Take a 2AFC task, where people also give a measure of confidence. In the framework of signal detection theory, the criterion values for each confidence level are simply additional criteria, placed on the same internal response axis as the (first order task) decision criterion.
Can I estimate the values of these additional confidence criteria from the data of a subject? If so, how? Would I simply take all trials with (say) confidence = 3, and calculate the criterion based on those trials alone?
There's actually a very simple answer to this question: calculate criterion of confidence level = 3 by counting as hits only those trials where the answer was correct (and the target was there), and confidence was equal or higher than 3.
This is also what Rahnev et al did in the following paper: Rahnev, D., Koizumi, A., McCurdy, L. Y., D’Esposito, M., & Lau, H. (2015). Confidence Leak in Perceptual Decision Making. Psychological Science, 0956797615595037. http://doi.org/10.1177/0956797615595037
