Experimenter Bias in Reporting Opinion Statements

Abstract

Rosenthal (1963) and Orne ( 1962) have emphasized that social psychological experimental results may be due to factors other than the independent variables, namely, experimenter bias and demand characteristics of experimental situations. Two groups of Es (introductory social psychology students) were instructed to carry out the Verplanck (1955) procedure in which the number of opinion statements offerecl by another person is recorded during three 10-min. intervals. E successively establishes a baseline for opinion statements, agrees with each opinion, and disagrees with each opinion. Group A was given Verplanck's reinforcement hypothesis: that agreement and disagreement with opinion statements act as positive and negative reinforcement, respectively, and the number of such statements should vary accordingly. Group B was given a balance hypothesis which stated opinions either established or disrupted balance. Since agreement balances the situation by establishing a positive bond, it reduces the need to give more balance-producing opinions. Disagreement, however, by disrupting balance, leads to an increased number of opinion statements. Es were free to choose their own Ss, experimental situation, and topics of conversation. It was predicted that data reported would vary according to the hypothesis given E. For 23 Es with the reinforcement hypothesis the mean number of opinion statements during baseline, agreement, and disagreement were 7.43, 9.74, and 7.96, respectively. For 23 Ss with the balance hypothesis (2 sets were omitted at random to achieve equal Ns) the values were 18.87, 18.35, and 22.57. Analysis of variance yielded an interaction berween groups and time periods (F = 15.61, df = 2/88, 9 < .01) of the form predicted (Fqu.drntle = 30.41, df = 1/44, p < .01). An effect occurred for time period (F = 7.76, df = 2/88, p < .01), but this appears to be due to the interaction. An unexpected difference berween the two groups warrants consideration. Group B reported more opinion statements than Group A (F = 89.01, df = 1/44, p < .01). A greater number from Group B chose the presidential campaign of 1964 for their main topic of conversation; and because this topic probably evoked strong opinions and chance for argument, it may account for the higher level of opinion statements in this group. Selection of this topic by Es may be considered part of their attempt to obtain results congruent with the hypothesis given them. The evidence for experimenter bia is open to various interpretations. For example, some data may have been fabricated by Es for presentation, selective recording may have occurred, or Es may have given cues indicating the correct response.