Psychology) (English E-Book) Personality Assessment - Perso

8/8/2019 Psychology) (English E-Book) Personality Assessment - Perso

http://slidepdf.com/reader/full/psychology-english-e-book-personality-assessment-perso 1/26

PERSONALITY ASSESSMENT

I: PERSONALITY TESTING

AND ITS CONSEQUENCES

5

If something exists, it exists in some quantity, and if it exists in some quantity, it can be measured.

—Edward Lee Thorndike (Thorndike, 1905; cited in Cunningham,1992, p. 35)

If you accept the conclusion of Chapter 4—that personality traits are realand everybody has them—then Thorndike’s quote, above, implies that thenext task is to measure them. Are you more or less friendly, for example,than the person sitting next to you? To answer this kind of question, or, morebroadly, for personality traits to be useful for the scientific understanding of

the mind, for the prediction of behavior, or for any other purpose, mea-surement is essential. The next two chapters explore “personality assess-ment,” the enterprise of trying to accurately measure characteristic aspectsof personality.

The Nature of Personality Assessment

Personality assessment is a professional activity of numerous research, clin-ical, and industrial psychologists, and it is a prosperous business. Cliniciansmay measure how depressed you are in order to plan treatment, whereaspotential employers would probably be more interested in measuring yourconscientiousness to decide whether to offer you a job. But there is more topersonality assessment than just measuring traits.

An individual’s personality consists of any characteristic pattern of behavior, thought, or emotional experience that exhibits relative consistency across time and situations (Allport, 1937). These patterns include the termscommonly thought of as personality traits that were discussed in Chapter 4,and personality psychology has a long, still active, and still fruitful traditionof conceptualizing and assessing such traits. 113



But these patterns also include other kinds of variables, includingmotives, intentions, goals, strategies, and subjective representations (the waysin which people perceive and construct their worlds; see Chapter 17). Thesevariables indicate the degree to which a person desires one goal as opposedto another, or thinks the world is changeable as opposed to fixed, or is gen-erally happy, or is optimistic as opposed to pessimistic, or is sexually attractedto members of the same or the opposite sex. All of these variables, and many others, are relatively stable attributes of the psychological makeup of indi-viduals, and any attempt to measure them necessarily entails personality assessment. As a result, assessment is relevant to an extremely broad rangeof research, including nearly every topic of interest to personality, develop-mental, and social psychologists.

Moreover, personality assessment is not restricted to psychologists. Theterm is usually used to refer to assessments by psychologists in research, inclinics, or in private industry; their results sometimes have important con-sequences for the individuals involved. But assessments by nonpsychologistsare even more widespread, and a good case could be made that they are evenmore important. These personality assessments are done by you, your friends, your family—and by me, in my off-duty hours—all day long, every day. Asmentioned at the beginning of Chapter 4, personality traits are a fundamentalpart of how we think about each other and about ourselves. We choose whoto befriend and who to avoid on the basis of our assessments of them—willthis person be reliable, or helpful, or honest—and they make the same judg-ments and choices concerning us. We even base our feelings about ourselvespartly on our beliefs about our personalities—are we competent, or kind, ortough? The judgments we make of the personalities of each other and our-selves are at least as consequential as those that any psychologist will evermake.

Regardless of whether the source of a personality assessment is a psy-chologist, an acquaintance, or a psychological test, the most important thingto know about it is the degree to which it is right or wrong. And regardlessof the source of a personality assessment, it must be evaluated according tothe same criteria. When professional personality judgments or personality tests are evaluated, the evaluation is said to be of their validity . When ama-teur judgments are evaluated, the term accuracy is generally used instead.Despite the difference in terms, the basis of the evaluation is the same eitherway. Two basic criteria are available: agreement and prediction (Funder,1987, 1995, 1999). The agreement criterion asks: Does this judgment agreewith other judgments obtained through other techniques or from other judges (whether professional or amateur)? The prediction criterion asks, Canthis judgment of personality be used to predict behavior?

The topic of this chapter and the next is personality assessment—how personality is judged by professionals and by amateurs. The remain-

CH. 5 PERSONALITY TESTING AND ITS CONSEQUENCES 114



der of this chapter considers the business of personality assessment andhow psychologists construct and use personality tests, as well as some of the consequences of testing. Chapter 6 considers personality assessmentby amateurs—people who are not trained or paid to be psychologists butwho nonetheless practice psychology every day.

The Business of Testing

Each year the American Psychological Association (APA) holds a conven-tion. It’s quite an event. Thousands of psychologists take over most of thedowntown hotels in a city such as San Francisco, Boston, or Washington,D.C., for a week of meetings, symposia, and cocktail parties. One of thebiggest attractions is always the exhibit hall, where dozens of high-tech, artis-tically designed booths fill a room that seems to go on for acres. These boothsare set up, at great expense, by several kinds of companies. One group is text-book publishers; all the tools of advertising are applied to the task of con-vincing college professors like me to assign their students to read (and buy)books such as the one you now hold in your hand. Another group is man-ufacturers of videotapes and various strange gadgets for therapy and research.Yet another group is psychological testers. Their booths typically distributefree samples that include not only personality and ability tests but also shop-ping bags, notebooks, and even beach umbrellas. These freebies prominently display the logo of their corporate sponsor: the Psychological Corporation,Consulting Psychologists Press, the Institute for Personality and Ability Test-ing, and so on. These goodies are paid for by what they are intended to sell:personality tests.

You can get a free “personality test” even without going to the APA con-vention. On North Michigan Avenue in Chicago, on the Boston Common,at Fisherman’s Wharf in San Francisco, and in Westwood in Los Angeles, Ihave seen people distribute brightly colored brochures on which are printed“Are you curious about yourself? Free personality test enclosed.” Inside thebrochure is something that in many ways looks like a conventional person-ality test—two hundred questions to be answered true or false (one itemreads, “Having settled an argument out do you continue to feel disgruntledfor a while?”). But as it turns out, it is really a recruitment pitch. If you takethe test and go for your “free evaluation”—which I do not recommend— you will be told two things. First, you are all messed up. Second, the peoplewho gave you the test have the cure: you need to join a certain “church” thatwill provide the therapy you so desperately need.

The personality testers who distribute free samples at the APA conven-tion and those who hand out free “personality tests” on North Michigan

THE BUSINESS OF TESTING 115



Avenue have a surprising amount in common. Both are seeking new cus-tomers, and both use all the usual techniques of advertising, including freesamples. The tests they distribute look superficially alike. And they both tradeon what seems to be a nearly universal desire to know more about person-ality. The brochure labeled “Are you curious about yourself?” asks a pretty irresistible question. The more staid tests distributed at the APA conventionlikewise offer an intriguing promise of finding out something about yourown or somebody else’s personality that might be interesting, important, oruseful.

But below the surface they are not the same. The tests peddled at theAPA convention are, for the most part, well-validated instruments that areuseful for many purposes. The ones being pushed at tourist destinations fromcoast to coast are potentially dangerous frauds. But you cannot tell which iswhich just by looking at them. To tell a valid test from an invalid one youneed to know something about how personality tests and assessments areconstructed, how they work, and how they can fail. Read on.

