8/13/2019 Try on 1935 Small

1/30

A T H E O R Y OF PSYCHOLOGICAL COMPONENTSAN ALTERNATIVE TO 'MAT HE MAT ICALFACT ORS 'BY ROBERT C. TRYON

University of California

I t is the possession of concepts, motives, and emotionswhich differentiates individualsinintellectual and personalitycharacteristics. And it is'universes ' of these psychologicaldeterminants which 'tests ' attempt to sample. Thesecom-ponentsare,however,sonumerous andcomplex,theexperi-mental study of their origins in the individual so difficultand laborious, their hereditary andenvironmental conditionsso resistanttoscientific analysis that some psychologists havesought various types of flight from reality into simplified,albeit fictitious, postulations of the causal determinantsofbehavior differences. Onetype of escape has been in theassumption of the existence of large psychological determi-nants such as 'general intelligence,' 'introversion-extrover-sion,' and other 'tr a it s. ' Serious studen ts of individualdifferences reject such alleged determinants; sincenoat temptis made to describe their psychobiological properties, orthe fundamental conditions of their emergence in theindi-vidual. Another typeofescape has recently presented itself,one considerably more dangerous than ' trait-mongering'becauseit is more artful. I referto therecent proposalstoaccept 'mathematical factors' as the determinants of indi-vidual differences. Here the escape has been into mathe-matics, and though it avoids the real problem in nolessadegree than does 'trait-mongering,' it has thesemblanceofbeing more scientific byvirtueof its mathematical content.The purposeofthis paper is first, toshow tha t 'mathem aticalfactors' represent yet another ignis fatuus, and second, topresent a systematic psychological envisagement of the de-terminantsof individual differences.

425

8/13/2019 Try on 1935 Small

2/30

426 ROBERT C. TRYONARBITRARYAND NON-PSYCHOLOGICAL CHARACTER OF

'MATHEMATICAL FACTORS'The acceptance of 'mental factors' educed by factor an-alysis1 from observed intercorrelations between psychologicaltests entails the belief in the validity of a certain set ofassumptions concerning the determinants of individual differ-ences. The general and broad group factors derived fromdata under these assumptions have no more validity than theassumptions themselves, and unfortunately, these latter arenot verified by psychological or genetic experiment. I havetherefore termed such factors 'mathematical fac tors' in orderbetter to differentiate them from factors subject to experi-mental variation and study.Before observing the assumptions made by mathematical-factorists, let us symbolize two current factor theories:

G theory P theoryxi = hg + k2su Xi = aifl +ftfc+ Tic + M ,*2 = k3g + kis2, x 2 = a2a + fab + y2 c + e2e.The Gtheory states t ha t individual differences in each of thetwo abilitiesxxandx2,are determined by one general factor,g,and the specifics. Spearman, its proponent, may admit a fewadditional non-specific factors. The P theory, where Pstands for parsimony, postulates the smallest number ofgeneral or group factors plus specifics leading to consistency.

1 See the following references on factor analysis:C. Spearman, The abilities of man, N. Y.: Macmillan, 1927.T. L. Kelley, Crossroads in the mind of man, Stanford University Press, 1928.H . Hotelling, Analysis of a complex of statistical variab les, / . Educ. Psychol.

1933. 24, 4i7-444> 498-52 0.L. L. Thurstone, Th e theory of multiple factors, Ann Arbor: Edwards Bros., 1933.L. L. Thurstone, The vectors of mind, PSYCHOL. REV., 1934, 41, 1-32.R. C. Tryon, M ultip le factors vs . two-factors as determiners of ab ilities, PSYCHOI.

REV., 1932, 39, 324-351-R. C. Tryon, So-called group factors as determiners of abilities, PSYCHOL. REV.,

1932, 39, 4O3-439-In the last two references, the writer presented a statistical critique of mathe-

matical factor theories emphasizing esp ecially the two-factor theor y. M y purposein those papers was primarily destructive, as I offered little in the way of an alternativeto mathem atical factors. Th e present article attemp ts to construct an alternativetheory.

8/13/2019 Try on 1935 Small

3/30

A THEORY OF PSYCHOLOGICAL COMPONENTS 4 2 7Kelley and Thurstone think of abilities in terms of the Ptheo ry. Spearman names his general factor, mental energy,whereas Kelley would call his setverbality, number, space, etc.The assumptions upon which these factor systems are basedand without which these factors could not be postulated maybe considered under the several heads indicated below.Num ber of factors.Both the Gand P systems assume avery small number of general or group factors, and ordinarilyas many specifics as tests. In defense of this assumption,mathematical-factorists generally resort to the rule of parsi-mony, that is, where a given set of intercorrelations may bedescribed in terms of any one of a large number of mathe-matical factor patterns (and this is always the case) whichdiffer in the number of factors in each, that particular patternis chosen which contains the fewest factors.

There is thus an infinite number of ways in which you couldrepresent the ability to produce a performance on [a] test. . . . Indealing with observed relations between variables, one factor prob-lem is to discover the lowest order of complexity that will accountfor the observed relations . . . (p. 4). W e want to know th esmallest number of factors by which any given table of coefficientscan be described (p. 16). Thurstone, Theory. . . . We will systematically choose the hypothesis involvingthe lesser number of unknown facto rs (p. 69). Ad opting thelaw of parsimony we will not resort to this system [a factor pa tterninvolving one general andtwogroup factors] if systems (49) and (50)suffice (p. 71). [(49) and (50) are pa tterns involving one generaland onegroup factor.] Kelley, Crossroads. The fact that any variable can be divided into the two factors,g and /, does not, of course, preclude it from being divisible in aninfinity of other ways (pp.viff.). [Spearman chooses the two-factor way because he considers it mathem atically a unique solution.]Spearman, Abilities.No one may object to mathematical-factorists employing therule of parsim ony in order to simplify their problem . Onemust realize, however, that the factors which come out of theanalysis arechosen from many possible sets, and that the ruleof choice is arbitrary. Now, it happens th at the 'la w ' of

8/13/2019 Try on 1935 Small

4/30

428 ROBERT C. TRYONparsimony is not a natural law, but a rule agreed upon amongmen to simplify their th in kin g. W ith reference to t h epsychobiological causes of individual differences, nature doesnot appear, however, to work parsimoniously but rather mostprodigally. As will be shown later, th e exp erim enta l evi-dence from psychological and genetics laboratories indicatesthat a very large number of causes determine mental differ-ences. Hen ce, the em ploy m ent of th e rule of parsim ony toselect out a parsimonious set of factors would appear todepict a fiction if such a depiction is urged as a representationof psychobiological causes.

