Download - o FACTOR THEORY OF PERSONALITY WITH - apps.dtic.mil · Cattell (1946, 1950, 1957, 1965, 1973 ) has devoted a major segment of his career to the development and refinement of his conception

Transcript

AR05-6

o FACTOR THEORY OF PERSONALITY WITH

PARTICULAR EMPHASIS ON CATTELL'S 16PF

-A LITERATURE REVIEW-

BY

CAPTAIN J.L. EAVES

1st PSYCHOLOGICALRESEARCH

UNIT DTICUNIT -SE19E2E E . 1 D9o

\TH4E L$NrTED STATES NATNlONALTHIALINFORMATIO

S1RIO

,a AUTHORISO-t TO I IE

,t~ODI),LCF AND SELL TI EO?2

AL 149 Depnrtment of Oefae -Page CtsificstonRevised May 86 DOCUMENT CONTROL DATA UNCLASSIFIED

Ia. A R Number lb. Establishment Number 2. Document Date 3. Tak Number

AR-005-064 PSRU-RN-2-89 March 19894. Title 5. Socunty Clausification 6. No. PagesFactor Theory of Personality with [P/8" proprate classfication in

box/s ie Secr(5). Cofidenoal(C)Particular Emphasis on Cattell's Restricted (R). Unclasfied (U)] 3116PF - A Literature Review 7. No. Raft

8. Author(s) 0]document Eltite[ abstract 7

Captain J.L. Eaves 9. Downgrading/Delimiting Instructions

10. Corporate Author and Address1 Psych Research UnitNBH 3-39 11. Office/Postion responsible for -Northbourne HouseTURNER ACT 2601 Sponsor .....................................................................................

AUSTRALIA Security .....................................................................................Dow ngrading ............................................................................

App roval ....................................................................................

12. Secondary Distribution (Of this document)

Approved for Public Release

Overseas enquiries outside stated limitations should be referred through the Document Exchange Centre, Oefence InformationServAce Branch. Department of Defence, Cambell Park, CANBERRA, ACT 260113s. This document may be announced in catalogues and awarenesa services available to ...........

No Limitations13b. Citation for other purposes (i.. casual announcement) may be [3 unrstricted or [ as for 13a

14. Descriptors 15. COSATI Group

Personality 0092BTesting

16PFLiterature ReviewFactor Theory

16. Abstract-- -f The concept of personality is widely recognised as being central inpsychology, yet its nature and the ways in which it can be defined andmeasured are questions on which psychologists are in considerabledisagreement. While theorists tend to disagree over definitions ofpersonality, they tend to agreee that in order to perform a systematicexploration of personality's relation to other variables, a definite setof personality factors need to be specified. Researchers engaged in thequestion of the number of factors in the personality sphere seem to divideroughly into three positions: Two factors, five-to-eight factors, and 13-to-18 factors. Cattell devoted a major segment of his career to thedevelopment of the Sixteen Personality Factor Questionnaire (16PF). Withthe continuing use of Cattell's 16PF by the Australian Army Psychology Corpscomes the need to review the available literature with a focus on Its use

1-

r Pace Classification

UNCLASSIFIED 2

This page is to be used to record information which is required by the Establishment for its own use butwhich will not be added to the DISTIS data unless specifically requested.

16. Abetract (Contd)in personnel selection. The 16PF purports to measure 16 distinctpersonality traits, yet there is no evidence of the convergent validityof the 16 scales apart from factor loadings and what evidence there is ofdiscriminant validity suggests the primary traits are not clearlydifferentiated. The psychometric adequacy of the test must bequestioned. (b. -

17. tmprint

18. Document Series and Number 19. Cost Code 20. Type of Report and Period Covered

21. Computer Programs Used

22. Establishment File Ref(s)

23. Additional Information (As required)

Research Note 2/89

FACTOR THEORY OF PERSONALITY WITH PARTICULAR EmFHASIS ON CATTELL' S 16FF

- A LITERATURE REVIEW -

by

Captain J.L. Eaves

March 1989

This Directorate of Psychology publication has been prepared by1st Psychological Research Unit and is authorized for issue by DPSYCH-A.

Lieutenant ColonelCamianing Officer

I let Psychological Research UnitNBH 3-44Northbourne House

ISSN 0156-8817 TURNER ACT 2601

Table of Contentsaa~

Table of CoTtenots

List of Tables iii

5 Abtract 3

Definition of Persoality 5

Cattell's Total Personality Cm-cept 7

*Aim8

Literature Rview 8

S r 24

Bibliography 27

/

bI

Acoesion For

ID71f . ;

Ufl'a ......

!Avai 1l it7 , qn,Dist

,By - -- t.,, o

:~tt '. 'eS1c~3

List of Tables

Table Pa

1 Fiw-to-Kight Factor Squivalerice Table 15

Abstract

The concept of personality is widely recognised as being centralin psychology, yet its nature and the ways in which it can be defined andmeasured are questions on which psychologists are in considerabledisagreement. While theorists tend to disagree over definitions ofpersonality, they tend to agree that in order to perform a systeenticexploration of personality's relation to other variables, a definite set ofpersonality factors needs to be specified. Researchers engaged in thequestion of the number of factors in the personality sphere seem to divideroughly into three positions: TNo factors; five-to-eight factors; and 13-to-18 factors. Cattell devoted a major segment of his career to thedevelopment of the Sixteen Personality Factor Questionnaire (16PF). Withthe continuing use of Cattell's 16PF by the Australian Army PsychologyCorps (AA Psych Corps) comes the need to review the available literaturewith a focus on it's use in personnel selection. The 16PF purports tomeasure 16 distinct personality traits, yet there is no evidence of theconvergent validity of the 16 scales apart from factor loadings and whatevidence there is of discriminant validity suggests the primary traits arenot clearly differentiated. The psychometric adequacy of the test must bequestioned.

The findings and views expressed in this report are the resultsof the author's research studies and are not to be taken as official policyor opinion of the Department of Defence (Army Office).

Definition of Personality

Th concept of personality is widely recognised as being centralin Psychology, Yet its nature and the ways in which it can be defined andmeasured are questions on which psychologists are in considerabled---greant. There are few words in the English language which have sucha fascination, even for the general public, as the term personality.According to Webster's Dicticnary (1961) there are a number ofalternatives:

The quality or state of being a person and not anabstraction, thing, or lower being.

The complex of characteristics that distirgis ap rticular individual or characterises him in hisrelationships with others.

The totality of an individual's emergent tendencies to actor behave especially self-consciously or to act on, Interactwith, perceive, react to, or otherwise meaningfullyinfluence or experience his environment.

The organisation of the individual's distinguishingcharacter traits, attitudes, or habits. (p. 1687).

' ~These alternative definitis comvey some of the contradictionsthat plague the area of personality. The first alternative represents arather global, philosophical point of view and one wtich psychologistsconsider too broad. The secc-Ix definition does provide a firmr foundationfor research, but exctly what characteristics or coplex ofcharacteristics should be ememined? At one time or another al ot everyhumn attribute has been studied as a distinguishing characteristic, forinstance, Sheldon's (1940) body-type theory of personality. The thirddefinition emphasises the study of behaviour. It represen the view thatIf psychology is to be scientific, it must only deal with what isobservable. Behaviour is observable, hence psychology must be the study ofbehaviour. However, cn could add sam confusion by asking whetherbehaviour is defined as including thinking. The fourth definition suggestsconsideratio of individual character traits or habits. These areconvenient map to denote mntal states that presumably underlie variousbehaviour patterns, yet are difficult to measure because they are hard todefine.

It Is obvious, therefore, that the word "personality" is used invarious senses. However, mot of these popular meanings fall under one oftwo hedi . The first usage equates the term to social skill oradrotnes. By this, Hall and Lidzey (1957) suggest that an Individual'spersonality is aessed by the effectiveness with which he is able toelicit positive reactions from a variety of persons under different

-6-

circumstances. It is in this se that schools which specialise in

glamrIsing the female intend the term when they refer to corses in"personlity training". Likewise. the teacher who refers to a student aspresenting a personality problem is probably indicating that his socialskills are not deqiate to maintain satisfactory relations with his fellowstudents and the teacher. The second uag ccmiders the personality ofthe in ividual to be described by the most outstanding or salientlmlpression which he creates In others. Thus, Hall and Lindzey (1957)

suggest a person may be said to bave n "ar esee persomality" or a"ubmi ersonalty"'. In each camse the observer selects an attributeor quality which is highly typical of the subject and wich is presmably

isan poItant part of the overall I mon wich he crat In others, andhis personality is Identified by this term. It is clear that there is anelement of evaluation in both usages.

