Irwin A. Berg

Personality and Group Differences in Extreme Response Sets

responses does not follow the laws of statistical probability. If a coin is flipped, for example, the probability of heads is .50; but as Goodfellow (6) noted, the probability is .80 that persons will call the first toss heads. This tendency to respond in a given direction is a source of contamination which often affects test reliability and validity. Indeed, the Minnesota Multiphasic Personality Inventory, the Kuder Preference Record, and other tests include built-in correction devices to guard against this human response tendency. In two excellent review articles Cronbach (3, 4) has observed that when situations are

Response Sets in a Multiple-Choice Test

WHEN responding to a difficult objective test item, the subject does not always select his option on the basis of his fund of information alone. As Cronbach (4) has noted, &dquo;... the score of a person is influenced by his set to react to the items in a certain way, apart from their content.&dquo; He gives a number of examples of such response biases in tests, such as acquiescence, the preference for &dquo;yes&dquo; or &dquo;agree&dquo; options as opposed to &dquo;no&dquo; or &dquo;disagree,&dquo; the tendency to work for speed versus accuracy, the predilection for extreme positions when options in a series are presented, etc. Such response sets, obviously, can have a marked influence on test reliability and validity. They are stable, as Berg (i) and others have shown; hence this consistency may contribute upon occasion to spuriously high reliability. Accordingly, most constructors of tests who have been aware of the problem have striven to minimize the influence of such biased responses. After carefully examining the manner in which bias was manifested in tests, Cronbach (3) concluded that the multiple choice form was relatively free from response sets. Yet there is some evidence which indicates that such may not be the case. Using extreme position response biases in a multiple-choice test of little structure, Berg and Collier (2) found significant group and personality differences when the number of such