Figure 4. Discrimination and generalization of similar odors. A, Three example MB recordings showing response patterns evoked by the odors PA, BA, and EL, averaged across all trials. Responses to BA and PA are similar and distinct from EL. B, Venn diagram showing overlap of significantly responding cells from all recordings (n = 24). Many neurons tend to respond to both PA and BA, whereas most EL-responsive KCs respond only to EL. C, Correlation scores for each pair of odors. Each point is the correlation score calculated from the average MB response to each odor within a single recording (n = 24 recordings). The correlation between PA and BA is significantly (see main text) greater than the other two. Dark gray box indicates 95% confidence interval; light gray box indicates 1 SD; black line is the mean. D, Euclidean distances between odor responses within recordings (circles). Distances are based upon average MB response patterns. Each point indicates the PA-BA distance (x-axis) versus either the PA-EL (orange) or BA-EL (brown) distance (y-axis). Statistics reported in D–F are results of a one-tailed binomial test. E, Linear classifier based on Euclidean distance distinguishes PA from BA (left) and BA from PA (right) above chance (dashed line at 50%) for all recordings. Black line is the mean. F, When the classifier is trained to detect PA, but tested with BA versus EL, BA is consistently chosen over EL (left). Likewise, when trained to BA, PA is chosen more often (right). G, Flies can discriminate PA and BA. Gray lines show choices of yoked control/experimental group pairs. Pairing PA with shock significantly reduces the number of flies choosing PA compared with un-shocked controls (paired t test, one-tailed). Black line indicates the mean across experiments. The label “PA: PA vs BA” indicates that flies in the experimental group were shocked to PA, then given a choice between PA and BA. Odor preference was calculated against PA in this case. Similar results were obtained in the reciprocal case, “BA: BA vs PA” (right). H, Flies generalize across PA and BA when tested against EL. Flies shocked to PA and given a choice between BA and EL chose BA less often (paired t test, two-tailed) than the PA-exposed-only control group (left). Right panel shows reciprocal experiment with flies shocked to BA. In only one case was the proportion of flies choosing PA greater in the shocked group than its yoked control (red).