A Note on the Measurement of Primary Memory Capacity

It is argued that the modification of the traditional Waugh and Norman (1965) method for the estimation of primary-memory capacity proposed by Watkins (1979) is not consistent with the assumptions of the modal two-store model. An alternative method is described that does not incorporate the assumption, implicit in the method proposed by Watkins (1979), of independence between the prob- abilities of recall from primary and secondary memory. It is shown that the most general version of the new method is prone to error because the number of parameters to be estimated equals the number of data points. Two simplified solutions are described, one based on a geometric decay function for the prob- ability of recall from primary memory and one based on the general buffer model described by Phillips, Shiffrin, and Atkinson (1967). It is shown that this method will give a valid estimate of primary-memory capacity; that is, when applied to simulated data with a known buffer capacity, it will generate an estimate of primary-memory capacity that corresponds quite closely to the known value.