Personality Tests

The personality testing companies that are so well represented at APA meet-ings do a good business. They sell their product to clinical psychologists, tothe personnel departments of large corporations, and to the military, forexample. You have most likely taken at least one personality test at somepoint in your life, and you are likely to take more.

One of the most widely used personality tests is the Minnesota Multi-phasic Personality Inventory, or MMPI.1 This test was designed for use inthe clinical assessment of individuals with psychological difficulties, but ithas also been widely used for purposes such as employment screening.Another widely used test is the California Psychological Inventory, or CPI;it is similar to the MMPI in many ways but is designed for use with non-disturbed individuals. Others include the Sixteen Personality Factor Ques-tionnaire, or 16PF; the Strong Vocational Interest Blank, or SVIB (used forvocational guidance in school settings); the Hogan Personality Inventory, orHPI (used by employers for personnel selection); and many more.

Many personality tests, including those just listed, are “omnibus” inven-tories designed to measure a wide range of personality traits. The NEO Per-sonality Inventory, for instance, measures five broad traits along with a largenumber of subscales (Costa & McCrae, 1997).


1By a tradition of mysterious origin, nearly all personality tests are referred to by theirinitials—all capital letters, no periods.



Other personality tests are designed to measure just one trait. There aretests to measure shyness, self-consciousness, self-monitoring, empathy, attri-butional complexity, nonverbal sensitivity, something called “Type A” (a hos-tile personality pattern that makes the person vulnerable to heart attack),something else called “Type C” (a passive personality pattern that suppos-edly makes the person vulnerable to cancer), and so on. No one has everdone an exact count, but there are at least thousands of such tests, and newones are introduced every day.

S-Data Versus B-Data Personality Tests

To use the terms introduced in Chapter 2, many personality tests—proba-bly most—provide S data. They ask you about what you are like, so the score you receive amounts to a summary of how you have described yourself. Theshyness scale consists of a bunch of questions that ask how shy you are, theattributional complexity scale asks you how complexly you think about thecauses of people’s behavior, and so forth. Other personality tests yield B data.The MMPI is a good example. It presents items—such as “I prefer a showerto a bath”—not because the tester is interested in the literal answer, butbecause answers to this item are informative about some aspect of person-ality (in this case, empathy—preferring the shower is the empathic response,for some reason) (Hogan, 1969).

Another kind of B-data personality test has been introduced very recently.The “implicit associations test” (IAT) measures how quickly participants canrespond to instructions to discriminate between terms that apply to “me” orto “others,” and between terms that are relevant, or not, to the trait being mea-sured (Greenwald, McGhee, & Schwartz, 1998). One recent study measuredshyness in this way (Asendorpf, Banse, & Mücke, 2002). Participants sat at acomputer keyboard and responded as quickly as they could as to whether terms(e.g., “self,” “them”) referred to “me” or “others,” and then to whether otherterms (e.g., “inhibited,” “candid”) referred to “shy” and “nonshy,” and finally did both tasks at the same time. The theory is that people who “implicitly”(i.e., not necessarily consciously) know they are shy will have faster associa-tions between “me” and “shy” than between “me” and “nonshy.” The samestudy also gathered more conventional, S-data self-ratings of shyness. Finally,the participants were videotaped as they interacted with an attractive individ-ual of the opposite sex, a task calculated to induce shyness.

The fascinating result was that “controlled” aspects of shyness, such ashow long people spoke, could be predicted by the S-data shyness scores. (Shy people speak less.) However, more uncontrolled or spontaneous indicatorsof shyness, such as facial expressions and tense body posture, were predictedmuch better by the IAT measure. This result suggests that people are only partially conscious of their own knowledge of their shyness, but that their

PERSONALITY TESTS 117



deeper, underlying knowledge can not only be measured, it can be used topredict behavior (this idea will be discussed further in Chapter 17).

Is intelligence a personality trait? Psychologists have differing opinionsabout this. Either way, it can be noted that tests of intelligence, or “IQ tests,”also yield B data. Imagine trying to assess IQ with an S-data test. Such a testwould include questions like “Are you an intelligent person?” and “Are yougood at math?” Researchers have actually tried this, but simply asking peo-ple whether they are smart turns out to be a poor way to measure intelli-gence (Furnham, 2001). So instead, IQ tests ask people questions of varyingdifficulty, such as reasoning or math problems, that they might get right orwrong. The more questions they get right, the higher is their (tested) IQ.These right or wrong answers comprise B data.

In my opinion, the distinction between S-data tests and B-data tests isimportant, but this distinction is rarely drawn in the conventional literatureon personality testing. More common is the distinction between projectiveand objective tests.

Projective Tests

Projective tests are based on a particular theory of how to see into someone’smind. The theory is this: If somebody tries to describe or to interpret a mean-

ingless, ambiguous stimulus—such as an inkblot—his or her responses cannotcome from the stimulus itself, because the stimulus does not, in truth, meananything. The answer he or she gives must instead come from (be a “projec-tion” of) the inner workings of the person’s mind (Murray, 1943). The answermay even reveal something the person does not know about himself or herself.

This is the theory behind the famous Rorschach inkblot test, for instance(Rorschach, 1921; Exner, 1993). The Swiss psychiatrist Hermann Rorschachdropped blots of India ink onto note cards, folded the cards in half, and thenunfolded them. The result was a set of symmetrical blots.2 Over the years,

uncounted psychiatrists and clinical psychologists have shown these blots totheir clients and asked them what they saw.

Of course, the only literally correct answer is “an inkblot,” but that isnot considered a cooperative response. Instead, the examiner is interested inwhether the client will report seeing a cloud, or a devil, or her mother, orwhatever. The idea is that whatever the client sees, precisely because it is notin the card, must reveal something about the contents of his or her mind.

The same logic has led to the development of numerous other projectivetests. The “draw-a-person” test requires the client to draw (you guessed it) a


2According to legend, Rorschach made many blots in this way but kept only the “best” ones.I wonder how he decided which ones those were.



person, and the drawing is interpreted according to what kind of person is drawn(e.g., a man or a woman), which body parts are exaggerated or omitted, and soforth (Machover, 1949). Large eyes might be taken to indicate suspiciousness orparanoia; heavy shading to mean aggressive impulses, and numerous erasuresto be a sign of anxiety. The Thematic Apperception Test (TAT) asks clients totell stories about various pictures (Morgan & Murray, 1935; Murray, 1943). Thethemes of these stories are held to be informative about the client’s motivationalstate (e.g., McClelland, 1975). If a person looks at a picture of two people andthinks they are fighting, for example, this might reveal a desire for aggression;if the two people are described as in love, this might reflect a need for intimacy;if one is seen as giving orders to the other, this might reflect a need for power.

