ABSTRACT
The ability to recognize objects or places despite variation in viewpoint seems fundamental to many adaptive behaviors. But do animals recognize objects or places in static pictures across changes in view? In Experiment 1, pigeons with prior outdoor experience were trained to locate an unmarked goal in 6 views of a scene presented on a touch-screen monitor. The goal location was fixed relative to landmarks in the scene, but 2-D vectors from landmarks to the goal varied across views. Pigeons learned the task, but showed poor transfer to novel views. Their performance resembled that previously seen in laboratory-raised pigeons with the same images, but differed from that previously seen in outdoor-experienced pigeons trained with more views of a different scene. Thus, stimulus and/or training factors, rather than outdoor experience, may determine degree of transfer to novel views. Experiment 2 explored pigeons' object recognition across changes in viewpoint. One group discriminated objects composed of identical parts and another group discriminated objects composed of different parts. Baseline accuracy was higher for the different-parts group, but transfer to novel views was comparable across groups. Both groups generalized across views, but accuracy decreased with rotation from the nearest training view. Implications for the use of pictorial stimuli to study cognitive processes in animals are discussed.
