Robust single-trial estimates of electrocortical generalized aversive conditioning: Validation of a Bayesian multilevel learning model

Abstract

Aversive conditioning changes visuocortical responses to conditioned cues, and the generalization of these changes to perceptually similar cues may provide mechanistic insights into anxiety and fear disorders. Yet, neuroimaging conditioning paradigms are challenged by poor single-trial signal-to-noise ratios (SNR), missing trials, and inter-individual differences in learning. Here, we address these issues with the validation of a steady-state visual evoked potential (ssVEP) generalization paradigm in conjunction with a Rescorla-Wagner inspired Bayesian multilevel learning model. A preliminary group of observers (N=24) viewed circular gratings varying in grating orientation, with only one orientation paired with an aversive outcome (noxious electric pulse). Gratings were flickered at 15 Hz to evoke ssVEPs recorded with 31 channels of EEG in an MRI scanner. The multilevel structure of the Bayesian model learning model informs and constrains estimates per participant providing an interpretable generative model. It led to superior cross-validation accuracy and insights into individual participant dynamics than simpler models. It also isolates the generalized effects of conditioning, providing improved statistical certainty. Lastly, the present report demonstrates that missing trials are interpolated and weighted appropriately using the full models structure. This is a critical aspect for single-trial analyses of simultaneously recorded physiological measures, because each added measure will typically increase the number of trials missing a complete set of observations. The present technical report validates a limited version of a learning model to illustrate the utility of this analytical framework. It shows how models may be iteratively built and compared in a modern Bayesian workflow. Future models may use different conceptualizations of learning, allow integration of clinically relevant factors, and enable the fusion of different simultaneous physiological recordings.