Count-based decision-making in mice: numerosity vs. stimulus control.

Numerical and temporal control of behavior is ubiquitous across many species of animals. Recent studies showed that in the presence of reliable discriminative stimuli, mice ignore temporal relations and probabilistic information but when discriminative stimuli become non-informative, the same mice can spontaneously start relying on previously experienced time intervals and probabilities. Similar dynamics do not readily generalize to counting behavior since the response-outcome contingency functions differ when reinforcement depends on the number vs. timing of responding. In the current study, mice (N = 32) learned to press two different levers 10 (few) or 20 (many) times, while the active lever was signaled by a light stimulus. The probability of the few/many trials was manipulated between groups. During testing, the informative value of light stimulus was eliminated by signaling both few- and many-levers. In a quarter of training trials, mice ignored the discriminative stimulus and adopted a numerical decision strategy (starting to respond on the few-option and then switching to the many-option in many trials) that was sensitive to probabilistic information. The frequency but not the probability-sensitive parametrization of switching behavior changed when the discriminative stimulus became non-informative in testing. These findings suggest that there is a relatively strong representational control over counting behavior even in conditions that afford strong stimulus control.