Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Fronto-Temporal Predictive-Coding Network Dynamics in Common Marmosets during Roving Oddball Task
Poster Session A, Wednesday, September 30, 11:00 am - 1:00 pm, Wangari Maathai
Felix Körber1, Alessandro Tavano1,2, Edoardo Pinzuti1,3,4; 1Goethe University Frankfurt, 2Max Planck Institute for Empirical Aesthetics, 3Martin Luther University Halle-Wittenberg, 4Leibniz Institute for Resilience Research
The predictive coding framework posits that mammalian brains make sense of the world by constantly comparing internal predictions with sensory input across hierarchical levels of abstractions and timescales. Recent evidence suggests that speech and language processes also rely on a predictive coding architecture, in which top-down predictions reduce computational load during comprehension, aid in providing context, or help to decipher ambiguous input. One neural marker of predictive coding is the Mismatch Negativity (MMN). The MMN is evoked by the violation of a sensory prediction, which produces a brief negativity increase in scalp potentials within the N1-N2 time range relative to predicted stimuli. MMN in the auditory domain can be elicited by various abstraction levels, ranging from pure-tone deviations to unexpected phonemes. Understanding how network dynamics influence expectations and mismatch negativity could, thus, elucidate mechanisms of low-level language processing. Studying human cortical network activity is constrained by both ethical and structural factors, thereby raising interest in finding potential model organisms. One promising candidate is the common marmoset. These small monkeys not only demonstrate complex vocal behavior but also facilitate neural recordings of otherwise folded-in brain structures, such as the auditory cortex, due to their lissencephalic brain architecture. We analyzed publicly available ECoG data from three common marmosets performing a roving auditory oddball task with tone trains of three, five, or eleven repetitions. Information network dynamics were estimated using spectral multivariate transfer entropy. The directed information flow between electrodes was studied during a 100 ms time window before the stimulus onset of the expected standard and unexpected deviant tones separately. Consistent pre-stimulus information transfer was observed from prefrontal to superior temporal regions. Notably, this information was carried in the alpha- and beta-bands, indicative of top-down predictions of upcoming stimuli. Differences in network dynamics before standard and deviant tones were found, suggesting implicit learning of statistical regularities. Our work provides preliminary evidence for a neural network underlying preparatory hierarchical predictions of auditory statistics in the common marmoset. The observed fronto-temporal network dynamics in marmoset monkeys appear to mirror the major components of proposed hierarchical predictive language processing networks in humans. Specifically, the beta band involvement further aligns with its proposed role in language processing for providing context and top-down predictions, suggesting that vocal animals such as marmoset monkeys can be studied as model organisms for human speech.
Topic Areas: Animal Communication and Comparative/Evolutionary Studies, Speech Perception