(A) Performance of a Bayesian classifier in determining which pattern was presented to the network during the recall epoch based on the spiking output. The mean accuracy of the classifier in the identification of either patterns A through D (cyan) or the null pattern (red) decreases as γ-aminobutyric acid receptor type A channel potentiation increases, consistent with the action of propofol at increasing plasma concentrations. At moderate concentrations of propofol, the classifier accuracy decreases to around 70% and completely loses the ability to reliably recognize previously learned patterns at plasma concentrations consistent with general anesthesia. SD for 100 trials is indicated by the correspondingly colored shaded regions. (B) Scatterplot demonstrating the relationship between polychronous group (PNG) size and classifier accuracy for all plasma propofol concentrations. The classifier fails when the group size decreases to less than 25% of baseline. (C) Receiver operating characteristic analysis demonstrates that at propofol concentrations (conc.) beyond 4 μM, increasing levels of false positives must be accepted to obtain positive identifications with high reliability.