Frank Guenther

Boston University
United States

Disorders at the intersection of language and motor control

At some stage of the language production process, the phonological content of an intended utterance must be translated into motor programs that generate rapid movements of the speech articulators that have been learned to efficiently articulate the linguistic message. In this talk I will discuss a detailed neurocomputational model of this interface that has been developed and experimentally tested in our laboratory over the past decade. Model simulations can be quantitatively compared to experimentally measured neural and behavioral variables, and model predictions have been tested in neuroanatomical, neurophysiological, and behavioral experiments involving speech in both neurotypical and disordered populations.

The model described in this talk is the Gradient Order DIVA (GODIVA) model, an extension of the long-established DIVA model of speech motor control. GODIVA quantitatively accounts for the neural processes underlying the temporary buffering and sequencing of the phonological material in an utterance prior to and during speech articulation. The model’s components are specified both mathematically (with equations governing neural activity and synaptic strengths) and neuroanatomically (in stereotactic coordinates). The model’s components include a phonological content buffer localized to left inferior frontal sulcus (IFS), a metrical structure buffer localized to pre-supplementary motor area (preSMA), a motor program map in left ventral premotor cortex (vPMC), a speech initiation map in SMA, and associated subcortical regions of basal ganglia-thalamo-cortical planning and motor loops whose functions in the model are inspired by prior theoretical proposals by Mink and others. Simulations of the model verify its ability to account for a range of behavioral and neuroimaging findings concerning speech buffering, sequencing, and learning of novel speech sequences.

The GODIVA model also provides a framework for understanding dysfunction in speech buffering and sequencing circuits. Apraxia of speech is accounted for by damage to either speech motor programs in left ventral premotor cortex and/or projections to this region from left pIFS. Phonemic paraphasias are accounted for by damage to left pIFS or its afferent projections. In keeping with prior theoretical work by Alm and others, stuttering is proposed to arise from impaired performance in basal ganglia-thalamo-cortical loops, resulting in impaired initiation of speech motor programs. Computer simulations of impaired versions of the GODIVA model have verified key aspects of these accounts, and ongoing neuroimaging and behavioral experimental studies are further testing the model’s explanations of these disorders.