You are viewing the SNL 2018 Archive Website. For the latest information, see the Current Website.


Poster D3, Friday, August 17, 4:45 – 6:30 pm, Room 2000AB

Identifying the Neural-Computational Correlates of Cognitive Control During Language Processing - Combined Activation Likelihood Estimation and Functional Imaging Evidence of Language Production

Nicolas Bourguignon1, Vincent Gracco2;1Department of experimental psychology, Ghent University, Belgium, 2Haskins Laboratories, New Haven

INTRODUCTION The relationship between language and cognitive control – i.e. the coordination of actions and thoughts in accordance with internal goals – is a highly debated issue in cognitive neuroscience. This debate has far-reaching implications insofar as both faculties are taken to reflect similar facets of human adaptive behavior such as creativity, generativity and intentionality. In this respect, recent functional imaging (fMRI) research has shown a marked functional anatomic subdivision between a fronto-temporo-parietal network involved in language processing (i.e. a language network) and a fronto-parietal and cingulo-opercular “multiple-demand” network for cognitive control tasks in the brain. While such findings are assumed to reflect an initial functional-anatomic separation between the two faculties, the possibility remains that both networks may interact in principled ways in certain language processing contexts. In the present study, we propose and test a dual-stream model of the cognitive control of language processing based on a computationally derived information-theoretic notion of control as the selection of responses amongst competing alternatives. Specifically, we propose that response selection takes place within a working memory stream (WMS) that monitors information held in verbal working memory and a lexico-semantic stream (LSS) involved in top-down retrieval of representations from long-term lexico-semantic memory. We further suggest that WMS and LSS closely follow the functional anatomic separation between the multiple demand network and the language network, respectively. Unlike a strict language vs. cognitive control dichotomy, this architecture predicts that the “language” network (i.e. LSS) should be involved in the controlled selection of language-relevant information when this information derives from lexico-semantic storage systems. METHODS We began with an activation likelihood estimation (ALE) analysis of 111 fMRI studies of working memory-monitoring (N=69) vs. lexico-semantic selection (N=42) to assess the proposed functional anatomic subdivision between WMS and LSS and to show that they closely follow the functional anatomy of the multiple-demand network and the language network, respectively. We then test the involvement of these networks’ core regions of interest (ROI) in the controlled selection of lexico-semantic representations in an fMRI study of confrontation naming and verb generation, using normed stimulus-related entropy indices of competition as predictors of cognitive control demands. RESULTS AND DISCUSSION Results from the ALE analysis reveal a functional anatomic dissociation between WMS and LSS and confirm their areal overlap with the multiple-demand network and the language network respectively. Furthermore, our ROI analyses of the confrontation naming and verb generation data show a clear involvement of LSS in controlled selection of lexico-semantic representations in concert with the cingulo-opercular sub-component of the multiple-demand network. By contrast, the fronto-parietal sub-component of the multiple-demand network does not covary reliably with selection competition in either confrontation naming or verb generation. Not only do these results confirm an involvement of the “language” network (LSS) in controlled selection of lexico-semantic information, but they highlight the need to reevaluate the role of the multiple-demand network in basic operations of cognitive control. These results have interesting implications for future efforts at integrating basic assumptions of neurocognitive models of language and adaptive behavior.

Topic Area: Control, Selection, and Executive Processes

Back