Interrelation betweenfactors. Proponents of th e G and Psystems assume that factors are uncorrelated, at least themathematical factors they postulate must show zero inter-correlations. We shall locate the coordinate axes, i.e., the principal factors

so as to maximize the sums of the squares of the projections of thetest on the axes. These axes will be orthogonal since the factorsare uncorrelated (p. 17). Thurstone , Theory.With reference to one of his fundam ental equations , Kelley says : The specific factors are e\ and et [in two tests]. Th e commonfactorisa. The independent variablesa, etandetare uncorrelated(P-39)- Kelley, Crossroads. First of all,gxis always the same for the individualx,whatevermay be the ability under consideration,a, b, c, etc. Contrariwise,his Sax, Sbx, sa, vary from one ability to another independentlyboth of each other and of the value gx (p. xv). SpearmanAbilities.T o assume zero interco rrelation is of co urse form ally per-missible but, as will be shown later, it is unsound if the factorsare presumed to represent psychological components.

Combination of factors.A nother ass um ptio n refers to th eway mathematical factors are made to combine in producingan indiv idual 's to tal score. A glance at th e G and P equa-tions shows that the score is assumed to consist of the simplealgebraic summation of his degrees of excellences in the sundryatomistic factors. We also make the assumption that the contributions of several

8/13/2019 Try on 1935 Small

5/30

A THEORY OF PSYCHOLOGICAL COMPONENTS 429independentfactors are summative in the individual's performanceoneach of the psychological tes ts (p. 6). Thurstone, Vectors.

Kelleyrefers to th e use of Tay lor's Series as the ra tionale for hisexpressingany test score as a sum mation of factors (p . 38). Kelley,Crossroads. Surprise may be felt that the measurement . . . , even if trulyenougha function of the two factors general and specific, should sosimplyconsist of merely thesum of these added together. . . . Theanswer to this question is that our proof has depended upon theusage of Taylor's theorem, according to which all mathematicalfunctions however complex can, in general, be expressed in theabove simple add itive form w ith some appro xim ation (p. xv ).Spearman, Abilities.We are here again confronted with another assumption whichno doubt is mathematically convenient and symbolicallyfeasible. B ut in the opinion of many psychologists, especiallyof the configurational school, an assumption of additivefactors would be considered clearly inconsistent with otherexperimental evidence. Likewise, the geneticist has dis-covered many instances of multiple genetic factors interactingnon-cum ulatively. M athematical factors therefore appear torepresent neither psychological nor genetic components.Weight of factors.Each mathematical factor possesses aweight in determining the individual's to tal score. Theseare thekweights of theGpatte rn , and th e Greek letter weightsof the P pa ttern . Th e assumption is made th at the weight('saturation' or 'loading') of a given factor in determiningeach individual's total scores is exactly the same for all indi-viduals, who differ only in their relative ranks in the factorvariable.

The coefficientsa\anda[weights oftwofactors in a given test]are essentially properties of the test. . . . (p. 2). [Thus each isconsidered constant for all individuals.] Thurstone, Theory. . . . Ci[weight ofafactor] is some constant not changing as wepass from member to member of the group in question (p. 38).Kelley, Crossroads. A s for the r's [weights of g and s], these vary for the differentabilities a, b, c, etc.,bu t, of course, do not vary for the differentindividuals (p. xv). Spearman,Abilities.

8/13/2019 Try on 1935 Small

6/30

4 3 0 ROBERT C. TRYONThe present writer believes that this new assumption has li t t levalidity with reference to factors having psychological sig-nificance. H e will at te m p t to show late r in this pa pe r t h a tsuch an assumption certainly does not apply to psychologicalcomponents .

Nature and origin of factors.When the ma them at ica l -factorist comes to the f inal problem of stating the psycho-logical n at u re an d th e origin of his factors, he usua lly confessesthat his factor methods do not definitely inform him of theseim po rtant par t icula rs . As an ancil lary proced ure, he ma ypick out those tests which app ear to be mo st heavily weigh tedby a given factor, note the content of these tests, and labelthe factor w ith a verba l ab strac tion wh ich be st defines th ecommon element of the manifest content of the selected tests.Th is nam ing procedure is highly subject ive. On e can no t besure that another psychologist noting the subject matter ofthe selected tests might not observe other common elementsand give a different na m e. E v en if one is relatively co nte ntwith the label given, the disconcerting fact often appears thatothertests in the analysis , thou gh no t satu rat ed by th e factorin quest ion, may contain subject matter sat isfying the sameabstract definition.

In 'The vectors of mind,' Thurstone appears to be concernedwith two problems: the giving of a psychological label to (i) theprincipal axes or factors educed by his method and (2) the clustersof tests which are revealed by the objective intercorrelations andgraphically indicated by points within or on the surface of ann-dimensional sphere. Regarding (1), Thurston e states th a t th elocation of the axes is arbi tra ry andwithout fundamental psychologicalsignificance (p. 27, italics mine). Bu t in (2), w e have somethingmore or less permanent in terms of which we may define psycho-logical categories and mental ab ilitie s (p. 27). [Apparently thedefining is done by discovering a verbal term best covering theoperations involved in the tests of a cluster.] It m ust be understood th at the captions [psychological names]. . . describing the general and group factors . . . are merelywords giving what the writer surmises to be the nature of thefactors. How these surmises are arrived at may be illustratedin Kelley's naming of factor /3as a verbal ' factor. H e discovered

8/13/2019 Try on 1935 Small

7/30

A THEORY OFPSYCHOLOGICAL COMPONENTS 431that this factor showed heavy positive weights in his Tests 1and 2,and weights of low value in Tests3and 5. No weight was found inTests 4, 6, 7, 8, or 9. Kelley writes: The appropriate verbaldescription of the factor can only be surmised by a study of thenine tests and the selection and naming of the element present inlarge amounts in Test1and 2, in slightly negative amount in Test 3,in slightly positive amount in Test 5, and absent in the remainingfive tests. The writer's characterization of this as a 'verba l'factoristhe result of such a procedure on his part (p. 107). Kelley,Crossroads.

Having ascertained which abilities involve g, our next step is tofind thedegreethat they involve it. . . . One purpose of this pro-cedure is that of throwing light on the intrinsic nature of the twofactors (p. 199). The best means for measuring the amount ofg in any ability, say a, consists simply in the correlation . ,rta (p. 200). In Chapter XII, Spearman examines many types oftest content with reference to the weights by which each type isdetermined byg,and in his ' Cardinal Conclusions' summarizes hisfindings by the statement that It [g] showed itself to be involvedinvariably and exclusively in all operations of eductive nature,whatever might be the class of relations or the sort of fundaments atissue (p. 411). Spearman, Abilities.Thus the psychological nature of a mathematical factor isinferential and moot. Further, what is its fundamentalorigin ? Towhat extent isitgeneticallyandenvironmentallydetermined? How may it be investigated experimentally?T o these questions,the factorist hasyet toprovideananswer.In fairness to mathematical-factorists, it should be saidtha t the more competent ones are fully awareof the natureoftheassumptions they employ,but itmust alsobesaid th atnonehasclearly listed all ofthese assumptions anddiscussedtheir psychological implications. We arehere notconcernedwith the factorists' algebra, bu t with their assumptions, which,I believe, do not have psychological or biological meaning.It maybeth at what is neededismore, rather than less, factoranalysis. Thefuture may seenew methods derivedinwhichsuch arbitrary restrictions as parsimonious selection, zerointercorrelations, etc., are relaxed. It may ultimately bethat mathematical factors will represent descriptions of

8/13/2019 Try on 1935 Small

8/30

4 3 2 ROBERTC.TRYONpsychological or genetic com ponents. To the present writer,however, the future along these lines looks black. The bestpromise for the study of individual differences lies in otherdirections.

