Landmark Averaging

In searching for a goal, pigeons often follow the dictates of more than one cue. The dictates of different cues are averaged in guiding searching behavior. Some examples from studies conducted in open laboratory rooms and touch-screen monitors are presented below to illustrate this point.

Open Field Tests in Laboratory Rooms

Training Phase

In these tests, the pigeons were trained and tested in a square arena in a lab room (Cheng, 1988).

The to-be-shifted landmark, a blue stripe, is shown above in its training position.

Transformation Test Phase

Performance on Control Tests. The following shows a video clip from overhead of a pigeon performing a control test using the above training setup. The target (food) is usually near where the bird is pecking in the picture, in front of the blue stripe on the back wall (light grey in the black and white video). But on this test, there is no reward to be found. Nevertheless, the bird searches in the right place on such tests.

Performance on Landmark Shift Test. Following this training, a test was conducted in which the blue stripe has been shifted 30 cm to the left. The pigeon's searching also shifts to the left, but not by the full 30 cm, suggesting the pigeon averages the dictates of both shifted and unshifted landmarks. Click on the picture for a short video clip of the bird's search during such a test. The markers at the beginning indicate the theoretical target locations according to the shifted stripe (left) and unshifted landmarks (right). Other experiments on landmark-based search have also found averaging (e.g., Cheng, 1989). When one landmark is shifted on a test, pigeons often follow its dictates only part of the way. In fact, pigeons (and humans) can average quite disparate dimensions of experience. Cheng, Spetch, and Miceli (1996) found that pigeons and humans averaged the dictates of elapsed temporal duration with the spatial position of a moving landmark.

Experiments with Touchscreens in Operant Chambers

Training Phase

In these experiments, pigeons pecked at a monitor screen in a computerized operant chamber (Spetch, Cheng, & Mondloch, 1992). The target location was near the top edge of the screen. A graphic landmark (rectangle in blue) was near the target. After preliminary training, the red target was made invisible.

Transformation Test Phase

On subsequently unrewarded tests, this landmark was shifted left (blue datapoints), remained where it usually was (green datapoints), or was shifted right (red datapoints). The extent of the horizontal landmark shifts are indicated by the distance between two Ts. The results showed that the pigeons shifted their peak place of searching with the horizontal landmark shifts, but not to the full extent of the landmark shifts. Thus, they were averaging the dictates of the shifted landmarks and the rest of the unshifted landmarks. The pattern of results resembles what was found in the search task in the tests in open-field arena (Cheng, 1988, 1989; Cheng & Sherry, 1992)