(A) Twenty simulations of a Bernoulli process where the response probability is 44% for all trials (identical to the average PRP in mice at 0.6% isoflurane) were performed. Each simulation consisted of the same number of trials as the experimental data from (B). Similar simulations were conducted for zebrafish using PRP of 0.45. Simulated (A) and experimentally observed (B) transition probability matrices are shown schematically above traces. The population response probability in simulations is not statistically different from experimental observations across mice (C) (U = 197.5, nsim = nexp = 20, p=0.95) or across zebrafish (D) (U = 2276, nsim = nexp = 72, p=0.34). Both mice (E) (U = 48.5, nsim = nexp = 20, p<0.0001) and zebrafish (F) (U = 300.5, nsim = nexp = 72 p<0.0001) have fewer state transitions than simulations of the Bernoulli process. Both mice (G) (U = 103, nP(U|R) = nP(U|U)=20, p<0.001) and zebrafish (H) (U = 501, nP(U|R) = nP(U|U)=72, p<0.0001) are more likely to stay unresponsive if they failed to respond in the previous trial (purple arrow in B) than if they were able to respond to the previous stimulus (red arrow in B). In plots (C–H) box plots show inter-individual differences in estimated parameters (central mark indicates the median, box shows interquartile range, whiskers extend to the most extreme data points). Statistical significance is shown by ***p<0.001 ****p<0.0001. Source data for all quantitative analyses described are available in the Figure 3—source data 1.