THE PSYCHOLOGICAL COMPONENTS OF INDIVIDUALDIFFERENCES

In view of the artificiality of the assumptions involved inthe postulation of mathematical factors and of their amorph-ous psychological nature and origin, it appears necessaryto turn elsewhere for an understanding of the determinantsof individual differences. I t is the writer's opinion th a t thesecomponents may be described and studied experimentally.The remainder of the paper presents a description of them.No claim is made that the treatment is other than tentative.Provisional as they are, however, such components possess theattributes of making psychological and biological sense.They derive in the main from facts collected in the psycho-logical and genetics laboratory and from the theories oflearning and of the gene. Much information m ay be securedfrom a shameless resort to introspection.Before a description of these determ inants is attem pte d, itis necessary to note the character of psychological measure-ments which elicit them. These are the objective tests ofability, aptitude, achievement, etc., having subject matterranging from problems requiring responses which dependupon relatively unspecialized training (such as those in so-called 'intelligence' tests) to problems requiring responseswhich depend upon a more special course of formalizededucation (such as those in achievement tes ts). In additionto these measures, there are the personality tests, ordi-narily having the form of standardized questionnaires orassociate's ratings. Some of these domains of behavior arerelatively easily defined, and item-situations satisfying thedefinitions are relatively easily selected, e.g., achievementdomains of spelling, grammar, reading, arithmetic, vocabu-lary, and the like; others are more difficult to define and pre-sent a greater problem of item selection. The significant fact

8/13/2019 Try on 1935 Small

9/30

A THEORY OFPSYCHOLOGIC L COMPONENTS 433to observe about mental measurement is that, having markedout by definition some domain for testing, the psychologistchooses as amethodofmeasurementone which indicates th a t heknows before giving the test to any subjects a great deal aboutthe nature of the factors which cause individual differencesin the domain. T he method is th a t of sampling behavior,and it definitely presupposes that for any defined domain thereexists a universeof causes, or factors, or components deter-mining individual differences. Each test-item attem pts to' t a p ' one or more of these components. Part of his verymethod is also that of choosing items which manifest lowintercorrelations, a procedure which shows that the test-maker postulates a large number of different components.Further, his method usually requires a selection of alargenumberof items, a procedure which shows that the test-makerbelieves that the components causing individual differences inthe defined domain occur in large num bers. The fact th at hesubsumes all the items under the same abstract definition ofdomain does not mean that he assumes the existence of anygeneraldifferentiae of ability in the domain. His inclusionwithin the same domain of all the diverse items put there in-volves no assumptions about the factorial bases of the re-sponses to the many items; his laying out of the commondomain derives from the mundane convenience of arbitrarilychoosing to study one a priori classification of reactions to-gether. The types of domains which he chooses to study aredictated by the needs of his cultural group and are of valueaccording to their personal, social and economic usefulnessor significance.The welter of components sampled by a given test maybest be understood by taking a specific type of test for illustra-tion . I have selected for this purpose Thornd ike's well-known measure of intellectCAVD (levels / toM), conceded tobe one of the best tests of the group verbal type. Thorndikeproperly proceeded by marking out by a definition four verbaldomains, namely, sentence completion (C), arithmetic {A),vocabulary (F), and directions (D ). For each of thesedomains, he chose many different items satisfying the defini-

8/13/2019 Try on 1935 Small

10/30

4 3 4 ROBERTC.TRYONtion of the dom ain. Consider the V part: it contains 50vocabulary items. Why this large num ber? In selectingsuch an array has not the test-maker tacitly granted thatthere are many factors determining differences in the knowingand naming of words, and th at only a copious sample of itemscan in an adequate manner cover the complex substrata ofword-knowing? The possibility of a general single verbalfactor functioning as a unity may occur to him as a mathe-matical after-thought, bu t when he builds his tes t he proceedson the basis of a theory of factorial com plexity. A test, then ,in its very construction plainly implies a multiplicity of factorsdetermining individual differences.

PRO PER TIES OF PSYCHOLOGICAL COM PONENTSWe shall review the properties of the psychologicaldeterminants of individual differences under the same rubricsused in the summary of the assumptions involved in thepostulation of mathem atical factors. The components in-volved in 'intelligence' and achievement domains will bedealt with first.Number of determinants. How many determinants areinvolved in the causation of differences in total score on, forexample, CAFD? Consider, for simplicity, only the V sec-tion. The universe of psychological determ inants sampled

here are the great hosts of concepts with appropriate nameswhich come under the domain label, 'voc abulary.' The Vsection samples this universe by presenting 50 multiple-choice vocabulary items, beginning with 'confess' and endingwith 'chas tity .' There are probably more than 50 conceptsinvolved in these 50 items, since the passing of each item re-quires not only knowing its conceptual basis and its linguisticmode, but, because of the multiple-choice character of theitem, also the conceptual bases and names of some of theincorrect 't e s t ' words. Individual differences in to tal Vscorewill be roughly proportional to the number of these conceptswhich the subjects have formed and named. The evidence asto the multiplicity of such conceptual components derivesprimarily from a consideration of the concepts elicited in

8/13/2019 Try on 1935 Small

11/30

A THEORY OFPSYCHOLOGIC LCOMPONENTS 4 3 5working on these items in the testing situation, and from astudy of the original learning of the concepts and their ap-propriate words. The subject taking this test is aware of thewealth of concepts which he must bring to bear on such alist of words. Every school boy knows of the multiplicity ofconditions involved in the learning of 50 such words as these;no unifying general verbal factor exists for him. In additionto the vocabulary section of CAVD, there are 50A's, 50 Vsand 50D's. The number of concepts which a subject musthave formed and have at command before he may earn a highscore on this test runs into hundreds. The final level of totalscore earned by different subjects is to a large extent de-termined by the total number of these possessed.Believers in general or group factors may immediatelyobject to this conception of determiners of mental differencesby protesting that they do not mean by 'factors' thosementioned above; whattheymean are the fundamental under-lying 'functional unities' which enable the individual toformthe welter of concepts measured by tests. We shall considerthis doctrine later in this paper. I t should be evident, how-ever, that the first and proper function of the psychologist isto study and describe the palpable psychological componentswhich determine mental differences. When he does this , heis immediately struck by the large number and complexity,and not by an organized, simplified array of psychological'unities. '

Interrelation between determinants.The concepts m eas-ured by 'verbal tests' appear tobe functionally independentof each othe r. By functional independence I mean th at thepossession and use of one concept necessary to pass one or afew items does no t necessarily predicate the possession and useof other concepts necessary to pass other blocks of items.In CAVD one may know the concept, 'confess,' but not thatof 'chastity'; one may possess the concepts necessary to pass'30 7 = ?,' bu t not those necessary to solve '20 = (2 /3)* ';one may form the concepts necessary to fill in the blanks in'T h e way to is by airplane,' bu t not thosein 'T h e old days are often with thepresent.'

