Stephenson Jung's Typology

8/2/2019 Stephenson Jung's Typology

1/22

10.1192/bjp.85.355.185Access the most recent version at DOI:1939, 85:185-205.BJP

W. StephensonMethodological Consideration of Jung's Typology

Referenceshttp://bjp.rcpsych.org/content/85/355/185.citation#BIBLThis article cites 0 articles, 0 of which you can access for free at:

permissionsReprints/

[email protected] obtain reprints or permission to reproduce material from this paper, please write

to this article atYou can respond http://bjp.rcpsych.org/cgi/eletter-submit/85/355/185

fromDownloaded

The Royal College of PsychiatristsPublished byon February 29, 2012http://bjp.rcpsych.org/

http://bjp.rcpsych.org/site/subscriptions/go to:The British Journal of PsychiatryTo subscribe to
http://bjp.rcpsych.org/http://bjp.rcpsych.org/


2/22

vieil viS!f;, COLLEGI' CITYTHE MAR2 41939JOURNAL OF MENTAL SCIENCE'

[P ublished by A uthority of the R oyal Medico-P sychologicalAssociation.]No. 355 [To.S3T] MARCH, 1939. VOL. LXXXV

Part I. Orig in al Articles.METHODOLOGICAL CONSIDERATION OF JUNG'S TYPOLOGY.

B y W . S TE PHEN SON,M .Sc., M .A., P h.D .From the Institute of Experim ental Psychology, O xford.(R eceive d N ovem be r 2?, 1938.)

INTRODUCTION.JU NG w rote his T ype P sychology in 1914, but does not seem to have developedany further its general thesis. The terms "introvert" and "extravert",

however, have slipped into everyday language, and experimental studies andquestionnaire inquiries on introversion-extraversion continue to be publishedin psychological journals. Ink-blots, am biguous figures, and even knee-jerksare called upon to measure this fundamental tendency of man ; inventoriesand self-rating sheets are employed widely to estimate it ; and yet it appearsto evade capture. There is as yet little or no evidence anywhere that acoherent tendency of the kind looked for does in fact exist.Underlying this previous work on introversion-extraversion, whetherexperimental or based on ratings, there has been the notion that a simpleone-to-one relation should be expected between the fundamental tendency onthe one hand, and any transient reaction of the person on the other. This, inmy opinion, is a vain expectancy. W e should see, rather, infinite complexityin the connections between introversion-extraversion and any personality. Inth e fo llo win g p ag es, th ere fo re, I pro po se th at the sim ple on e-to -o ne re latio nsh ipshould be replaced by a statistical one, which tries to take account of theh ig hly ramifie d w ay s in w hic h in tro ve rs io n-e xtra ve rs io n s ub se rv es a p ers on ality .This conception leads us to a simple representation of Jungian types, and doesjustice, I think, to Jung's m ode of thought.

To see his system of typology in perspective it is necessary to range furtherafield than into Jung's book, and in what follows I have drawn freely uponthe w ider body of psychology and psychotherapy.

LXXXV. 13This One

JG9S-FSF-GW5K


3/22

l86 METHODOLOGICAL CONSIDERATION OF JUNG'S TYPOLOGY, [Mar.,

TYPES AS UNIVERSALFUNCTIONS.The word " type " has as many meanings as almost any word in psychology.

One, purely classificatory, scarce needs mention. All women with blond hairhave been regarded as of a type on that account ; men with convex or withconcave facial profiles have also been regarded as types. Some people nibblethe loose crumbs on the table at dinner, and others brush them into a neatpile ; very soon we begin to talk about crumb-picking and crumb-brushingtypes. Typification of this kind is purely classificatory, and its value topsychology cannot be m uch greater than attaches to the classification of booksin a library into all those of folio, quarto and octavo sizes respectively. It isbecause Rorschach testing (Beck (1937)) involves so much of this kind ofelementary classification into types that some of us feel dubious about itspretensions. Typification of this kind is usually gratuitously invested withproperties that far transcend the m ere classification, which itself is harm lessenough. Very soon blond women are all regarded as potential Mae Wests ;crum b-nibblers are felt to be m ean and depressive, crum b-brushers profligateand manic ; and people who pay attention doggedly to the white spacessurrounding the Rorschach ink-blots, after the blots themselves have beenexhausted, are considered to have an inner capacity for strong will and determination (Beck (1937)). The detail that individuals have in common ispresumed to be indicative of more general tendencies, of the very essence oftheir personalities. It usually takes psychologists a few decades to marshalfacts which show how little there is in these ad hoc suppositions and wilfula dd itio ns to th e o rig in al c la ss ific atio ns .W hen psychiatrists classify their patients into reaction types (Hendersonand Gillespie (1927)), or subnorm al types of various kinds, the classification isbased on far more than one manifest detail that the cases of a type have incommon. Each type is a compilation of many traits and symptoms, and anyindividual of the type will have many of these, or sim ilar ones, in his make-up.Thus psychopathic inferiors (Norwood East (1936)) tend to be eccentric,stubborn, hostile, reckless, quarrelsom e, irritable and excitable ; they are givento depression, confusion, ideas of persecution and delusions ; and they arecranks, eccentrics, chronic drunkards and sexual perverts. But any particularpsychopathic inferior will not have all these qualities ; it is sufficient if someof them , in the characteristic way of the type, are dom inant in his personality.It seems to me that Jung's types are methodologically of the latter kind.Jordan's description of a typical extravert woman (see Jung (1923), page 195),which Jung amplifies, consists of a long list of traits, strung together withliterary skill ; but the description could represent a real woman. Otherwomen of this same type need not have all this long list of traits dominant intheir personalities, but they m ust have som e, or m any like them . The "typical"woman, indeed, is little more than a compilation of all the qualities that enter