The idea common to all these tests is interesting and reasonable, and inter-pretations of actual responses can be fascinating. A large number of practicingclinicians swear by their efficacy. In one survey, 82 percent of clinical psychol-ogists reported using the Rorschach at least occasionally (Watkins, Campbell,Nieberding, & Hallmark, 1995). However, objective research data on the valid-ity of these tests—the degree to which they actually measure what they are sup-posed to measure—is surprisingly scarce (Lilienfeld, Wood, & Garb, 2000).

It is probably fair to say that only two projective tests have a backgroundof evidence that comes even close to establishing validity according to the stan-dards by which personality tests conventionally are evaluated. One of them isthe TAT (McClelland, 1984). This instrument measures “implicit motives,”meaning motivations concerning achievement, intimacy, power, and othermatters that the participant himself or herself might not be aware of. The TATcontinues to be used in current research concerning motivations connected




with complex thinking (cognitive complexity), the experiences in life one findsmost memorable, and other topics (Woike, 1995; Woike & Aronoff, 1992). Itis possible that projective tests and more conventional objective tests (see nextsection) get at slightly different aspects of personality. One group of researchershas proposed that motives measured by the TAT reflect what people want,whereas traits as measured by questionnaires predict how these motives willbe expressed (Winter et al., 1998). For example, the TAT might reveal that aperson has a great need for power, whereas a more conventional test mightreveal how he or she will go about trying to obtain power.

The other projective test that seems to have a modicum of validity is theRorschach, when it is scored according to one of two specific techniques(either Exner’s Comprehensive System [Exner, 1993] or Klopfer’s system[Klopfer & Davidson, 1962]). According to a recent, comprehensive reviewof research using the Rorschach, the correlation coefficient between scoresgarnered from one of these systems and various criteria relevant to mentalhealth averaged about .33 (Garb, Florio, & Grove, 1998). As you will recallfrom Chapter 3, this means that a dichotomous (yes-no) diagnostic decisionmade using the Rorschach will be correct about 66 percent of the time.3

To again use the terminology introduced in Chapter 2, all projective testsprovide B data. They are specific, directly observed responses to particular stim-uli, whether inkblots, pictures, or instructions to draw somebody. All the dis-advantages of B data therefore apply to projective tests. For one thing, they areexpensive. It takes around 45 minutes to administer a single Rorschach andanother 1.5 to 2 hours to score it (Ball, Archer, & Imhoff, 1994). Compare thisto the time needed to hand out a pile of questionnaires and score them witha machine. This is a serious issue because it is not enough for the Rorschachto have some small (and even surprising) degree of validity. For its continueduse to make any sense, it should provide extra information that justifies itsmuch greater cost (Lilienfeld et al., 2000). But despite hundreds of studies, noevidence has been found that the Rorschach provides information above andbeyond that provided by easier, cheaper tests such as the MMPI (Hunsley &Bailey, 1999; Parker, Hunsley, & Hanson, 1999).

The even more fundamental difficulty with projective tests is that, per-haps even more so than other kinds of B data, a psychologist cannot be surewhat they mean. What does it mean when somebody thinks an inkblot lookslike genitalia, or imagines that an ambiguous picture portrays a murder, ordraws a person with no ears? The answer and the validity of the answerdepend critically on the test interpreter (Sundberg, 1977). Two different


3The main point the authors of this review intended to make was that the validity of the MMPIis even higher (the parallel r .55 in their analysis). But in the process they seem to haveprovided one of the most convincing demonstrations to date that the Rorschach has morethan zero validity.



interpreters of the same response might come to different conclusions, unlessa standard scoring system is used. Of the projective tests, only the TAT isconsistently scored according to a well-developed system (e.g., McAdams,1984). While, as was mentioned above, scoring systems have been developedfor the Rorschach (Exner, 1993; Klupfer & Davidson, 1962), not everybody uses them. Most of the remaining projective tests are interpreted accordingto the predilections of the individual examiner.

The survival of most projective tests into the twenty-first century is some-thing of a mystery. Even literature reviews that show projective tests to havesome degree of validity tend to conclude that other, less expensive techniqueswork as well or even better (Garb, Florio, & Grove, 1998, 1999; Lilienfeld etal., 2000). As the measurement expert Ann Anastasi wrote more than twodecades ago, “projective techniques present a curious discrepancy betweenresearch and practice. When evaluated as psychometric instruments, the largemajority make a poor showing. Yet their popularity in clinical use continuesunabated” (Anastasi, 1982, p. 564).

Perhaps they endure because clinicians have fooled themselves into think-ing they are valid. One writer has suggested these clinicians may lack “a skillthat does not come naturally to any of us: disregarding the vivid and compellingdata of subjective experience in favor of the often dry and impersonal resultsof objective research” (Lilienfeld, 1999, p. 38). Perhaps, as others have suggested,the validity of these tests is beside the point; instead they simply serve a useful,if nonpsychometric, function of “breaking the ice” between client and therapistby giving them something to do during the first visit. Or, just possibly, there isa real validity to these instruments in actual clinical application, and in the handsof certain skilled clinicians, that cannot be duplicated by other techniques andthat has not been captured successfully by controlled research studies.

Objective Tests

The tests that psychologists call “objective” can be detected at a glance. If atest consists of a list of questions to be answered yes or no or true or falseor on a numeric scale, and especially if it uses a computer-scored answersheet, then it is an objective test. The term comes from the idea that thequestions that make up the test seem more objective than the pictures andblots of projective tests.

VALIDITY AND THE SUBJECTIVITY OF TEST ITEMS

It is not clear that the term objective is really justified, however. Consider thefirst item of the famous MMPI. The item reads “I like mechanics magazines,”which is to be answered true or false (Wiggins, 1973). The item may seem




objective, compared to a question like “What do you see in this inkblot?”But the appearance may be misleading. For instance, does “like” mean inter-est, fondness, admiration, or tolerance? Does liking such magazines requirethat you regularly read them? Are Popular Mechanics and High Fidelity mechanics magazines? How about Computer World ? Are only popular mag-azines included in the classification, or does the item also refer to trade journals of professional mechanics, or the research literature produced by professors of mechanical engineering? This example is not an unusually prob-lematic item; it is rather typical. And it illustrates how objectivity is harderto attain than it might seem at first.

It is hard to escape the conclusion that the items on objective tests, whileperhaps not as ambiguous as those that constitute projective tests, are stillnot absolutely objective. Writing truly objective items might be impossible,and even if it were possible, such items might not work. Think about it fora moment: if everybody read and interpreted an item in exactly the sameway, then might not everybody tend also to answer the item in the same way?If so, the item would not be very useful for the assessment of individual dif-ferences. In some cases, the ambiguity of objective items might not be a flaw;the interpretation of an item might have to be somewhat subjective in orderfor responses to it to be informative for personality assessment.

Harrison Gough, the inventor of the CPI, included on his test a scalecalled “commonality,” which consists solely of items that are answered in thesame way by at least 95 percent of all people. He included it to detect illit-erates who are pretending to know how to read and individuals deliberately trying to sabotage the test. The average score on this scale is about 95 per-cent, but an illiterate answering at random will score about 50 percent (sinceit is a true-false scale) and therefore will be immediately identifiable, as willsomeone who (like one of my former students) answered the CPI by flip-ping a coin—heads true, tails false.

