ABSTRACT
Using these statistics for more fine-grained analyses of operator performance also leads to a plausible and cohesive account of human behavior. For example, studies have consistently shown that the operator shifts the detection criterion when the base rates are changed, but the size of the shift is presumably insufficient to cause the decision process to be “optimal” (i.e. to maximize the percentage of correct decisions). In other words, the decision process is “conservative” (e.g. Creelman and Donaldson 1968; Macmillan and Creelman 1990). In laboratory studies of watchkeeping (i.e. the vigilance paradigm), the measures have been used to “establish” that changes in the detection rate are due to changes in the operator’s willingness to make a detection response under some conditions (e.g. relatively slow paced detection tasks) and losses of sensitivity under others (e.g. relatively fast paced detection tasks with a memory load; Parasuraman 1979). These accounts are interesting and plausible, and certainly do not raise any special concerns about the validity of the model that gives rise to them, even though it could easily have been otherwise. If the model was untenable, one might expect a more inconsistent or a more confusing pattern of results.
