Changing change detection: improving the reliability of measures of visual short-term memory capacity.

The change detection paradigm is a popular way of measuring visual short-term memory capacity. Using the paradigm, researchers have found evidence for a capacity of about four independent visual objects, confirming classic estimates that were based on the number of items that could be reported. Here, we determine the reliability of capacity measures found by change detection. We derive theoretical predictions of the variance of the capacity estimates and show how they depend on the number of items to be remembered and the guessing strategy of the observer. We compare the theoretically derived variance to the variance estimated over repeated blocks of trials with the same observer and find close correspondence between predicted and observed variances. Also, we propose a new version of the two-alternative choice change detection paradigm, in which the choice is unforced. This new paradigm reduces the variance of the capacity estimate substantially.