4/22

I939-] BY w- STEPHENSON, M .A. 187into all the persons classified into a type ; each individual contributes onlysome of the qualities, and few have all the qualities in their personality. Thedescription of personalities that so distinguishes Jung's work is at first sightlittle m ore than a kind of literary com posite photography.Actually it is much more. Each reaction-type or personality-type isthought of as dependent upon a few inner and fundam ental tendencies, usuallyof a constitutional, innate, or relatively permanent nature. W hatever formthe m anifest behaviour of a psychopathic inferior m ay take, however transientmay be his particular reactions, something akin to emotional instability,uncontrolled and innate, underlies all. The overt behaviour m ay take differentforms, but the inner and more fundamental make-up of the individuals of sucha type is the same in all. Typology of this kind, and the supposed fundamentaltendencies that subserve a type, is exam ined m ethodologically in the follow ingsection.

BAS IC FORMULAT IONS (a ) MATHEMATICAL .In all typology of the kind just discussed the initial concern is w ith theovert behaviour or characteristics of individuals. The psychotherapist's

diagnosis begins with the observation of such raw data. For convenience 1propose to represent the data as follows :data = (a, b, c, d . . . n) . . . (i).

Thus in Jungian typology, a, b, c . . . could be the personality traits thatrequire looking at from the standpoint of a theory of introversion-extraversion ;or they could be scores obtained by a person in certain tests which are meantto measure introversion-extraversion. L ikew ise in Spearman psychology theoriginal data may be the marks obtained by a person in some intelligence tests.Quite generally (i) stands for data, to be considered for any reaction type, orp erso na lity ty pe .Each item of the data is merely a transient reaction of the individual ; butwe have the view that each is explained, or subserved by, or dependent upon, amore permanent underlying tendency that gives it part at least of its essentialcharacter. That is, we have to contemplate some kind of relationship betweenthe items of data on the one hand, and one or more underlying and possibly-fundamental tendencies on the other.S pearm an so ug ht to re presen t this rela tio nsh ip m ath em atically .If we consider the simple case of one presumed underlying tendency (Y),which subserves all the data, Spearman asks us to contemplate a connectionbetween it and Y by equations of the kind :

*=/ (Y)\b=f'(yy- (2).e=f(Y}}. etc.


5/22

l88 METHODOLOGICAL CONSIDERATION OF JUNG'S TYPOLOGY, [Mat.,Here the same Y is involved in each item of data, but the functions /, /', /*m ay be differen t.The quantities and qualities under consideration at (i) are relativelynon-inferential ; they are raw phenom ena, although often already very greatlytampered with, owing to the inevitable formation in our minds of concepts.(Traits are often so tampered with.) For the main part all the data are statem ents about a person that involve relatively little technical or scientific knowledge. W hen we say, for example, that M rs. Smith has palpitations, headache,stabbing pains over the heart, and anxieties ; that she sleeps badly and fearsthat she has cancer ; that she lacks self-confidence, and has nervous mannerisms ; then we are dealing with data much of which could be described byalmost anyone, including Mrs. Smith herself. Data of this kind are contemplated by (i) above.But when the psychiatrist diagnoses Mrs. Smith as a case of anxietyneurosis he is no longer speaking the language of data, but, however vaguely,the language of " constructs " (Brown, 1936). Mrs. Smith might say or guessthat she is such a case ; but the psychiatrist does not depend on hearsayor guessw ork ; the full im plications of his diagnosis, and of the term ' ' anxiety-neurosis ", can only be appreciated by psychiatrists trained alike.

