Can hierarchical models display parallel cortical dynamics? A non-hierarchical alternative of brain language theory.

Résumé En

Muding, Dubarry and Alario (henceforth MDA) provide in their target article (2015) a much-needed synthesis of MEG evidence related to language production. The outcome of this work is both important and surprising: Important because it allows for the first time to have an integrative view on the available data concerning the spatiotemporal dynamics of language production processes; and surprising because it highlights a sequence of brain activities which considerably diverge from the traditional way many MEG evidence in the field is reported. While most MEG studies, especially those involving picture naming (e.g., Salmelin et al., 1994; Levelt et al., 1998; Soros et al., 2003; Hulten et al., 2009; Liljestrom et al., 2009; Laakson et al., 2012), typically are cited to support a sequential, serial-like processing structure between different word production components (e.g., Indefrey & Levelt, 2004; Indefrey, 2011), the current review demonstrates that the latter is not always the optimal interpretation of the data. Instead, the data seems to support cascading and parallel dynamics underpinning speech preparation in the brain. Although I fully agree with the latter important demonstration, I do wish to comment on certain theoretical aspects of the article. In particular, I question whether it would not be more revealing to address these MEG data from the perspective of hierarchical versus non-hierarchical model architectures, rather than the serial versus cascaded/parallel processing dynamics the authors focus on.