Hi, I'm currently working on some data from an associative-memory study. Participants have to learn associations between stimuli from two different classes (1 and 2). As stimulus from class 1 is presented participants have to select the corresponding stimulus from class 2 (trough sliding). After response selection participants have to rate their confidence (binary, 'low confidence', 'high confidence') concerning their response. After the confidence rating feedback ('true', 'false') and the correct response (the correct matching of S1 and S2) are provided. Participants are instructed to guess during the first block. There are four retrieval sessions (including the first block).
I would like to perform an analysis regarding the metacognitive capabilities of the participants. Therefore I would like to apply a Signal Detection Theory (SDT) approach:
Hits: high confidence correct responses
False alarms: high confidence incorrect responses
Correct rejections: low confidence incorrect responses
Misses: low confidence correct responses
Since meta d' is affected by the first-order performance it is wise to quantify metacognitive efficiency which is the ratio of meta d' and d'. And here is the problem: I can't apply SDT for the first order task (which is the pure memory performance without the confidence ratings). The only score I can get is the recall rate.
I would be very grateful if you could share your ideas on this. Thanks!
Update: There are 12 pairs to memorize during each run. So everytime stimuls from class 1 is presented there are 12 stimuli from class 2 from which one is the correct one. Is it valid to say that everytime a participant is selecting the correct answer he/she also rejected the remaining 11? This would mean that hit rate = correct rejection rate and false alarm rate = miss rate.
- Badges
- Science topic
- Similar topics
- Psychology
- Mental Processes
- Learning