Even more so, the functions (/) at (2), and the presumed fundamentaltendencies such as Y , are no longer data, but constructsm atters for technicaltreatment. They are inferential and assumptive in nature.To make a practical beginning, Spearman connected the data of (i) tofundamental tendencies by a simple linear mathematical relationship, which Irepresent as follow s : a = Y+ (3)Here one item (a) of the data is linearly related to one underlying tendency Y,the values n and t being constants. If this same tendency (Y) enters into allthe data in this same fashion, then for one person under consideration wecould write : o = Yt',c = 7 Y + ',

. . . etc.,where Y is constant in all the data, but where the mathematical constants and ,etc., m ay be different for each.It is not necessary here to go into all the details and consequences of thisparticular mathematical representation. It is sufficient to draw attention tothe general method and its attendant conceptions. The mathematics can bepursued to its logical end, and we can seek to determine how far the originaldata square with the mathematical theorems. If agreement is found, then some


6/22

I939-] BY W - STEPHENSON, M .A. 189scientific justification is found for the subsequent language of " constructs "by which we " explain " the original data.Thus, when Spearman concerned himself with character traits, he foundwith Webb (1919) that they did in fact satisfy the simple consequences offormulae such as (3). The common underlying tendency in this case wascalled " w " ; in its purely mathematical setting it was called z0-factor, butit w as conceived of as the fundam ental integrating tendency (Spearm an, 1930)in personality, the underlying basis of character. Essentially the same treat -m cnt underlies S pearm an's fam ous -factor.A person with high g tends toscore highly in all tests of intelligence ; the common underlying tendency isg-factor, and this is conceived of in the language of constructs as general(cortical) energy, or as eduction. Each of the well-known Spearman factors,as well as the same factors dressed up anew by later workers such as Thurstone(1935) and Kelley (1928), have this kind of methodology behind them. Ineach case a very sim ple functional relationship is contem plated that connectsan item of data (i) to underlying tendencies (2). So simple is the functionalconnection, indeed, that the two, data and fundamental tendency, can beregarded as tending to be directly proportional to one another.*

Now Jungian psychology contemplates data such as (i). And no less thanin Spearm an psychology it has postulated certain fundamental tendencies that"explain" or underlie the data. Jung, however, has nowhere formulatedthese tendencies mathematically, nor have any functional connections beenconsidered that connect the inner tendencies to the transient reactions. Evento suggest to him that his m ode of thought could be represented m athem aticallym ight stir up at m ost a sm ile of condescension.In any case in Jung's typology there appear to be five major tendencies,Introversion-extraversion (I-E), and Sensory (s), Thinking (T), Intuitive (u),and Feeling (F) tendencies. I have not the slightest doubt but that theseJungian tendencies could be reduced to mathematical formulations, and thata school of Jungian factorists could build up an edifice of scientific evidenceon these five foundation stones. Indeed one might expect connections somewhere between Feeling (F) and the Spearman factor of general emotionality(e, Burt (1938)) ; likewise between Sensory (s), Thinking (T), and Intuitive(u) tendencies and the Spearman g-factor.f M oreover, introversion-extra-version is a tendency for the individual to act one way or the other, amechanism that can be inserted or disconnected at will, and one might no lessexpect connections between Spearm an w-factor and Jungian I-E ; very possiblyextremes of either introversion or extraversion are related to inadequate w.That is, the underlying fundamental tendencies in the two psychologiescould well supplement each otherfor in truth both merely embrace the* Thi s n ee ds q ua li fic at io n, b ut i s s ub st an tia lly t ru e in p ra ct ic e.t B y this I do not m ean that g w ill em brace Jung's s, T a nd u com pletely ; it is certain thats pe ci al S pe arma n f ac to rs ( gro up f ac to rs ) c ou ld b e f ou nd t o c ov er th es e t hr ee J un gia n te nd en ci es ,an d tha t all w ou ld h ave at le ast so me g -co nte nt.


7/22

190 METHODOLOGICAL CONSIDERATION OF JUNG'S TYPOLOGY, [Mar.,classical differentiation of individual psychology into cognitive (g, s, T , u),feeling (e, F ) an d d irection al (w , I-E ) tend en cies.As I have said, however, Jung's tendencies have yet to be representedmathematically, and this in turn used as the basis of systematic experimentalwork. I doubt very much, however, whether Jung would contemplate the kindof functions involved in factor analysis, even supposing for a moment that hewas disposed to try to reduce his typology to mathematics. Things are notquite so simple, he would say, as a direct and simple proportional relationbetween any trait of a type, and the underlying tendencies subserving thepersonality. Som ething far m ore com plex is involved.

