Processing of deviant information in inference and descriptive tasks with simultaneous and serial presentation

Abstract

The weighting of deviant scores was studied in simultaneous and serial number-averaging tasks. In the Descriptive Task subjects were instructed to estimate the mean of each set of numbers; in the Inference Task subjects were instructed to use sample information to infer a population mean; and in the Discounting Task subjects were instructed to disregard a deviant or unrepresentative number when inferring a population mean. Results for the Discounting Task showed that a deviant number within a sample was discounted with either simultaneous or serial presentation. In the Descriptive Task, a deviant number was weighted at least as much as other numbers. In the Inference Task there was a tendency to discount deviant numbers, but this tendency was much less with serial presentation than with simultaneous presentation. The processing of deviant information thus differed between descriptive and inference tasks, but this difference was attenuated with serial presentation. © 1976.