While the diversity in ordinary use of the word personality mayseem considerable, it is overshad by the variety of meanings with whichthe psychologist has endowed this term. Soe theorists list the traitsconsidered to be of primary Importance In describing the individual andsugest that personality ccnists of these. Other definiticns placeprimary emasis upon the integrative or organisational function ofpersoality. Such definitions suggest that personality is the orgaisationor pattern that is given to the various discrete behaviours of theindividual, or else they suggest that the organisation results from thepersonality which is an active force within the individual. Persmality isthat which gives order and conrue to all the different kinds ofbehaviour in which the individual engages. Other theorists have chosen toemphasise the function of personality in mediating the adjusment of theindividual to daily activities. In that way, perscnlity consists of thevaried and yet typical efforts at adjuset which are carried out by theindividual.

lore time could be spent dealing with the problem of definingpersonality, however, it sems evident that no substantive definiton ofpersonality can be applied with generality. Hall and Lirndzey (1957)cclluded that the my in wich a given individual will define personalitywill depend coupletely upon his particular purpose. That is, personalityis defined by the particular empirical concepts; which are a part of thetheory of personality used by the observer. Personality consistsconcretely of a set of values or deriptive terms which are used todescribe the individual being studied according to the variables ordimensions which occupy a central position within the particular theoryutilised. The definition problem can be pursued in much more detail, Vetit is evident that while theorists tend to disagree over definitions ofpersonality, they tend to agree that in order to perform a systemticexploration of personality's relation to other variables, one must be ableto specify the universe of traits that comprise personality. MoCrae andCosta (1985) go on to say that, without a definite set of personalityfactors, research efforts are fragmented.

An analogy with cheistry, referred to by Mershon and Corsuch(1986) helps explain why a set of agreed upon personality factors isneeded. A well-defined ta=my of personality would d lineate therelations between personality factors an the periodical table of the

-7-

elem ts delineates interrelationships of chemical elmits. Thedifference between an alphabetical listing of the elemnts and theperiodical table is that the table allows chemists to express all capoun sin a common vocabulary. That is, any chemical compound can be viewed as anamalgam of bisic elements, and compouds that are similar in their elementshave other important similarities. It is hoped that a tamnmy of

A personality wold Introduce similar clarity Into the study of personality.

Cattell'a Total Perscality Concept

Cattell (1946, 1950, 1957, 1965, 1973 ) has devoted a majorsegment of his career to the development and refinement of his conceptionof the noral personality sphere and the primary instrument for itsmiesmument, the Sixteen Personality Factor QuestionrAire (16FF). Cattell(1957) describes the 16PF as the psychologist's aswer, in thequestionaire realm, to the demand for a test giving fullest information inthe shortest time about most personality traits. He goes on to say that itis not merely concerned with same rro concept of neurotIcism or"adjustment," or same special kind of ability, but sets out to coverprecisely all the main dimensions along whtich people can differ, accordingto basic factor analytic research.

In applied fields such as clinical, educticnal and industrialpsychology, Cattell suggests that the practitioner usually has a feelingthat certain traits, for example, emotional stability, will be soreimportant and predictive than others. However, Cattell (1957) goes on tosuggest that In doing this the practitioner may overlook some other,unconsidered, personality dimenmion that is equally important. Cattell(1957) therefore, suggests that the best my to begin is to take cognizanceof the TOTAL pers ality, In all its dimensions.

Despite the use of a coprehensive plan and externivepsychometric studies, Cattell's methods and results have not been totallyaccepted by other personality psychlogists, for example, Eysenck andKyseeck (1969) and Bss and Firn (1987). Furthermore, many of thedetractors used the seame psychometric procedure, that is, factor analysis,to cast doubt on the conception of the normal personality that Cattell usedto comtruct the 16PF. Very little evidence regarding Cattell's views ingeneral, and the 16PF in particular, is seen to be ambiguous.Investigations either support him unequivocally or they find practically novalue in the results of his vast research programme. There seem to be nomiddle ground. Invariably the camps are composed of Cattell's students andcolleagues on one hand, and almost everyone else on the other. Thispolarisatin sem due in good part to Cattell the man; as Berg (1967)states:

Cattell has little patience with soft-headed psychologists,whom he has characterised as refugees from the naturalsciences; he ha persisted In recating everything frompersonality traits to psychometric theory into his amsystem, with its particular esoteric vocabulary; and he isgenerally argumentative and critical of the work of other

qt

,t

psychologists, especially those who do not support hisvisa. (p. xi).

Of course, Cattell ' own contributions are volmircus and he ranks with hischief competitors, d ilford and Eysenck, as one of the most productivepsychometric researchers in this area of specialisation.

Aim

The purpose of this article is to review and summarise theavailable evidence and opinions surrounding the factor theory ofperso-allty with particular ehasis on Cattell's 16PF in an attempt toasce its predictive validity as a masuremnt of personality.

Literature Review

Number of Factors in the PersoraI_ S"re

There is a continuing controversy over the appropriate number ofsonality factors Different theorists and researchers (Eysenck and

Eysenck, 1969), have proposed differing numbers of factors as beingwncessary and sufficient to describe normal personality. New personality

factors and scales to measure them are continually being proposed (Mrshonard Gorsuch, 1986). Attempts to integrate and amalgamate the new factorsw•.th existing factors are rare, although inspection and factor descriptionsugest many factors "in different system differ in name only. In clinicalas well as research areas, the lack of agreed-upon personality factors is ahandicap. Mrshon and Gorsuch (1986), strongly state that wat is neededis mn useful information, where uselessly specific data on one handand overly general information on the other hand can be avoided.Researchers engaged in the question of the number of factors in thepersonality sphere seem to be divided into roughly three positions: Twofactors, five-to-eight factors and 13-to-18 factors.

The Two-Factor Position. Peterson (1965) and Eysenck (1971) aretwo major critics of Cattell's 16 Personality Factor system, maintainingthat there are only two normal personality factors worth study.Furthermore, they advocate virtually the same two factors but underdifferent names. The two factors Peterson (1965) proposed wereIntroversion-Extraversion and Adjustment. In developing his rationale fortwo factors of personality, Peterson (1965) discussed reductionism, choiceof factor level, "descriptive efficiency" of factors and factor invarianceas criteria for factor selection. These criteria are primarily based onfactor analytic theory. According to Peterson (1965), reductionism was amajor advantage of a two-factor system of personality; he argued that whenpeople talk about personality, their statements are gereral and revolvearound global concepts or categories such as introversion-extraverslon andadjustment. He cited W.tzel's 1963 study in which subjects, peers and

rents provided ratings on introversion-extraversion and adjustment andthese ratings were then correlated with the subjects' corresponding second-order factor scores on the 16PP (Cattell, Eber and Tatsuoka, 1980). The

correlation of Intro.ersion--ctraversion with Catatll's Exvia was r-.61 andof , just Mnt with Cattell ' Anxiety was r-. 45. Petersmn concludedWetzel's subjects' ratings were organisina Information similarly to that inthe 16 Fl, but more simply.

In his diussion of the generality and scope of personalityfactors, Peterson (1965) believed the choice of factor level ws animportant consideration. Factors vary in omplexity. The area of humanabilities illustrates broad and narrcm factors. Spearman (1904)investigated the very broad general intelligence factor. Tharstne (1938)brok this broad factor dow into seven to nine primary mental abilities.Continuing the process of identifying and specifying factors moreprecisely, Gu.lford (1975) identified thirteen intelligence factors.G.nrmlly, narrow factors can be suheud under a broader factor; broadfactors can be split into narrower factors. In a given situation, "theproces has to stop when a single general factor emerges at the top of thehierarchy, and instrmtal specifics apear at the bottom" (Peterson 1965,p. 48), or when orthogoal factors appear. Another my to describe a"narrow" factor is as a prizmary or first-order factor, and another way todescribe a "broad" factor is as a second- or third- or more-order factor.It was Peteracr's contention that two broad factors, Introversion-Extraversion and Adjustment, summarised available data in an infozativeyet concise way. He believes a broader factor wuld blur description bybeing too general, and narrower, or lower-order factors would confuse byproviding too much specific information.

Descriptive efficiency was another major criterion for factoranalysis cited by Peterson (1965). Exaination of variance plots of actualstudies, in which magnitude of the variance is plotted against successivefactors, Peterson argued, show that after the first two or three factorsare extracted, there is little variance remaining. He believes thatfactors extracted after the crucial first few have little descriptivepower, and are therefore inefficient according to Peterson's use of thedescriptive efficiency criterion. This also supports Peterson's use of twofactors to describe personality.

