Robinson, M.D., Johnson, J.T. & Robertson, D.A. (2000). Process versus content in eyewitness metamemory monitoring. Journal of Experimental Psychology: Applied, 6, 207-221.

This article builds off of past research on the accuracy-confidence (AC) relationship in eyewitness testimony.  Past research has found that the accuracy-confidence relationship is weak, although more recently some have challenged this conclusion

Past research by Robinson and Johnson has shown that the AC relationship is stronger in recall than it is in recognition.  In the present research the authors wanted to understand whether process factors (e.g. processing speed) or content factors (e.g. memorial quality, vividness) were better predictors of the AC relationship.

Studies 1 and 2
Method

Subjects watched a shorted video depicting a theft.  Twenty-four hours later (study 1) or 5 minutes later (study 2) subjects returned and were tested on their memory for the film.

Subjects in Study 1 answered either a four alternative forced choice test or a cued recall test.  Subjects in Study 2 answered either a four alternative forced choice test, an old/new recognition test in which 25% of the items were "old", an old/new recognition test in which 50% of the items were "old", or a cued recall test.

Additional Measures included:

Confidence on a 9 point scale
Response time (process measure)
Subject's estimate of their response time (process measure)
Subject's estimate of their effort (process measure)
Memorial Vividness (content measure)
Results

As in their previous research, the AC relationship was stronger in recall than in recognition.  This was both because recall subjects were more confident in their accurate answers and because recall subjects were less confident in their inaccurate measures.

Memorial Vividness was a better predictor of confidence than were any of the process measures.  With one exception, memorial vivideness was a better predictor of accuracy than any of the content measures.

Moreover, in a survey study they report, subjects reported that they are more likely to use vividness as a cue to memory accuracy than any of the process measures.  This suggests that people have a naive theory of memory accuracy.

They argue that vividness is a a better cue to accuracy in recall than in recognition and that this is why the AC relationship is stronger  in recall than in recognition.

To establish this mediational role the authors sought to establish (1) the IV (recall vs. recognition) predicts the mediator (vividness-accuracy) (2) the mediator (vividness-accuracy) predicts the DV (confidence-accuracy) and (3) the mediator still predicts the DV even when the IV is controlled for.

All three of these criterion were met, suggesting that the strength vividness-accuracy correlation predicts the differences between recall and recognition.

Experiments 3A and 3B
The purpose of these experiments was to make process variables (i.e. reaction time) more salient to individuals.  In Experiment 3A subjects were asked to estimate their retrieval time right before making their confidence judgments (reaction time salience).  In Experiment 3B subjects estimated their reaction times and then received feedback about whether their estimates were correct.  Some subjects received accurate feedback and some inaccurate feedback.

Basically, the RT salience and RT feedback manipulations had no effect on confidence or accuracy.

Discussion and What Not
 
First off, these results confirm Robinson & Johnson's earlier work showing that the AC relationship is stronger in recall than in recognition.  One reason for this may be that base rate accuracy is higher in recognition resulting in low confidence accurate responses.  But this doesn't explain why recognition also resulted in a number of high confidence inaccurate responses (comment: although wouldn't this primarily be a function of how one selects distractors?)

Second, the authors point out that although much eyewitness testimony research uses recognition, it is probably the case that recall memory is as important or more important than recognition memory in forensic settings.

Third, the authors point out that some theories of meta-memory argue for fluency as being and important determinant of confidence and that others argue for the importance of experiential details.  With regard to the AC relationship the authors argue that it is experiential detail that provides the best account of the AC relationship and that fluency has very little role.

University of 
Arkansas
Department of 
Psychology
Graduate Program 
in Experimental Psychology
Lampinen 
Lab
False Memory 
Reading Group
False Memory Reading 
Group Summer 2001