HIGGINS, C.L.; BUTLER, P.J.; STRAUSS, R.E.*: Discrimination of foraging paths produced by different search models

The analysis of search paths plays a key role in optimal foraging theory. We developed a simplified model of resource acquisition in which we randomly dispersed N points (“food items”) of equal value within an arena, and considered only the search path followed to find and consume the N items (assuming no satiation or learning) by computer simulation. For a given point configuration, two deterministic models, the globally optimal shortest and longest paths, provide the lower and upper bounds on path length. We characterized the lengths and shapes of search paths produced under five probabilistic and two other deterministic models: random choice, Pearson random walk, Levy random walk, reciprocal-distance preference, inverse-squared-distance preference, trajectory-directed search, and nearest neighbor. Each resulting search path was characterized geometrically by its total length, distribution of step lengths, vector autocorrelation function, distribution of angular deviations, and number and spacing of path intersections. We used linear discriminant analysis and nonlinear multilayer perceptrons (artificial neural networks) to discriminate paths produced by the different models based on these path descriptors. This basic procedure can be extended in various ways and can potentially be used to classify observed search patterns of foragers according to underlying behavioral models. The results suggest that organisms can attain near-optimal performance using biologically realistic search tactics.