Another of Peterson's (1965) factor analytic criteria s factorinvariance. "Invariance" refers to the cortancy of factor content fro oneanalysis to the next. Where a factor for numerical skill, for instance,awpears, this should be regularly In evidence whenever the same or asimilar test battery is used with samples of the population. UsingCattell's data, Peterson (1965) concluded that the 16PF factors are veryweak when examined from the standpoint of this criterion. Peterson thentested the invariance of the two IW'F higher-order factors wi'ch helabelled "Introversicn-Fztraversion" and "Adjustment". These two factorswere quite invariant, and therefore preferable, according to Peterson.

Virtually synonymous with Peterson's Introversion-Dxtraversionn Ajustmet respectively are the two factors advanced by Eysmck &

Kysenck (1969) as necessary and adequate in describing personality, namelyExtraversion and Neuroticlsm. Eysenck (1971) maintained that the 16PF,rather than measuring 16 primany factors as claimed, (Cattell, Eter andTatsuoka, 1980), measures two. He bssed his claim on the fact that thefive 16IF factor scores measuring extraverslcn correlate highly, and

-10-

likwise the five IOP? factor s corms mesuring neuroticisi correlate veryhighly. Tron this he concluded that the 1P measures only two factors,and the test item which make up the other six factor were otherwiseallocated on a coletely raIsk bss to the Cattell factors. Eysenck,White and Soueif (1969) separately factor analysed approximately 100 itefrom the Guilford, Cartell and IVsenck questicnnaires. There areconsiderable similarities between the personality descrlptons given by thefaCtor-analysis based systms of Cattell, Guilford and Eysenck; thesesimilarities, hoever, appear only in the higher-order factors calledExtraversion-Introversion ad Neuroticiam-Stability by Eyeenck, and E via-Inva and Adjustment-Anxiety by Cattell. While the factors extracted atthis level from sets of questions contributed by these three authors arevirtually identical, there is little agremnt on primary factors. TheImplication Is that certain 16PF scales can be dropped without loss of anyessential information. Yet Cattell is quite specific In his claii:

The primary factors give one t infomation and we wouldadvocate higher strata contributors only as supplementaryocepts..... It is a mistake, gener lly to work at the

asaidary level only, for one certainly lome a lot ofvaluable information present Initially at the primary level,(Cattell, Eber and Tatsuo ta, 1980, p.111).

KEyenckc's position is equally clear, maintaining that second-order factorsare far more meaningful psychologically (Eysenck, 1967).

As Eysenck and Zysenck (1969) have pointed out, higher-orderfactors shg considerable agreenmt betwen the two models. As the sawereference explains, Eysenck believes that the distinction betmaen primariesand higher-order or superfactors is useful, although the allocation of afactor to a particular order (first, second or third, etc) is not. Thisdistinction between primary and higher order factors is loosely tied upwith his distinction between 'T' and 'C' factors, that is, tautologicalfactors (primary) and complex factors (higher order), wiich combine severaldifferent primarles on the basis of their intercorrelations.Psychologically, Eysenck believes tautological factors are of less Interestthan are complex factors, If only because logically the Items in T factorsare conected through similarity; the discovery of such a factor is hardlysurprising and any nmiber of such factors could be artificially created bysimply rewriting a particular questionnaire Item in several slightlydifferent ways. C factors are mad. up of dissimilar It, is of Itte.where the finding of a correlatio does not involve a tautology;sensitivity and nervouess are exmples of factors same ay towards the Cend of the continuum. Such factors are usually law easy to identify andae, and sam subjectivity is attached to such efforts, however, Eysenck

believes them to be much more Interesting psychologically.

Comrey and Duffy (1968) adinistered the Eysenck PersonalityInventory, the Cattell 16P and the Comrey Personality Inventory to 272volmteers. In developing his personality system, Oomrey worked withFactored H=Kpgneoue Item Dimensions (FHID) which consist of Item writtento measure a specific factor and wich have been sham to do 3o by factoranalysis. Eysenck and Cattell's factor scores mare correlated with scoresover mgeneous item groups which define the Coerey tes factors. This

-11-..

mstx wa factor analysed to relate the Eysenck and Cattell factor scoresto the factor structure underlying the Coey test. 7e EymeackNeurticiam, Comy N, amd Cattll's secori-order Anxiety factors appearedto umtch. h Eyseeck Introversion and the Comry Sies factors alsowatched. 7he 16 Cattell primary factors overlapped but did not match withthe Coorey factors. To the extent that the respective factors occupy theem sp, that is the sam hyperplane, the question arises as to wichset is to be c"" idered more useful and maningful psychologically, theCattell factors or the Coarey factos. 71h aw to this question nodobt depens on one's orientation and purposes. Ieferring to tvo an(1961), coniderable heterogeneity of content of Items appears on theCatell factors. 7hCcorey factors, on the other hand, are defined bysubstantially correlated F!I , which In turn consist of tm which arehomgeneo s, both with respec: to apparent content and inter-itemcorrelations. The Item defining a caorey factor therefore exhibit greaterhomogeneity thn those of the Cattell factors. Although these facts do notprove it, they suggest that the corey factors are more unitary than theCattell factors. If the relatively uncorrelated Item n the Cattaellfactors in fact define unitary factors, item statistics should be publishedwhich demonstrate this. Cattell has published a diagram (Cattall andTsujlcka, 1964) showing how it is theoretically posible for twoumnorrelated items to measure the same unitary factor, but he h ritpublished item data Ainich show that the bulk of his item do in factonstitute actual examples of such a phencmenon.

Another critical Investigation of the 16PF was the Item factoranalysis reported by Howarth and Br'm (1971). With a Canadian sample theyfound ten out of eighteen factors rotated could be interpreted and, ofthese, half could be regarded as having similar significance to the 16PFfactors when studied in a 10-factor Varimax rotation. Of the 34 highloadings on these factors, 18 belong to the 'proaprte Ixdividual 16Ffactors. Five factors have no representation at all, and the rmaining fewhave little or doubtful relation. 7 authors concluded that" ..... Cattell's questionnaire factor system has been developed on the basisof inadequate invest gation of primary factors .... " (p. 138). Eysenck(1971) seized upon these results and strongly stated conclusions of Nmarthand Browne's study and used them as a basis for a broad denmciation of the16F, including criticism of the MinaeWota Iltiphasic PersonalityInventory (DWI) and other competitors of his cma Eysenck PersonalityInventory (BI) for good mesure. Regarding the 16PF, he concluded that. ....... the investigator using this scale is in fact getting sixteenmeasures of doubtful anlnes, a h -aich are almost certainly nm-univocal .... " (p. 88).

Adcock and Adcock (1977) analysed a large sample of New Zealandsubjects in order to provide further evidence with regard to the validityof the 16PF (Form k). Eighteen factors suggested by the Screi Test wererotated to oblique simple structure and matched as in the 1972 Cattellpaper. In same cases the watch was better than that obtained by Cattell,but no match can be found for N or M, while nine Q4 variables, nine from Cand five fron 0 c together to define the beat represented factor, whichis clearly general emotionality or arniety. A separate C factor did,however, appear. While confirng the general factorial structure of the16FF this study suggests that for thIs New Zealand sample it is not an

-12-

adequate INzsUring instrument in its present form. In a crom-culturalcontext there seem to be a challenge to the reliability of the lapstructure across populations. Citing German and British (Saville andBlinihon 1976) stuies, Eyseack (1982) stated: "Relicaticr of Cattell'sfactor analytic studies in different countries have usually failed toproduce factors even reasonably similar to his Amerlcan solutins." (p.239).