BASIC FORMULATIONS ( 6) STAT IST ICAL.In the above paragraphs I have purposely expressed the relationshipsunder consideration in terms of only one fundamental tendency, but of coursesim ilar mathematical formulations can be given for several underlying tendencies. The hypotheses, and mathematical theorems, are dealt with veryadequately in modern factor analysis. But other relationships besides mathematical ones may be used to relate the raw data of (i) to one or several underlying constructs. Jung, like most psycho-analysts, only contemplates adescriptive, historical, relationship ; the individual's reactions depend upon

his life, or even his racial, history. This mode of explanation is satisfactoryas far as it goes, but we should also try to explain each particular situationin its more immediate context. Even in the course of a psycho-analysis thechanges and processes at work in the individual here and now, his mode ofthinking, feeling and direction in the present state of his development, areclearly worth at least as much scientific regard as what has happened to him inthe past. That is, we have to consider the personality as it is now, and ourexplanations should be historical as well as historical (Lewin (1936)).Spearman " constructs ", e.g., general mental energy, are of this ahistoricalkind. We need some kind of representation of the relation between data andconstructs, of a kind that will take due account of the system and ideas oftypology outlined by Jung. The thesis I w ish to elaborate is that purely-statistical connections can be used, instead of m athematical ones, to relate theactive present personality to its underlying tendencies.Statistical m ethods are involved in the basic form ulations already describedin the previous section ; but these pertain to the persons as units of a statisticalpopulation, and not essentially to the dependency of an item of data upon anunderlying tendency or tendencies. Nor is it necessary to point to the view,held by Spearm an and Thom son alike (see Spearm an (1928)), that any constructsuch as g can be regarded as having a sub-structure of innumerable smallercomponents ; these components are additively related, and the concern is


8/22

I939-] BY w- STEPHENSON, M .A. igistill essentially mathematical rather than statistical. Statistics deal withfrequency distributions, and 1 propose to postulate a frequency distributionfor each person, for the whole population of his data (i), where each item ofdata deviates quantitatively from an average, ow ing to the joint effect of avery large number of causes, each of which is of small influence only. Thedata, that is, them selves form a frequence distribution; this describes the presentpersonality ; and the thesis is that the underlying fundam ental tendencies are suchhighly com plex causal agencies, that no sim ple m athem atical relationship canpossibly represent the w ay in w hich they ram ify into the personality data.That is, instead of the function (2), we have to contemplate a relationshipy = Kf> .... (4).where x is the quantity given to an item of data, and where K and


9/22

IQ2 METHODOLOGICAL CONSIDERATION OF JUNG'S TYPOLOGY, [Mar.,

Y et, if the form ulations given above have any point, there is surely a profounddifference between the modes of explanation of data in the two casesin theone case a one-to-one relationship is contemplated between a factor (e.g., g)and the overt trait or item of datum. This factor is usually identified w ith anunderlying tendency, that explains the traits (Prof. Burt, however, regardsfactors as purely statistical constructions, and is very careful about makingsuch identifications). In the other case the relationship between a trait andits subserving tendencies is considered to be so com plex that only a statisticalrepresentation of it could do it justice ! The one-to-one relationship and thestatistical are in this respect the very poles apart. Nor does this concernthe mode of "explanation" only, for as we shall see, the factor methods in thetwo cases are very different. The units are not the same in C-techniqueand r-technique. Factors in the one case merely represent unanalysed typesas defined later, w hereas in the other case the concern is w ith analytical fundam ental tendenciesor else w ith m ere m athem atical and statistical artifices.T he tw o, ty pes an d tend en cies, are v ery d ifferent th in gs as d ifferen t as F alstaffand his glandular m ake-up.

I LLUSTRAT IONSOF THESE BASIC FORMULATIONS .The general descriptions of personality to be found in Jung's (1923) book

can now be considered in terms of the above basic formulations and considerations.The typical introvert woman, to give a simple example of a descriptionquoted by Jung, . . .' . . . has quiet manners and a character not easy to read ; she isoccasionally critical, even sarcastic . . . bad temper is sometimesnoticeable, but she is neither fitful nor restless, nor capricious, norcensorious, nor is she a ' nagging ' woman. . . . She is sympathetic. . . their passions and emotions are so strong . . . They love toomuch, but they also hate too much "... and so on.

In Appendix I a list of some 500 traits is described which were selectedfrom descriptions of the above kind given on many pages of Jung's book.From this longer list I selected at random a lesser number, 176 in all, and theseare to constitute our data ((i) p. 187). If need be we could repeat the followingprocedures with any other sample of traits drawn from the 500, or any larger" population ".