8/13/2019 Try on 1935 Small

12/30

4 3 6 ROBERT C. TRYONParadoxically, however, though such concepts are func-tionally independent in a given individual, when arrays of

different individuals are studied objectively, one frequentlyfinds a positive correlation between th e presences and absencesof different concepts sampled by a test. Thu s it appears tha tthe individuals who tend to pass one group of hard items in atest tend to pass other groups of hard items involving func-tionally quite different concepts. Likewise individuals whofail in the first tend on the average to fail in the second. I tis such positive correlation between different items or differentblocks of items which has been termed 'internal consistency,'a phenomenon which has misled many psychologists to educefrom it the existence of a common 'functional core' or factordetermining success. Such positive correlation may arise,however, from a number of 'external' agencies, and not fromcommon conceptual content. One such agency is systematicdifference among individuals in the general quality of theeducational and cultural medium in which the different con-cepts are evolved. To illustrate , if individual A is reared in ageneral educational medium which presents superior oppor-tunities and demands for learning verbal concepts, he willbe superior in the knowledge of a large number of quitedifferent verbal concepts, even though these concepts areentirely independent with respect to content and form andwith respect to the specificenvironm ental fields in whichhe evolves them . Individual B, if reared in a general culturalmedium manifesting inferior opportunities and demands foreduction of verbal concepts, will analogously manifest de-ficiencies in the knowledge of many diverse blocks of suchconcepts. Thus, functionally quite independent conceptsmay display correlation in a test situation {i.e., internalconsistency) because of a correlation between the environ-mental fields in which they evolved, and not because of anypsychological unity in the individuals.2

1Ifonechooses to call the general agencies which occasion the correlation betweenenvironmental fields, 'general' or 'group' factors, such appellations may, I presume,be valid. It should be noted, however, tha t such a meaning is not the one usuallygiven to such factors by factor analysts.

8/13/2019 Try on 1935 Small

13/30

A THEORY OFPSYCHOLOGIC L COMPONENTS 4 3 7Combination of determinants.It would seem naive psy-chologically to assume that the components of mental differ-

ences combine by simple addition. To be sure, one adds upthe item responses of an individual to secure his total score,but from this it cannot be inferred that the numerous con-cepts which he employs in producing these responses havecombined psychologically by addition in his solution of anyparticular item or a group of them . I need not bore thereader by presenting Gestalt experiments which show that thewhole is not necessarily the sum of its par ts. In a vocabularytest the subject brings to the problems a large number ofconcepts which he may employ in diverse and complex ways.In solving some problems he may employ previously formedconcepts; for other problems such as those in many 'intelli-gence' tests, he may evolve for the first time quite new con-cepts which, though grounded upon previously formed oldconcepts, may manifest new emergent properties not char-acteristic of the old. The interaction of concepts used in thesolution of such complex problems as those presented inCAVD, for example, is doubtless multiform. To envisagethem, or the results of their use, as the expression of a simpleaddition of some sort of elementary psychic 'stuff' would beto indulge in an aprioristic fiction.

Weight ofdeterminants The number of concepts and themanner in which they combine in producing a total responseon a test is closely tied up with the problem of the weight ofeach concept in producing the tota l score. Consider thecase of two individuals who earn the sametota l score. Oneindividual might employ 25 functionally independent con-cepts in getting this score, the other many hundreds. Th ismay be illustrated in a test involving arithmetic problems;in earning the same total score one subject may employalgebraic equations, the other simple rule-of-the-thumbarithm etic operations. Th e weight of each of the concepts indetermining the total score of each individual would be quitedifferent, in fact, quite indeterminate because of their dy-namic non-additive character.Nature of determinants.Mathematical factors, it may be

8/13/2019 Try on 1935 Small

14/30

438 ROBERT C. TRYONrecalled, suffer as a major defect from the fact that theirpsychological na ture remains unknown. N o such perversitycharacterizes psychological de term inants . In the domainsof cognition, it has been suggested above that they are con-cepts and conceptual relations. The study of them has been amajor problem in the experimental psychology of perception,learning and thinking. In addition to the experimentalattack, introspection provides ano ther approach . In suchtests as CAVD, and other group verbal types which, in theirmore difficult levels at least, are primarily measures of con-cepts formed in school situations, the psychological nature ofthe concepts sampled m ay best be understood from a study ofthe experimental and introspective psychology of reading,vocabulary knowledge, arithm etic and algebra. Briefly itmay be said tha t a concept has the following aspects or proper-ties:it possesses ( i)form (orprocess), this ranging from simplecontiguous relations or associations between stimuli or eventsto highly complex types of relations or abstractions; (2)con-tent (or subject matter)which refers primarily to the specificmeaning or significance it has for the experiencer; (3)reten-tivity,which refers to its stab ility or persistence through tim e,the degree of which is conditioned by frequency, recency,intensity, e tc.; and (4)degreeofgeneralityor transfer. Statingthe features of concepts in the above ways is only one mannerof describing the m ; there are doubtless others. Th is much istrue, however: if a concept is studied in terms of its form,content, retentivity, and degree of generality, a reasonablygood description of it as a functional psychological entity inthe subject's adjustment to his environmental field will beobtained.

The degree of generality of concepts sampled by tests isof special interest here for the reason that this property ofconcepts determines, in part, the intercorrelation betweentest items and between tests. Consider, for instance , theconcepts measured by 've rbal te sts .' Some concepts mayhave the most limited application, as, for example, 'hoof,'making for successful response in but a few item-situations;others may have the most general application , as, for example,

8/13/2019 Try on 1935 Small

15/30

A THEORY OFPSYCHOLOGIC L COMPONENTS 4 3 9certain basic verbs like 'make,' resulting, if known, in a suc-cessful response in a wide variety of verbal situa tions. Con-sider the negligible effect of ignorance of such concepts as, forexample, 'bite,' 'cook,' 'howl' on a total score in a test ascompared with the disastrous effects of ignorance of suchconcepts as 'g o ,' 'b e ,' 'd o .' The difference lies in the degreesof generality of each. Verbal tests sample hosts of such con-cepts, these differing widely in their several degrees of gener-ality . Quite different verbal tests usually sample many ofthe same ones, especially those of the more generalized types,a fact which partially accounts for the positive correlationsusually found.Origin ofdeterminants A final consideration is a state-ment of the origin of the psychological components determin-ing mental differences. For this, we must tu rn to studies onthe evolution of concepts, on insight, and on reasoning.8Briefly stated, concepts come into being in the followingmanner: when the subject is faced by a problem-situation re-quiring for solution a given concept or group of concepts, ifthe psychologicaliel s properly organized, one may evolvethe necessary concept or concepts adequate for the solutionof the problem. By the 'psychological field' is mean tboththe presence and sentience of an adequate environmentalexternal field and the presence of an adequateconceptualbackgroundin the light of which the external field is observed.By being 'properly organized' is meant that the externalfield features and the conceptual background must be experi-enced in a certain form or relation. The whole relationalset-up must be 'r ig h t' or the concept may not evolve. I tturns out to be pretty much of an all-or-none matter, asexperiments on insight show. Even though these conditionsare adequate, the concept m ay not come off. A furtherrequirement is that the experiencer must have an adequate

Of special relevance are the following references:N . R. F. Maier, Reasoning in human s. I. On direction, / . Comp. Psychol. 1930,

i o , 115-143;'I I. T he solution of a problem and its appearance in consciousness, / .Comp. Psychol. 1931, 12, 181-194.

