Figure 5
Conditioned response as a function of training with varying stimulus discriminability.
(A) Sample trajectories at training units 150 and 300 (rows) of mice trained with a constant SSIM of either 0.32 (left) or 0.04 (right), with the local curvature represented in color (see Methods). (B) Scheme of the swimming pool, divided into 6 ‘regions of interest’, defined in terms of their locations relative to the CS+ image. Using data on the path length and cumulative curvature from each sub-region, pattern vectors for each swimming trial were calculated, pooled and fuzzy-clustered using principal component analysis (see also Fig. S5 ). Regions are symmetric and allow combined access to CS+ and CS− (regions 3 and 4), visual access to CS+ or CS− before making a choice (regions 2 or 5) or visual access to CS+ or CS− after making a choice (regions 1 or 6). (C) Group average cluster maps from the cluster analysis of the mice's swimming trajectory as a function of their training; the data from correct (left) and incorrect (right) swims are shown separately. Color designates the distance to the center of mass for each of the 8 identified clusters, piled up as rows (color-bar on the right). (D) Average sum of all clusters per training unit (line plots; mean ± S.E.M) shows the evolution of conditioned responses as a function of training for correct and incorrect swims. Asterisks depict differences between groups (Two-way repeated measures ANOVA tests followed by Bonferroni's post hoc tests, *P < 0.05). (E) Point-to-point subtraction between the above cluster maps from (C) is shown on black background. Color-bar on the right. Black vertical lines in panels C and E arise from insufficient data to estimate the average clusters for the given training unit. (F–H) Group comparisons of conditioned responses at the end of training, as measured by the difference from SSIM0.04. Asterisks depict differences against the reference (Wilcoxon test, *P < 0.05, **P < 0.01); lowercase letters depict across-groups differences (one-way ANOVAs, followed by Bonferroni's post hoc tests, or Kruskal-Wallis tests, followed by Dunn's multiple comparison tests, P < 0.05; see also Methods).
