An Update of Findings Regarding Vigilance and a Reconsideration of Underlying Mechanisms

Abstract

According to a signal detection analysis, the behaviour of an operator can be considered in terms of sensitivity to signal/noise differences d' and by the criteria for responding as if it is a signal - that is the ammount of positive evidence needed to make a response, or the readiness to make a response at a given level of evidence b . Some studies hae indicated that the percentage of true signals detected by the observer is directly related to the probability of signal and non-signal occurrance (Baddeley). Further more the effects of prior expectation of target probability was examined (Baddeley and Col-)> when an expectation of a high target probability was given in practice and was then followed by a lower target probability in the test period, the criteria for making a response b increased. In the reverse case, B decreased. By this reasoning, assuming that an uninstructed subject may carry the assumption to an experiment that there will be an equal number of targets and non-targets, and the actual probability is lower than this, B should decline - that is that the subject will begin to miss more targets as the experience with the task goes on. You will see a vigilance decrement. Baddeley 1969) went on to argue therefore that if this expectation was modified, then no such decrement in performance would occur. ** It is important to distinguish, therefore, between the types of processes we are talking about. Apparent perceptual failures to respond to targets in this type of framework are therefore primarily determined by a decision making strategy underconditions of ambiguity between targets and non-target occurences. This must surely be quite different from the kind of readiness to respond argument, that is to process priority signals rapidly after their occurence** Vickers et al conducted an experiment in which the probability of a target occurance was systematically decreased over some 600 trials (on a block of 100 trials basis - from .5 to 0.05). In contrast to traditional vigilance tasks they found that the process of adjustment produced two effects. At the initial decrease in objective probability subjects increased their readiness to respond positively so detection and false positive rates increased. As information accumulated over subsequent trials, this bias was reduced downwards to more accuratley reflect the actual probabilities. In Vickers et al's analysis of the usual relationship in vilance tasks, the vigilance decrement may therefore reflect this scaling down of an initially over- positive bias on the first 'few' trials - based on prior expectation of equal signal/non-signal probability.