Like the Baddeley theory, the ACT-R theory claims that timing of the recall is important to memory performance and that with the passage of time memory chunks can decay to the point where they are no longer available for recall. However, in addition it claims that the activations of the memory chunks have a strong influence on the timing of recall (Retrieval Time Equation 3.10). Thus, there is a feedback loop between timing and activations with higher activations yielding shorter retrieval times and shorter times yielding higher activations. Since there has not been research that has delved into this interaction between timing and recall, Anderson, Bothell, Lebiere, & Matessa (in press) performed an experiment that focused on this issue. They looked at memory span for digits, presented at the typical rate of 1 per second. They varied list lengths from 3 to 12 digits to get a good range of performance and measured both the timing and the accuracy of recall. They also looked at both forward and backward recall to manipulate the delay between presentation and recall. In forward recall the first-presented digits are recalled first while in backward recall the last-presented digits are recalled first. This creates very different delays for the recall of digits that are studied in the same serial position. They also controlled the grouping of the digits by presenting them visually segregated into units. All the groups were of length 3 except for the last which was varied from 2 to 4 to accommodate various list lengths.
A series of boxes would appear on the screen-one box for each group. The number of items that would appear in each box was made obvious by the number of spaces in the box. Thus, the subject knew immediately the number of and structure of the items to be studied. However, while they were studying them they did not know whether they would be tested in the forward or the backward condition. This was done to test the same memory structure in different orders. The items were presented one at a time in the appropriate spaces within the appropriate boxes. When one digit appeared, the other disappeared and so only one digit was visible at a time. As soon as the last digit disappeared a signal appeared telling subjects the direction in which to recall the digits. The cursor would either move to the first slot in the first box (for forward recall) or the last slot in the last box (for backward recall). As each digit was typed the cursor moved to the next slot (or previous slot in the case of backward recall). The subject could skip over positions by typing a space or terminate the recall by hitting the return key. However, the subject could not back up and change recall of a digit.