Signal detection theory: Considerations for general application.

Abstract

While there exist a number of papers describing the theory of signal detection, it appears that many psychologists are not aware of the ease with which signal detection theory can be applied, the range of applications possible, or the limitations of signal detection theory. This paper briefly summarizes the assumptions of signal detection theory and describes the procedures, the limitations, and practical considerations relevant to its application. A worked example of an application of signal detection theory to the study of cognitive processes is included.

Using the Theory of Signal Detection to Improve Ad Recognition Testing

“…Therefore, given equal memory across two subjects for the ads in a certain recognition test, a subject with B'= -20% is twice as prone to say "yes" on a given recognition test as a subject with B'H = -10%. Funher, though B'H is computed across all ads presented in a test session, it can be used to ascertain recognition scores adjusted for response tendencies for individual ads because it is reasonable to assume that the decision criterion (B'H) employed by a single subject in a given measurement is stable (Pastore and Scheirer 1974), that is, the response bias remains constant.…”

“…measurement error or (2) the subject is performing the discrimination task and then giving a contrary response, that is, saying "no" when he or she should have said "yes" (and vice versa) on the basis of discrimination ability. In other words, the subject is able to perform the discrimination task, but is malingering (Pastore and Scheirer 1974).…”

“…Because of this complementary relationship between the responses, only two responses customarily are used to summarize a subject's performance-the hit and false alarm ratios. The performance of a subject in the recognition test depends mainly on two factors, the subject's ability to perform the task and the subject's motivational state and response tendencies (Pastore and Scheirer 1974). The experimenter can affect the subject's response tendencies and motivations by changing payoffs and/or by changing prior odds.…”

“…When the assumptions do not hold, other statistics with meanings similar to d' and P are calculated from a subject's responses. For the description of the procedures for testing the assumptions and the statistics that should be used when either or both assumptions do not hold, see Pastore and Scheirer (1974). For a graphic demonstration of the independence of d' and p, see Klatzky (1980, p. 249-51).…”

“…Pastore and Scheirer (1974) suggest it is reasonable to assume that the criterion used by a single subject during any measurement session (block of trials) is stable. In other words, in a given set of trials, the response bias should remain stable.…”