on Representational Similarity Analysis (RSA). This method has major advantages compared to conventional approaches since it can capture fine-grained dynamic neural computations in the brain, revealing the representational content of the neural response rather than overall regional involvements of brain areas. Coupled with a novel spatiotemporal searchlight algorithm,, we can now conduct these analyses on a large scale, whole-brain basis. This makes it possible to closely relate dynamic patterns at the neuronal level to patterns derived from higher-level cognitive theories.
The cross-linguistic NEUROLEX research programme is closely integrated with these novel methods, generating unique data about the spatio-temporal distribution of different types of linguistic processing, millisecond by millisecond as the speech is heard. We focused in particular on the analysis of grammatical morphemes (such as the English past tense [-ed]), and were able to track the exact timing of the processes triggered by such a morpheme in the core left hemisphere language areas. Strikingly, we were able to show that the same types of dynamic processes, located in left fronto-temporal regions, could be seen in a very different language like Mandarin, pointing to the potential universality of key human left hemisphere specialisations for combinatorial processing.
We expanded this approach cross-linguistically to contrast specifically linguistic forms of complexity (derivational and inflectional processes of word-formation) with non-linguistic sources of complexity reflecting processing competition between different lexical and phrasal interpretations. Our research in English shows that these forms of complexity engage different processing systems across the two hemispheres, with parallel results for Polish, Russian, and Italian - languages which are morphologically much richer, but share with English similar word-formation mechanisms. Arabic, in contrast, offers a radically different lexical environment, with fundamentally different mechanisms of word formation and where key grammatical morphemes serve multiple linguistic functions. Despite these major differences, the neuroimaging data shows a very similar underlying distribution of language processing function to the left hemisphere system on the one hand and the bilateral system on the other. Together with the results for Mandarin, these outcomes provide strong support for the dual neurobiological systems approach that the Neurolex programme set out to investigate.
At the start of this project, analysis methods for magneto- and electro-encephalography (MEG and EEG) were very limited in their ability to characterise the dynamic spatiotemporal changes in the brain that underlie language comprehension. To address this crucial limitation, we developed a novel multivariate data modelling and analysis method based on Representational Similarity Analysis (RSA). This method has major advantages compared to conventional approaches since it can capture fine-grained dynamic neural computations in the brain, revealing the representational content of the neural response rather than overall regional involvements of brain areas. Coupled with a novel spatiotemporal searchlight algorithm,, we can now conduct these analyses on a large scale, whole-brain basis. This makes it possible to closely relate dynamic patterns at the neuronal level to patterns derived from higher-level cognitive theories.
The cross-linguistic NEUROLEX research programme is closely integrated with these novel methods, generating unique data about the spatio-temporal distribution of different types of linguistic processing, millisecond by millisecond as the speech is heard. We focused in particular on the analysis of grammatical morphemes (such as the English past tense [-ed]), and were able to track the exact timing of the processes triggered by such a morpheme in the core left hemisphere language areas. Strikingly, we were able to show that the same types of dynamic processes, located in left fronto-temporal regions, could be seen in a very different language like Mandarin, pointing to the potential universality of key human left hemisphere specialisations for combinatorial processing.
We expanded this approach cross-linguistically to contrast specifically linguistic forms of complexity (derivational and inflectional processes of word-formation) with non-linguistic sources of complexity reflecting processing competition between different lexical and phrasal interpretations. Our research in English shows that these forms of complexity engage different processing systems across the two hemispheres, with parallel results for Polish, Russian, and Italian - languages which are morphologically much richer, but share with English similar word-formation mechanisms. Arabic, in contrast, offers a radically different lexical environment, with fundamentally different mechanisms of word formation and where key grammatical morphemes serve multiple linguistic functions. Despite these major differences, the neuroimaging data shows a very similar underlying distribution of language processing function to the left hemisphere system on the one hand and the bilateral system on the other. Together with the results for Mandarin, these outcomes provide strong support for the dual neurobiological systems approach that the Neurolex programme set out to investigate.