Dynamics of count-based decision-making in mice

Numerical competence is an important cognitive function and is widely seen in a diverse group of species (Nieder, 2020). Studies conducted with mice have shown that mice can not only learn different quantities, but they can also incorporate the level of endogenous uncertainty (Çavdaroğlu & Balcı, 2016) as well as probabilistic information (Berkay et al., 2016) regarding the previously learned quantities into their decisions. Recent studies have revealed similar results for time-based decisions in addition to showing that mice represent different intervals and can later utilize them in a model-based fashion when new experimental conditions require it rather than learning the optimal strategy by trial-and-error (Tosun et al., 2016). Despite the psychophysical similarities between interval timing and numerical functions, whether mice can use previously learned numerical information in an abrupt fashion when there is an ambiguous situation has not yet been studied. Here we investigated if and how mice (N = 32) use the numerical and probabilistic information that they acquired when faced with an ambiguous situation. In the task that we used here, reward was presented either after 10 (few) or 20 (many) presses, the last of which had to be on the correct lever. We also manipulated the relative frequency of few- and many-response trial types to investigate the effects of probability. Our results showed for the first time that, different from time-based decision-making literature (Tosun et al., 2016) showing an emergent decision strategy that takes into account the task history in case of discriminative stimulus ambiguity, count-based decisions prioritize task history over stimulus associations even in the absence of ambiguity. Keywords: count-based decision-making, numerical switch behavior, ratio- and intervalschedules