Presentation
Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Poster Slams
‘Language regions’ are artifacts of averaging
Poster E65 in Poster Session E, Saturday, October 8, 3:15 - 5:00 pm EDT, Millennium Hall
Bangjie Wang1,2, Sarah Aliko1, Jeremy I Skipper1; 1University College London, London, UK, 2The University of Texas at Dallas, Richardson, Texas, USA
Neo-localizationist (e.g., ‘dual-stream’) models of the neurobiology of language suggest that a small number of anatomically fixed brain regions are responsible for language functioning. This observation derives from centuries of analyzing brain lesions causing aphasia and is supported by decades of neuroimaging studies. The latter rely on thresholded measures of central tendency applied to activity patterns resulting from heterogeneous stimuli and participants. We hypothesize that these methods obscure the whole brain distribution of regions supporting language processing. Specifically, we suggest that ‘language regions’ are input regions and connectivity hubs. The latter primarily coordinate other peripheral regions whose activity is highly variable, making them likely to be spatially averaged out when thresholding is applied. We conducted three analyses to provide support for this model. First, we did a ‘meta-meta-analysis’ to demonstrate that ‘language’ activates the same regions in neuroimaging studies. This consisted of 85 individual meta-analyses of language representations (e.g., phonemes, words, sentences) and associated processes (e.g., speech, semantics, syntax), derived from the brainmap.org (N=28) and neurosynth.org (N=57) databases. Each meta-analysis was thresholded at p < .01 FDR corrected for multiple comparisons using a minimum cluster size of 50 voxels and combined by count. Second, we developed a meta-analytic network connectivity hub metric. Specifically, degree centrality was calculated via 165,953 voxel-wise co-activation meta-analyses across 14,371 studies, thresholded at p < .01 FDR corrected and combined by count. The resulting map was converted to z-scores to identify hubs, using 3.5 standard deviations from the mean as a cutoff (p < 0.0005). Third, the neuroquery.org database was used to conduct meta-analyses of ‘verbs’ (N=662) and ‘nouns’ (N=889) to demonstrate that language is distributed across the whole when specific linguistic categories are not indiscriminately averaged over. The 45 meta-analyses that survived thresholding all roughly activated the same regions regardless of language representation or associated linguistic process, including the bilateral superior temporal gyrus (STG), posterior middle temporal gyrus (pMTG), and posterior inferior frontal gyrus (pIFG). There was surprising overlap between these same regions and those constituting highly central regions or hubs. Finally, verbs and nouns activated a whole brain distribution of activity, with verbs producing greater activity in pre- and primary motor, somatosensory, and visual motion regions whereas nouns produced greater visual and fusiform activity (among other regions; using z-scores greater than 3.72 or p < .0001). These results suggest that ‘language regions’ and ‘the language network’ are an artifact resulting from the indiscriminate averaging of heterogeneous word categories are linguistic processes. This is not to say these regions do not perform functions. Indeed, given that we have demonstrated that they are primarily highly central connectivity hubs (explaining why they are left over after averaging), one such role would be to coordinate the whole-brain distributions of regions necessary for processing the complexities of language in the real world. That ‘language regions’ are connectivity hubs also explains why they likely result in gross disorders like aphasia that also affect ‘nonverbal’ functioning.
Topic Areas: History of the Neurobiology of Language, Speech Perception