I consider here experiments that offered arrays of artificial flowers under controlled conditions to foraging pollinators. The provisioning behavior of bumble bee workers allowed researchers to study long foraging sequences without satiation of the experimental subjects.
Typically, laboratory experiments test bees' preferences on arrays of two or more colors of artificial flowers. The frequencies of the colors are varied to see if frequency affects the proportion of visits a color morph receives. Frequency-dependent (common or rare morph) and frequency-independent (bias to one morph independent of frequency) preferences can occur simultaneously. Figure 12.1 shows the range of potential relationships between frequency and visitation, along with the resultant expected fitness relationships.
Gendron (1987) discussed statistical tests for frequency dependence Comparing observed and expected numbers of visits with goodness of fit tests is suitable only for very small data sets, due to the risk of Type I errors. Two authors have developed methods specifically for analyzing frequency-dependent behavior.
Manly's model (1973) can be used to measure selection when morph frequencies change during an experiment, and this model is frequently used for predator-prey experiments. However, it does not fit some types of data well, as the probability of a morph being selected is constrained to be a linear function of frequency (Gendron 1987), and may be invalid if selectivity varies with learning (Greenwood & Elton 1979). I have adopted Greenwood & Elton's (1979) model,
F (VA)b (VA)b + (1 - A)b, which relates the availability of a morph in a two-morph system (A) to those eaten (F) using two parameters, frequency-independent (V) and frequency-dependent (b) components. This model is constrained by being
Was this article helpful?