ASYMPTOTIC LEARNING IN PSYCHOPHYSICAL THEORIES 1
R. Duncan Luce · British Journal of Statistical Psychology · 1964
The major types of models that have been proposed to account for the psychophysical data that are obtained when the stimulus differences are small are described briefly. Because it is clear that contingency variables, such as presentation schedules and payoffs, as well as the physical stimuli affect the response probabilities, recent models have included a trial dependent decision mechanism in addition to the (usually) static sensory one. Such models all appear to be special cases of two very general, but mathematically distinct, families of models, which are formulated in eqns. 1 and 5. Hypotheses about how the subject selects the response‐bias parameters of the decision process are examined. The assumption that they are chosen so as to maximize the expected payoff leads to incorrect predictions for special cases of both families. The alternative hypothesis studied is that the bias parameters are altered from trial to trial on the basis of information feedback according to one or another stochastic learning model. Primary attention is paid to the asymptotic expected values predicted for these parameters. Several such learning processes are described, their relations to static psychophysical models outlined, and their ability to explain data discussed.