Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Neural Dynamics of Inception in Spoken Language Comprehension: Evidence from Intracranial Recordings
Poster Session B, Wednesday, September 30, 4:30 - 6:30 pm, Wangari Maathai
Judah Huberman-Shlaes1, Julien Dirani2, Eliza Reedy1, Raouf Belkhir1, Jiahao J. Chen1, Arka N. Mallela3, Jorge Gonzalez-Martinez1, Bradford Z. Mahon2; 1University of Pittsburgh Medical Center Department of Neurological Surgery, 2Carnegie Melon University, 3Rush University Medical Center Department of Neurosurgery and Rush Epilepsy Center
During spoken language comprehension, there is a moment when a listener converges on the intended concept—for example, “hearing the sour yellow fruit” leads to the activation and retrieval of LEMON. The neural basis of this moment of semantic ‘inception’ is difficult to study without affecting it: requiring a behavioral report by participants indicating when they arrive at the target concept can alter the underlying processes. Here, we take a different approach by “inception-norming” spoken definitions to estimate when, during an unfolding utterance, listeners typically arrive at the target concept. This allowed us to identify the spatiotemporal fingerprint of the moment during language processing when the listener arrives at the target concept using intracranial recordings. Eight individuals undergoing sEEG monitoring with frontal and temporal electrode coverage completed a trimodal naming paradigm (picture, sound, and spoken definition). Neural responses to picture and sound trials were analyzed in 100 ms post-stimulus locked epochs. We first defined ‘amodal contacts’ as contacts that responded significantly, regardless of whether the cueing stimulus was a sound (sound of a dog barking) or a picture (picture of a dog). Contacts were classified as amodal if they showed significant changes in z-scored local field potential (LFP) activity in both picture and sound modalities using one-sample t-tests. We then separately, and independent of the above amodal contact search analysis, defined contacts for which LFP activity while hearing spoken definitions was predicted by Inception Likelihood Values (ILVs). ILVs were derived from a healthy control cohort, who normed the auditory definition stimuli using a gating procedure. ILVs are thus an empirical estimate of the probability that a listener will align to the target concept at each word in the definition. For each contact, linear regression tested whether ILV predicted time-resolved LFP activity, identifying contacts whose neural dynamics track the moment of inception. Finally, we quantified overlap between the set of ‘amodal contacts’ and the set of ILV-predicted contacts. Statistical significance was assessed using permutation testing, with false discovery rate (FDR) correction for multiple comparisons. Overlap between amodal contacts and ILV-predicted contacts was significant in the 300–400 ms post stimulus (17% overlap, q = 0.014) and 400–500 ms post stimulus (16% overlap, q = 0.026) windows. Spatially, overlapping contacts were distributed across temporal cortex and inferior frontal gyrus. No significant overlap with ILV-predicted contacts was observed for contacts responsive only to picture trials (q = 0.90) or only to sound trials (q = 1.00), or to contacts that responded to neither modality (q = 1.00), in any window post stimulus presentation. Our approach enables direct observation of the spontaneous emergence of semantic inception during language processing, in the absence of explicit behavioral reporting. This convergence emerged specifically in the 300–500 ms window, consistent with the expected timing of lexical-semantic retrieval of words. Ongoing work is testing the contribution of pulvinar-cortical connectivity in driving the cortical-cortical connectivity, and how electrical stimulation to ‘inception contacts’ can facilitate or disrupt behavioral measures of lexical semantic access.
Topic Areas: Speech Perception,