Abstract
Previous research has demonstrated that it is difficult for participants to incidentally learn and use statistical information about one feature of an object (e.g., color) to detect a change in another feature (e.g., shape; Beck et al., 2008). For example, if a red object is most likely to change shape, participants are unable to incidentally learn and use this color probability information to become better at detecting shape changes to red objects. On the other hand, participants are able to incidentally learn and use statistical information about location information to detect feature changes. For example, if a color change is most likely to occur in the far left column of a grid, participants are able to use this information to improve color change detection in that location. In the current set of experiments, we tested what conditions are required for participants to learn and use statistical information about features (color and shape) to improve change detection performance. Contrary to previous research, we found that participants are able to learn and use shape probabilities to detect color changes and color probabilities to detect shape changes, but only when all of the features of the objects on a display are unique (no repetition of shape or color in a single array; Experiment 1). However, if the objects in the array are not unique (e.g., two objects of the same shape or color may appear in a single array), participants are unable to learn and use color probabilities to detect shape changes, or vise versa (Experiment 2). This inability to incidentally learn and use statistical regularities when there is feature repetition in an array is likely not due to intertrial noise (Experiment 3) or chunking strategies (Experiment 4).
Meeting abstract presented at VSS 2012