7tlm areas of controversy have resulted in confusion toadequately ae- the 16F after 34 years of se. This indicates the needfor further application and analysis on the broadest scale possible, sothat acme idea of the replicability ad invarlance of factors may beobtained. Cross-cultural stuies should ccrnitute a vital part of suchwork. In attempting this, Aodel-anle, Abdel-Satter and hdek (198)performed a study to determine the factorial structure of the 16pF in two

Egyptian smpls of mle and female university students in an attempt totest the replicability of Cattell's 16FF with Egyptian subjects. In theirstudy they decided to administer the 16PF in addition to a les factoriallycmpllcated invemtory, the ymenck Personality Qusticrmare, (EpQ),(Mseck and BYsrxk, 1975). M1e NQ with its broader dimensionsm e usedas a set of markers leading tmud a better interpretatlon of the factorresolution of the 16WF in Egyptian subjects. The factorial analysis of theintercorrelatons betkeen the subscales of the 16F in this study did notreveal more than two significant factors. The corresponded to Eysenc'sNeurotlcism and Utraversion which In turn corresp to the first two ofCattell's (1967) second-order factors, Is Anxiety and &tvla. 7he resultsof this study are congmrent, either totally or partially, with previousr wrk done by oer investigators (Costa and McCrae, 1976). It seemsuiotresearch studies are only emphatic about two of the factors In the 16PFalbeit second-order factors, yet Cattell (1972) insists that the dependenceon the second order factors is a relatively inefficient procedure because,he believes, - information will certainly be missed. Prediction, hesuggests, carnot be carried out using secn-order factors to the sameextent as with first-order factors. Nevertheless, the concepts ofrePlicability and invariance of factors are highly relevant to the validityof any given theory or test of personallty. Tis repllcabillty of factors,their invariance sten smaple parameers change, is a basic requiremet forany reliable and valid maeaurement of personality. Cattell 's argument infavour of the definition and use of the primary source traits ofpersonality structure is mkekaed by lack of agreemet as to theirinvariance, replicability and generalizability. 11w primary factors whichhave been unearthed and used in personality research, such as Cattell'sfactors, fail because they are in fact not unitary, univocal combinationof items measuring the same fundamental trait. as is assumed by Cattell;they are not invariant across such differences as sex, age or education.Instead, ftsenck and KEyssck (1969) sugests that they appear to be halfarbitrary, half accidental conglomerations of items sharing functionalequivalence only to a limited degree.

Five-to- icht Factor Position. It is evident that the task ofprimary factor extraction is not nearing completion as some writers claim.In fact it has rdly begun . Proponents of a five-to-eight factorpositlon include Norman (1963), McCre and Costa (1985), Aoerth (1976),C~ey (1973) and Bedcr (1961). Each researcher factor analysed ne or

-13-

ore mjor data pools and coclued that five-to-eight factors werezuoassar ard sufficient to describe persoality, as opposed toCatte11's 16. Their factors inclued Mcmcavesion ad Anxiety but thesetwo wee not sufficient in their view. Following the lead of Alport and0fert (1936) an of Cattell (1946), Norm (1963) as that "naturall9n ug such as Enlish wold have evolved term for all fundamntalIndividual differences. An analysis of language would, therefore, providea c'90rw-ive model of personality traits", (tCrae and Costa, 1985, p.710). Nomn used peer rzination rating methods similar to Cattell's(1947) and found Only five factors In his factor analysis. VxCrse andCosta (1985) eained the Nora factors; they proposed changes inI ntrpre ttion of two factors, but also reded the use of five.FHAcrth (1976) rotated Cattell's (1947) unrotated factor matrix and tried10, sevn and five factor soluti. He conclde that a seven factorsolution w the best and that Cattell had overfactored. Another approachusing five-to-eight factors wm talkn by Digman and Tekmoto-Chock (1981).They re-analysed eight sets of data from Cattell, Digmen, Fiake, Norman andTupes and Christal. Digmen and Ta1eoto-Chock's purpose ws to determinehow many factors would be found if the Identical model of factor analysisand the Identical rotation method were used in anamlysing all the studies.In all cases five-to-elght factor solutions were preferred. It wasconcluded that:

The number of factors involved in these studies is aninteresting rnaber and suggests the psychological processeswtich are involved in the rating of personalitycharacteristics... various c ideratins of the encoding,storage, and retrieval processes in rmor are involved.The nmber five, nerhaps even seven as the upper limit,probmbly represents limits on hwman information processing(Digeun and TaWWto-Chock, 1981, p.165).

In developing his personality factor system, Comrey (1973) woriedwith FIBID. These FIHDs are smes*mt analagous to Cattell's "parcels" ofItems. Comaey does not directly factor the Items, but factors the FJBDs.He repor eight factors for personality. As can be seen fz Table 1, thefive-to-eight factors arrived at by different researchers are quitesimilar. It is interesting to note that factors aside fro xmtravrsionan:d Anxiety ce up almost as frequently as these two factors on thevariou researchers' list. One such factor is Agreeableness (Norman)/Cooperativenes (Ham th)/ COapliance (Digimn and Tabwmot-Chc). Anotherc n factor Is Ccmcentiouneaee (Norman; ft;Crae and Costa)/ Superego(iwrth)/ Will to Achieve (Dimn and Takmxoto-*ock). This similarity offive-to-eight factors tends to provide further support for the more thantwo facto position.

Further support for the five-to-eight factor position waprvided by Becker (1961) who factored the 16WF and concluded that ". .... atbest the 16W is measuring only eight factors ..... that is not to imply thatthere My not be more factors then eight In the personality questionnairedbuain, bit rather that, If they exist, they have not been measured withsufficient reliability or indeperdence for related factors to be revealedby the tests as presently scored". (p.402).

-14-

Inspection of Table 1 strc ly s tggests the consensus of factorsfound by diverse investigators using widely varying tecmiques and saples.Norman, Mcre and Costa, Hoerth, Digrn and TakIto-Chock, Comrey, andBecker, all concluded that five-to-eight factors described their data.Thu they hold that Peterson ad senk had too few factors. A five-to-eight factor personality systm, according to Digi ad Tak to-Chock(1961), my be momt only found because that is how people think. Theybelieve that this Is another eample of the seven plus/miim two ruleregarding processing bits of Informtion.

Other researchers (Goldberg, 1982; Peabody, 1984) have chieflyconcerned with the repLeentativeess and caiprehuulveness of the

five-factor model with respect to the natural language of traits, andothers have sought to provide a theoretical basis for the taxoy (Hogan,1983). 3cCrae and Costa's (1987) major concern has been the convergent anddiscriminent validity of the dimension of the five-factor model acrossinstrwments and observers. If the five-factor model is a reasonablerepresentation of human personality, it should be recoverable frquestionnaires as well as F adjectives and fran observer rating as wellas fran self-reports. 7his line of research adresses substantivequestins from the methodological perspective developed in the pest fewyears. One of the stru t arguments in favour of the five-factor modelhas been its appearance in both self-reports and ratings. Norman (1963)reported the structure in peer ratings. Goldberg (1982) shoed parallelstructures in both ratings and self-reports. As early as the 1960's,Cvergence across observers was also dmontrated (Borgatta, 1964).However, with a few exceptions, these studies used only adjective-ratingscales and few attempts were o to mare adjective factors with

dquest na that are more widely used in personalityresearch.

In a recent publication Mcrae and Costa (1985) examined the

zbzlence between adjective and quasticonnare formats to see if thesubstantive dimension of personality would be obtained n each. 7he

adjective-ratin instrument w an ectension of am devised by Goldberg(1982); the questionnaire w the Neurotician, Extraversion and Openness toEmperience (NMD) Inventory (HoCrae and Costa, 1983), which measures threebroad dimesions Identified in analyses of standard personality measures.Self-reports on five adjective factors were compared with both self-reportsand spouse ratings on the Inventory dimsions of neuroticim , extraverslonand openness to experience. In brief, the study showed that a version ofthe five-factor model could be recovered from the adjectives; that therewere clear c s for neuroticism and extraversion dimensionsacross the t; instruments; and that validity coefficients above .50 couldbe obtained with both self-reports and spouse ratings.

Fran the reviem above It is evident that fron time to time thecharge has arisen that the number of factors in the 16FF and, byimplication, n other tests In the 16PF series, is not sixteen.Pesearchers such as Canrey and Duffy (1968), Rysenck (1972), Guilford(1975), and Horth and Browne (1971) have concluded that fewer factors'equately explain the covariation amog the primary scales of the 16PF.But they have represented a broad range of opinion as to what the numbershould be. If the charge Is true, then the implication is that certain

f6

, , , , i.+i+

I - ++ -

+ !+++ i

UI

J

,,+.+..Li

4.

U,1,j

(

-" -4.-.

+ " .1+.

I £"

if ",+

h -,+'

_,-

+ + ++

+im m +t

S

++

+i)

al

I' . 'i

I 11

-16-

1WF scales can be dropped without loss of any essential information.Cattell's (1972) position is unyielding, stating that the number ofprimaries encapsulated in the 16PF series is no fewer than the statednmber. In an attempt to support the premise that the number of dimemnionsin the 16PF is Indeed 16, Cattell and &w (1986) presented evidence frantwo independent approaches and more than 50 separate studies.

