Figure 1.

Training mice on a reversal paradigm based on 2-AC frequency discrimination. A, Schematic of the reversal paradigm using the tonecloud task for mice (top). Animals are first trained on one contingency, e.g., to pair a low-frequency tonecloud with reward on the right referred to as “low-right” (“learning”). Once they reach the performance criteria, the training contingency is reversed, requiring the same animal to now pair a low-frequency tonecloud with reward on the left, or “low-left” (“reversal”). The trial structure (bottom) shows the sequence of events in a single typical “correct” trial in the task. B, Example learning curves of mice trained in the reversal paradigm (black, low-left → low-right, n = 3 and red, low-right → low-left, n = 3), where 0 denotes the start of training on the reversed contingency. The performance of the animals is smoothed over three sessions for this plot. C, Animals require a comparable number of sessions to reach performance criteria during initial learning as during reversal.