Two experimental results often reported in support of perceptual interpretations concerning the influence of set on perception are critically examined: (a) the relation between word frequency and recognition threshold, and (b) the so-called compromise reactions between set and stimulus, Alter elimination of certain methodological artifacts (e.g., introduction of a temporal forced-choice method instead of the ascending-limits method), both phenomena disappear; the influence of set on perception appears to be wholly a matter of response bias. Howes and Solomon (1951) and Solomon and Postman (1952) found high negative correlations between frequency and recognition threshold of English words and nonsense syllables. Bruner and Postman (1949) obtained much higher thresholds for incongruous playing cards (e.g., red spades) than for normal cards (e.g., black spades).Since the study of Goldiamond and Hawkins (1958) the explanation of these phenomena in terms of response probabilities has been widely accepted. The latter interpretation, however, does not rule out the possibility of genuine perceptual effects resulting from implicit sets (Dember, 1960). Zajonc and Nieuwenhuyse (1964) made a further test of the response bias hypothesis by introducing very stringent recognition criteria. In this experiment the negative correlation between frequency and recognition threshold decreased with increasing stringency of the criterion used (viz., pseudorecognition, one correct identification and three successive correct identifications). Nevertheless, even in the most stringent criterion condition a very reliable effect of word frequency was found.These results are complemented by the so-called compromise responses in the playing cards experiment by Bruner and Postman (1949). For instance, after presentation of (incongruous) red spades many Ss gave prerecognition answers such as "brown spades," "purple," "rusty black," "black with red edges," i.e., names of colors that might be considered mixtures of red and black.These two experimental results are repeatedly quoted in support of perceptual theories concerning set-perception interactions. In the present study it is hypothesized that they must be attributed to certain methodological artifacts.EXPERIMENT I The word frequency effect was supposed to disappear after the introduction of the following two procedural modifications: (a) a temporal forced-choice method for measuring recognition thresholds, and (b) a very strict recognition criterion (three successive correct identifications). ProcedureNonsense syllables (CVCs) of equal meaningfulness were used as stimuli. In each trial of the pre training phase, E showed one syllable to the Ss (30 undergraduate students) while asking first to spell and thereafter to pronounce it. Frequency was varied by presenting some syllables 25, 10, 5, and 2 times, respectively. For each frequency there were two syllables. Counterbalancing these syllables against frequency was not considered essential, first, since their selection was restricted by the r...