These are interesting and clever uses for a commonality scale, but itsproperties are mentioned here to make a different point. Gough reportedthat when individuals encounter a commonality item—one being “I wouldfight if someone tried to take my rights away” (keyed true)—they do not say to themselves, “What a dumb, obvious item. I bet everybody answers it thesame way.” Instead, they say, “At last! A nonambiguous item I can really understand!” People enjoy answering the commonality items because they do not seem ambiguous.4 Unfortunately, such items are not very useful forpersonality measurement, because the items are ones on which few peoplediffer. A certain amount of ambiguity may be a good thing for an item on apersonality test (Gough, 1968; Johnson, 1981).


4Another item from the commonality scale reads, “Education is more important than mostpeople think.” Paradoxically, almost everybody responds true.



WHY SO MANY ITEMS?

If you look at an objective test, one of the first things you will notice is howmany questions it asks. This number may be very large. Some of the shorter

personality tests have around a dozen items, but most have more, and someof the most famous personality tests (e.g., the MMPI, the CPI) have hun-dreds. To complete a test like the MMPI can take an hour or more, and afairly tedious hour at that.

Why so many items? The answer lies in the principle of aggregation,which was explained in Chapter 3. The answer an individual gives to any onequestion might not be particularly informative; it might vary according toexactly how he or she interprets the item or other extraneous factors. In theterminology used in Chapter 3, a single answer will tend to be unreliable . But

if a group of more or less similar questions is asked, the average of the answersto these questions ought to be much more stable, or reliable . All other thingsbeing equal, you can make a personality test more reliable simply by mak-ing it longer. If the items you add are as good indicators of the trait beingmeasured as the items you already have—something easier said than done,frankly—then the improvement in reliability can be estimated rather pre-cisely using a calculation called the Spearman-Brown formula .5

As you will recall from Chapter 3, a reliable test is one that gives close tothe same answer, time after time. However, you will also recall that while reli-

ability is necessary for validity, it is no guarantee. The validity of anobjective test depends on its content . The crucial task in test construction, then,is to write and select the right questions. That is the topic of the next section.

Methods of Objective Test Construction

There are three basic methods for constructing objective personality tests:the rational method, the factor analytic method, and the empirical method.Sometimes, a mixture of methods is employed, but let me begin by consid-

ering the pure application of each.


5If you must know, here are some formulas. First, the reliability of a test is measured in termsof “Cronbach’s alpha” according to the following formula: If N is the number of items in thetest, and p is the average correlation among all of the items, then the reliability (alpha) N p/[1 ( p(N 1)) (Cronbach, 1951). The “Spearman-Brown formula,” just mentioned,predicts the increase in reliability you get when you add equivalent items to a test (Brown,1910; Spearman, 1910). If k n1/n2, the fraction by which the number of items is increased,then the reliability of the longer test is estimated by

alpha (longer test)

In both formulas, alpha is the predicted correlation between a score on your test and a scoreon another test of equivalent content and length.

k alpha (shorter test)

1 (k 1) alpha (shorter test)



THE RATIONAL METHOD

Calling one method of test construction “rational” does not mean the oth-ers are irrational. It simply means that the basis of this approach is to comeup with items that seem directly, obviously, and rationally related to what itis the test developer wishes to measure. Sometimes this is done through care-ful derivation from a theory of the trait in which the researcher is interested.Thus, test developer Douglas Jackson (1971) wrote items to capture the“needs” postulated years earlier by Henry Murray. Other times, the processof writing the items is less systematic, reflecting whatever comes to mind assomething relevant to ask. The data that are gathered are S data (see Chap-ter 2), or direct and undisguised self-reports.

An early example of a test constructed by the rational method is drawnfrom World War I. The U.S. Army discovered, not surprisingly, that certainproblems arose when individuals who were mentally ill were inducted as sol-diers, housed in crowded barracks, and given loaded weapons. To avoid theseproblems, the U.S. Army developed a structured interview consisting of a listof questions that a psychiatrist could ask each potential recruit. As the num-ber of inductees increased, however, this long interview process becameimpractical. There were not enough psychiatrists nor was there enough timeto interview everybody.

To get around these limitations, a psychologist named R. S. Woodworth(1917) proposed that the questions of a typical interview could be printedon a sheet, and the recruits could check off their answers with a pencil.Woodworth’s list, which became known as the Woodworth Personality DataSheet (or, inevitably, the WPDS), consisted of 116 questions, all of whichwere deemed relevant to potential psychiatric problems. The questionsincluded “Do you wet your bed?,” “Have you ever had fits of dizziness?,” and“Are you troubled with dreams about your work?” A recruit who responded yes to more than a small number of these questions was referred for a morepersonal examination. Recruits who answered no to all the questions wereinducted forthwith into the army.

Woodworth’s idea of listing psychiatric symptoms on a questionnairewas not unreasonable, yet his technique raises a variety of problems that canbe identified rather easily. For the WPDS to be a valid indicator of psychi-atric disturbance, and for any rationally constructed, S-data personality testto work, four conditions must hold (Wiggins, 1973).

First, each item must mean the same thing to the person who fills outthe form as it did to the psychologist who wrote it. For example, in the itemfrom the WPDS, what is “dizziness,” exactly?

Second, the person who completes the form must be able to make anaccurate self-assessment. He (only men were being recruited at the time the




WPDS was administered) must have a good enough understanding of whateach item is asking about, as well as the ability to observe it in himself. Hemust not be so ignorant nor so psychologically disoriented that he cannotreport accurately on these psychological symptoms.

Third, the person who completes the form must be willing to report hisself-assessment accurately and without distortion. He must not try to deny his symptoms (e.g., in order to get into the army) or to exaggerate them (per-haps in order to stay out of the army). Modern personality tests used forselecting employees encounter this problem frequently (Rosse et al., 1998).

Fourth and finally, all of the items on the form must be valid indicatorsof what the tester is trying to measure—in this case, mental disturbance. Doesdizziness really indicate mental illness? What about dreams about work?

For a rationally constructed test to measure accurately an attribute of personality, all four of these conditions must be fulfilled. In the case of theWPDS, none of them probably were (although given how inexpensive admin-istering the WPDS was, and how expensive psychiatrists’ time was, if it caught just a few serious cases, it may have been cost-effective anyway). In fact, mostrationally constructed personality tests fail one or more of these criteria. Onemight conclude, therefore, that they would hardly ever be used anymore.

Wrong. Up to and including the present day, self-report questionnairesthat are little different, in principle, from the WPDS remain the most com-mon form of psychological measurement device. Self-tests in popular mag-azines are always constructed by the rational method—somebody just thinksup some questions that seem relevant—and they almost always fail at leasttwo or three of the four crucial criteria for validity.

Rationally constructed personality tests appear in psychological journals,too. Such journals present a steady stream of new testing instruments, nearly all of which are developed by the simple technique of thinking up a list of questions that seem relevant. These questions might include measures of health status (how healthy are you?), self-esteem (how good do you feel about yourself?), or goals (what do you want in life?).

