Behavioral scoring: freezing and residualized change

Freezing was defined as the absence of all movement aside from breathing and ear twitching, not including sleeping or resting. Behavior was scored manually from videos by an experimenter blind to experimental conditions. The total amount of CS-induced freezing was expressed as a percentage of total time spent freezing during each 20-s CS. Our initial plan was to operationalize post-acquisition conditioned fear as mean freezing over the first 3 tone-alone trials 24 hours after acquisition (i.e., the first 3 trials of extinction), post-extinction conditioned fear as mean freezing over the last 3 trials of extinction, and post-reinstatement conditioned fear as mean freezing over three CS presentations 24 hours after reinstatement. This plan was contingent on the statistical demonstration that variance between trials for these 3-trial blocks was random with respect to all independent variables and therefore ignorable. As detailed at the beginning of the Results, this turned out to be true only for the 3 trials at the end of extinction. Consequently, this was the only block of trials that we averaged over, and we limited analysis of acquisition and reinstatement behavior to the first CS-alone presentation following each of these sessions.

In addition, we operationalized within-session extinction and reinstatement as residualized change in freezing between initial post-acquisition freezing and these respective time points. Residualized change is a difference score defined by Y – Y’, where Y’ is the regression of a subsequent freezing score, Y, on the initial freezing score, X. This partials out the variance attributed to individual differences in acquisition learning, rendering the residualized change scores linearly independent from (uncorrelated with) initial conditioned freezing levels. While this could also be accomplished by including initial freezing as a covariate in models that predict subsequent freezing behavior, regressing out acquisition variability prior to running subsequent models has the advantage of simplifying model output and interpretation and enabling us to plot and visualize extinction and reinstatement scores that are “uncontaminated” by differences in acquisition. Thus, these scores offer an individualized, relativistic measure of change, e.g., the residualized extinction score reflects how well an individual rat extinguishes relative to how well it is expected to extinguish given its initial level of fear.