Abstract
In typical recognition memory tests, participants memorize a number of items and then get tested by targets presented together with foils. On each test trial, participants have to make a single decision as to whether the target is there (yes-no task) or which one is the target (n-AFC task). However, the use of a single decision on each trial can hide some important information and can favor binary thinking about recognition as either remembering or guessing. Wu and Wolfe (2018) recently suggested a method to measure spatial memory capacity in which participants click on test locations until they find a target. Here, we used this method to study visual long-term memory for objects. Our participants (N = 35) memorized over 300 target pictures. In test trials, each of these targets was tested against 1, 3 or 7 of new foils. The participants were instructed to find the target within as few clicks as possible. We analyzed the probability distribution of finding a target as a function of click’s serial order. To this end, we implemented a simple signal detection model. We calculated the sensitivity measure d’ based on the probability of the first click being correct (given the number of alternatives). We then simulated multi-click recognition as a noisy familiarity ranking algorithm. On each simulation, we picked N random samples from the noise distribution (mean=0) for N foils and one sample from the signal distribution (mean=d’) for the target. All samples were put together and ranked by magnitude (familiarity). The rank of the target sample defined how many clicks an observer would make before finding the target. Overall, this model showed very good fit to the data (r = .97). We conclude, therefore, that a single parameter of continuous memory strength is sufficient to account for multi-alternative and multi-try recognition.