Abstract
It is suggested that, if the relevance of the information given to a judge with regard to the predictions he is asked to make is not controlled, it may serve as a confounding variable in studies of the prediction process. A possible method for dealing with this variable is outlined and evidence for its validity is presented. 58 judges predicted the responses of 8 targets to interest test items under two conditions: (1) where the information given them was relevant (valid) for the predictions to be made and (2) where the information was not relevant. The judges were highly accurate when the information was relevant but performed at chance level when the information was irrelevant.

This publication has 4 references indexed in Scilit: