Learning about environmental geometry: An associative model.

Abstract

K. Cheng (1986) suggested that learning the geometry of enclosing surfaces takes place in a geometric module blind to other spatial information. Failures to find blocking or overshadowing of geometry learning by features near a goal seem consistent with this view. The authors present an operant model in which learning spatial features competes with geometry learning, as in the Rescorla-Wagner model. Relative total associative strength of cues at a location determines choice of that location and thus the frequencies of reward paired with each cue. The model shows how competitive learning of local features and geometry can appear to result in potentiation, blocking, or independence, depending on enclosure shape and kind of features. The model reproduces numerous findings from dry arenas and water mazes.

Absence of overshadowing between a landmark and geometric cues in a distinctively shaped environment: A test of Miller and Shettleworth (2007).

“…This prediction was verified with a computer simulation based on Equation 2, with λ set at 100 and with the remaining parameter values the same as those recommended by Miller and Shettleworth (2007). The left--hand panel of Figure 2 shows the predicted changes in associative strength as training progresses for the landmark and for the geometric cues at the correct and incorrect corners.…”

“…An alternative explanation for the failure to observe overshadowing when a landmark is situated near a goal in an environment with a distinctive shape has been proposed by Miller and Shettleworth (2007). This account is based on the supposition that learning about geometric cues adheres to the same principles that apply to all other cues.…”

“…Approaches to the apex were assumed to reduce the associative strength of A and X, approaches to the incorrect corner were assumed to reduce the associative strength of I and X, and approaches to the correct corner were assumed to increase the associative strength of C and X, and L if it was present. To predict behavior during a trial, Miller and Shettleworth (2007) suggested that the probability of a rat heading for a particular corner after being released into the pool is determined by the overall associative strength of the corner in question, divided by the sum of the associative strengths of all three corners. The probability of the experimental group heading for the correct corner is expressed in Equation 1, where each pair of brackets in the denominator expresses the overall associative strength for each of the three corners, and V represents the associative strength of the cue identified by the subscript.…”

“…The model has also been used with some success to explain why the control by geometric cue in a kite--shaped environment is potentiated by making the color of the walls of the correct corner different to the color of the other walls , but see Horne and Pearce (in press) for an alternative explanation for this finding. The purpose of the present paper is to test further the model of Miller and Shettleworth (2007) by evaluating two specific predictions that it makes.…”

“…The right-hand panel of Figure 2 shows the predicted effects of conducting test trials in the triangular arena without a landmark after increasing amounts of Figure 2. The results from computer simulations based on the associative model of geometry learning by Miller and Shettleworth (2007). Left-hand panel: The predicted acquisition of associative strength by some of the cues in the triangular arena for the overshadowing, and the one-(1-LM) and two-landmark (2-LM) control groups that were used in the experiments.…”