Cattell's 16PF. There are many res for disagreements agresearchers regarding the nmber of dimensons underlying the 16PF, andCattell and Krug (1986) believe two are paramoumt; (a) differences inbreadth of varable sampling, and (b) lrxnlsstencies in applying acorrect, objective test for the number of factors among factor analysts.Before a presentation of the evidence for the dimenionality of the 16w isdecribed, a brief review of these two issum is necessar.

At a- level, factor analysis ca correctly be understood asn gore t a reductive technique that transform one set ofobservations to a smaller set of reference aas. At this simple level ofapplication, there is no reason to presume that the resulting comntsrepresent either prirmary PersnalitY characteristics or underlyingbehaviour dimmicm . But when factor analysis Is applied to a wll-defined, carefully stratified, and properly sailed data set and whenexperimental results are subjected to independent reconfirmation overobservational facets, -over cultures, and over experimenters, then, attelland Krug (1986) believe that the factors one extracts fron data matricesbegin to correspond to important psychological crutructs. In short,Cattell and Krug are sugesting that it is not the application of factoranalysis alone that results in source traits, but the application of factoranalysis within a r1garus and comprehmnsive programme of scientificexperimentation that does so. They stress that one cannot factor analyseJust any data set and argue convincingly that the results truly reflect thedimesionality or structure of humn personality.

In the case of the 16LW, the original sampling of variables --

drm fr the total universe of personality description - the personalitysphere (Cattell, 1972). The first analyses were exploratory, hypothesis-generating studies. They were not guided or limited by preconceivednotiCs about the structure of personality. Although mny of the factorsthat were found bore sane resemblance to concept ropoased by earliertheorists (for example, fretcher's cyclothymic vs schizothymic dichotomy inwhat Cattell's researchers indexed as Factor A, Freud's ego strength InFactor C), Cattell believes that many were new concepts that had, at leastin their pure state, formerly eluded personality theorists (Cattell, 1973).

Other theorists began from a different perspective. Guilford andhis asociates sought hypothesised compoents of introversion-ectraversion(Guilford, 1975). Eysenck (1972) started with itma taken from the

SGulford inventories, and Comrey (Ccmy and Duffy, 1968) began byanalysing itm in the NMI. In each case, Cattell insists that thevariable definition procedure resulted in a more limited sampling than thatafforded by the total personality sphere concept in wich the 16PW warooted, therefore suggesting that it is not surWising tdLt otherresearchers should conclude that fewer dimersilzis need to be considered(Cattell and Krug, 1986).

-17-

If the difference with regard to variable sampling is a source ofmany disagreemts, Cattell suggests that the failure of many factoranalysts to apply correct objective tests of the number of factors mayrepresent an even more significant Impediment to consesus. Smith andVetter (1982) recognised this issue as a serious problem, asking 'iendifferent factor analyses involve basically the same set of variables,should they not isolate at least the same number of factors in variousstudies ?" (p. 304). To this question Cattell gives an emphaticaffirmative.

Cattell's unfailing ability to -%ce support for the primarydimensions of his personality sphere is nowhere more apparent than in hisre-analysis of 69 Cattell and Guilford parcels that originated in adissertation project by one of Guilford's students (Cattell and Gibbons,1968). While the original Varlmex solution mas ,re consistent withGuilford's system (Gibbons, 1966) the results of Cattell's re-analysisconfirmed all 14 of the 16WF factors represented in the data set. Itshould be noted, hcwever, that Gibbons did not concur in this colusion:he is quoted by Howarth & Bronm (1971) as follows, "I have tried tointerpret the results Fran the Cattell-Gibbons oblique rotation and havefailed to get a meaningful Interpretation". Two of Cattell's students,Bundsal aid Vaughn (1974) concluded that the results of their item factoranalysis of the 16F " ....... i'm essentially the expected factor pattern.The study did, howemver, indicate that four of the sixteen factors (G, M, N,Q1) were probably in need of revision and further research (p. 223)".

It is obvious that the controversy over the number of factors inthe personality sphere is continung. In fact, Cattell (1957) suggeststhat in a clinical sernse there are a further seven factors of smallvariance and rather unstable reappearance, and with one exception, Cattellsuggests that these also appear 'norm'al, and a part of his totalpersonality sphere. Thus, what a clinician recognises em abrmal can beeither a statistically deviant score on a normal factor, or, alternatively,a score on acme factor the very existence of which is demontrable only inpathological investigations, as is recognised on sme 1WI scales. In astudy by Cattell and Bolton (1969) a set of mutually exclusive exaustiveparcels of NMI items and the 16 personality factor scales from the Cattell16FF were factored together on 217 normal and 40 abnormal adult Air ForceMen. This study found that some NMI scale scores, notably schizophrenia,anxiety, psychasthenia and social introversion can be predicted withappreciable efficiency from the 16WF. A result by Ebr (1966)i* epIdently supports these conclusions. The results are also conistentwith the findings that the directIons in which the 16PF have been foundmst potent in clinical practice (Cattell and Scheier, 1961; Cattell, Tatroard Koalos, 1964) are those of neuroticism (including anxiety),psyckasthenia ard the separation of te shizoeris. The cormerseprediction, that is, OPWI to 16PF scales m not possible.

Validity of Test Measurement

It has been demostrated that the controversy over the number offactors in the personality sphere is continuing. The studies reviewed upto this point have used factor analytic theory ard technology to arguetheir respective positions. In compiling such a review, mention must be

-18-

made of the notion of the validity of the test measurement and themisunderstandings associated with it.

Te validity of a test concerr whtat the test measures and howWell it does so (Arnstasi, 1982). Fundamentally, all procedures fordetermining test validity are concerned with the relationships betweenperformance on the test and other indepe ently observable facts aboutbehaviour characteristics under consideration, that is, a test's validitymust be determined with reference to the particular use for which the testis being considered. In the American Psychological Association'spublication "Standards for Mdicational and Psychological Tests and Manuals"(1985) the procedures for determining test validity are classified underthree principle categories: content, criterion-related and constructvalidity.

Content validity essentially Involves the systematic enminationof the test content to determine whether it covers a representative sleof the behaviour din to be measured (Anastasi, 1982). Thus, contentvalidity is built into a test fra the outset throug the choice ofappropria t t . It might thus appear that mere Inspection of thecontent of the test should suffice to establish its validity for such apurpose. This solution is not as simple as it appears to be. onedifficulty is that of adequately sampling the item universe. The behaviourdin to be tested must be systematically analysed to iake certain thatall major aspects are covered by the test item and in the correctproportions. This does create some misuderstandings within thepersonality field as is evidenced by the different samplir of variables.Mention has already been ma of the different sampling variablesconcerning Cattell, Eysenck and Guilford, a reason Cattell believes thatthe other researchers conclude that fewer diimions need to be considered.

Criterion-related validity indicates the effectiveness of a testin predicting an individual's behaviour in specified situatior. For thispnpose, performance on a test is checked against a "criterion", le adirect and independent measure of that which the test is designed topredict. The criterion measure against wich test scores are validated maybe obtained at approzxiately the same time as the test scores or after astated interval (Anastasi, 1982). 7he American Psychological Association'stest "Standeads-" (1985) differentiate between concurrent and predictivevalidity on the basis of these time relations between criterion and test.The term "predictio" can be used in the broader sense, to refer toprediction from the test to any criterion situation, or in the more limitedsense, of prediction over a time interval. It is in the latter sense thatit is used in the expression "predictive ability". The informationprovided by predictive ability is moet relevant to tests used in theselection and classification of personnel.

The construct validity of a test is the extent to which the testmay be said to measure a theoretical construct or trait (Wallen, 1956).Focussing on a broader, more enduring and more abstract kind of behaviouraldescription than the previously discussed types of validity, construct-validation requires the gradual accumilation of information from a varietyof sources. Any data throwing light on the nature of the trait underconsideration and the conditions affecting its development and

-19-

mnifestations are vital for this validity. In a thoughtful analysis ofconstruct validation, Campbell (1960) points out that in order todemostrate ctruct validity it nust be sham not only that a testcorrelates highly with other variables with which it should theoreticallycorrelate, bit also that it does rot correlate significantly with variablesf which it should differ. In an earlier article, Campbell and Fiske(1959) described the former process as co ,rgent validation and the latteras discrimirnat validation. Discrisinant validation is especially relevantto the validation of personality tests in which irrelevant variables mayaffect scores in a varlety of ways.

Validity of the 16W

Ion, %abucha and Dinur (1981) exmnined the validity of the 16PFa predictor of the success of female military officers. Three

independent criteria were measured:

a. officers' ratings;

b. peer ninatios; and

c. final course grades.

