Encoding of spatial information in images of an outdoor scene by pigeons and humans

Abstract

Pigeons and adult humans searched for a 2-cm 2 unmarked goal in digitized images of an outdoor scene presented on a touch-screen monitor. In Experiment 1, the scene contained three landmarks near the goal and a visually rich background. Six training images presented the scene from different viewing directions and distances. Subsequent unreinforced tests in which landmark or background cues were removed or shifted revealed that pigeons' search was controlled by both proximal landmarks and background cues, whereas humans relied only on the proximal landmarks. Pigeons' search accuracy dropped substantially when they were presented with novel views of the same scene, whereas humans showed perfect transfer to novel views. In Experiment 2, pigeons with previous outdoor experience and humans were trained with 28 views of an outdoor scene. Both pigeons and humans transferred well to novel views of the scene. This positive transfer suggests that, under some conditions, pigeons, like humans, may encode the three-dimensional spatial information in images of a scene.Many organisms have been shown to use visual landmarks to remember and locate a goal area (see Collett, 1992, andGallistel, 1990, for reviews). Use ofvisual landmarks is demonstrated experimentally by manipulating the visual cues between the opportunity to encode the spatial information and subsequent search tests in which the subject attempts to locate the goal. Two types of evidence indicate that a subject uses a particular landmark: (I) the subject shifts its search location in response to landmark shifts, and (2) accurate search is disrupted by the absence of the landmark. In a series of experiments using a laboratory task in which food is hidden on the floor of a spatial arena, Cheng and his colleagues (e.g., Cheng, 1988Cheng, , 1989Cheng, , 1994Cheng & Sherry, 1992) have clearly demonstrated that pigeons use visual landmarks to locate a hidden goal, and they have identified several principles of pigeons' landmark-based search.Pigeons' landmark-based search has also been investigated in a touch-screen task in which pigeons search for an unmarked goal on the surface of a monitor (Spetch, Cheng, & Mondloch, 1992;Spetch & Mondloch, 1993;Spetch & Wilkie, 1994 85search task has yielded results that are remarkably similar in certain ways to those found in real-space tasks. In particular, near landmarks are weighted more heavily than are far landmarks (Cheng, 1989;Spetch & Wilkie, 1994), and landmarks near an edge ofthe search space exert more control in the dimension parallel to the edge than in the dimension perpendicular to the edge (Cheng & Sherry, 1992; Spetch et aI., 1992). Recent work (Spetch, Cheng, & MacDonald, 1996; Spetch et aI., 1997) has revealed that, on both the touch screen and the laboratory floor, pigeons can use the configuration of an array of landmarks to locate a goal. In both environments, however, they respond to expansions of the array by maintaining the absolute training distance from individual landmarks in the array, rather than adju...

Recognition of static and dynamic images of depth-rotated human faces by pigeons

“…They concluded that the ability of these avian species to recognize equivalence between pictures and real objects is severely limited when a behavioral task requires a relatively precise correspondence. Although there is evidence that pigeons transfer discrimination from objects to pictures (e.g., Cabe, 1976;Delius, 1992;Watanabe, 1993) and that pigeons may have the ability to use 2-D cues to perceive three dimensionality (Cook & Katz, 1999;Reid & Spetch, 1998;Spetch, Kelly, & Lechelt, 1998), little is known about avian recognition of 2-D images of complex natural objects. In Experiments 2 and 3, we explored this issue.…”

Rufous hummingbirds' memory for flower location

“…Indeed, the superior performance in experiment 2 may be the result of increased salience of flower locations owing to the presence of a vertical dimension in the arrays. Fox squirrels, Sciurus niger, show excellent spatial memory in three dimensions when tested in the field (Jacobs & Shiflett 1999), pigeons, Columba livia, appear to attend to three-dimensional relationships when shown multiple two-dimensional views of a scene (Spetch et al 1998) and rats, Rattus norvegicus, in a three-dimensional maze appear to give priority to goal heights when learning goal locations (Grobéty & Schenk 1992).…”

Outdoor View Recognition Based on Landmark Grouping and Logistic Regression

“…This idea continues to be supported by recent research in this field [13,20,33], which encourages its use as inspiration for the development of robot localization algorithms.…”