Jeremy J. D. Greenwood

Analysing Experiments on Frequency-Dependent Selection by Predators

(1) Predators sometimes eat disproportionately more of the more abundant forms of their prey, thus promoting diversity in the prey population. (2) Such selection may be investigated by exposing a series of prey populations with different relative frequencies of the various types to a series of predators. (3) We use a simple model to describe the outcome of such a series of trials, relating the relative frequencies of the various prey types eaten to the relative frequencies available. Frequency-independent and frequency-dependent components of selection are separated in the model. (4) Previous criticism of the model does not appear to be valid. An alternative model due to Manly seems a priori unrealistic and potentially misleading. (5) We have investigated the fit of our model to all the available data. In most cases it is good. Manlys model is also a good fit to most of the data, except at extreme prey frequencies. (6) Experiments of the type considered need to be carefully designed if the results from different experiments are to be comparable. (7) They are more useful than most of the other types of experiment and observation that have been used for investigating frequency-dependent selection. (8) Appendix 1 describes two techniques for fitting our model to the data from experiments with two prey types. (9) Appendix 2 describes an approximate technique for fitting our model to the data from experiments with more than two prey types and illustrates it with a worked example.

Frequency-dependent selection by seed-predators

This paper is concerned with selection of seeds by animals that search for and respond to seeds as individual items. It also considers mutualistic frugivory. Techniques and results of studies of frequency-dependent selection of food are discussed. Frequency-dependent selection may be constant in direction, only its strength varying with the frequencies of the various types of food available, or it may vary in direction according to frequency (apostatic selection). Frequency-dependent selection is common, though only some experiments show apostatic selection. Selective behaviour is often variable within a single experiment, making replication important. Selection tends to be anti-apostatic when the prey are crowded. In principle, frequency-dependent seed-predation could be important in enhancing or reducing vegetational diversity and within-species variation - and in causing the convergence or divergence in appearance of coexisting species. I discuss the evidence for

Food selection by chicks: Effects of colour, density and frequency of food types

Abstract Previous studies have shown that birds, including 3-day old chicks, show frequency-dependent selection when feeding on a food population made up of two types. Two experiments using 5-day old chicks are described here. In one, foods of three colours were presented in their three pairwise combinations, at five different relative frequencies. It was found that measures of selection were transitive: that is, selection of red versus yellow could be predicted from selection for red versus green and for green versus yellow. Selection was not frequency-dependent. In the second experiment, total food density, as well as relative frequency, was varied. Selection was density-dependent, being strongest at the middle one of the three densities. It was also irregularly frequency-dependent, being weaker at one (inter-mediate) frequency than at the other four, among which there was little variation.

Frequency-dependent selection by predators: comparison of parameter estimates

We consider a simple model of frequency-dependent selection of food by predators. It has two parameters: b measures the frequency-dependent component of selection; V measures the frequency-independent component. If an experimenter obtains different parameter estimates under different conditions, he wishes to assess their significance. We give methods for testing the null hypotheses that both b and V are constant under various treatments, that b only is constant, and that V only is constant. We show how, should any of these hypotheses prove acceptable, revised parameter estimates may be obtained.