Behavioral Detection Theory: Some Implications for Applied Human Research
Dianne C. McCarthy University of Auckland
Since its introduction some 25 years or so ago, signal-detection methodology has enjoyed wide application in many areas other than its psychophysical origins. For example, its impact has been felt in such diverse realms as vigilance situations, personality differences, memory, animal learning, clinical psychology, military target detection, weather forecasting, reaction times, subliminal perception, risk taking, medical decision making, and quality control in industry, to name but a few.
The practical advantages of the theoretical independence of its two measures -- discriminability and response criterion -- have been largely responsible for the popularity of signal-detection theory (SDT) over the years. But nowhere is the importance of its application more apparent than in the assessment of the detection performance of subjects suffering various organic insults (such as closed-head injuries, schizophrenia, and epilepsy). Clearly, some of these insults would be expected to affect only one of these measures; for example, damage to a sense organ would presumably change the discriminability measure but leave the response-criterion measure unchanged. SDT has been applied in these areas but since the necessary independence of the traditional SDT measures has been questioned ( McCarthy & Davison, 1984), doubt can be cast on the meaning of the results obtained using these indexes in many applied human areas.
The signal-detection procedure necessarily involves choice between two responses "A" or "B," "Yes" or "No," "Respond" or "Don't respond," or other pairs of well-defined, mutually exclusive, alternatives (Fig. 11. 1). Davison and his coworkers have shown how an empirically based quantitative formulation of choice provides a framework into which the effects of