N .R. F. M aier, An aspect of human reasoning, Brit. J. Psychol 1933,24,144-155.E. C. Tblman, Purposive behavior in animals and men, N . Y .: Century, 1932.

8/13/2019 Try on 1935 Small

16/30

4 4 0 ROBERTC TRYONgenetically determined somatic constitution. As an i l lustrationof these conditions, the writer recalls the situation in whichhe is confronted with the problem of developing the concept,'probable error,' in an elementary statistics class. Before thestudent can evolve this concept, a conceptual backgroundcomprising certain basic concepts must first be evolved inhim, then an external field consisting of black-board m aterialand verbal details must be arranged in a certain form and inrelation to the conceptual background. Then the studentmay 'click,' or he may no t. If he does no t, the conceptualbases are examined and clarified, and different external fieldsare arranged for him in the hope that one of these may work.With certain students, of course, either the instructor isunable to organize the right psychological field for them, or,even if properly organized, they may native ly lack the somaticcomplexity necessary for the execution of this particular typeof insight. With reference to the concepts measured in sucha test asCAVD, the question of how these are formed shouldproperly be referred to teachers of vocabulary, reading,arithmetic, and algebra. In vocabulary education, theproblem appears to be, first, that of presenting the subjectwith adequate psychological fields so that he may 'insight'the innumerable concepts necessary, and second, that ofteaching him to affix the proper name to each of these con-cepts. The V element of CAVD measures the number ofsuccesses which the subject has attained in many pastsituations.

UN IVERSA LITY OF PSYCHOLOGICAL COM PONEN TSThe confinement of the above discussion to the determina-tion of mental differences in cognitive domains, such as those

measured by CAVD and similar tests, has been an unneces-sary restriction. For the psychological determinants of indi-vidual differences in 'personality' domains, whether measuredby questionnaires or ratings, have characteristics analogousto the conceptual components described above. In anydefined personality domain, whether it be neuroticism, angers,fears, annoyances, interests, attitudes, cooperation, deceit,

8/13/2019 Try on 1935 Small

17/30

A THEORY OF PSYCHOLOGICAL COMPON ENTS 4 4 1persistence, etc., there exist in individuals motivational andemotional dispositions or sets for action and feeling. A ' per-sonality test' samples by its items a defined universe of thesedispositions, which appear to be vast in number, functionallyindependent of each other, though, like conceptual compo-nents, often correlated by virtue of certain generalized ex-ternal and biological agencies, and the y combine in more com-plex ways than by simple sum mation. Dem ands for brev ityprevent a treatm en t, similar to tha t given above to conceptualcomponents, being presented here. For an understanding ofthe psychological nature and origin of these sets and disposi-tions, one must turn to the experimental study on the originand na ture of the emotions, drives, aversions, etc. In thesedomains of behavior, though not yet submitted to the inten-sive program of measurement as have been the cognitive,any attempts to educe general or group mathematical factorsfrom intercorrelations will result in the same fictitious sim-plifications as have been proposed in the realms of abilities.

PRO PERT IES OF THE FUNDAMENTAL CAUSES OFCAPACITIES AND PREDISPOSITIONS

Those who embrace the conception of broad general orgroup factors may admit that the above psychologicaldescription of the causes of mental differences is cogent, andpossibly acceptable to them . B ut they will probably objectto such an interpretation as not constituting the whole casefor factors. W ha t they mean by such factors as g, verbality,number,space, etc., are the more 'fundamental unities' whichdetermine the individual's capacityor predisposition to formthe concepts and sets measured by cognitive tests, question-naires, and ratings. They will claim th at both the psycho-logical and mathematical interpretations could be advocatedwithout con tradiction . The cogency of such a claim dependsentirely upon what is meant by fundamental unities. Ifthe meaning is that the general or broad group factorsrepresent the fundamental causesof differences in capacitiesand predispositions, then the available experimental evidenceon the nature of such causes precludes the postulation of

8/13/2019 Try on 1935 Small

18/30

ROBERT C. TRYONgeneral or broad group factors. There seem to be two typesof fundamental causes, hereditary factors and environmentalfields. We must now consider the properties of these causeswith special reference to completing our own systematictreatment, and incidentally to showing that general and groupfactors may not be interpreted as these.Hereditary determinants.Is it not possible th a t thegeneral factors proposed by the mathematicians representspecific hereditary unitary components? May not, forexample, Spearman's g, which supposedly is a major deter-minant of differences inCAVD scores (let us recall Holzinger'sdeclaration:'CAVDis full ofg'), represent a stable hereditaryvariable which saturates in different degrees all cognitivetraits? M ay not broad group factors also be hereditary?The reply to these queries is simply that there is no evidenceto support an affirmative answer. To the geneticist, a factoror component is agene,or more properly, alocusin a chromo-some. A number of genetic studies have been m ade onquantitative structures which, like mental abilities, are moreor less normally dis tributed. The geneticist's conclusionsare that for such characters, no simplified set-up such as thatpostulated by the Gand P theories occurs genetically, butthat multiple, independently assorting genes, or Mendelianfactors, are operative.Most of the evidence on the properties of genetic factorshas derived from the study of animal and plant structures;little work has been done on behavior. These properties,experimentally observed, may be summarized as follow^, ( i)As to thenumber of genetic factors, there are doubtless thou-sands of them in the 24 pairs of chromosomes of human beings,for in the fruit-fly, Drosophilamelanogaster, with only fourpairs of chromosomes, a conservative estim ate is 4,000. (2)With respect to interrelationbetween them , they manifestindependence in their d istribution to progeny, th a t is, differentfactors in different chromosomes assort independently, differ-ent factors in the same chromosome lose correlated associationby 'crossing-over.' (3) As to combination and interaction, inthe case of normally distributed characters, where blocks of