Th results shoed that factors E, H, L, M, N, Q1, Q2 and Q3 ofthe 16PF were correlated significantly with at least one of the criteria.Although the findings generally suRported the interpretation of many of the1WP scales, as rendered by Cattell, Eber and Tatsuoka (1980), the outcmsuggested an alternative interpretation of Factor M. Factors E and Mstood out as statistically significant predictors of most criteria. FactorZ w-- strogly correlated with the officers' ratings of self-confidence andsignificantly correlated with peer rizination as well as with success inofficer school. Factor M ws also related to Characteristic A of the peernomination scores, assessed by self-confidence and success in the course.Factors C, H, L, N, Q1, Q2 and Q3 were associated significantly with atleast one of the criterion measures, and Factors F, G, I, 0 and Q4 were notrelated to any one of the criterion variables.

The ai striking aspect of the results is that each of thecomInents of Cattell's QIV second stratum factors (Cattell, Eber andTatwidi, 1980) me significantly correlated with at least one of thecriterit measures. This broad factor, composed of the E, L, M, Q1 and Q2scales, was interpreted as reflecting a subduedness-indepadence continuum.A person high on this factor Is described by Cattell as "independent,radical, autistic, projective and a law to himslf" and as having "ageneral temperetal independence in the broadest sense" (p. 119). Sinceboth the officers' course and the assesemt-centre activities whichpreceded it are leadership programs, it follow that characteristics suchas assertiveness, self-confidence and independence could have affected themer In which candidates were viewed and evaluated by their peers and

superiors.

If one turns attention to the particular source traits composingthis second-stratum factor, it is aparent that Factor E reveived thestrongest support for the interpretation given to it by Cattell. The low

-20-

correlations of Factor E with "Cooperaticn" and "Reaspibility" scoressuggest the existence of conver nt and discrmmint validities of thefactor (Campbell and Fiske, 1959) amd thus strer~ten l Cattell'sinterpretation of it. In sum, this study generally supported the constructvalidity of second-order factor QIV. Fox, Haboucha and Dinur (1981) alsoSuggest that Cattell's interpretation of scales E, H, L, Q1 and Q2 senplausible. In addition, It is suggested that these findin offer evidencethat the 16F can be of practical value in applied contexts.

Saville ard Kznroe (1987), howver, question the reliability ofthe 16PF. Fundcamentally, if a measure is unreliable, no matter how wellaccepted it is, or how uch confidence users have in it, it cannot posmesmuch validity, since its reliability sets a ceiling on the amount ofvalidity It can have. Reliability, of course, cannot Imply validity, batwithout reliability there can be no validity. If the 16PF, therefore, hasunreliable scales, that is, if they are ade up of random error variance ornoise, the use of the 16F s be questicmed. Saville and Kmroe (1987)reported the alternate form reliabilities of the 16PF. 7he alternate formreliability coefficients indicate the extent to vich two different formor editions of the same test measure the same characteristic. For scalesto be psychometrically acceptable the coefficients are expected to be about0.6 or more. Lower values than this mean wide bands of error indicatingtests should not be used for individual decision making. For the 16FF,Saville and Munroe (1987) repor that the reliability coefficients of Itsscales fall within the range, 0.25 to 0.73, with a median around 0.5. Se12 out of the 16 scales of the 16FF fail to reach an alternate formreliability of 0.6; six are as low as 0.4 or less. In fact, Saville anMmre (1987) state that many 16PF scales (B, G, L, M, N, Q and Q2) simplyfail to cae up to ccnentional sta rxards of reliability. They ask, where,for example, wald a pezso fall on taking 3 B of the 16PF If they wr

I .-aveage mn Form A? In fact, on Factor N, Saville an~d ltmune (1987) sugjgest

that a tolerance of plus or miu 3.46 stem is needed for 95% conidnc.Th~at is, a band covering th complete range from the aeaxnd to the ninthStan.

Boyle (1988) is quite emphatic with his clais that reliabilityis not an Issue with the 16PF. He states, "the reliability of the 16PF canbe Improved to any extent desired .... by having the applicant completeadditional fm of the questionmaire. Reliability is NOT an issue withthe 16IF, since use of more than a single farm results in quite reliableprofiles" (p. 11). Cattell (1973) has always advocated the use of combinedforum, believing the use of parallel forms Increases the reliability bydoubling the nmbers of it administered. Boyle (1988) chooses to usethe qualifier "quite" when referring to the reliability of the 16PF afterthe use of a than one form of the 16PF. "Quite" is an ambiguous adverb,'ere the colloquial use Sugsts uncertaliity. Should Boyle hae beenusing the word advisedly, how many forw of the 16PF are needed to confirmits reliability? Also, why are short versions of the 16FF, of debatablereliability, readily available and widely used ?

It is clear from the research done on the definition andmeasureent of personality that there is still consIderble controversy asto the number of factors required to adequately describe bian personality.Curiously, even the EI can be used to make a case for the 13-to-18 factor

-21-

position. Fifteen factors have been extracted from the EPI by Howarth andBrowne (1972). They actually concluded that the EPI should be scored forprimary factor scales in order to present a more detailed picture ofextraversion and neuroticism. Eysenck himself (Eysenck and Eysenck, 1969)extracted 12 factors from the EPI, but continues to advocate that only thetwo higher-order factors are useful and replicable.

The majority of studies reviewed here have used factor analytictheory and technology to argue their respective positions, yet factoranalytic approaches really have not been convincing as it seems differentnumbers of factors can be extracted for different purposes. Cattell,Guilford, and Howarth and Browne firmly maintain that 13-to-18 factors arepresent in and necessary to describe personality. Proponents of both thetwo factor and the five-to-eight factor position would argue that this isoverfactoring, therefore unnecessary and meaningless. However, even morethan 18 factors have been suggested. Cattell concluded that 23 factors arenecessary to describe normal adult personality (Cattell, 1973). So, itseems, 18 is not the upper limit of those proposed, and the controversyover the number of factors in the personality sphere continues.

In reviewing the data on the 16PF, there is little doubt thatsignificant contributions to the factor theory of personality, bothconceptually and technically, have been made by Cattell. However, there isconsiderable doubt about the construct validity of the 16FF. This doubtinvolves both its convergent and its discriminant validity. Although the16PF purports to measure 16 distinct personality traits there is noevidence of the convergent validity of the 16 scales apart from factorloadings, and what evidence there is of discriminant validity suggests thatthe primary traits are not clearly differentiated. Mien coupled with therelatively low internal consistency of the scales and their modest test-retest reliabilities, the psychometric adequacy of the test must bequestioned.

AA Psych Corps' Use of the 16PF

The AA Psych Corps introduced the 1967 revised version of the16PF as its principal measure of personality functioning (DPSYCH-A, 1985).The Australian Army requires that an assessment of personality be made by apsychologist before soldiers can be selected for certain specialistcategories of emloyment, as well as service for the Antarctic, However,

as suggested by Wrn (1986), no interpretative guidelines were offered toArmy psychologists tasked with selection, except that attention was drawnto the Handbook for the 16PF (Cattell, Eber and Tatsuoka, 1980). Thisomuission is of some concern especially in view of the fact that Armypsychologists do not necessaily receive formal training in the use of the16PF as part of their supervision. For this purpose, Warn (1986) proposeda method by which psychologists can use the 16PF to assist in assessment ofan applicant. In doing this he referred to a type placement approach wherethe psychologist compares an applicant's scores on the primary fictors tothe means for the group for which entry is being considered. For instance,If the applicant wished to join the Military Police, his scores ok eachfactor would be contrasted with those for the group of current (and perhapspost) military police. One drawback of the type placement approach is thatstandards oL performance are not considered. The criterion being used is

-22-

simply membership of a special group by having passed a course to qualify.To continue the example, the type placement approach would fail toackncwledge that some military police might perform better than others,either in training or in later service.

Warn (1986) suggests a perfornance approach would overcome theabove objection by contrasting the applicant's scores against those of thesubgroup who have displayed some degree of excellence. The contribution ofeach factor of the 16PF towards this superior performance is indicated byits weight on a specification equation. After entering Individual factorscores into the specification equation, the psychologist obtains a scorethat indicates the extent to which the applicant approximates the criteriongroup. An example of a specification formula for a group is provided inthe 16PF Handbook for success as a patrolman. Patrolman Success = -. 47A -.35F -. 35L +.23Q2 +.23Q3 + 9.41 (Cattell, Eber and Tatsuoa, 1980, p.165).Hcever, the psychologist would require a valid and measurable criterion ofsuperior performance before the performance approach could be effective.

