A study of judgements of relative likelihood

Abstract

This thesis is a study of the behaviour of a person using the terms 'likelihood', 'certainty','confidence', etc (collectively referred to as 'likelihood judgements’), in describing event s of uncertain outcome. The study is a psychophysical one in that, given that people may be observed to make statements involving these terms, consideration is given to the way these statements are determined by characteristics of the outcome judged. In general, the question of the meaning of the terms to the subject, is avoided and no reference is made to other forms of behaviour with respect to uncertain outcomes. In the first place, the behaviour of a person judging the likelihood of an isolated outcome is discussed. Consideration is given to methods that have been employed to 'scale' the judged likelihood of an outcome. Evidence is considered that the value of an outcome on the judged likelihood scale, is positively related to its mathematical probability. Consideration is then given to the practice of describing likelihood judgements in terms of 'psychological probability'. It is suggested that this usage is based on the practice in psychophysics, of describing a stimulus dimension and its judgemental correlate, as ’objective’ and ‘subjective’ aspects of the same dimension. In this case, psychological and mathematical probabilities would be best regarded as incommensurable. Consideration is then given to the behaviour of a person when asked to judge which of two uncertain outcomes is the more likely to occur. A number of terms and concepts are introduced for the description of patterns of such judgements. Consideration is given to the consequences of the occurrence of intransitivities in relative likelihood judgements. Finally, experimental evidence is reported, consistant with the conclusion that, for at least some people, such intransitivities do occur.