RELATIVE PREFERENCES FOR VARIOUS BIVALUED RATIO SCHEDULES

Widely cited experiments on optimal foraging have used bivalued distributions as representing environmental stochasticity, characterizing these in terms of their arithmetic means. In contrast, research on free-operant choice has established that organisms prefer variable patterns of food delivery, relative to fixed patterns with the same mean values. To explore such departures from linear averaging, specifically with respect to bivalued alternatives, pigeons were given choices between a fixed-ratio (FR) schedule of food delivery that required 15, 30, or 60 responses and bivalued variable-ratio schedules with an arithmetic mean of 60.5 or 60. A bivalued schedule of 1 and 120 was preferred almost exclusively over each of the FR values. With a bivalued schedule of 15 and 105, there was a shift of preference, most notably in the FR-15 condition, but in no case was linear averaging a good predictor of the birds' choices. Geometric averaging fared better, but even this failed to represent the apparent salience of the minimum value of the bivalued schedule in some conditions.