Busey, T. A., Tunnicliff, J., Loftus, G. R., & Loftus, E. F. (2000). Accounts of the confidence-accuracy relation in recognition memory. Psychonomic Bulletin & Review, 7 (1), 26-48.

Purpose of the present line of research:

To determine if the underling processes responsible for a person making an accurate or inaccurate selection of an item on an old/new recognition test are inherently the same as those used to determine the confidence at which those decisions are made.

Specifically the authors wish to determine if a single memory trace is used to make both recognition judgements and confidence judgements.

There are two types of confidence judgements that can be made:

  1. Prospective confidence judgements are given at the time of learning and a person is reporting how confident they will be to make a memory judgement based on what they have learned. (These types of confidence ratings are often referred to as Judgment of Learning (JOL))
  2. Retrospective confidence judgements are given after all memory judgements have been made on an old/new recognition test.
General ways of looking at confidence and memory judgements:
  1. Trace Access theory (Burke, MacKay, Worthely, & Wade, 1991): Memory judgements on old/new recognition tests and confidence rating for these memory judgements (both prospective and retrospective) are all based on one underling memory trace for a presented item.

    2. This hypothesis predicts that as accuracy increases so should the level of confidence for those memory judgements. So accuracy and confidence judgements should be positively correlated with one another.

  2. Accessibility hypothesis (Koriat, 1995) – the ability to retrieve information for a memory judgement determines the confidence rating given to a memory judgement. Cues that allow for easier access to memory information will lead to increased confidence. These cues can be intrinsic cues (those related to the processing of the target item) and or extrinsic cues (those related to the study conditions)
  3. Multidimensional model of recognition and confidence
Propose that two processes are operating, which differentially influence confidence and recognition. There are two variables which feed into the system for these studies P = duration and R = rehearsal. The strength of the memory trace is determined by both P and R. Certainty is only determined by the amount of R. Accuracy on recognition test is only determined by the strength of a memory trace. The amount of confidence given to that memory judgment is based on both the strength of the memory trace and the amount of certainty that a person feels. A simple way of looking at this is to assume that the amount of rehearsal has double the impact on confidence ratings, because rehearsal influenced both the strength of the memory traces and the certainty that a felt that the traces were encoded properly.

To test this multidimensional model the authors vary two independent variables that they assume will dissociate the results of confidence and accuracy. The two independent variables are the amount of time that a person has to rehearse a given target item at acquisition and either the duration a target item is presented at acquisition or the luminance at which an item is presented at acquisition. In study three the luminance of the items given on the recognition test is also manipulated in addition to the amount of time given for rehearsal and the luminance of the item at acquisition.

 Basic experimental design:

120 pictures of bald men were presented in grayscale on a computer monitor. Five levels of exposure duration (230 to 930 msec in logarithmically equal steps) . There were two levels of the rehearsal variable (for 15 secs after presentation of the stimulus participants either, silently rehearsed the face or they silently did math problems).

Participants viewed 30 faces and then took an old/new recognition test of 60 faces. Immediately after the recognition test, another 30 faces were displayed followed by another 60 item old/new recognition test.

Procedure

View stimulus pictures – rehearse or do math problems – make a prospective confidence judgement

Results of Study one and two

Study 1 participants had two independent variables (rehearsal or no rehearsal – duration of stimuli at acquisition)

Study 2 participants had two independent variables (rehearsal or no rehearsal – level of luminance of stimuli at acquisition)

The results of study 1 and 2 showed that the prospective confidence judgements by accuracy plots dissociated for rehearsal and no rehearsal. It appears that knowing that you rehearsed the information influences prospective confidence judgements more than it helps with subsequent recognition judgements. Levels of retrospective confidence did not differ for amount of rehearsal. So from these first two studies it appears that levels of prospective confidence are based on a multidimensional model but not levels of retrospective confidence.

Study 3

Study three had three independent variables (rehearsal, luminance at study and luminance at test).

Results of study 3 showed that levels of prospective confidence dissociated based on if a person was allowed to rehearse an item at acquisition replicating studies 1 and 2. Levels of retrospective confidence were influenced by the brightness of the recognition items at test. Levels of retrospective confidence were higher for those items that were presented at test at higher levels of luminance. However, accuracy was best when target items were presented at the same luminance at both acquisition and test.

The results of these three studies show that accuracy and confidence for face recognition can be dissociated. Participants will place more importance on cues that do not aid their level of accuracy as much as they expect. This allows for levels of both prospective and retrospective confidence to be dissociated from accuracy on old/new recognition tests.

 


 
University of Arkansas
Department of Psychology
Lampinen Lab
False Memory Reading Group
False Memory Reading Group Fall 2000