8/13/2019 Try on 1935 Small

19/30

A THEORY OF PSYCHOLOGICAL COMPONENTS 443multiple independent genes provide the factorial basis,thegenes of the block do not inevitably combine by simplecumulationinthe formingofthe phenotype but frequentlyinmore complex waysas is the case with duplicate and com-plementary factors. (4) As to the nature of the gene, itmanifests adefinite contentof an as yetunknown biochemicalconstitution which provides thebasisfor thedifferent char-acters,adegreeofstabilitywhich, exceptforrare m utation,isone of permanence, and a degreeofgenerality with respecttotheextentofeffect ithason different tissuesof thesoma.(5) As to theoriginofgenes and gene blocks, the process is oneof mutation, supplemented by fragmentation, duplication,polyploidy,etc.The vas t num bers of somatic structures in the body possessas their genetic basisanexceedingly complex matrixofgenesinteractingincomplex ways, hence sucha simplified factorialpattern as that suggested by mathematical-factorists is notconsistent with a proper conception of Mendelian factors.It follows, therefore, that a mathematical factor cannotbeconceived of asan hereditary factor. May it not then beargued that a mathematical factor, instead ofbeingasinglegene,is ablockofgenes, this block behaving asastable un ityaffecting many tissuesandhenceagreat varietyofbehaviors?I have found no evidenceinDrosophilastudies th atagroup ofgenes functions togetherin a stable unvarying equilibriumaspartial determiners of a variety of different tissues. Sincedifferent behaviors appeartobring into play different tissues,4and since thereis noevidenceto indicate that astable blockof genes operatesas a common genetic matrixindeterminingdiverse tissues, no basis exists for assuming that agroupofgenes constitutes ageneralorgroup genetic factor.

A certain risk is involvedinour eductionofthe propertiesof the genes determining differences in mental abilitiesandtemperamental characters from those known to determinedifferences instruc tures. Crucial evidence must come fromgenetic studies of behavior differences. Such evidence is4 In this connection,see the interesting treatiseby J. A. Gcngerelli, Brain fields

andthelearning process, Psychol. Monog. 1934, 45, No. 4 p. 115.

8/13/2019 Try on 1935 Small

20/30

ROBERT C. TRYONvirtually non-existent, for psychologists have been slow toseize upon the techniques of the geneticists and to apply themto a study of the Mendelian bases of behavior. The writerhas been making an attempt along this line for a number ofyears. His experiment aims at a discovery of the na ture ofthe genetic factors determining individual differences in maze-learning in ra ts . Here we have behavior. The results todate shed sufficient light on the problem to warrant a briefstatement of them.

The rats are reared and run through a maze under condi-tions of rigorous environmental contro l. Ind ividual differ-ences in ability to learn the maze are wide and consistent.This fact itself seems to indicate the operation of multipleindependently assorting genes. The evidence is not crucial,however, because of the possibility that the differences inlearning might be determined by environmental variationsimpossible to control. For crucial evidence it is necessary toproceed with selective breeding for maze-brightness and maze-dullness. Such selective breeding has now been continued tothe eleventh generation, with the result that we possess tworaces of rats, one breeding relatively true for maze-brightness,the other for maze-dullness. There is negligible overlapbetween the two races.5 The significant fact appears th a tthe effects of selective breeding are quite gradual, a phe-nomenon indicating that we are here working with a complexblock of genes determining this specific ability. Selectivebreeding results in the gradual sorting into one pile, as it were,of all the genes determining capacity for superiority in thismaze abilitythe bright raceand into another pile, of allthe genes determining inferior capacitythe dull race.Crucial tests have been made, namely, the breeding of brightrats to dull, the Fi of which were then inbred, giving the F 2 .Among these progeny arose rats of all degrees of ability,indicating clearly recombination effects among the manygenes involved in the functional matrix determining this

R. C. Tryon, The genetics of learning ability in ratsa preliminary report,Univ.Calif.Publ. Psychol., 1949, 4, 71-85- Graphs of results secured since the publi-cation of the preliminary report appear in F. Moss, (Ed.) Comparative psychology,N.Y.: Prentice-Hall, 1934, Chap. XIII, 'Individual differences' by R. C. Tryon.

8/13/2019 Try on 1935 Small

21/30

A THEORY OFPSYCHOLOGIC L COMPONENTS 44 5ability. One is led to conclude from all these results th a t thegenes determining capacity in this problem-solving behaviorare exceeding numerous and complex in their production ofthe phenotype.Evidence on the genetic determination of the activity ofrats has recently been published by Rundquist,6 who foundgradual effects of selective breeding for active and inactiverats. The effects here were somew hat more rapid than in themaze experiment, suggesting that there are probably fewergenes involved in determining differences in activity than inmaze ability.Returning to human-beings and mental tests, the abovetre atm en t may shed some light on the role of genetic factors indetermining the capacities for the evolution of concepts. I tmust be kept in mind that such a test asCAVD, for example,is a sample of many diverse conceptual operations, involvingthe participation of many different somatic tissues (brainfields, sensoria, musculatures, etc.) brought into action bymany past and present environmental fields. In the evolu-tion of the many concepts sampled by the test, great massesof somatic tissues are and have been involved, and hence thedifferent gene blocks that determine the relative efficienciesor capacities of these tissues constitute the fundamentaldeterminan ts of this ability. Thu s it is th at a tes t, insofar asheredity affects the performance, indirectly samples a largegroup0/genes. Different tes ts may sample different groups.Environmental fields. The second class of fundamentalcauses of individual differences in abilities and o ther behaviorsis that of externalenvironmental fields. In order to describethe properties of these fields, one must hold genetic varianceconstant. Let us specify this treatm en t by asking the ques-tion: What are the properties of the environmental fieldswhich determine differences in CAVD score between twoindividuals of exactly the same genetic constitution? Theseproperties m ay be described under the heads considered above.As tonumber,there are countless ways in which environmental

6E. A. Rundquist, Inheritance of spontaneous activity in ratt, / . Comp.Psychol.1933. 16,4I5-438.

8/13/2019 Try on 1935 Small

22/30

4 4 6 ROBERT C. TRYONfields may differ for identical genotypes, each way having itsown unique effects on behavior. I t is to be recalled th a t thepsychological field not only includes the external situationpresented to the individual, but also the conceptual back-ground. Among identical genotypes differences in this con-ceptual mass are determined by differences in previouslypresented external fields. Thus the number of present andpreviously presented external fields that may produce differ-ences in performance among identical genotypes is infinitelylarge. In mental tests such as the Binet, an effort has beenmade by a standardization of test procedure to control theimmediately presented external field, and by a selection of'common every-day' content to hold constant differences inpreviously presented fields, thus leaving as the sole source ofvariance the hereditary differences am ong individuals. Sonumerous are the fields which affect performance, that suchefforts have been attended by little success, as is evidencedby the wide divergence in I.Q. discovered among some iden-tical twins reared apart.7 As to the interrelation betweenenvironmental fields, they appear to manifest physical in-dependence. For example, the specific external conditionspresented the individual in endeavoring to get him to formvocabulary concepts are arranged for him independently ofthose in which he develops arithmetic concepts. Likewise,the specific stimulus conditions arranged for his evolving agiven verbal concept are physically independent of thosearranged for other verbal concepts. The problem of thecombinationof external fields is one of great com plexity.Thatfieldsshould combine either physically or psychologicallyby summ ation is inconceivable. Certainly the disparatestimuli which make up a given external field are not experi-enced by the individual as a summation but ordinarily as apattern; indeed, this fact was the basis of our choice of theterm, field. The physical nature of external fields may bedescribed in terms of the forms and patterns of physicalenergy presented to and perceived by the experiencer. W ith

7A sum mary of the evidence on iden tical tw ins reared apart appears in G .Schwesinger, Heredity and environment, N . Y .: Macm illan, 1933, 209 -233.

