How the possibility of error affects falsification on a task that models scientific problem solving

Document Type

Article

Publication Date

1-1-1986

Abstract

This paper reports an experiment in which groups of four subjects attempt to solve two related rules when they know that 0–20 per cent of the feedback they receive may be in error. Groups were randomly assigned to one of three strategy conditions: confirmatory, in which subjects were urged to propose experiments consistent with their hypotheses; disconfirmatory, in which subjects were encouraged to propose experiments at variance with their hypotheses; and a control strategy, in which subjects were merely urged to test their guesses by proposing more experiments. In a previous study using the same rules and strategy conditions, but no error, Gorman et al. (1984) found that disconfirmatory groups performed significantly better than groups in other conditions. In the study described here, this performance difference did not arise. About 60 per cent of the time groups immunized their hypotheses against disconfirmation by classifying disconfirmatory results as error. About 40 per cent of the time groups correctly recognized that there was no error, but spent so much of their time replicating experiments to make sure that there was no error that they failed adequately to test their hypotheses. The few successful groups used a strategy that combined disconfirmation with replication. Detailed examples of how groups handled error are discussed, as are the implications of these results for future research. 1986 The British Psychological Society

Publication Title

British Journal of Psychology

Link to Full Text

Share

COinS