For example, research has addressed the differences between college stu-dents who follow optimistic or pessimistic strategies in order to motivatethemselves to perform academic tasks (such as preparing for an exam). Opti-mists, as described by this research, motivate themselves to work hard by expecting the best outcome, whereas pessimists motivate themselves by expecting the very worst to happen unless they work hard. Both strategiesseem to be effective, although optimists may have a more pleasant time of it(Norem & Cantor, 1986) (these strategies are considered in more detail inChapter 17). For purposes of this chapter, the question is, how are optimistsand pessimists identified? The researchers in this study used an eight-itemquestionnaire, on which participants were asked to respond using an




11-point scale ranging from “not at all true of me” (1) to “very true of me”(11) to questions such as:

• I go into academic situations expecting the worst, even though I

know I will probably do OK.• I generally go into academic situations with positive expectations

about how I will do.

• I often think about what I would do if I did very poorly in an acade-mic situation.

The first and third items listed are “pessimism” items (higher scoresreflecting a pessimistic strategy) and the second is an “optimism” item (ahigher score reflecting an optimistic strategy) (Norem & Cantor, 1986, p.

1211).By the definitions I have been using, this is a rationally constructed,

S-data personality test. And in fact, it seems to work fairly well at identify-ing students who approach academic life in different ways. So clearly testslike this can be valid, even though the four cautions raised earlier shouldalways be kept in mind.

A slightly different case is presented by the technique used in someresearch to measure “personal strivings.” In this research, participants areasked to list the “objective you are typically trying to accomplish or attain”

(Emmons & King, 1988, p. 1042). This measure is rational in the sense thatit directly asks about what the researchers are trying to find out. But unlikethe rational method considered previously, this measure is open-ended.Instead of answering a set of printed questions by marking true or false, orby choosing a point on an 11-point scale, the participant writes his or herstrivings on a blank piece of paper. This assessment technique, like otherrational approaches, will only work to the degree a participant can and willreport accurately what he or she is trying to accomplish. (Research indicatesthat most participants can and do accurately report the strivings that guide

much of their lives [Emmons & McAdams, 1991].) This technique gets aroundsome of the problems found in other rationally constructed questionnairesthat ask participants to answer questions that may be ambiguous or of dubi-ous relevance. Yet it poses a different problem: that of figuring out the rele-vance and decoding the ambiguity of what the participant has written.

THE FACTOR ANALYTIC METHOD

The factor analytic method of test construction is an example of a psycho-logical tool based on statistics. Factor analysis is a statistical method for find-ing order amid seeming chaos. The factor analytic technique is designed toidentify groups of things—such as test items—that seem to be alike. Theproperty that makes these things alike is called a factor (e.g., Cattell, 1952).




Factor analysis might seem esoteric, but it is not much different fromthe way people intuitively group things. Consider the rides at Disneyland.Some—the Matterhorn ride, for example—are fast, scary, and enjoyed by teenagers. Others—the “It’s a Small World” ride—are slow, include sweetsongs, and are enjoyed by young children. Because certain properties seemto occur together (e.g., the slow rides are usually accompanied by sweetsongs) and others rarely occur together (e.g., few rides include sweet songsand are scary), the groupings of properties can be seen to constitute factors.The properties of being fast, scary, and appealing to teenagers all go togetherand form a factor that you might call “excitement.” Likewise, the properties of being slow, having sweet music, and appealing to young children form a factorthat you might call “comfort.” You could now go out and measure all the ridesat Disneyland—or even anticipate the score of a new ride—according to bothof these factors. And then, you could use these measurements for the pur-poses of prediction. If a new ride gets a high “excitement” score, you wouldpredict long lines of teenagers waiting to ride it. If it gets a high “comfort”score, you can predict long lines of small children and frazzled parents.

To use this technique to construct a personality test, researchers beginwith a long list of objective items of the sort discussed earlier. The items canbe from anywhere; the test writer’s own imagination is one common source.If the test writer has a theory about what he or she wants to measure, thattheory might suggest items to include. Another, surprisingly common way to get new items is to mine them from old tests. (Items from the MMPI, inparticular, have a way of reappearing on other tests.) The goal is to end upwith a large number of items; thousands are not too many.

The next step is to administer these items to a large number of partici-pants. These participants are recruited in any convenient manner, which iswhy they are often college students. Sometimes, for other kinds of tests, theparticipants are mental patients. Ideally, they should represent well the kindof people with whom you hope, eventually, to use this test.

After your large group of participants has taken the initial test that con-sists of these (maybe) thousands of items, you and your computer can sitdown together and do the factor analysis. The analysis is based on calculat-ing correlation coefficients (see Chapter 3) between each item and every otheritem. Many items—probably most—will not correlate highly with much of anything and can be dropped. But the items that do correlate highly (or evensomewhat) with each other will begin to group together. For example, if aperson answers true to the item “I trust strangers,” you might find that heor she is also likely to answer true to “I am careful to turn up when some-one expects me” and to answer false to “I could stand being a hermit.” Sucha pattern of likelihood or co-occurrence means that these three items arecorrelated. The next steps are to read the items, decide what they have incommon, and then name the factor.




The three correlated items just listed, according to Cattell (1965), arerelated to the dimension “cool versus warm,” with a true-true-false patternof responses indicating a “warm” personality. (Cattell decided on this labelsimply by reading the items, as you just did). The factor represented by theseitems, therefore, is “warm-cool,” or, if you prefer to name it by just one pole,“warmth.” These three items now can be said to form a warmth scale. Tomeasure this dimension in a new participant, you would administer thesethree items, as well as whatever other items in your original list correlatedhighly with them, discarding the rest of the thousands you started with.

Factor analysis has been used not only to construct tests, but also todecide how many fundamental traits exist—how many out of the thousandsin the dictionary are truly essential. Various analysts have come up with dif-ferent answers. Cattell (1957) thought there were sixteen. Eysenck (1976)believed there are just three. More recently, McCrae and Costa (e.g., 1987)claimed five; theirs is perhaps the most widely accepted answer at present.These five traits—sometimes called the “Big Five”—are extraversion, neu-roticism, conscientiousness, agreeableness, and openness (see Chapter 7).

At one point some psychologists hoped that factor analysis might pro-vide an infallible mathematical tool for objectively locating the most im-portant dimensions of personality and the best questions with which tomeasure these dimensions. They were disappointed, however. Over the years,it has become clear that the factor analytic technique for constructing per-sonality tests and for identifying important dimensions of personality is lim-ited in at least three important ways (Block, 1995).

The first limitation is that the quality of the solution you get from a fac-tor analysis will be limited by the quality of the items you put into it in thefirst place—or, as they say in computer science, “GIGO” (“garbage in,garbage out”). In theory, a factor analysis requires an initial set of items thatare fairly representative of the universe of all possible items. But how are yousupposed to get that? Although most investigators do the best they can, it isalways possible that some types of items are overrepresented in the initialsample of items and that other, important types are left out. If either of thesethings happens—and there is no way to ensure that both do not—then theresults will provide a distorted view of which factors of personality are really important.

