On the social psychology of the psychological experiment: the experimenter's hypothesis as unintended determinant of experimental results.

Abstract

Scientists are aware of the fact that they are imperfect instruments in the quest for lawful relationships (Wilson, 1952). Errors of ob servation and of interpretation have been discussed systematically from the time of the discovery of the personal equation among the astronomers Bessel, Kinnebrook, and Maskelyne, et al. (Boring, 1950). A lively in terest in these problems is to be found today among medical researchers and particularly among those working with drugs. Various techniques, such as the double-blind method, have been developed in which neither patient nor physician is to be aware of the nature of the sub stance ingested by the patient (Beecher, 1959). One purpose of this technique, of course, is to avoid errors of observation3 and interpreta tion3 in both subject (patient) and experimenter (physician). For many of the sciences, there seems to be little danger that the act of observation itself may change the object of study, if the object be macroscopic (Reichenbach, 1951). For the behavioral sciences, how ever, when humans or animals are the object of study, the act of ob servation may very well change the object of study. Research in the assessment of personality has shown that the personality and behavior of the assessor (observer) can change the response of the subject (Mas ling, 1960). The interviewer in the public opinion survey has been very systematically studied for his effect upon his respondents (Hyman, et al., 1954). The experimental psychologist, working in his laboratory rather than the clinic or the field, has been less systematically investi gated. Nevertheless, studies have shown that different psychological experimenters (Es) may obtain statistically significantly different data from comparable human subjects (Ss) (McGuigan, 1961; Mulry, 1962; Pflugrath, 1962). Further evidence suggests that different Es may ob tain statistically significantly different data from comparable ?s even