Limited evidence for probability matching as a strategy in probability learning tasks

Probability matching is not the default decision making strategy in human and non-human primates

“…Despite being sub-optimal, probability matching and over-matching are reported in simple two-alternative ( binary ) prediction tasks with adult human participants 1 – 3 . These behaviours are not restricted to decision tasks; they also occur in other domains such as in language learning—i.e., when learners encounter “different ways of saying the same thing” in their input, they tend to reproduce those variants at the observed proportions 4 – 6 .…”

“…One possible interpretation of the conflicting results in the literature is that probability matching is not necessarily the product of a single process, but rather the result of a mosaic of mechanisms which produce similar behavioural outcomes even though they might operate in different circumstances 28 , 46 , 47 . However, the interpretation of the literature is complicated by the fact that behaviour on probability learning tasks is very sensitive to the details of the task and, for adults at least, its framing 3 . Given this sensitivity to task design and implementation, it is particularly problematic that the methods used often differ across species or age-groups, meaning that directly comparable studies are relatively rare.…”

“…Similarly, the details of the reinforcement procedure affect behaviour in animal experiments: experiments using guidance or correction procedures where, following incorrect responses, the correct response is highlighted and the animal continues responding until the correct response and the reward is provided, result in less maximising behaviour. This difference is attributed 3 , 52 to greater task difficulty in guidance/correction reinforcement paradigms where the trial boundaries are more opaque; however, it could equally be due to a reduced reward for maximising when the reward will eventually be obtained anyway, and the absence of the equivalent manipulations with humans makes this finding difficult to interpret.…”

“…For instance, experiments with adult humans typically do not feature concrete rewards (getting a trial right is the only reward), whereas experiments with animals and children more often or always feature actual rewards. And to give another example, experiments with humans (particularly children, to a lesser extent adults) tend to be relatively short compared to the long trial sequences provided to animals-this is despite the fact that there is evidence that the presence of rewards and more trials both tend to boost over-matching or maximising behaviours 1,3,16,36,[48][49][50][51] . Furthermore, potentially informative manipulations have not been systematically applied across species or age-groups, resulting in an incomplete picture of which factors influence behaviour across groups and potentially obscuring commonalities across species and age-groups.…”

Children’s value-based decision making

“…Humans can recognize and use probabilistic relationships in their environments as early as infancy 10 , 11 . Indeed, probabilistic learning during early childhood has been implicated in a number of different processes, including language learning, category formation, and social behavior 12 – 14 .…”