Acoustic similarity and interference in short-term recognition memory: Some results of a decision-theory analysis

The following three studies of single-probe recognition memory set out to show the effect on the signal-detectability measures of d' and β (Tanner and Swets, 1954) of variations in the acoustic similarity of interfering material, which may either precede or follow the item to be remembered (proactive or retroactive interference --PI or RI). The first experiment studies a situation employed by Wickelgren (1966a), who reported that acoustically similar RI substantially reduced d'. It is shown that this effect could have been due to biases in Wickelgren's original designs, and that when a bias-free design is used, the fall in d' is only of borderline significance.

To investigate this problem further, a design was evolved in which two items were presented for memorizing, which varied in acoustic similarity to each other, and (after a distracting task) a probe was presented with one of three questions: Was this the first item of the pair? Was it the second? or, Did it occur in either position? In the first case, recognition-memory with RI of varying acoustic similarity was being studied, and as in the first experiment, it was found that similarity slightly reduced d'. With the second question, PI effects were being studied, and here negligible differences were found. With the third type of question, a “location-free” test, no effects of similarity were found. The last result rules out Posner's (1967) “acid-bath” explanation of similarity effects in interference: an explanation in terms of “differentiation” (or “filtering”) was also invalidated by the results of a third experiment, in which the same effects were found even though similarity varied only between stimulus items and interference, and not between these and the probe. Wickelgren's (1966b) associative model appears to have least difficulty in accommodating these results, though even this needs certain ad hoc assumptions to be able to do so.