The type placement and performance approaches are bothrecaanended by Cattell and Krug (1986), who believe that the occupationalpsychologist would do well to build up a file on the sample group. Notonly would the measures of the local sample increase the validity of thecomparisons between group and individual applicant but Cattell also seessuch research as necessary for the compilation of more universal norms.

The advantage of using the statistical approach outlined above ispthat the psychologist need not rely on guesswork in order to determine the

relative significance of individual factor scores. Another statisticalmethod is the use of cut-off or critical scores on relevant factors. Forexaimple, a psychologist might find that a low score on factor G ispredictive of problem and consequently decide that any applicant scoringbelow the cut-off is to be screened out (Cronbach, 1970). Alternatively,specific hypotheses might be generated for a structured interview whereeach can be evaluated by integr ting real life content with the 16PFpsychometric findings (Cattell's depth psychametry approach (Cattell,1983)).

The Army psychologist is provided with none of these statisticaltools and thus there is a requirement for a more interpretative approachinvolving an analysis of the underlying dynamics of the factors. Warn(1986) outlines such a method suggesting that the interpretative method canbe regarded as complementary to the statistical. By adopting aninterpretative as distinct from statistical method the psychologist is ableto search the primary factors for an individual pattern, yet this stillleans on the insight of the individual psychologist.

Warn (1986) offers a routine approach to the Army psychologistfor the interpretation of the 16PF. In brief, Warn suggests:

a. the psychologist needs to make a note on the level of distortionthat is suggested by the faking or MD sten; and

b. the second order factors need to be examined since it is fromthese that the psychologist can begin to hypothesise about the

-23-

applicant's personality. Once a suiory description has beenformulated, the individual complexities of each second orderfactor can be explored by examining their primry factorcomosition.

At this point, Warn suggests that the profile structure can becompared against the soldier's service record and personal history. Thepsychologist needs to hypothesise about the behavioural implications ofextreme scores on primary factors as well as for each of the second orderfactors. These extrapolations from the profile, Warn believes, can bechecked against the behavioural records. Exaination of the service recordand personal history might require a reworking of the hypotheses derivedfrom the 16PF profile, and discrepancies need to be noted so that they canbe explored at the interview stage. The interview allows the psychologistto further explore the patterns suggested by the 16PF. Real life instancesneed to be gathered in order to flesh out the more analytical structure ofthe profile. Thus, after the interview, Warn believes the psychologistwill have a mass of firsthand impressions and profile assessments which canbe organised into a coherent write-up.

Warn's (1986) routine approach for the Army psychologist'sInterpretation of the 16FF is quite functional; the doubt, however, aboutthe overall validity of the 16FF is still a problem. Although Warn'sstarting point for interpretation is the second-order factors, the primaryfactors are considered subsequently. Because of the problems with theprimaries, noted earlier, there needs to be a clear warning thatconsiderable caution should be exercised in interpreting the primaries.This does add to the responsibility of the individual psychologist. Aswell as this, O'Gorman (1986) expressed a general concern about theinterpretative approach advocated by Warn (1986), suggesting that it runsthe risk of confounding personality description with personalityprediction. Personality description involves characterising theindividual's behaviour in terms of the concepts of a selected theory ofpersonality. Personality prediction involves forecasting the likely futurebehaviour of the individual in a given situation. It is important to makethe distinction, as O'Gorman suggests that the basis for evaluating thevalidity of the two differ. In the case of personality description,validity is evaluated in terms of the accuracy with which the concepts havebeen applied. Comparisons with judgements of those who know the individualwell or with salient features of the individual's life history can be used.For personality prediction, on the other hand, validity depends on anincrease in the efficiency of decision making. Because of this difference,an accurate description of personality does not imply an accurateprediction of future behaviour. The interpretative approach outlined byWarn will help in formulating accurate descriptions of personality.However, the problem of prediction remains.

O'Gorman (1986) believes the problem can be dealt with in twoways. One is to establish empirical relationships between 16PF scores andaspects of performance in the actual situations which are the focus ofprediction. However, there are few data of this sort currently available.In their absence, O'Gorman suggests the use of the 'analytic approach',described by Stern and colleagues (Wiggins, 1973). This involves aconceptual analysis of the criterion situation followed by the formulation

-24-

of explicit statements of the theoretical relationship between test scoresand the features of the criterion identified in the conceptual analysis.This analysis would employ a qualitative rather than a quantitativemethodology and draw on the observations of those with immediate experienceof the criterion. Hever, as O'Gorman succinctly puts it "it is only whenthe assessor is armed with the conceptual analysis and the theoreticallybased linkages that one can expect the personality description provided bythe interpretative approach to lead to improved prediction," (p. 4).

In sum, It must really be asked whether any self-report test cando justice to the subtleties of human personality and behaviour, but if the16PF is to be used the Army psychologist should always interview theapplicant after sighting the 16PF profile. This enables specifichypotheses from the 16FF data to be explored with particular emphasis onsecondary factor scores. This is not to imply that there are not morefactors than these in the personality questionnaire domain, but ratherthat, if more exist, they have not been measured with sufficientreliability or independence to be revealed by the questionnaire.

Swaary

1. The concept of personality is widely recognised as being centralin psychology, yet its nature and the ways in which it can bedefined and measured are questions on which psychologists are inconsiderable disagreement.

2. tile theorists tend to disagree over definitions of personality,they tend to agree that in order to perfom a systematicexploration of personality's relation to other variables, adefinite set of personality factors needs to be specified.

3. Researchers engaged in the question of the number of factors inthe personality sphere seem to divide roughly into threepositions: Two factors, five-to-eight factors and 13-to-18factors.

4. Cattell devoted a major segment of his career to the developmentand refinement of his conception of the personality sphere andthe primary instrument for its measurement, the SixteenPersonality Factor Questionnaire (16PF).

5. The 16 PF purports to measure 16 distinct personality traits, yetthere is no evidence of the convergent validity of the 16 scalesapart from factor loadings and what evidence there is ofdiscriminant validity suggests the primary traits are not clearlydifferentiated. When coupled with the relatively low internalconsistency of the scales and their modest test-retestreliabilities, the psychometric adequacy of the test must bequestionned.

6. The Australian Army Psychology Corps uses the 161F. Since thepsychometric adequacy of the questionnaire is in doubt,

-25-

psychologists must use caution when interpreting the 16FFprofiles.

m

-27-

References

Abdel-Khalek, A.M., Abdel-Sattar, I., & Budek, M.H. (1986). Thefactorial structure of the 16 PF & EPQ in Egyptian samples: Apreliminary study. Personal and Individual Differences 7, 16.5-72.

Adcock, N.V., & Adcock, C.J. (1977). The validity of the 16PFpersonality structure: A large New Zealand Sanysee itemanalysis. Journal of Behavioural Science g, 227-237.

Allport, G.W., & Odbert, H.S. (1936). Trait-names: Apycholexical study. Psychological Monograhs, 47, 2111-171.

American Psychological Association. (1985). Standards foreducational and Psychological tests and manuals. Washington,D.C: American Psychological Association.

Anastasi, A. (1982). Psychological testing (fifth ed.). NewYork: MacMillan.

Becker, W.C. (1961). A comparison of the factor structure andother properties of the 16PF and the Guilford - Martinpersonality inventories. Educational and PsychologicalMeasurement 21, 393-404.

Berg, I.A. (1967). Response set in personality assessment.Chicago: Aldine Publishing Company.

Borgatta, E.F. (1964). The structure of personalitycharacteristics. Behavioural Science 9 8-17.

Boyle, G.J. (1988). A guide to the military use of the 16PF inpergonnel selection. Melbourne, Australia: Department ofDefence (Aray Office), Headquarters Third Military District.

Bundsal, C.A., & Vaughn, D.S.A. (1974). A contrast of thepersonality structure of college students found in thequestionnaire medium by items as compared to parcels. TheJournal of Genetic Psychology 125, 219-224.

Buss, A.H., & Firm, H.M. (1987). Personality: evolutionaryheritaae and human distinctiveness. Hillsdale, New York:Erlbaum.

Campbell, P.T. (1960). A typology of tests projective andotherwise. Journal of Consulting Psycholocy 2, 207-210.

Campbell, P.T., & Fiske, D.W. (1959). Convergent and discriminantvalidation by the multitrait-multmethod matrix. PsychologlcalBulletin , 81-105.

Cattell, R.B. (1946). Description and measurement of personality.New York: World Book.

-28-

Cattell, R.B. (1947). General pywholxzv Cmbridge, MA-:Sdi-art.

