Menger, N. S.; Kotchoubey, B.; Pavlov, Y. G.

Reversal learning offers a window into how associations are acquired, updated, and overwritten. Because olfactory inputs bypass much of the thalamus and are tightly linked to emotion, we examined whether humans can flexibly form and subsequently reverse aversive associations to smells. Thirty healthy adults underwent an olfactory reversal conditioning protocol in which one neutral odor (CS+) was followed by a 90 dB aversive sound (US) and a second odor (CS-) was not. After five blocks the contingencies were reversed. Throughout 300 trials we collected ratings of pleasantness and intensity together with autonomic physiological indices (skin conductance, ECG, photoplethysmography, respiration), facial EMG, and 64-channel EEG. Contrary to expectations, pleasantness, intensity, and all autonomic or facial muscle measures failed to differentiate CS+ and CS- either before or after reversal (all p > .01). Event-related potentials, alpha suppression, heart rate and pulse wave responses likewise showed no CS specificity. Only multivariate classifiers: trained (i) on the time-domain EEG signal and (ii) on alpha-band activity, separately, distinguished CS+ from CS- at above-chance levels in late post-stimulus intervals. These neural signatures did not translate into overt physiological or behavioural differences. The pattern suggests that, at least with neutral odors and an auditory US, olfactory fear learning is subtle, spatially variable across the cortex, and easily masked in group-level averages. Our findings highlight both the promise of multivariate EEG for detecting fragile olfactory associations and the challenge of eliciting robust conditioned responses with cross-modal (odor-sound) pairings.