The current set of studies was intended to investigate the possible existence of 2 memorial processes responsible for the acceptance of items on recognition tests. These processes are similar to those suggested by Mandler (1980) and Dallas and Jacoby (1981). There is an initial memory process that determines the overall familiarity of an item. If an item seems familiar it will be accepted as old on the recognition test. If enough evidence cannot be derived to determine if an item is old or new, a second process is engaged that attempts to recall the actual occurrence of the recognition item. If a person can recall the item, it is accepted as old. If the person cannot recall the items, it is rejected as new. 

Rotello and Heit intended to demonstrate that a second process, recall-to-reject, is engaged as more time is allowed for a recognition decision to be made. Recall to reject is analogous to the recollection rejection parameter posited by Fuzzy Trace theory and mathematically estimated by Conjoint Recognition. By both accounts a person is able to retrieve a memory trace of what happened at test. If this memory trace mismatches the recognition test item, that recognition item will be rejected as old (e.g. "I know it wasn't cat because it was dog).

Rotello and Heit make the added assumption that the likelihood of the recall-to-reject process to be engaged will increase as the time allowed to make the memory judgment increases. Based on these assumptions they predict a pattern of responding to related distractor items that differs from both targets and unrelated distractors. It is assumed that early on, related distractor items should be accepted based on the familiarity process, but as more time is allowed for a memory judgment these items should be increasingly subject to rejection, based on the recall-to-reject process. 

Rotello and Heit (1999) found no support for the recall-to-reject process in single item recognition tests. The present set of studies was intended to test for the use of a recall-to-reject process on associative pair recognition tests. 

Experiment 1 

Method and Procedure 

3,600 words were randomly organized into 1350 pairs. The word pairs were then assigned to one of 50 study-test blocks. Each block consisted of 27 pairs of words presented one at a time for 4 sec. 

 Immediately following the presentation of 27 pairs of words was a test list of 27 word pairs. Each test list was composed of 9 intact pairs from the just studied list of word pairs, 9 rearranged pairs, and 9 pairs made up of completely new words.

A response signal technique was used.  Participants were trained to respond to the test item when a response cue sounded even if this involved guessing.  This technique is useful for assessing how much processing has occurred after a specified amount of time.  Six test lags (100, 300, 500, 750, 1,200, and 2,000 msec) were used to cue the participants' memory judgments. Because there were 9 instances of each type of test item there was 1 instance of each test item at each response time. However, there three remaining instances that were assigned to 100-, 500-, and 1,200 msec for even presentation blocks and 300-, 750-, and 2000 msec for the even numbered presentation blocks. 

Results 

Percentage of hits and false alarms were presented in the results but the data points of interest were the D_L score calculated for targets, related distractors and unrelated distractors. D_L is a measure of sensitivity similar to d' ( for a more detailed explanation of the analyses used in this study pease visit the summary for Rotello & Heit (1999)). Therefore, the assumption was that as response time increased there would be a nonlinear trend for D_L corrected acceptances of related distractors to be lower in shorter response lags, increase as a function of response lag and then to decrease as response lags continues to increase. As can be seen in Figure 2 this pattern was obtained for related distractor items. In addition to this result there was the additional finding that these data fit a non-monotonic function better than a monotonic function.

What does this mean? It simply means that the acceptance of related distractor items as measured by D_L fit a non-monotonic function that initially increased and then decreased, supporting the existence of a recall-to-reject process. 

In the second experiment the researchers wanted to illustrate that the associative nature of the recognition test was resulting in this recollect to reject. Rotello and Heit, 1999 did not observe the recall-to-reject process, but a single item recognition test was used in this series of studies. Experiment 2 investigated the potential that a paired associate recognition test brings about the recall-to-reject process not the paired nature of the presentation of stimulus words. 

Experiment 2 

Method and Procedure 

Again a similar method to the one used in experiment 1 was used to present word pairs at study. However, the types of items presented at test were altered slightly. In addition to 7 intact pairs, 7 rearranged pairs, 7 new pairs there were 7 synonyms of studied words, and 7 completely new words. These test items were presented at one of 7 different response lags (100, 300, 500, 750, 1,000, 1,250, and 2,000 msec). 

Results 

Replicating the results of study one, the non-monotonic model fit the associative recall data better than the monotonic model. And replicating the 1999 article, the item recognition data did not fit the non-monotonic model any better than the monotonic model. Synonym distractors for words presented at study did not evoke the recall-to-reject process as the response lag increased in duration. 

Experiment 3 

Method and Procedure 

Participants studied list of unrelated words prior to being each recognition memory test. For half of the recognition memory tests participants were respond positively to words from the first list and to respond negatively to new words and words from list 2. For the remaining half of the recognition tests participants were instructed to respond negatively to new words and words from list 1 and positively to words from list 2. Again there were 7 different response lags (100, 300, 500, 750, 1,000, 1,250, and 2,000). 

Results 

D_L scores were computed for target list items (items from the lists that were to be responded positively to), non-target list (items from the lists that were to be responded negatively to) and new items (non-presented distractor items). Again, the non-monotonic function fit the data provided for the non-target list items better than the monotonic function. 

Discussion 

These researchers feel that the results of these three studies illustrate that the recall-to-reject process functions with associative recognition test, but not item recognition test. In addition, they believe that this explanation of the recall-to-reject is the appropriate secondary process, opposed to a recall-to-accept process. Their data were not able to support such a recall-to accept process. In addition they, acknowledge the possibility that a single process model could account for these data but these researchers advance the notion of the recall-to-reject process.