If we wished to base our m ethodology on m athem atical lines (r-technique),so as to demonstrate a factor that could be called I-E , which in turn could beregarded as a construct concerned with introversion-extraversion as a fundamental tendency, we should proceed as follows : A sample drawn from apopulation of women (say 100 or 200 women of a particular age) would be


10/22

I939-] BY w- STEPHENSON, M .A.estimated for each personality quality in turn, e.g., for manners, criticalm indedness, tendency to sarcasm , restlessness, strength of emotion . . . etc.,etc. The estimation would have its own assumptions ; in particular it wouldbe based on postulates about controlled or averaged situationsi.e., w e m ightask, what would A's manners be like if all other influences militating againstthese were controlled ? Or we might estimate on the basis of an average ofher displays of good manners. Each quality, such as " manners ", in effectbecomes a variable w ith respect to a population of women ; the differentqualities are correlated and factorized ; and if any factor or factors appear,that is if the traits correlate one with another, then evidence for the underlyingfunction I-E has been obtained.

In such a case, an extremely introvert person would tend to have all thetraits of the function in an exaggerated degree in her personality. She wouldnot only be quiet in her manners, but ascetic in them ; her character would beinscrutable ; she would be given to bouts of acerbatic criticism and to bittershafts of sarcasm ; she would be sympathetic to the point of saintliness, yether passions would be as strong as a raging fire . . . and so on, each trait inexaggerated degree. On some views* this caricature of a woman would becalled the " typical " introvert !Apart from the difficulty of contemplating any such person, an account ofthis kind does not fit w ith the conception that not all persons of a type needhave all the qualities of that type dominant in their own personalities (seep. 186). Two women might both be introverts, but the one might havetraits a, c, e, f , g, k . . ., etc. clearly in her make-up, whereas another mighthave a, b, e, f , i, l . . ., etc. It is sufficient that each has some traits of thetype clearly marked in her personality, but not necessarily that each has allin h er m ake -u p.There is a further difficulty that no one could sensibly deal with 176 traits,each estimated separately for a large sample of women ; and still more, whenattempts have been made hitherto along such lines no factors emerge ot anyco heren cy o r co gen cy .Let us see, then, what the other approach brings in its wake. Everypsychiatrist knows what it is to write up a description of the personality orother features of his cases ; in his time he " writes up " many patients in case

histories. -technique does much the same thing, only in a more schematicway. Instead of a literary description, such as Jung or any psychiatristwould give about a person Pj, we describe the person in terms of all n traits inthe form of a statistical distribution of them. That is, we give each of the ntraits a mark or score on some basis, to represent its relative dominance ordescriptive significance in the personality P!. But we postulate that thesemarks will tend to be normally distributed about the common core of the

* For exam ple, the view that regards " typical " persons as the extrem es of a norm al distribution.


11/22

194 METHODOLOGICAL CONSIDERATION OF JUNG'S TYPOLOGY, [Mar.,personality P,how far this is justified is a matter depending on fact, andpractical exigencies. That is, most of the traits, of a large sample, w ill notbe very critical in P, ; only a few traits will tend to be extrem ely characteristicof P, in the positive sense ; and only few will tend to be as extremely characteristic in the negative sense. The various traits w ill have different degrees of" significance "* of this kind ; but these will tend to be normally distributedabout the average for the personality P,. Thus " greed " might be so characteristic of P, that no other trait could so adequately describe him ; ' ' g enerosity ' 'will thereby tend to be as strongly characteristic negatively. The former traitm ight be given a score +3 (in standard terms) and the latter 3. The trait" gluttonous ", on the other hand, might play little part in his personalityhe is greedy for money, but no glutton for food ; he is neither a gormand noris he abstemious ; food does not matter, is not significant one way or the otherin his personality, and this trait m ight be given score o.Statistical description of this kind is done in the follow ing way :The estimator, it is presumed, knows a great deal about the personality tobe described, either as the result of analysing him , or because of long acquaintance with him . W ith this individual in mind, the estimator (in the study-to be reported below this was always myself) picks up the pack of cards whichlist the 176 traits (see Appendix I), and begins an evaluation of each trait forits relative significance in the personality under consideration. The cardsare first thoroughly shuffled and then sorted out into piles, just as they mightbe in a game of cards, except that now each card is deliberated upon and placedon the pile m ost suited to the particular trait's place in point of significance in thepersonality concerned. For practical purposes, because the use of integersfacilitate the calculations that have to be m ade later, the following frequencydistribution is made use of :

Score : 0123456789 ioFrequency : 2 4 io 20 32 40 32 20 io 4 2The two traits which, according to one's judgm ent, are m ost characteristic

of the person appraised are placed on pile io, that is, each is given io marks ;the next four most characteristic traits are given 9 marks each . . . and so on.The tw o traits m ost negatively characteristic of all in the personality are givenscore o each ; the next four are given score i each . . . a nd so on. In the end,then, each trait obtains a mark to stand for its significance in the personality ;and these marks are distributed as shown above. I repeat that one's object isto give a description of P,, as it is m ore or less intuitively appraised, the description having a statistical form . Traits so distributed oner a great deal of information about the person, and I am prepared to pit a statistical distribution

* For f ur th er i nf orma tio n a bo ut th is c on ce pt, s ee S te ph en so n ( 19 38 ).