A second limitation of the factor analytic approach is that once the com-puter has identified a cluster of items as being related statistically, a humanpsychologist must still decide how they are related conceptually . This processis highly subjective, and so the seeming mathematical rigor and certainty of factor analysis are to some extent an illusion. The items used earlier, forexample, could be named “warmth,” but they could just as easily be called“sociability” or “interpersonal positivity.” Any choice between these labels isa matter of taste as much as of mathematics or science. And disagreements




over labels are common. For example, the very same factor called “agree-ableness” by some psychologists has been called “conformity,” “likability,”“friendliness,” or “friendly compliance” by others. Another factor has beencalled, variously, “conscientiousness,” “dependability,” “super-ego strength,”“constraint,” “self-control,” and even “will to achieve.” These labels are allreasonable, and there are no firm rules for choosing which one is best (Berge-man et al., 1993).

A third limitation of the factor analytic approach is that sometimes thefactors that emerge do not make sense, sometimes not even to the psychol-ogist who did the analysis. Years ago, the psychologist Gordon Allport com-plained about one such failure to make sense of a factor analysis:

Guilford and Zimmerman (1956) report an “unidentified” factor, called C2,

that represents some baffling blend of impulsiveness, absentmindedness, emo-tional fluctuation, nervousness, loneliness, ease of emotional expression, and

feelings of guilt. When units of this sort appear [from factor analyses]—and I

submit that it happens not infrequently—one wonders what to say about

them. To me they resemble sausage meat that has failed to pass the pure food

and health inspection. (Allport, 1958, p. 251)

Matters are not usually as bad as all this. But it is not unusual for thepersonality dimensions uncovered by factor analysis to be difficult to nameprecisely (Block, 1995). It is important to remember that factor analysis is astatistical rather than a psychological tool. It can identify traits or items thatgo together. As in the example just described, however, figuring out the psy-chological meaning of this grouping will require some difficult psychological thinking.

Factor analysis continues to have important uses. One, which is describedin Chapter 7, is to help reduce the long list of personality traits from the dic-tionary down to an essential few that are really useful. Another importantuse is for the refinement of personality tests, in conjunction with the othertechniques of test construction. It is useful to know how many different clus-ters of items are in a new personality test because some tests turn out to mea-sure more than one trait at the same time. A routine step in modern testdevelopment, therefore, is to factor analyze the items (Briggs & Cheek, 1986).

THE EMPIRICAL METHOD

The empirical strategy of test construction is an attempt to allow reality tospeak for itself. In its pure form, the empirical approach has sometimes beencalled “dust bowl” empiricism. The term refers to the origin of the techniqueat Midwestern universities (notably, Minnesota and Iowa) during theDepression, or “dust bowl,” years of the 1930s. Intentionally or not, the termalso serves to remind one of how “dry” or atheoretical this approach is.




Like the factor analytic approach described earlier, the first step of theempirical approach is to gather lots of items. The methods for doing this canbe just as eclectic—or haphazard—as described earlier.

The second step, however, is quite different. For this step, you need tohave a group of participants who have already and independently beendivided into the groups you wish to detect with your test. Occupationalgroups and diagnostic categories are often used for this purpose. For exam-ple, if you wish to measure the aspect of people that makes them good andhappy religious ministers, then you need at least two groups of participants—happy, successful ministers and a comparison group. (Ideally the com-parison group would be miserable and incompetent ministers, but moretypically the researcher will settle for people who are not ministers at all.)Or, you might want a test to detect different kinds of psychopathology. Forthis purpose, you would need groups of people who have been diagnosed assuffering from schizophrenia, depression, hysteria, and so forth. A group of normal people—if you can find them—would also be useful for comparisonpurposes. Whatever groups you wish to include, their members must be iden-tified before you develop your test.

Then you are ready for the third step: administering your test items toall of your participants. The fourth step is to compare the answers given by the different groups of participants. If schizophrenics answer a certain groupof questions differently from the way everybody else does, those items mightform a schizophrenia scale. Thereafter, new participants who answer ques-tions the same way diagnosed schizophrenics did would score high on yourschizophrenia scale. Thus, you might suspect them of being schizophrenic,too. (The MMPI, which is the prototypical example of the empirical methodof test construction, was built using this strategy.) Or, if successful ministersanswer some items in a distinctive way, these items might be combined intoa minister scale. New participants who score high on this scale, because they answer the way successful ministers do, might be guided to become minis-ters themselves. (The SVIB was constructed this way.)

The basic assumption of the empirical approach, then, is that certainkinds of people have distinctive ways of answering certain questions on per-sonality inventories. If you answer questions the same way that members of some occupational or diagnostic group did in the original derivation study,then you might belong to that group, too. This philosophy can be used evenat the individual level. The developers of the MMPI published an atlas, orcasebook, of hundreds of individuals who took this test over the years (Hath-away & Meehl, 1951). For each case, the atlas gives the person’s scoring pat-tern and describes his or her clinical case history. The idea is that a clinicalpsychologist confronted with a new patient can ask the patient to take theMMPI, and then look up those individuals in the atlas who scored similarly




in the past. This might provide the clinician with insights or ideas that he orshe might not otherwise have gotten.

Items are selected for empirically derived personality scales solely on thebasis of whether they are answered differently by different kinds of people.Then, the scale is cross-validated: If the test can predict behavior, diagnosis,or category membership in new samples of participants, it is deemed ready for use. At each step of the process of test development, the actual contentof the items is purposely ignored. In fact, empirical test constructors of theold school sometimes prided themselves on never actually reading the itemson the tests they developed!

This lack of concern with item content, or with what is sometimes calledface validity, has four implications. The first is that empirically derived tests,unlike other kinds, can include items that seem contrary or even absurd. AsI mentioned a few pages back, the item “I prefer a shower to a bath” whenanswered true is correlated with empathy. Similarly, the item “I like tallwomen” tends to be answered true by impulsive males. Why? According tothe empirical approach to test construction, the reason does not matter inthe slightest.

Consider some other examples. People with psychopathic personalities(who have little regard for moral or societal rules and lack “conscience”) tendto answer false to the item “I have been quite independent and free fromfamily rule.” Paranoids answer false to “I tend to be on my guard with peo-ple who are somewhat more friendly than expected.” One might surmise they do this because they are paranoid about the question itself, but they alsoanswer true to “I believe I am being plotted against.” Even though the rela-tionship between item content and meaning seems counterintuitive, a faith-ful adherent to the empirical strategy of test construction is not supposed toeven care.

Here are a few other examples (all taken from the excellent discussionin Wiggins, 1973): “I sometimes tease animals” is answered false by depres-sives; “I enjoy detective or mystery stories” is answered false by hospitalizedhysterics; “My daily life is full of things to keep me interested” is answeredtrue by dermatitis patients; “I gossip a little at times” is answered true by people with high IQs; “I do not have a great fear of snakes” is answered falseby prejudiced individuals, whose prejudice apparently extends even to rep-tiles. Again, in each case the indicated people are more likely to answer in theindicated direction; they do not always do so.

