Beyond Broca: neural architecture and evolution of a dual motor speech coordination system

Abstract

Classical neural architecture models of speech production propose a single system centred on Broca's area coordinating all the vocal articulators from lips to larynx. Modern evidence has challenged both the idea that Broca's area is involved in motor speech coordination and that there is only one coordination network. Drawing on a wide range of evidence, here we propose a dual speech coordination model in which laryngeal control of pitch-related aspects of prosody and song are coordinated by a hierarchically organized dorsolateral system while supralaryngeal articulation at the phonetic/syllabic level is coordinated by a more ventral system posterior to Broca's area. We argue further that these two speech production subsystems have distinguishable evolutionary histories and discuss the implications for models of language evolution.

Keywords: auditory; evolution; motor speech; precentral gyrus; sensorimotor control; speech production.

Figure 1

Brain maps from a range of studies implicating a dorsal precentral gyrus region in speech processing (red arrows). (A) PET study showing activation associated with repeating words minus saying ‘crime’ after hearing a reversed word. (B) Functional MRI study showing activation associated with both auditory (listening to nonsense sentences) and motor (silently rehearsing nonsense sentences) speech. (C) Functional MRI study showing activation associated with listening to syllables (heat map) and producing the same syllables (outline). (D) Myelin (left) and functional MRI task activation (right) map from the Human Connectome Project database showing area 55b. (E) Schematic diagram of the dual stream model of speech processing showing two frontal regions associated motor speech-related processes.

Figure 2

Neurosurgical and stroke-based evidence for dorsal and ventral speech coordination systems. (A) Map of locations eliciting speech arrest (see text) during direct cortical stimulation. Colours and numbers represent the results of a cluster analysis. The yellow (2) and green (1) clusters have density distributions centred on the dorsal and ventral precentral gyrus, respectively. From Lu et al. (B) Map of white matter connections predictive of speech repetition deficits in post-stroke aphasia implicating both dorsal and ventral connectivity. From Baboyan et al.

Figure 3

Surface-rendered map of functional and anatomic regions. Inflated brain map showing the dPCSA and vPCSA regions of interest in relation to the dorsal laryngeal motor cortex (dLMC), ventral laryngeal motor cortex (vLMC), area 55b, and maximal probability maps for cytoarchitectonic areas (colour-shaded areas) from the Jülich-Brain atlas, supplemented by a thresholded probability map of premotor area 6. dPCSA and vPCSA regions of interest are estimated from functional imaging activation peaks on the precentral gyrus in Rong et al., which align with coordinates from the speech arrest stimulation study (see Supplementary Fig. 1). The dLMC and vLMC were defined by performing an activation likelihood estimate (ALE) meta-analysis of funtional MRI studies that explicitly refer to laryngeal motor cortex in their results (see Supplementary Table 1). ALE values were thresholded at P < 0.00001 (cluster family-wise error correction at P < 0.00001 using 4000 permutations) and the resulting clusters were projected onto the fsaverage surface along with the vPCSA and dPCSA regions of interest. Area 55b was extracted from the Human Connectome Project’s (HCP) Multimodal Parcellation 1.0 atlas. Maximal probability maps for motor areas 4a and 4p, somatosensory areas 3a, 3b, 1 and 2, opercular areas Op6 and OP4, and area 44 were all extracted from the Jülich-Brain atlas (version 2.9). The probability map for premotor area 6, which is largely unmapped in the Jülich-Brain atlas, was projected onto the fsaverage surface, then thresholded to show vertices that were at least 40% likely to be area 6. See Supplementary material for more information.

Figure 4

Spectrotemporal receptive field maps. Distribution of mappable STRFs (brain image) and plots of the STRF maps associated with the two lateral frontal STRF regions (left, A and B), one in dorsal precentral gyrus and one on the pars triangularis. The dorsal precentral site exhibits significant tuning to the pitch-related region of STRF space (black oval). Coloured dots correspond to the coordinates of three separate fMRI study activations, two which map auditory–motor speech responses (the slightly more posterior green and blue spheres^,) and one that mapped laryngeal motor cortex with the coordinate on ‘gyral portion’ (see text). Figure from Venezia et al.

Figure 5

An automated meta-analysis of functional imaging studies on the term ‘syntactic’ (heat map) using neurosynth.org. Note the relation to area 55b (yellow outline). See Supplementary material for more information about methods and analysis.

Figure 6

Functional connectivity analyses for dorsal versus ventral precentral speech areas. (Left) Resting state functional connectivity map for the dPCSA (cool colours) and vPCSA (warm colours). Whole brain maps were generated for each region of interest (defined in Fig. 3) in each participant (n = 137) using multi-echo independent components regression. Voxelwise paired t-tests were then performed to determine brain regions where resting functional connectivity was significantly higher to either dPCSA or vPCSA in the group. Family-wise error was controlled at P < 0.01 using threshold free cluster enhancement (TFCE) with 5000 permutations. The resulting t-map was projected onto the fsaverage surface using registration fusion with Advanced Normalization Tools. (Right) Task fMRI-based, functional connectivity maps. Beta time series functional connectivity analyses (n = 34) were conducted using data from 400 trials of a receptive competing speech task (‘Competing’ condition). Data were projected onto a cortical surface model of the MNI152 template and surface-node-wise estimates of functional connectivity (least-squares-separate beta time series; Pearson correlation) were obtained using the average time series within dPCSA and vPCSA regions of interest defined as in Fig. 3 as seeds. A second-level contrast t-map (dPCSA connectivity versus vPCSA connectivity) was obtained and thresholded at FDR-corrected P < 0.01, then projected onto the fsaverage surface using registration fusion with Advanced Normalization Tools. See Supplementary material for more information about methods and analysis.

Figure 7

Schematic depiction of the dual speech coordination system model. Input labels (arrows) are intended to reflect a bias toward the named input type, not exclusivity, and are not intended to reflect anatomical locations of white matter pathways. For detailed arguments regarding the morphosyntatic pathway, see Matchin and Hickok. Anatomical localizations are approximate. Location of orofacial motor cortex ventral to dLMC is based on a meta-analysis from Guenther.