Sources of Error and Bias Response Styles

Most personality questionnaires consist of a series of statements that the respondent must answer either true or false or rate on a scale (e.g., from strongly disagree to strongly agree). As anyone who has taken such a test knows, the items are often ambiguous and sometimes of dubious relevance. The question, "Are you devoted to your work?'' might be interpreted in several ways. Some respondents might compare their devotion to work with their commitment to family. Some might compare their own devotion to that of their co-workers. Retired or unemployed respondents might not know how to respond. Even with the sin-cerest cooperation, respondents may not give the response the test developer intended.

Further, some respondents may not be sincerely cooperative. They may respond carelessly or at random simply to be finished with the task. Or they may wish to present a flattering picture of themselves to the tester. One of the most troubling discoveries of personality psychology was that laypeople are exquisitely sensitive to the social desirability of items and can, if so instructed, fake most personality tests.

Another common problem is acquiescent responding. It was discovered long ago that individuals differ in the tendency to agree with statements, regardless of content. So-called yea-sayers interpret items in ways that allow them to endorse most of them; nay-sayers find something in most items to which they object. If all the items are keyed in the same direction—that is, if true or agree responses are always indicative of the trait—then scale scores will confound measurement of the trait with measurement of acquiescent tendencies. Two such scales might show a positive correlation even if they measured very different traits, because both might also measure acquiescent tendencies.

This particular response style can be controlled quite effectively by creating scales with balanced keying: Half the items are scored in the positive direction, half in the negative. For example, we might measure conscientiousness by including the item "I often fail to keep my promises,'' and giving points for conscientiousness if the respondent disagrees. In responses to a balanced scale, acquiescent tendencies cancel themselves out, leaving a purer measure of the trait.

Similar strategies have been developed for dealing with other response styles. For example, random responding can be detected by including a set of items that virtually no one would endorse if they were paying attention and cooperating (e.g., "I keep an elephant in my basement''). Endorsing several such items would suggest random responding, and test results should be considered invalid. Cooperative respondents, however, may find the inclusion of such "trick questions'' offensive. An alternative way of detecting one common form of random responding is by looking for a string of identical responses on an answer sheet, which may indicate thoughtless, repetitive responding merely intended to finish the questionnaire. This is an unobtrusive measure of random responding.

The greatest attention has been paid to the problem of socially desirable responding. Many scales have been devised in the hopes that they could identify individuals who responded on the basis of the desirability of an item rather than its accuracy as a description of their personality. Researchers routinely include such scales in construct validity studies to estimate the discriminant validity of the scales of interest from socially desirable response tendencies. Unfortunately, however, no good measure of desirable responding per se has yet been developed, and most research suggests that attempts to correct for social desirability do more harm than good.

The root of the problem is that statements have both substantive and evaluative meanings. Anyone who wished to appear in a good light would endorse the item "I always try to do my best''—but so would highly conscientious individuals who are scrupulously honest in their responses. It is impossible to determine from the response alone whether the individual really has desirable characteristics or is presenting a falsely favorable picture of him- or herself.

Two general strategies appear to be useful for dealing with this problem. First, in most cases it appears that respondents are more truthful than psychologists anticipated. Even though they can endorse desirable items when instructed to do so, test takers normally do not, when asked to be honest and accurate. Research volunteers have little incentive to distort their responses, and clients in counseling and psychotherapy should be convinced by the assessor that accurate responding will be in their best interest. Mutual respect and trust between test administrators and test takers is usually the best basis for assuring valid results.

Personality Assessment

However, in some cases there may be good reasons for mistrusting self-reports. The responses of prison inmates who describe themselves as saints when being evaluated for parole should be regarded with considerable skepticism. In these cases, the most appropriate tactic may be to obtain observer ratings from knowledgeable and impartial informants. The current availability of validated observer-rating questionnaires (such as Form R of the NEO Personality Inventory) makes that approach feasible.

None of these approaches to scale construction or administration eliminates all the limitations of personality assessment by questionnaire. The inevitable ambiguity of items and respondents' imperfect knowledge of themselves or the individuals they rate mean than personality measures lack the precision that we admire in the physical sciences. The data in Table I show that our assessments are on the right track, but they can also be interpreted to show that our measurements are far from perfect. Both self-reports and observer ratings are useful tools that give valuable information about personality, and either is acceptable for use in research on groups. For the intensive understanding of the individual (e.g., in psychotherapy), it is desirable to obtain both self-reports and informant ratings, and all inferences about personality traits should be considered provisional, subject to revision or refinement as new information becomes available.

The Power Of Charisma

The Power Of Charisma

You knowthere's something about you I like. I can't put my finger on it and it's not just the fact that you will download this ebook but there's something about you that makes you attractive.

Get My Free Ebook


Post a comment