12/22

IQ39-] BY w- STEPHENSON, M .A.against most literary ones, and to gain as much insight from the one as fromth e o th er.Unlike literary accounts, statistical ones can be audited. If several peopleare estimated in the above statistical fashion, they can be correlated one withanother, and the correlations subjected to factorial analysis. G ranted certainassumptions, the frequency distributions for a number of persons P, . . . Pvas variables can be correlated amongst each other for the traits a, b, c . . . nas a " population ".

One or two assumptions require mention. The form of the distribution isarbitrary, and can only be a first approximation to the actual facts. The factthat a distribution with the same mean and standard deviation is used forall persons alike seems odd until the concept of significance is understood.The trait x that is most significant in P,'s personality is just as important forthat personality as the trait y (that is most significant in P2) is in its ownperson ality (P 2). S ig nifican ce h as no th in g n ecessarily to d o w ith q uantificatio nof a trait in terms of individual differences w ith respect to a population ofpersons. I have given some attention to this concept in another paper(Stephenson, 1938), and here it is perhaps sufficient to say that this is a conceptespecially called for as a result of the increasing attention now given to thecorrelation of persons, instead of to the correlation of tests and traits, etc.The significance of a trait is what it is relative to other traits in a given personality, in the light of hypotheses and the like of a psychological nature, andnot of a statistical nature. W hen we estimate a trait like " honesty " normally,it is for individual differences, that is, how honest the person is relative toother persons. In the present case no such individual differences are anyw heredirectly involved ; nor are they necessarily involved even indirectly.

Other assumptions are as follows : Each trait is granted a certain tangibility as an item in a statistical population of traits. For practical purposesthis assum ption is warranted. A gain in m aking his appraisals the psychologistcan be granted an expertness in keeping with his technical knowledge ; but thesam e personalities can be appraised by different psychologists and the resultstreated for the familiar errors of estimation. There can always be a propercontrol of subjectivity introduced in data by one psychologist.For the purpose of this study I have described statistically som e 46 persons,all directly known to memy adult friends, psychoneurotic patients I havehad access to, and members of my own family. But I do not w ish to press thevalidity or objectivity of the appraisals I have made ; it is sufficient for thepresent purpose if the data are regarded as of illustrative value only to showhow it can be done, and how to treat the results. In point of fact I have chosenpersonalities purposely to bring out as clearly as possible the connectionsbetween my thesis and Jung's typology. The 46 men and women are inno sense a random sample of individuals ; nor is there any methodologicalreason why they should be, for in work of the present kind, where persons


13/22

METHODOLOGICAL CONSIDERATION OF JUNG'S TYPOLOGY, [Mar.,

are used as variables, they can be selected on other than purely samplinggrounds.

CORRELATIONSAND THEIR ANALYSIS .For my present purpose it is not necessary to calculate all the inter-correlations (i ,035 in all) for the 46 persons. E ighteen of the persons are extravertedin tendency ; 16 are introverted ; the others have little inter-correlationamongst themselves, or w ith other persons. They are of no clear type. Forconvenience I give only the correlations for the two groups of persons am ongstthemselves respectively. In Table I the 18 extraverts have been inter-correlated, with the results shown. All correlations are product-moment. In

Table II the sixteen introverts are intercorrelated. It is not necessary to giveanother table of correlations, which shows correlations between persons inTable I and those in Table II, and in which all the correlations tend to benegative.The correlations in the two tables I and II are shown in rearranged form ,w hich brings out the essential sim ilarities betw een particular groups of correlations. Grouping of this kind is readily made empirically, or can be made asthe result of previous factor analysis. Table I shows four groups, for persons1-5, 6-IO , 11-14 an


14/22

I939-] BY w- STEPHENSON, M .A.DEFINITION OF "TYPE" .

