The information provided by the absence of cues: insights from Bayesian models of within and transgenerational plasticity.

Empirical studies of phenotypic plasticity often use an experimental design in which the subjects in experimental treatments are exposed to cues, while the subjects in control treatments are maintained in the absence of those cues. However, researchers have virtually ignored the question of what, if any, information might be provided to subjects by the absence of the cues in control treatments. We apply basic principles of information-updating to several experimental protocols used to study phenotypic plasticity in response to cues from predators to show why the reliability of the information provided by the absence of those cues in a control treatment might vary as a function of the subjects' experiences in the experimental treatment. We then analyze Bayesian models designed to mimic fully factorial experimental studies of trans and within-generational plasticity, in which parents, offspring, both or neither are exposed to cues from predators, and the information-states of the offspring in the different groups are compared at the end of the experiment. The models predict that the pattern of differences in offspring information-state across the four treatment groups will vary among experiments, depending on the reliability of the information provided by the control treatment, and the parent's initial estimate of the value of the state (the parental Prior). We suggest that variation among experiments in the reliability of the information provided by the absence of particular cues in the control treatment may be a general phenomenon, and that Bayesian approaches can be useful in interpreting the results of such experiments.