Detection-theoretic analysis of the observer-based psychophysical procedure.

The observer-based procedure, used in research applications to measure the hearing sensitivity of young infants, is analyzed within the framework of a two-stage (infant-judge) detection model involving a minimum number of assumptions. The model is taken to develop convergence theorems useful for estimating the number of judges and/or trials required to achieve a desired level of accuracy using the procedure. The model is also used to consider ways in which bias in estimates, known to be associated with the procedure, might be evaluated and reduced. Finally, a method is proposed by which the results of the analysis and different procedural variations designed to improve estimates can be evaluated empirically.