Invariant recognition memory spaces for real-world objects revealed with signal-detection analysis

Recognition memory refers to distinguishing between previously experienced and novel events. Apart from the objective quality of stored memories, recognition depends on the retrieval context produced by all items (foils) presented together with actually memorized items (targets) and causing confusion. Memory models often conceptualize target-foil confusability via distances in psychological spaces where greater confusability originates from shorter inter-item distances. Here, we test whether recognition spaces change when other foils are added to the retrieval context or target memory strength is changed. Using signal-detection modeling, we found that separately measured distances, d'-s, from each individual foil to the target provide a good linear prediction of those distances for all foils being presented together against that target. Those predictions stay accurate even when the absolute distances are scaled up or down due to memory strength change. This suggests strong metric invariance of spaces used for recognition decisions under variable retrieval contexts.