The cortical dynamics of speaking: present shortcomings and future avenues

International audienceA central question in the cognitive neuroscience of language production concerns the cortical activation time-course by which different word production components become available to the speaker. According to the dominant neurocognitive model of word production, this is achieved in a sequential fashion where each of the linguistic components involved in speaking has its specific time-course and dedicated processing centre in the brain [Indefrey, P., & Levelt, W. J. M. (2004). The spatial and temporal signatures of word production components. Cognition, 92(1-2), 101-144. doi:10.1016/j.cognition.2002.06.001]. Language production would initiate with lexico-semantic processes reflected in posterior brain regions and slowly move upwards through the representational formats until an articulatory programme is activated in anterior brain regions. In this opinion piece, we raise three critical issues aimed at questioning the strict feedforward and serial conceptualisation of the spatiotemporal dynamics engendering word production. Concretely, we address the following three topics: (1) Are the data supporting the model or the model supporting the data? We conclude that, despite the evidence supporting the notion of some local sequentiality underpinning speech preparation in the brain, there is insufficient explicit spatiotemporal evidence to support the notion that such local sequenciality by itself is enough to explain how the brain computes language; (2) What is the role of top-down processing on the spatiotemporal dynamics? We argue that if top-down processes such as goal-directed behaviour and attention can proactively modulate the linguistic system, the spatiotemporal correlates of word production components should likewise be regarded as context-depend and speaker-adaptive; (3) Can we map psycholinguistic stages onto the brain in a one-to-one fashion? We advocate to start considering, above and beyond the feedforward anatomical connectedness that underpin traditional sequential hierarchical models, other types of neural communication such as neural coherence, feedback projections and reverberatory brain activity in order to mechanistically explain how basic brain properties sustain language production