During visual search for samples of varying proportions of familiar, natural food grains displayed against a complex gravel background, pigeons exhibited "matching selection," a tendency to overselect the more common grain. The matchingselection effect was decreased at low levels of stimulus/background contrast and reversed when the grains were highly conspicuous. The results were consistent with the hypothesis that stimulus detectability should be enhanced by recent experience with a particular grain type, but they showed no convincing indications of a corresponding effect on the response criterion. An explanatory model, termed the attention threshold hypothesis, argues that the mean latency of discovery can be minimized by selectively attending to one stimulus type at a time and switching to a more generally receptive state when the rate of discovery falls below a threshold value. The model appears to account for the fact that the response rate was highest toward samples containing a single grain type and decreased as the relative proportions approached equality. Additional consequences of the adoption of this theoretical perspective were explored in some detail. Among other results, the evidence suggests that the switching threshold might be chosen so as to optimize the rate of food discovery.Animals that are ecological generalists, that is, apimals that forage for a variety of food, commonly exhibit a selection bias that depends on relative frequency (Curio, 1976;Krebs, 1973). Food types uncommon in the environment tend to be disproportionately underrepresented in the diet, whereas more abundant foods are consumed in excess. A selection bias could result from a number of passive factors, such as heterogeneity in the spatial distribution of food types or changes in food accessibility with density (Murdoch & Oaten, 1974). In animals that conduct a visual search for food, however, Tinbergen (1960) hypothesized that the bias may be attributable to active processes. Recent expe-I wish to thank D.