A second implication of this lack of concern with item content is thatresponses to empirically derived tests are difficult to fake. With a personal-ity test of the straightforward, S-data variety, you can describe yourself theway you want to be seen and that is indeed the score you will get. But becausethe items on empirically derived scales sometimes seem backward or absurd,




it is difficult to know how to answer in such a way as to guarantee the score you want. This is often held up as one of the great advantages of the empir-ical approach.

The psychologist interpreting a person’s responses to an empirically derived personality test does not really care whether the person told the truthin his or her responses. The literal truth does not matter because the answersto the questions on such a test are not considered important for their ownsake. They matter only as indicators of what group the person belongs to.Thus, empirically derived personality tests give B data instead of S data.

The third implication of the lack of concern with item content is thateven more than tests derived through other methods, empirically derivedtests are only as good as the criteria by which they are developed or againstwhich they are cross-validated. If the distinction between the different kindsof participants in the derivation or validation sample was not drawn cor-rectly, the empirically derived test will be fatally and forever flawed. Forexample, the original MMPI was derived by comparing the responses of patients who had been diagnosed by psychiatrists at the University of Min-nesota mental hospitals. If those diagnoses were incorrect in any way, thendiagnostic use of the MMPI will only perpetuate those errors. The theory behind the SVIB is that if the participant answers questions the way a suc-cessful minister (or member of any other occupation) does, he or she toowill make a successful minister (or whatever) (Strong, 1959). But perhapsthe theory is false—perhaps new ministers who are too much like the oldministers will not be successful. Or perhaps some of the “successful” minis-ters in the derivation sample were not really successful. Such difficultieswould undermine the SVIB as a source of vocational advice.

A more general problem is that the empirical correlates of item responseby which these tests are assembled are those found in one place, at one time,with one group of participants. If no attention is paid to item content, thenthere is really no valid reason to be confident that the test will work in a sim-ilar manner at another time, in another place, with different participants.The test must be continually validated and revalidated over time, for differ-ent geographic regions, and for different kinds of participants (e.g., those of different ethnicity or gender mix than the original derivation sample). A par-ticular concern is that the empirical correlates of item response might changeover time. The MMPI was developed decades ago and was revised only recently (Butcher, 1999). The revision will take a long time to revalidatebefore psychologists can be confident it works the same way the old test did.

The fourth and final implication of the empirical approach’s lack of con-cern with item content is that it can cause serious public relations and evenlegal problems—it can be difficult to explain to a layperson why certain ques-tions are being asked. As was mentioned during the discussion of rationally designed tests, face validity is the property of a test seeming to measure what




it is supposed to measure. A similar property is content validity, in whichthe content of the test matches the content of what it is trying to predict. Forexample, the WPDS had content validity for the prediction of psychopathol-ogy; the MMPI does not. Most psychologists would consider this differenceto work to the advantage of the MMPI, but it also has caused problems. Alack of content validity not only can lead to skeptical reactions among thepeople who take the test, but also can raise legal issues that have becomeincreasingly troublesome in recent years.

The original MMPI, for example, contained questions about religiouspreference and health status (including bowel habits). According to somereadings of antidiscrimination law, such questions can be illegal in appliedcontexts. In 1993, Target stores had to pay a two-million-dollar judgment to job applicants to whom it had (illegally, the court ruled) asked questions suchas “I am very strongly attracted to members of my own sex,” “I have neverindulged in unusual sex practices,” “I have had no difficulty starting or hold-ing my urine,” and “I feel sure there is only one true religion” (Silverstein,1993). (These items all came from a test called the Rodgers Condensed CPI-MMPI Test, which included items from both the MMPI and the CPI.)

As Target found to its dismay, it can be difficult for users of the MMPI,CPI, or similar tests to explain to judges or congressional investigating com-mittees that they are not actually manifesting an unconstitutional curiosity about religious beliefs or bowel habits, and that they are merely interested inthe correlates of the answers to these items. Target may have asked these ques-tions not because its management cared about sexual or bathroom habits, butbecause the scale scores built from these items (and many others—the test they used had 704 items) showed validity in predicting job performance, in this caseas a security guard. Indeed, reports from tests like these typically include only total scores and predictions of job performance, so the responses to the indi-vidual items might not have even been available to Target’s personnel depart-ment. Nevertheless, in many cases it would not be difficult for an interestedtest conductor to extract applicants’ responses to individual questions and thento discriminate illegally on the basis of religious or sexual orientation or healthstatus. For this reason, developers of new tests, such as the MMPI-2, areattempting to omit items of this sort, while still trying to maintain validity.

Developers and users of empirically derived tests have come almost fullcircle over the last fifty years or so. They began with an explicit and some-times vehement philosophy that item content does not matter; all that mat-tered was the external validity, or what the test items could predict. As thesocial and legal climate has changed over time, however, developers of empir-ical tests have been forced, to some degree against their will, to acknowledgethat item content does matter after all. Regardless of how their responses areused, individuals completing any kind of personality test are revealing thingsabout themselves.




A COMBINATION OF METHODS

In modern test development, a surprisingly large number of investigators stilluse a pure form of the rational method: they ask their participants the ques-

tions that seem relevant, and hope for the best. The factor analytic approachstill has a few adherents. Pure applications of the empirical approach are raretoday. The best modern test developers use a combination of all threeapproaches.

A good example is the way Douglas Jackson developed the Personality Research Form (known as the PRF, of course). He came up with items basedon their apparent relevance to the theoretical constructs he wished to mea-sure (the rational approach), administered them to large samples of partic-ipants and factor analyzed their responses (the factor analytic approach), and

then correlated the factor scores with independent criteria (the empiricalapproach) (see Jackson, 1967, 1971). Jackson’s approach is probably close toideal. The best way to select items for a personality scale is not haphazardly,but with the intent to sample a particular domain of interest (the rationalapproach). Factor analysis should then be used to confirm that items thatseem similar to each other actually elicit similar responses from real partic-ipants (Briggs & Cheek, 1986). Finally, any personality measure is only asgood as the other things with which it correlates or can predict (the empir-ical approach). To be worth its salt, any personality scale must show that it

can predict what people do, how they are seen by others, and how they farein real life.

Purposes of Personality Testing

If we can assume that a personality test has a modicum of validity, then afurther question must be considered: How will this test be used? This is animportant question that some psychologists, engrossed in the techniques anddetails of test construction, may forget to ask. The answer one reaches haspractical and ethical implications (Hanson, 1993).

The most obvious uses for personality tests are those to which they areput by the professional personality testers—the ones who set up the boothsat APA conventions—and their customers. The customers are typically orga-nizations such as schools, clinics, corporations, or government agencies thatwish to know something about the people they encounter. Sometimes thisinformation is desired so that, regardless of the score obtained, the personwho is measured can be helped. For example, schools frequently use tests tomeasure vocational interests to help their students choose careers. A clini-cian might administer a test to get an indication of how serious a client’sproblem is, or to suggest a therapeutic direction.




