Fig. 2: Startle- and sound scaling underlie the phenomenon of PPI.

a Startle responses for one rat (same rat as Fig. 1c, f) across different startle sounds (x-axis) and prepulse sounds (colors) with a constant 100 ms delay. x-axis (and all subsequent dB references) indicates dB above background. Warmer colors indicate louder prepulse sounds. Bars indicate standard error of the mean. The normalized cross-validation (CV) error for this rat was 1.20; the median CV error across rats was 1.33 with an interquartile range of 0.26. b Diagram showing how a baseline startle curve (black) could be scaled via sound scaling (blue), startle scaling (red), or both startle- and sound scaling (magenta). c Distribution across all rats and experiments of the difference between the normalized cross-validation error of the model with both startle scaling and sound scaling, and the normalized cross-validation error of the model with only startle scaling. Positive error differences indicate that the two-parameter model had lower error, and the dotted vertical line shows error difference of 0. d Sound scaling versus prepulse sound (top) and startle scaling versus prepulse sound (bottom) for an example rat (the same rat as Fig. 2a) from an experiment that varied prepulse level. e Sound scaling versus delay (top) and startle scaling versus delay (bottom) for a different example rat from an experiment that varied delay. For d and e, dotted lines indicate linear regressions. f Distribution of linear regression slopes across all rats of sound scaling versus prepulse (top left), sound scaling versus delay (top right), startle scaling versus prepulse (bottom-left), and startle scaling versus delay (bottom right). Dotted lines show a slope of 0.