8/13/2019 Try on 1935 Small

23/30

A THEORY OF PSYCHOLOGICAL COM PONENTS 4 4 7reference to the origin of external fields in the individual'sexperience, these may be classified into the adventitious fieldsin which the individual accidentally finds himself, thoseformally and systematically prepared for him by his societalculture, and those which he seeks out himself. Such com-plexity faces us in our envisagement of external environ-mental fields th at it would be egregious oversimplification tobelieve that general or group factors proposed by mathe-matical factorists may be conceived as these.

In the determination of the psychological components ofability and temperament, genetic determinants and environ-mental fields often interact in such ways that, among humanbeings at least, it is ordinarily impossible experimentally tohold one constant while studying variation in the other. Thereason for this difficulty is that, though physically indepen-dent variables, these two types of causes often manifest acorrelation. Th us in an economic and educational caste orclass system, which often results in a tendency of inbreeding,different types of genotypes tend to become confined todifferent types of environmental fields. Even underlaissexjaireconditions correlation may appear: individuals possessingsuperior genetic constitutions for special excellence in certainclasses of environmental fields tend to seek out and providethemselves with special superior fields befitting their innatelysuperior predispositions.

VARIABILITY OF THE INDIVIDUALThe above view of the conceptual components of mentaldifferences illuminates a number of important problems inmental measurement which are given a fictitious solution bymathem atical-factorists. One such problem is th a t of the

so-called 'chance error' of an individual's tes t score. Themathematical-factorist believes that an individual's responseon, say, a given vocabulary item, X\, in the V element is thesum of truth, Xm, plus anerror, E\. Stated equationally andusing deviations, Xi xm -\- e\, where the true factor is thesummation of all the systematic factors extracted by factoranalysis, and the error is 'chance.' The form of the equation

8/13/2019 Try on 1935 Small

24/30

4 48 ROBERT C. TRYONis supposed to hold for all the other test items, xt, #,- ,the true factor being thesamefor all items, but the errors allbeing different and independen t. The summation of allthe items is supposed to result in the cancellation of th e chanceerrors (since uncorrelated), leaving in the total score thetru th , except for a minimal total chance error. Th is concep-tion of truth and error is, however, psychologically absurd.Suppose, for example, that in the V element in CAVD onepasses the item, 'confess,' and fails 'chas ti ty .' W ha t is theidentical core of 'truth' in both of these responses, and whatare the two residual chance errors? Th e truth-error postula-tion is quite legitimate, mathematically,and many of theformulae referring to the reliability of measurement may bededuced from it,8 but it has no psychological meaning.The error of an individual's total score may be estimatedfrom one of these formulae. Th is error represents the vari-ability among the total scores of a given individual on manytests exactly comparable in con tent and form. Under thetruth-error doctrine, this variability is supposed to be due tominimal chance errors, each score diverging from some core oftruth in the individual because of one such erro r. All of thisis decidedly mysteriousand quite unnecessary. One hasonly to recall how 'comparable tests' are constructed and tokeep in mind the above treatment of the conceptual compo-nents of ability in order to understand why an individualdiffers in his scores on com parable te st s. His va riabil ity isdue,not to some mysterious undefinable error, bu t to the factthat though each test samples the same universe of conceptsthe samples may differ in many particulars.9 W ith respectto those concepts in which the tests differ, an individual maypass those in one test and fail those in the other. In general,

See L. L. Thurstone, The reliability and valid ity of te sts, A nn Arbor: EdwardsBros., Inc., 1931; and R. C. Tryon, The reliability coefficient as a per cent, withapplication to correlation between abilities, PSYCHOL. REV., 1930, 37, 140-157.

' There is an increasing tend ency for writers to term Vr as t h e ' theoretical validity*ofagiven measure who se reliability coefficient isr. Since Vr is the correlation betweenthe given sample and the universe of measures from which it is drawn, the valueindicates how valid a measure of the universe the given samp le is. For a discussionof this point, see H. Hartshorne, and M. May, Studies in deceit, N. Y.: Macmillan,1930,Book II , pp. 118 .

8/13/2019 Try on 1935 Small

25/30

A THEORY OF PSYCHOLOGICAL COMPON ENTS 4 4 9the error of an indiv idua l's score will be significantly affectedby the correlation between these residual concepts. Thevariability is due to 'chance,' not in the sense that the indi-vidual is unsystematically in error, but in the sense that thetest-maker has by chance put in one test some items requiringthe use of some concepts not required by items in the othertes t. In fact, two com parable tests give correlations less th anunity for the same reason that two non-comparable tests do,namely, because in some degree they measure two somewhatdifferent conceptual complexes. Only when two com parab letests contain so many items that they both cover adequatelythe whole universe of concepts will an individual tend to earnthe same total score on both tests.

RATIONALE OF THE CORRELATION BETWEENPSYCHOLOGICAL DOMAINS

Com munity {'overlap') of multiple psychological com-ponents.In the foregoing pages, we have dealt primarilywith the causes of mental differences in a single sample froma defined universe of psychological determ inants. A properconception of these components leads us directly to a descrip-tion of how they operate to produce a correlation betweendifferentpsychological dom ains. I t has been the study of theintercorrelations between mental measures which has leadmathematicians to postulate their general and group factors.Consider the case, for instance, of the correlations betweenCAVD and other group 'intelligence' tests, such as Otis,Alpha, Terman, National etc . In a fairly random sample ofindividuals homogeneous in age, the true intercorrelationscluster around .90. On the other hand, correlations betweensuch verbal tests and mechanical, non-verbal, and perform-ance tests, and measures of 'narrow' verbal domains, suchas immediate verbal memory, etc., are very much lower,though ordinarily positive. To a mathematical-factorist,these positive coefficients indicate the presence of a singlegeneral factor and/o r a few broad group factors. Thepsychological interpretation, however, is that the magnitudeof each positive coefficient indicates roughly the extent to

8/13/2019 Try on 1935 Small

26/30

45 ROBERT C. TRYONwhich the two tests in question sample similar universes ofconceptual components. CAVD shows high correlation withOtis, Alpha, and similar tests, not because some mysticgeneral factor saturates them all heavily, but because all thesetests sample to a marked extent the same complex welter ofconcepts determining vocabulary, reading, and arithmeticabilities. Instead of a single common factor, there are in-numerable common conceptual components, which possess theproperties indicated earlier in this paper. CAVD shows alow correlation with non-language tests and tests of narrowverbal domains because each test tends only to a limiteddegree to sample the same domains of conceptual componentssampled by the others.

