hide
Free keywords:
-
Abstract:
Anticipatory appetitive behaviours (e.g., approach) based on learned predictions of future rewards are generally adaptive, and are hence prevalent in many animals. However, these innately programmed responses frequently interfere with subjects’ performance in scientific experiments, amongst other contexts [1]. What might be a nuisance to the experimenter also provides valuable insight into the progress of subjects’ learning, their natural propensities, and the self-control they need to exert to perform the task correctly – as such, it is important to consider them when accounting for behaviour. As a case in point, we model the performance of mice in a temporal discrimination task administered in a series of studies by Paton and colleagues (e.g., [2][3]), in which animals have to categorise time intervals as short vs. long relative to a fixed, implicit, threshold. A salient feature of behaviour is the enduring high frequency ( 34% of trials on average) of premature responses before the completion of the interval is signalled, despite this resulting in missed rewards and time penalties. We ascribe this propensity to a form of ’Pavlovian misbehaviour’ [1] in which the mice fail to suppress an urge to act in the face of evolving predictions of future reward. We formalize this in a model that aims to integrate parsimoniously the different factors that influence behaviour in this task. We view this fine-grained analysis of behaviour as a necessary prelude to fully understanding the rich pattern of dopaminergic [2] and striatal [3] measurements also reported in this task.