Slide Slam P7 Sandbox Series
Supplementary motor area first predicts reaction time in stereotyped word articulation in large-scale intracranial EEG
Latane Bullock1, Kiefer Forseth1, Nitin Tandon1,2,3; 1Vivian L. Smith Department of Neurosurgery, McGovern Medical School, Houston, TX, USA, 2Texas Institute for Restorative Neurotechnologies, University of Texas Health Science Center at Houston, Houston, TX, USA, 3Memorial Hermann Hospital, Texas Medical Center, Houston, TX, USA
Single-word production, especially in the context of picture naming, has long been used to probe the neural networks underlying our uncanny ability to retrieve and produce specific lexical items from a large vocabulary. To characterize the articulation-related portions of these networks, we used intracranial recordings from a large cohort (134 epilepsy patients) performing multiple repetitions in which they articulated a single word, ‘scrambled,’ on average 66 times. The 25,000+ electrodes (subdural grids as well as depth electrodes) provided comprehensive coverage of language-dominant hemisphere. Using just one word eliminates numerous semantic, lexical, and phonological confounds intrinsic to the analysis of the production of a range of words. Behavioral data showed an average within-patient reaction time of 1137ms with standard deviation of ~300ms. Global mean broadband gamma (BGA) power dynamics, derived using a surface-based mixed-effects multilevel analysis (SB-MEMA), were used to create a 4D representation of cortical dynamics during this single-word production. From the SB-MEMA, we isolated suprasylvian regions of interest (ROIs) to investigate further. Gamma power in supplementary motor area (SMA), preSMA, precentral sulcus, and inferior frontal sulcus reach significance earliest, followed by the central sulcus and IFG. Finally, motor regions (subcentral gyrus and postcentral gyrus) are engaged to produce the appropriate motor gestures. Using these high-temporal resolution dynamics at hand, we sought to explain reaction time variability. We modeled reaction time as a function of BGA across trials at various time points across the pre-articulatory window. Activity of the supplementary motor area was the earliest predictor of reaction time, reaching significance just 250ms after stimulus onset. In contrast, activity in other early-onset ROIs, such as IFS and precentral sulcus, are not predictive of reaction time until 300ms after stimulus onset. In a final set of analyses, we used linear mixed-effects analyses to model reaction time per trial as a function of whether or not the trial was preceded by a ‘scrambled’ trial (PRIMED), the number of cumulative SCRAMBLED trials, and the number of trials since the previous SCRAMBLED trial, and implant type (SEEG or SDE). Each of these predictor variables were significant. Only PRIMED and implant type had a notable effect size (87 and 134ms, respectively). To identify a cortical correlate of the reaction time decreases in the PRIMED condition, we tested each active electrode in each ROI for BGA power divergence between the two conditions. 13 of 637 recording sites showed a ‘preference’ for either the PRIMED or UNPRIMED conditions when time-locking to stimulus onset. Taken together with the finding that SMA best predicts reaction time, reaction time decreases in the PRIMED condition may be primarily driven by subcortical structures with sparse cortical effects discernible with priming. SMA-mediated cortical-basal ganglia-thalamus loops may be initiated sooner for PRIMED trials that require a re-execution of a motor program as opposed to motor program selection and execution in UNPRIMED trials. Overall, this work highlights the key role of the SMA in speech initiation and timing and elaborates in detail the global networks involved in overlearned articulation.