Learning and Memory in Mimicry. I. Simulations of Laboratory Experiments

An understanding of the dynamics of mimicry requires the modelling of the behaviour of predators in the wild. Current knowledge about the behavioural expression of learning and the dynamics of forgetting is insufficient for the construction of a definitive model of such behaviour. In particular, there is insufficient information on the response of a vertebrate when subjected to the presentation of a single identical conditioned stimulus paired with two (or more) unconditioned stimuli of different intensity or even of opposite effect, which in the present situation can be regarded as `model' and `mimic'. A general algorithm of learning and forgetting, based on the behavioural model of Bush and Mosteller, is proposed; it is applied as a linear operator in Monte Carlo simulations of the behaviour of a predator confronting a mixture of models and mimics. The algorithm is varied in detail: learning may be cumulative or instantaneous, constant or variable according to the strength of stimulus, and towards a continuously distributed or two-state (0 and 1) asymptote; forgetting may be cumulative or instantaneous, constant or variable according to the strength of stimulus, and dependent on time or on the occurrence of external events. Thirty different behaviour systems arise from rational combinations of the various learning and forgetting rules, including as special cases those behavioural models already proposed in the literature on mimicry. A standard experimental technique is the presentation to predators of a constant number of prey with varying proportions of models and mimics: we term this a reciprocal frequency experiment, and simulate its outcome for all thirty rational models. The results of such experiments, when appropriately transformed, will yield straight lines or curves according to the behavioural model employed. Models with all-or-none features (instantaneous learning or forgetting) tend to yield straight lines: curves tend to appear when the model assumes gradual or cumulative learning and forgetting. The result is dominated by the mode of learning; forgetting plays a secondary part. The method will therefore discriminate well between `switched' and `gradual' modes of learning and forgetting, but only if the experiments are carried out over a wide range of palatabilities for both the model and the mimic, and with other adequate design features. It is also necessary to design the experiment to distinguish between the dynamics of short-and long-term learning, which may well be different. Experiments to date permit no definitive or even tentative choice between the behavioural models, and do not necessarily, contrary to the literature, unambiguously support the models of discontinuous learning and forgetting proposed by J. E. Huheey. Experiments by M. L. Avery do, however, suggest tentatively that the dynamics of the short-term assessment of strategy by birds confronted with an intimate mixture of models and mimics may be adequately modelled as a process of rapid all-or-none switching between high and low attack probabilities, at least if the model and mimic are strongly aversive and strongly appetitive. It appears that there is a major gap in the understanding not only of mimicry systems but of the dynamics of learning and forgetting as motivational drivers in vertebrates.