Failure to demonstrate short-cutting in a replication and extension of Tolman et al.’s spatial learning experiment with humans

Successful demonstrations of novel short-cut taking by animals, including humans, are open to interpretation in terms of learning that is not necessarily spatial. A classic example is that of Tolman, Ritchie, and Kalish (1946) who allowed rats to repeat a sequence of turns through the corridors of a maze to locate a food reward. When the entrance to the corridors was subsequently blocked and alternative corridors were made available, rats successfully selected the corridor leading most directly to the food location. However, the presence of a distinctive light above the goal, in both the training and testing phases, means that approach to the light as a beacon could have been the source of successful short-cutting. We report a replication of the experimental design of Tolman et al. with human participants who explored geometrically equivalent virtual environments. An experimental group, who followed the original procedure in the absence of any distinctive cues proximal to the goal, did not select the corridor which led most directly to the goal. A control group, who experienced a light above the goal during training and testing, were more likely to select a corridor which led in the direction of the goal. A second control group experienced the light above the goal during training, but in the test the location of this cue was shifted by 90° with respect to the start point of exploration. This latter group responded unsystematically in the test, neither selecting a corridor leading to the original goal location, nor one leading directly to the relocated light cue. The results do not support the hypothesis that travelling a multi-path route automatically leads to an integrated cognitive representation of that route. The data offer support for the importance of local cues which can serve as beacons to indicate the location of a goal.