Sometimes the motivation for testing is a little more selfish. An employermay test an individual’s “integrity” to find out whether he or she is trust-worthy enough to be hired (or even to be retained), or may test to find outabout other personality traits deemed relevant to future job performance.The Central Intelligence Agency (CIA), for example, routinely uses person-ality testing when selecting its agents (Waller, 1993).

Reasonable arguments can be made for or against any of these uses. By telling people what kind of occupational group they most resemble, voca-tional-interest tests provide potentially valuable information to individualswho may not know what they want to do (Schmidt, Lubinski, & Benbow,1998). On the other hand, these tests are based on the implicit theory thatany given occupation should continue to be populated by individuals likethose already in it. For example, if your response profile resembles thoseobtained from successful mechanics or jet pilots, then perhaps you should consider being a mechanic or a jet pilot. But this approach, although it seemsreasonable, also could keep occupational fields from evolving and keep cer-tain individuals (e.g., women or members of minority groups) from joiningfields from which they traditionally may have been excluded. For example,an ordinarily socialized American woman may have outlooks or responsesthat are very different from those of the “typical” garage mechanic or jetpilot. Does this mean that women should never become mechanics or pilots?

Similarly, many “integrity” tests administered in pre-employmentscreenings seem to yield valuable information. In particular, they often pro-vide good measures not so much of integrity, but of the broader trait of con-scientiousness. That is, people who score high on integrity scales often also

PURPOSES OF PERSONALITY TESTING 135

“Here’s a little test I like to give prospective employees, Phil:

let’s see if you can grab this twenty-dollar bill before I can.”



score high on the trait of conscientiousness, which is highly correlated withlearning on the job and with performance in many fields (Ones, Viswesvaran,& Schmidt, 1993; see Chapter 7 for a more detailed discussion).

Still, many of these integrity tests ask questions about minor past offenses(e.g., “Have you ever taken office supplies from an employer?”), which putsnonliars into the interesting dilemma of whether they should admit pastoffenses, thereby earning a lower integrity score for being honest, or deny them and thereby earn a higher integrity score for having lied. Liars, by con-trast, experience no such dilemma; they can deny everything and yet earn anice, high integrity score.

A more general class of objections is aimed at the wide array of person-ality tests used by many large organizations, including the CIA, major auto-mobile manufacturers, phone companies, and the military. According to onecritic, almost any kind of testing can be objected to on two grounds. First,tests are unfair mechanisms through which institutions can control individ-uals—by rewarding those with the institutionally determined “correct” traits(such as high conscientiousness) and punishing those with the “wrong” traits(such as low conscientiousness). Second, perhaps traits such as “conscien-tiousness” or even “intelligence” do not matter until and unless they aretested, and in that sense are “constructed” by the tests themselves (Hanson1993). Underlying these two objections seems to be a more general sense,which I think many people share, that there is something undignified or evenhumiliating about submitting oneself to a test and having one’s personality described by a set of scores.

All of these objections make sense. Personality tests—along with otherkinds of tests such as intelligence, honesty, and even drug tests—do func-tion as a part of society’s mechanism for controlling people, by rewardingthe “right” kind of people (e.g., those who are intelligent and honestand don’t do drugs) and punishing the “wrong” kind. It is also correct tonote the interesting way in which a trait can seem to spring into existence assoon as a test is invented to measure it. Did traits like “parmia” or even “self-monitoring” (see Chapters 6 and 7) exist before there were scales to mea-sure them? Perhaps they did, but you can see the argument here. Finally, itis undeniably true that many individuals have had the deeply humiliatingexperience of “voluntarily” (in order to get a badly needed job) subjectingthemselves to testing, only to find out that they somehow failed to measureup. It’s bad enough to lack the right skills or relevant experience. But to bethe wrong kind of person? That hurts.

On the other hand, these criticisms are also overstated. A relatively minorpoint to bear in mind is that, as we have seen in the preceding sections of thischapter, personality traits are not merely invented or constructed by the processof test construction, they are also to an important degree discovered . That is,




the correlates and nature of a trait measured by a new test cannot be presumed,they must be discovered empirically by examining what life outcomes or otherattributes of personality the test can predict. For example, psychologists didnot know until after doing research that integrity tests are better measures of conscientiousness than of integrity, and this finding forced them to changetheir conception of what they incorrectly had assumed were honesty tests. Thefact that the interpretation of psychological tests depends critically on the gath-ering of independent data undermines the argument that personality traits, orany other psychological measurements, are mere social constructions.

A more basic and important point is that the criticisms that viewpersonality testing as undignified or unethical are criticisms that, when con-sidered, appear rather naïve. These criticisms seem to object to the idea of determining the degree to which somebody is conscientious, or intelligent,or sociable, and then using that determination as the basis of an importantdecision (e.g., employment). But if you accept the fact that an employer isnot obligated to hire randomly anybody who walks through the door, andthat the employer only uses good sense in deciding who would be the bestperson to hire (if you were an employer, wouldn’t you do that?), then youalso must accept that traits like conscientiousness, intelligence, and socia-bility are going to be judged. The only real question is how . What are thealternatives? One alternative seems to be for the employer to talk with theprospective employee and try to gauge his or her conscientiousness by howwell his or her shoes are shined, or hair is cut, or some other such clue.(Employers frequently do exactly that.) Is this method an improvement?

PURPOSES OF PERSONALITY TESTING 137

“Give me a hug. I can tell a lot about a man by the way he hugs.”



You may argue that you would rather be judged by a person than by acomputer scanning a form, regardless of the demonstrated invalidity of theformer and validity of the latter (Ones, Viswesvaran, & Schmidt, 1993).Although that is a reasonable position, it is important to be clear about thechoice being made. The choice cannot be for personality never to be judged—that is going to happen even if all of the tests are burned tomorrow. The only real choice here is this: How would you prefer to have your personality judged?

Summary

Any characteristic pattern of behavior, thought, and emotional experiencethat exhibits relative consistency across time and situations is part of an in-dividual’s personality. These patterns include personality traits as well aspsychological attributes such as goals, moods, and strategies. Personality assessment is a frequent activity of industrial and clinical psychologists andresearchers. Everybody also performs personality assessments of the peoplethey know in daily life. An important issue for assessments, whether by psychologists or by laypersons, is the degree to which those assessments arecorrect. This chapter examined how psychologists’ personality tests are con-structed and validated. Some personality tests comprise S data and otherscomprise B data, but a more commonly drawn distinction is between projective tests and objective tests. Projective tests try to gain insight into per-sonality by presenting participants with ambiguous stimuli and recordinghow the participants respond. Objective tests ask participants specific ques-tions and assess personality on the basis of how the participants answer.Objective tests can be constructed by rational, factor analytic, or empiricalmethods, and the state of the art is to use a combination of all three meth-ods. Some people are uncomfortable with the practice of personality assess-ment because they see it as an unfair invasion of privacy. However, becausepeople inevitably judge each other’s personalities, the real issue is how per-sonality assessment should be done—through informal intuitions or moreformalized techniques.


Documents

Psychology) (English E-Book) Personality Assessment - Perso