ALTERNATIVES TO SELF-REPORT
ALTERNATIVES TO SELF-REPORT
Although there are some in our field who seem intent on minimizing the problems associated with self-report measurement of personality, others are willing to accept the pervasiveness and damage of intentional distortion and self-deception. Regardless of perspective, we persist in our reliance on ‘objective’ self-report because we perceive no viable alternatives. We con- sider here a class of alternatives that will make the skin of the average I/O psychologist crawl: Projective Techniques. Please don’t touch that dial just yet.
We will, in fact, focus our attention on a class of projective techniques known as implicit measures. For I/O psychologists, the term ‘projective technique’ is associated with inkblots and drawings that are scored in an
P ERSONALITY IN I NDUSTRIAL /O RGANIZATIONAL P SYCHOLOGY 133 utterly subjective fashion and that probably reveal more about the tester than
they do about the tested. The concept, however, is not irrational. A projective measurement system is designed to elicit the construct in question. The test situation is typically ambiguous. This ambiguity is intentional. It provides a sort of void to be filled by the personality of the respondent. We present only
a brief overview of a subset of the existing implicit measures (for a more comprehensive review, see Fazio & Olson, 2003). The best known implicit measurement technique is the Implicit Associa- tion Test or IAT (Greenwald, McGhee, & Schwartz, 1998). In an IAT, response latency is used as a measure of the ease with which respondents’ pair attributes and targets. For example, the IAT measuring racial attitudes marks the ease with which respondents use target/attribute pairs such as black/pleasant, white/unpleasant, black/unpleasant, and so on. A shorter latency reflects more facility with the pairing in question. One can imagine
a similar approach to the measurement of conscientiousness (e.g., organized/ competent vs. organized/compulsive), agreeableness (friendly/virtuous vs. friendly/credulous), or adaptability (dynamic/opportunity vs. dynamic/ chaos).
A similar implicit technique involves priming. Racial attitudes might be assessed with a ‘prime’ comprised of a picture of a person of a particular race followed by an adjective for which the respondent must find a synonym. Respondents who take less time to find a connotation for a negative adjective in such a situation than they do for a positive adjective are thought to hold more racist attitudes because the adjective in question was already primed (for that person) by the picture.
Another implicit measure has been developed by an I/O psychologist (of all people). Larry James and his colleagues have developed what he calls a conditional reasoning measure (James, 1998). A conditional reasoning measure is predicated on the notion that inductive reasoning is, to some degree, conditional upon personality. Thus, if a person is presented with conflicting information and is asked to make a decision based on that information, the person will tend to attribute more importance and reason to the information that is consistent with his or her own personality. This process is, in turn, based on what James calls justification mechanisms. These are the cognitions, often implicit, with which we rationalize a course of action that is consistent with our predispositions. Thus, it is the predispositions that drive us to prefer a particular course of action, and it is the justification mechanisms that allow us to feel that this course of action is rationally driven.
There are other tasks involving sentence completion, description of expectancy-consistent vs. expectancy-inconsistent behaviors, and even ten- dencies as esoteric as letter preference (see Fazio & Olson, 2003 for a review). All of these approaches have two things in common. First, they do not involve asking the respondent to provide his/her standing on the construct of interest. Second, they are based on hypotheses about how goals, attitudes,
134 I NTERNATIONAL R EVIEW OF I NDUSTRIAL AND O RGANIZATIONAL P SYCHOLOGY 2005 and beliefs manifest themselves. It is this second commonality on which we
wish to focus. The development of these methods began with efforts to understand how a given construct would produce a very specific behavior. Once a construct/behavior pairing was hypothesized, a methodology was developed that would allow the pairing to emerge. Specifically, a stimulus was identified that would produce different behaviors depending on one’s standing on the construct of interest.
This is very different from the development of self-report measures. Consistent with the lexical tradition, self-report items are generated based on the degree to which they look like the construct of interest. We know a conscientiousness item measures conscientiousness because it looks like a measure of conscientiousness (e.g., it contains facets of conscientiousness, adjectives of conscientiousness, etc.). We hope that this stimulus produces the appropriate response (e.g., selection of the most accurate response option), but we have little in the way of an understanding of internal processes to support this stimulus–response connection.
What we do have is reliability, and it is probably this that has led us to prefer self-report. While self-report measures generate relatively large inter- nal consistency reliabilities and test–retest reliabilities, implicit measures generate notoriously low reliabilities of all types (Cunningham, Preacher, & Banaji, 2001).
We suspect that these low reliabilities, coupled with the difficulty in devel- oping and administering implicit measures, have led us to disregard implicit measures. To this we say, who better than I/O psychologists to build a better (implicit) mousetrap? Who better to isolate and remove the sources of random error that have plagued existing implicit measures? There is reason to believe that such an effort might be worthwhile. In spite of low reliabilities, implicit measures have generated bivariate and incremental correlations with relevant criteria that exceed those typically reported for self-report measures (James, 1998; James et al., 2000). If random error could be reduced substantially, the sky is the limit on predictive validity.
Unfortunately, we have been content to trade random error (e.g., internal inconsistency, temporal instability) for systematic error (intentional distor- tion, self-deception). We have, in short, chosen to look for our keys under the streetlight even though we dropped them two blocks back. It is our hope that our field will end its rationalization of the use of self-report in personality research. Implicit measurement may not be the answer, but, until we recog- nize that self-report isn’t the answer either, we are unlikely to go far in our understanding of personality in the workplace.
One final alternative to self-report is worth mentioning, even though it doesn’t qualify as implicit in the strict use of the term. There has been a great deal of work done on the neurological and neurochemical bases of personality. In addition to Eysenck’s work on extraversion, there has been considerable research organized around the BAS and BIS mentioned earlier.
P ERSONALITY IN I NDUSTRIAL /O RGANIZATIONAL P SYCHOLOGY 135 There seems no point in regurgitating the jargon of neuropsychology (see the
Biological Substrates section of Revelle, 2003 for a review). It suffices to say that traits such as aggression, hostility, agreeableness, conscientiousness, positive/negative affect, introversion/extraversion, impulsivity, and many others have been linked to particular brain centers and particular neurochem- icals. Although the human brain is too complex for anyone to suggest that a given aspect of personality has a single physiological cause, it should interest us to know that the possibility of physiology-based measurement of person- ality exists. The measurement of extraversion via physiological reactivity is, perhaps, farther along than the measurement of other traits, or than measure- ment through other means. Nevertheless, physiology, along with any basis for measurement that avoids the problems of self-report, should be of interest to us.