Cattell, R.B. (1950). Personality: a swtemtic theoreticaland factual totuy. New York: McGraw-Hill.

Cattell, R.B. (1957). Personality and motivation structure andmeasurement. London: Harrap.

Cattell, R.B. (1965). The scientific analysis of personality.London: Penguin.

Cattell, R.B. (1967). The scientific analysis of personality.Harmondsworth, Midilesex: Penguin.

Cattell, R.B. (1972). The 16PF and basic personality structure:A reply to Eysenck. Journal of Behavioural Science 1,169-187.

Cattell, R.B. (1973). Personality and mood by cmestionnaire. SanFrancisco: Jossey-Bass.

Cattell, R.B. (1983). Structured personality learninq theory.New York: Praeger.

Cattell, R.B., & Bolton, L.S. (1969). Mat pathologicaldimensions lie beyond the normal dimensions of the 16PF? Acomparison of MMPI & 16PF factor domains. Journal ofConsulting & Clinical Psycholoy. 33, 18-29.

Cattell, R.B., Eber, H.W., & Tatuocka, M.M. (1980). Ilandbiok forthe 16 personality factor qaestionnaire. Champaign, Illinois:IPAT.

Cattell, R.B., & Gibbons, S.D. (1968). Personality factorstructure of the combined Guilford & Cattell personalityquestionnaires. Journal of Personality & Social Psychologv ,107-120.

Cattell, R.B., & Krug, S.E. (1986). The number of factors in the16PF: A review of the evidence with special emphasis onmethodological problems. Educational & PsvchologicalMeasurement, 46, 509-522.

Cattell, R.B., & Scheier, I.H. (1961). Predicting anxiety fromclinical symptoms of anxiety. Psychiatrika QuarterlySupplement, 14-126.

Cattell, R.B., Tatro, D., & Komlos, E. (1964). The diagnosis andinferred structure of paranoid and non-paranoid schiz, from the16PF profile. Indian Psychological Review, L 52-61.

Cattell, R.B., & Tsujioka, B. (1964). The importance of factor-trueness aid validity, versus homogeneity and orthogonality, intest scales. Eduational & Psycholocical Measurement, XXIV,

Comrey, A.L. (1973). A first course in factor analysis. NewYork: Academic Press.

-29-

CcArey, A.L., & Duffy, K.E. (1968). Cattell & Eysenck factorscores related to Ccmrey personality factors. MultivarlateBehavioural Research 4, 379-392,

Costa, P.T., & Mcdrae, R.K. (1976), Age differences inpersonality structure: A cluster analysis approach.J. Geroni 31, 5, 564-570.

Cronbach, L.J. (1970). Essentials of Psychological testing (3rded.). New York: Harper.

Digman, J.M., & Takemoto-Chock, N.K. (1981). Factors in thenatural language of personality: Re-analysis, comparison, &interpretation of six major studies. Multivariate BehaviouralResearch 16 149-170.

DPsych-A (1985). Introduction of revised 16PF form A + form B, AAPsych Corps Admin Instruction 5/85 (Revised). Canberra,Australia- Department of Defence (Army Office), Directorate ofPsychology - Army.

Eber, H.W. (1966). Toward oblique simple structure: Mxplane.Multivariate Behavioural Research 1. 112-125.

Eysenck, H.J. (1967). The bioloical basis of personalit.Springfield, IL: Charles C. Thomas.

Eysenck, H.J. (1971). On the choice of personality tests forresearch and prediction. Journal of Behavioural Science 185-89

Eysenck, H.J. (1972). Primaries or second-order factors: Acritical consideration of Cattell's 16PF battery. BritishJournal of Social and Clinical Psvcholoqvg 1, 265-269.

Eysenck, H.J. (1982). Personality, aenetics and behaviour:selected papers. New York: Praeger.

Eysenck, H.J., & Eysenck, S.B.G. (1969). Personality, structureand measurement. London: Routledge & Kegan Paul.*

Eysenck, H.J., & Eysenck, S.B.G. (1975). Manual of the Eysenckpersonality auestionnaire. San Diego, CA: Author.

Eysenck, H.J., White, P.O., & Soueif, M.I. (1969). Factors In theCattell personality inventory. In H.J. Eysenck & S.B.G.Eysmenck, et al. (Eds.), Personality structure and measurement.San Diego, CA: Robert Knapp.

Fox, S., Haboucha, S., & Dinur, Y. (1981). The predictivevalidity of the sixteen personality factors questionnairerelative to three independent criterion measures of militaryperformance. Educational and Psychological Measurement,515-520.

Gibbons, B.D. (1966). Functional psychological testing.

Multivariate Behavioural Research, L 140-161.r

-30-

Goldberg, L.R. (1982). From ace to zombie: Some explorations inthe language of personality. In C.D. Spielberger & J.N.Butcher (Eds.), Advances in personality assessment,(pp. 203-234). Hillsdale, NJ: Erlbaum.

Guilford, J.P. (1975). Factors and factors of personality.Psycholocical Bulletin, 8, 802-814.

Hall, C.S., & Lindzey, G. (1957). Theories of personality. NewYork: Wiley.

Hogan, R. (1983). Socioanalytic theory of personality. In M.M.Page (Ed.), 1982 Nebraska Symposium on motivation: Personality- Current theory and research. (pp. 55-89). Lincoln Neb:University of Nebraska Press.

Hacrth, E. (1976). Were Cattell's 'Personality Sphere' factorscorrectly identified in the first instance? British Journal ofPsycholocy 2 213-230.

Howarth, E., & Browse, J.A. (1971). An item factor analysis ofthe 16PF. Personality, 2, 117-139.

Howarth, E., & Browne, J.A. (1972). An item factor analysis ofthe Eysenck personality inventory. British Journal of Socialand Clinical Psychology, 1 162-174.

Levonian, E. (1961). Personality measurement with items selectedfrom the 16PF questionnaire. Educational and PsychologicalMeasurement 1, 937-946.

McCrae, R.R., & Costa, P.T. (1983). Joint factors in Self-reportsand ratings: Neuroticism, extraversion and openness toexperience. Personality and Individual Differences 4,245-255.

McCrae, R.R., & Costa, P.T. (1985). Updating Norman's "AdequateTaxonomy": Intelligence and personality dimensions In naturallanguage and in questionnaires. Journal of Personality andSocial Psychojy, 49 710-721.

McCrae, R.R., & Costa, P.T. (1987). Validation of the five-factormodel of personality across instruments and observers. Journalof Personality and Social Psychology, 81-90.

Mershon, B., & Gorsuch, R. (1986). Number of factors In thepersonality sphere. (Unpublished).

Norlnan, W.T. (1963). Towrd an adequate taxonomy of personalityattributes: Replicated factor structure in peer nominationpersonality ratings. Journal of Abnormal and SocialPsychology.66, 574-583.

O'Gornan, J.G. (1986). Warn's cuide to the use of the 16PF. (FN738-2-4). Canberra, Australia: Department of Defence (ArmyOffice), let Psychological Research Unit.

-31-

3Peabody, D. (1984). Personality dimensions through traitinferences. Journal of Personality and Social Psychology 4384-403.

Peterson, D.R. (1965). Scope and generality of verbally definedpersonality factors. Psychological Review 72 48-59

Saville, P., & Blinkhorn, S. (1976). Undergraduate personality byfactored scales. Windsor, UK: NFER.

Saville, P., & Munroe, A. (1987). The relationship between thefactor model of the occupational personality questionnaires

and the 16PF. Canberra, Australia: Proceedinas of the 22ndAnnual Conference of the Australian Psycholofical Society.

Sheldon, W.H. (1940). The varieties of human physiaue: Anintroduction to constitutional psychologyv. New York: Harper.

Smith, B.D., & Vetter, H.J. (1982). Theoretical approaches topersonality. Englewood Clifts, NJ: Prentice-Hall.

Spearman, C. (1904). "General Intelligence" objectivelydetermined and measured. American Journal of Psyc holoqT ,201-293.

Thurstone, L.L. (1938). Primary mental abilities. PsychometricMonographs, 1.

Wellen, R.W. (1956). Clinical psycholoy: the study of person.New York: McGraw-Hill.

Warn, J.R. (1986). A gulde to the use of the 16PF qu.estionnairein personnel selection: A military context. (FN 738-2-4).Canberra, Australia: Department of Defence (Army Office), 1stPsychological Research Unit.

Webster, R.K. (1961). Dictionary. New York: Wiley.

Wiggins, J.S. (1973). Personality and prediction: Principles ofpersonality assessment. London: Addison-Wesley.

~I

II