To study the effects of hunting by searching image on the dynamics of the prey population, we created a virtual population of 200 digital moths (Bond & Kamil 1998). Our initial population contained equal numbers of each of three distinctive moth types. As in our earlier work, the moths were presented singly on cryptic backgrounds to blue jays in an operant chamber. To obtain a realistic "virtual ecology," however, four different blue jays were used as predators, with each bird seeing a randomly-selected quarter of the moth population each day. Detected individuals were considered "killed" and were removed from the moth population. Those that were overlooked were allowed to breed, in proportion to the relative abundance of their moth type, bringing the population up to its previous level the following day. Each day thus constituted a generation. Our only experimental intervention was to set the initial numbers of the three morphs.

Moth images used in the virtual ecology experiment. Moths 1, 2, and 3 constituted the initial prey population. Moths 4 and 5 were novel forms that were introduced in later stages of the experiment. Note that Moth 5 is considerably different in appearance from the other four.

This methodology defines a fixed phenotype or "coexistence" procedure(Kassen 2002), in which the population of digital moths consists of a set of asexually reproducing clones of invariant appearance. They do not mutate, and each generation is brought up to a constant size based on the relative numbers of surviving individuals. The population dynamics of the different morphs and their asymptotic levels of abundance are, thus, the main dependent variables. Over the course of 50 generations, the numbers of the three morphs rapidly achieved a characteristic equilibrium that was independent of initial relative abundances and resistant to subsequent perturbation.

In the first 50-generation run, three moth types were initially set to equal abundances. Within 10-15 generations, they had shifted to characteristic equilibrium levels. In subsequent runs, one of the less abundant moths at equilibrium was initially set high to perturb the system. In each case, the previous equilibrium levels were rapidly regained, demonstrating stable system dynamics.

Additional analyses demonstrated that the equilibrium was a result of "apostatic selection" (Clarke 1962, 1969), a negative feedback between prey abundance and detectability. If jays tend to concentrate their searching efforts on more common moth types, we would expect that the proportion of individuals of a specified type that were discovered (="taken") by the birds would be higher than expected on a random basis for common forms and lower than expected for rarer ones. Thus, there should be a significant linear relationship between the deviation from randomness ("proportion taken minus proportion present") and the proportion present in the population. This relationship was confirmed for all three moth types (see below), providing the first direct demonstration of the dynamic relationship between searching image, apostatic selection, and prey population stability.

We then tested the effects of apostatic selection on novel morphs, moth types that the birds had not seen previously (Bond & Kamil 1998). When small numbers of each a new moth type were introduced into the population, they were not initially detected by the jays and their abundance rapidly increased. In the first case (Moth 4), the jays ultimately took notice of the new prey items and drove their numbers down, establishing a new equilibrium state. The second novel form we tried had a different outcome, however. Moth 5 rapidly increased until it completely dominated the population. Even after 100 generations, only two of the four jays had learned to find it, and if we had not continuously reintroduced small numbers of the original three forms, they would all have been driven to extinction (see below).

From these results, the stability of the observed configuration of morphs appeared to be a function of subtle relationships between the appearance of mutant forms and that of the pre-existing moth types. Whether this would prove to be the case in a more realistic system in which prey phenotypes were free to evolve was not clear. To explore this question, we needed to extend our procedures to enable the evolution of prey appearance, and for that, we needed to develop a functional virtual genetics.