Sentence construction in health and neurological disease: an individualized approach using magnetoencephalography and contextualized models of meaning
Language is capable of conveying a seemingly unlimited amount of ideas. The possibility to flexibly combine words into sentences underlies this power. Using recent advances in linguistics & neuroimaging, our goal is to elucidate the neural computations underpinning sentence construction in healthy controls and patients with aphasia.The brain can extract the meaning of a single word in 400 milliseconds. Which neural computations are needed to integrate this single word knowledge into an unfolding sentence? By combining probabilistic measures of sentence construction -derived from contextualized models of meaning- with neural signals measured at the millisecond resolution, we can characterize the dynamic neural activity patterns in single participants while sentences unfold. Sentence construction is often impaired in patients with aphasia. Contextualized models can determine which words spoken by patients are less likely to occur in the same context when healthy controls are speaking. We identify which pathological brain changes relate to impaired sentence construction. Lastly, contextualized models are perhaps capable of completing the sentences of patients. This would constitute a new therapeutic tool to restore communication. In summary, we outline an individualized approach capturing the real time construction of sentences in the brain and we use the resulting knowledge to reconstruct speech in patients with aphasia.