A technical definition is required, however, of the word " type ", and Ipropose to regard any number of persons P, . . . PiV as persons of one type ifth eir inter-co rrela tio ns sa tisfy the th eo rem o f tw o fa cto rs, on e fac to r b eing comm onto each trait and the other specific to each. That is, the well-known theorem oftwo factors, upon which Spearman has founded his theory of two factors incognitive abilities, is here turned to use in typology. In the theory of twofactors the scores supplied by any person in a number of tests are held to bedivisible into two parts, one general to all the scores, and the others specificto each. By specific is meant " uncorrelated ", i.e., of the nature of randomerrors. In the present theory of typology likewise, it is maintained that inthe case of persons of one type the scores for significance given to any trait intheir personality descriptions are divisible into two parts, one common (/) toeach trait, and the other specific (e) in each (i.e., a random error). That is, allthe postulates and consequences of the Spearman theorem of two factors areapplied to the present case of a population of traits, for personalities as variables.T hus, if a num ber of persons, correlated am ongst them selves for a populationof traits which describe them, satisfy the tetrad-criterion (Spearman, 1927),or a Spearman factor analysis, then the necessary evidence has been obtainedfor maintaining that the traits which have been scored for significance areeach divisible into two parts, one common (/) and the others specific (e) to eachtrait ; the persons so correlating will by definition be called persons of a type.Consider, for instance the persons in group A. These persons satisfy thetheorem of two factors amongst themselves t can be demonstrated either byco nstru ctin g th e fam iliar tetrad -d ifferen ces, or by co nd uctin g a straig htfo rw ardfactor analysis based on the assumption that the correlations do in fact satisfysampling conditions for zero tetrad-differences. In the latter case the /-saturation is calculated for each person in terms of the other persons of the same(for the moment, presumed) type, using the Spearman (1927) formula 21 ;viz., for person A, _ (A zA" i TAI~ \T-2A I

First residuals are then calculated ; thus for the two persons i and 2,ri2.i = ri2 ri/-ru/ ..... (6).

If these residuals for all the persons of the (still presumed) type are insignificant as judged by the probable error of the differences involved, then it ispermissible to regard the type as " proved ", and no longer as merely presumed.If the residual differences (6) are not insignificant, then the presumed type isnot " proven ", and other hypotheses w ill be necessary. I place the word" proven " in inverted commas to remind the reader that the evidence is


15/22

198 METHODOLOGICAL CONSIDERATION OF JUNG'S TYPOLOGY, [Mar.

35 S 1

" *sIH.H-^~*j0t i0s-&*5iy'!

pSDU-*Q**V . [T]

S0SP- 5HJ1

St00 ri:''H,ci

HQC4"t*1-1

ino >OfiO 'H-r>-'- o ' ""lO^'NCMON'N'S'N'sI'v|i^fO'^)t*^.oo

foCîH oo O r*-^.1-

m ro rooO

00mDO

O 'S NT^m o o Mf*"JN IO M"^ 00 Ô ^1-

t^ oo N i ooÔ io..NI

....IO 1 ....4 Jv '3 uB 0 jg 8o VE u0 C' .C S1 1V ft

^ in * '- 3^0 - 2; ^JT 'S ^


16/22

I939-] BY W. STEPHENSON, M.A. 199

6ersonalityraits.TypeHThinking).r-iS

.0-S'A""*3ft-

-*\

g._H'A

Eveg,a1 "11f.^HH^NHHOOOM^CI-AE3Q

N rO A t>. MO(rOO-|t^-i-i-Ô^vSr^^ÔOfON

C -^-ootôo.2rO

-S -O PO ^-Sm " 2 eo O fi"3l/v

l-( PO GVVf

2, *." S i iri-PO1^ô^::;;::::::::"1

U "H N IO O f^ OO i^1 O *^ (Sfî


17/22

200 M ETHODOLOGICAL CONSIDERATION OF JU NG'S TYPOLOGY, [M ar.,permissive only ; the same absence of significant residuals is consistent withm an y o th er factor h yp otheses.The four groups in each of Tables I and II are types according to the abovedefinition. The /-factor saturations for each group, calculated from (5) above,are shown in Table III. These give insignificant residuals ; for example, fromTable I, r12 is 0-703 ; Table III gives the /-saturations of persons i and 2 as0-883 and 0-778 respectively. Thus,

12.( 12 2t= 0703 (0-883 X 0778)

The P.E . of this residual is 0'026, so that it is not a significant difference.This applies also to all residual intercorrelations between persons of one type.

TABLE III.Show ing Factor-saturations (Oblique) for Types A to H . Personsof the one type only are used in calculating the saturations of persons oft ha t t ype .

Persons.It3456789IOii1213H15ib1718Extravertive.Factor-saturation.

Perenne(Obliqueactors.)TypeA. ' Type.883778775745714899853854Type. ; Type.3132333435363738757

39649....:

4765:1785427643770 , 44938]45822!4675973.Introvertive.Factor-saturation.

(Obliqueactors.)TypeF..844775Type F. I Type G . TypeI732

;679804760697709765

6 95 58 35 33 88 17 35 67 15 07

(T he decim al point is om itted in each saturation coefficient. T he saturationfor person i is 0-883.)


18/22

IQ39-] BY w- STEPHENSON, M .A. 201Attention must be drawn to two corollaries. One is that types as defined

above may correlate one with another. It can only be a matter of fact whethera " pure " type exists, i.e., " pure " in the sense that it has no correlation withother types. Persons in group A are of one type, and group B are another, butthe persons of the one type correlate positively with those of the other. Theother corollary is that the approximation of other persons (Pa . . . P)oany type (X) other than their own can be calculated in terms of persons ofthe type (X). This is done by adding the persons (Pa . . . Pn), one at a time,to the correlation matrix for the persons of type X , and thereafter calculatingthe ^-saturation of each.