Correlation between independent environmental fields.Thisoverlap of numerous psychological components is not, how-ever, the sole source of positive intercorrelation betweentests. Often it appears tha t two tests which on analysisseem to sample in no degree identical psychological com-ponents nevertheless give positive intercorrelations. Indeedit is customarily observed that any table of intercorrelationsbetween a diversity of cognitive domains rarely shows asingle coefficient tha t is no t positive. From such facts asthis, some psychologists conclude that here, finally, is definiteevidence of a single general factor. This conclusion is,however, a too simple eduction, for there are more cogentreasons for such universal positive correlation. I t is verydoubtful whether any two cognitive tests of however differentsubject matter could be given to a random population withouteliciting an overlap of some conceptual, emotional, andmotivational dispositions. But over and above this overlapthere are other explanations of positive intercorrelation.Consider the case of the correlation between a vocabularytest and an arithm etic com putation test. Even if the com-putation test be so administered that no verbal commands orwords of any sort were employed, a positive correlation wouldappear between these two measures. No one could seriouslydoubt, however, that the subjects, or at least some of them,would bring to bear on the two tests some common concepts,

8/13/2019 Try on 1935 Small

27/30

A THEORY OF PSYCHOLOGICAL COMPONENTS 4 5 1emotions and motivations which would account for some ofthe positive correlation.But another source of this correlation is that mentionedearlier in our explanation of 'internal consistency,' namely,the phenomenon of correlation between the environmentalfields in which the different psychological components areoriginated. How this phenomenon operates to occasion apositive correlation between the illustrative vocabulary andcom puta tion tests should be eviden t. Individuals who, inthe home, school, or general educational medium, are formallypresented with the special environmental fields which elicitsuperior verbal concepts, tend alsoby the operation of thesegeneral cultural media to be exposed to the special environ-mental fields which elicit the formation of concepts basic tosuperior arithm etic com puta tion. Analogously, a generallyinferior cultural medium would lack many special environ-mental fields necessary for the formation of both verbal andcomputational concepts, the consequence being that theindividual suffering such lacks would fail to develop muchknowledge and skill in bo th achievement domains. In asocial system where families and individuals suffer widedifferences in general social, economic, and educational status,a positive correlation occurs between the qualities of themany special environm ental fields necessary for the evolutionof qu ite different blocks of concepts. Such phenomena wouldproduce a positive correlation between verbal and computa-tional concepts, even though for the individual they would beactua lly functionally qu ite independent. Thus, two testssampling these two quite independent blocks of conceptswould show a positive correlation. The manifest inter-correlation between physically independent environmentalfields due to such general agencies as educational opportun ityand culture appears to be an important source of the generalpositive intercorrelation universally observed between quitediverse cognitive measures. Such a phenomenon is a general'factor' of a sort, but it is not a general psychological func-tional un ity or 'c o re ' within the individual. The generalcommon factor lies not within him, but in his cultural back-ground.

8/13/2019 Try on 1935 Small

28/30

452 ROBERT C. TRYONCorrelation between independent gene-blocks.The secondphenomenon that accounts in part for the general positive

correlation between diverse mental measures also lies backin the fundam ental causes of menta l differences. I t is thecorrelation between independent gene-blocks. It resultsfrom one type of assortative mating among parents . Malestend to mate with females of their own generaleconomic,educational and cultural status . As a consequence men andwomen of genetic superiority for achievement in quite differentdomains tend to interbreed, as well as do those of diverseinferiority. On the genetic side, what happens, then, is th a tindividuals possessing blocks of genes determining superiorcapacity for one type of achievement unite with individualspossessing blocks for superior capacity in another. The iroffspring, in consequence, tend to receive genes for superiorcapacity inbothtypes of achievement. Assortative mating ofindividuals inferior in diverse gene-blocks results in a distri-bution to their offspring of inferior genes for capacity indiverse abili ties. Whereas th e genes affecting differentabilities are physically independent, this type of assortativemating produces a positive correlation in their phenotypicexpressions. Here the positive correlation between differentabilities is occasioned by no general genetic factors in theindividual; the common factor lies one or more generationsback.The correlations between environmental fields and betweengene-blocks are not of high magnitude. T ha t they are lowis evidenced by the fact that the actual objective correlationbetween measures possessing a negligible overlap of-psycho-logical components is qu ite low. Crucial evidence of theoperation of such agencies can apparently be secured only inthe animal psychological laboratory, where control of en-vironmental fields and systems of breeding is possible.Ordinarily environmental control and random breeding areinstituted as a m atter of routine in the animal laboratory. I tis therefore of interest to discover that in the few and as yetinadequate studies made on laboratory rats, where thecorrelations between environmental fields and between inde-

8/13/2019 Try on 1935 Small

29/30

A THEORY OF PSYCHOLOGICAL COM PONENTS 4 53pendent gene-blocks are avoided, there appears no evidenceof a significant positive correlation between such diversedomains as maze-learning, problem-box learning, and visual-sign learning,10tasks which apparen tly involve environmentalfields so different as to elicit a negligible overlap of conceptualcomponents. I t is to be concluded tha t if the correlationbetween independent environmental fields and between in-dependent gene-blocks were zero, tests involving no commu-nality of psychological components would show intercorre-lations of zero.

SUMMARYTheories which postulate 'mathematical factors' as thedeterminants of individual differences involve assumptionsth at do not possess psychological or genetic validity. As analternative, a theory of psychological components is presented.These determinants are considered to be concepts, and moti-vational and emotional dispositions. Th ey exist in large

numbers, manifest functional independence, combine incomplex ways not necessarily summative, display relativelyindeterminate weights in producing a reaction, possess form,content, retentivity, and degree of generality as definitivefeatures. Each component originates in a psychological,relational field consisting of an external environmental fieldand a conceptual background, and requires as a necessarycondition for emergence an adequate somatic constitution.Psychological 'tests' are samples of such components from ana prioridefined universe or domain of com ponents. Variationof an individual on different forms of a test is due not to amysterious factor of 'chance' but largely to non-identity ofthe components sampled by each form. The fundamentalcauses of the capacities to form the innumerable componentsare gene loci and environmental fields, these manifesting greatmultiplicity, independence, complex interaction, indetermi-nacy of weighting, and other definitive features. Knowledgeof these psychological components and of their fundamental

10For a summary of this type of evidence, see F . M oss, (Ed.) Comparativepsychology, Chap. XIII, 'Individual Differences,' pp. 440. and M. Tomilin, andC. P . Stone, Intercorrelations of measures of learning in the albino rat, / . Comp.PiychoL, 1934,17,73-88.

8/13/2019 Try on 1935 Small

30/30

454 ROBERT C. TRYONcauses provides a rationale for the intercorrelation betweentests or behavior samples. Positive correlation between twotests is conditioned by the extent of community of compo-nents,by the degree of correlation between the environmentalfields in which the components sampled by the two testsoriginate, and by the degree of correlation between the gene-blocks determining the somatic tissues brought into actionby the two tests.[MS.received November 14, 1934]