Recognition tests of visual information storage.

Using recognition rather than recall test procedures gives results which are inconsistent with the main features of the visual trace model for short-term memory. Recognition efficiency does not deteriorate over time, although rapid decay is postulated by the model. Transposed letter rows are detected as often as rows which reappear at the same position in the test array, suggesting that the storage medium is not directly visual. The number of letters stored appears far less than are presented in the array, but consistent with the perception of a single row. Correspondingly, the response times tend to group into faster reactions, in recognition problems where single row responding is appropriate, and slower reactions which imply that what is reported depends upon decisions based upon inference.