Many organisms take time before approaching unfamiliar objects. This caution forms the basis of some well-known assays in the fields of behavioral ecology, comparative psychology and animal welfare, including quantifying the personality traits of individuals and evaluating the extent of their neophobia. In this paper we present a mathematical model which identifies the optimal time an observer should wait before approaching an unfamiliar object. The model is Bayesian, and simply assumes that the longer the observer goes without being attacked by an unfamiliar object, the lower will be the observer’s estimated probability that the object is dangerous. Given the information gained, a time is reached at which the expected benefits from approaching the object begin to exceed the costs. The model not only explains why latency to approach may be repeatable among individuals and vary with the object’s appearance, but also why individuals habituate to the stimulus, approaching it more rapidly over repeated trials. We demonstrate the applicability of our model by fitting it to published data on the time taken by chicks to attack artificial caterpillars which share no, one, or two signaling traits with snakes (eyespots and posture). We use this example to show that while the optimal time to attack an unfamiliar object reflects the observer’s expectation that the object is dangerous, the rate at which habituation arises is also a function of the observer’s certainty in their belief. In so doing, we explain why observers become more rapidly habituated to “weaker” stimuli than “stronger” ones. 

The data presented here were originally collected and interpreted by Skelhorn et al. (2016). Multicomponent deceptive signals reduce the speed at which predators learn that prey are profitable. Behavioral Ecology, 27(1), 141–147. doi:10.1093/beheco/arv135.  Here we use these data to demonstrate the applicability of our Bayesian model, and in so doing we help explain why observers become more rapidly habituated to “weaker” stimuli than “stronger” ones. 

In short, Skelhorn et al. (2016) presented naive domestic chicks with palatable pastry models of snake-mimicking caterpillars that varied in their resemblance to snakes. These models either had eyespots or not and were set in a resting or defensive posture in a 2 x 2 factorial design. All chicks were presented with just one caterpillar type and there were 9 chicks per treatment combination. Presentations of the caterpillars to the chicks were conducted over 4 consecutive trials which were completed in 2 days. The primary endpoints was the time (in seconds) taken for a given chick to attack the pastry bait. 

The data provided lists the treatment (1 = resting posture, no eyespot;  2 = resting posture, eyespots; 3 defensive posture, no eyespot; 4 = defensive posture, eyespot), the trial (1-4), the latency of the chick to attack (in seconds) and the individual chick ID.

The R code used to analyze and plot the data are presented in the Supplementary Materials of Sherratt et al. (2023) The optimal time to approach an unfamiliar object: A Bayesian model. Behavioral Ecology.