Martin & Jones (1979) suggested a modification of Waugh & Norman's (1965) procedure for measuring primary memory, one that would take into account negative recency in secondary memory. Specifically, they proposed that secondary memory representation could be estimated from a function selected as approximating negative recency effects observed with delayed recall procedures. It is argued here that there are problems with this particular approach, and an alternative solution is proposed. This solution follows directly from predictions of the two-store model underlying the Waugh & Norman procedure. It is illustrated with an example.