Slide Slam N15
Differential selectivity of the left and right hemisphere language regions for non-linguistic processing
Hannah Small1, Benjamin Lipkin1, Josef Affourtit1, Alvincé Pongos1, Evelina Fedorenko1; 1Massachusetts Institute of Technology
The language network, which supports comprehension and production beyond perceptual input and output motor processes, is lateralized to the left hemisphere (LH) in most individuals. However, highly reliable—albeit lower in magnitude—responses to language are also observed in the homotopic right hemisphere (RH) areas (e.g., Mahowald & Fedorenko, 2016). The relative contributions of the LH and RH language network to language and other cognitive functions remain debated. The degree of a brain region/network’s functional specialization can critically inform its functions (e.g., Kanwisher, 2010). Chai et al. (2015) used dynamic network modeling to show that the RH language regions constitute the flexible ‘periphery’ of the language network: they co-activate with different regions at different times. In contrast, the LH language regions consistently co-activate with one another and thus constitute a stable ‘core’. Given the hypothesis advanced by Bassett et al. (2013)—whereby the inter-regional coupling flexibility of a given brain region over time is inversely related to its degree of functional specialization—RH language regions should be less functionally specialized for language compared to the LH language regions. Yet, studies that probed a relatively restricted range of non-linguistic functions did not observe reduced specialization for language in the RH language regions (e.g., Fedorenko et al., 2011; Deen et al., 2015). Here, we comprehensively evaluate the degree of functional specialization in the LH and RH language regions using data from 34 experiments (68 conditions) across a total of 761 participants. Each participant performed a language localizer task, which robustly identifies language-responsive cortex (Fedorenko et al., 2010; Braga et al., 2020), and different subsets performed diverse non-linguistic tasks (n across experiments ranged from 11 to 715; average: 44.7±20.4, median: 18). The non-linguistic tasks spanned a broad range of perceptual, cognitive, and motor functions, including visual and auditory processing of both social and non-social stimuli, hand and face motor control, music perception, numerical cognition, executive function tasks, categorization, visual event semantics, and computer code comprehension. Averaging across the different non-linguistic conditions, we observed stronger responses in the RH language regions compared to the LH language regions, as predicted by Bassett et al. (2013). Furthermore, we observed a high correlation across the non-linguistic conditions in how strong of a response they elicited in the LH vs. the RH (r=0.63; p<0.001): if a condition elicited a low response in the LH, it also tended to elicit a low response in the RH. However, a finer-grained examination of the different conditions revealed interesting differences across domains. In particular, whereas social stimuli (e.g., faces and bodies), music conditions, and visual meaningful events elicited stronger responses in the RH language regions, categorization and computer code comprehension elicited stronger responses in the LH language regions. Tasks taxing numerical cognition and executive functions elicited similarly low responses in the LH and RH language regions. Thus, reduced selectivity in the RH language regions is not ubiquitous. Understanding why certain non-linguistic functions elicit stronger responses in the RH language network may inform the nature of its contributions to language and cognition more generally.