The response signal method of Reed (1973) was used to study the time-course of list membership recognition after 2 sec of uncontrolled rehearsal, with lists of one, two, and four consonants. Fourteen specific hypotheses about the time course of this process were derived from various theories (Anderson, 1973; Anderson & Bower, 1973; Baddeley & Ecob, 1973; Corballis, Kirby, & Miller, 1972; Kirsner, 1972; Murdock, 1971; Sternberg, 1966, 1969; Theios, Smith, Haviland, Traupmann, & Moy, 1973; etc.) and additional assumptions about the effect of the response signal. When members of the to-be-learned lists are drawn from a small population of highly confusable items, as in the current experiment, list membership recognition appears to follow the model of Theios et al. (1973). Latency functions of signal lag appear to be particularly useful in differentiating among hypotheses which predict similar speed-accuracy tradeoff functions. The lag by positive-negative interaction for latencies of correct responses is highly significant for lists of one consonant, a result predicted by the hypothesis derived from the model of Theios et al. and incompatible with hypotheses derived from exhaustive search and single-threshold strength models.