Thus, by inspection, or by use of prior analysis into additive factors, thecorrelations can be grouped into clusters, and in so far as each cluster satisfiesthe theorem of two factors it may be held to constitute a type. A minimumof four persons is sufficient to define a type ; any significant residuals confinedto only one pair amongst a considerable number of persons would be disregarded in any first mapping out of the persons into types, although anysuch " specific " residuals may point to another type, embracing a smalln umbe r o f in div id ua ls .

REPRESENTATIONALAND FRACTIONALFACTORS.In rendering a factor account of Tables such as I and II it is essential todistinguish betw een tw o different kinds of factor analysis. O ne, just illustrated,merely seeks to sort out existing types and to represent them by factors(usually oblique, i.e., correlated with one another) ; the other seeks to analysethe relations between such types. Analysis of this latter kind is in terms ofo rth ogo nal facto rs, an d h itherto facto r-an aly sis h as atte nd ed alm ost ex clu siv ely

to this slicing up of variables into fractional factors which are additive anduncorrelated with each other. Thus Prof. Burt, contemplating Table I, wouldagree that the groups A to D might be regarded " empirically " as four types(Burt, 1938) but he would add that a more " rational " procedure wouldbe to subject the table to fractional analysis, and to identify the orthogonal factors with " types ". Such " types ", however, are methodologicallyof the same nature as underlying tendencies, and they are not overt, lived,p ers on ality ty pe s.A typical fractional analysis into additive factors gives results of the kindshown in Table IV , where Table I is subjected to analysis. The factors areorthogonal, and substantially sim ilar results would issue from a Burt (1938),Spearman (1927), Thurstone (1935), or other factor analysis. Analysis of thisfractional kind might offer information about a type, but in no way representsany such type itself. Types as I have defined them are existing, unanalysed,personalitiesthey may be compared to a Mae West, w ith just such hair,features, figure, attitudes, etc., of a total personality. But analysis into

LXXXV. 14


19/22

20 2 M ETHODOLOGICAL CONSIDERATION OF JUNG'S TYPOLOGY , [M ar.,fractional factors is like typification in terms of blond hair, or eyes, or otherp arts of su ch w hole perso nalities.T ABLE IV . S howin g F ac to r-satu ratio ns (O rtho go nal) fo r a F ractio nal A naly sisof Table I. This analysis m ight be m ade in m any ways, by Buri, H olzinger,T hu rsto ne, o r o th er m eth ods, w ith very sim ila r resu lts.

Extravertive.

Persons.123456789

IOii1213M15io1718Factor-saturation.

(Orthogonalactors.)FactorI. Factor II. Factor III. FactorV.247

84527073206779426369921268l405 371961360 36 668399 273 720^414 292 568403 32821'502 114 21834576 075 24576564 083 in1053! 59 44 i5938822!759f730 1

CONCLUSION.Typology of the above kind depends upon the size of the population oftraits employed. Only the two arch-types, introvert-extravert, would be

distinguishable if a sample of 20 traits is used ; but for the 176 traits hereconsidered a finer differentiation is possible, into sensory, intuitive, feelingand thinking sub-types. A still larger population of traits would give riseto a still finer definition of sub-types. Jung says somewhere that there arepossibly thousands of types. The above system of factor representationcertainly could envisage as many, but they would branch from the arch-typeslike members in a genealogical table. The larger the population of traits, thelonger and more ramified can be the family tree.The traits in -technique are purely statistically ordered with respect to apersonality, but the definition of a type involves the use of a mathematical


20/22

I939-] BY w- STEPHENSON, M .A . 203theorem. This theorem, however, is used only for demonstrating a type, andnot at all for " explaining " it. It seems to me to be important to distinguishbetween mere demonstration or isolation of types, and their explanation.Factorists in the past have gone ahead too quickly to dissect and explain theirspecimens long before they have caught them. -technique calls for thespecimens to be caught first. It can then perhaps be seen more clearly thanhitherto what the analysis into fractional factors amounts to, and how littlesuch analysis may help one to explain anything. For it is open to questionw hether fractional analysis into a few m ajor factors, each presum ed to indicatean inner fundamental tendency, offers any help in explaining a personality. Inthe above paper I have gone to the other extreme, admittedly, to oner insteada statistical explanation of it. But this, based as it is on the complexity ofhuman relations, is not inadequate ; it provides a basis for quantification oftraits independently of individual differences ; and the data with which itoperates are those of tangible whole personalities, and not of analytical partsof them. It is submitted that this statistical description does reasonablejustice to Jungian typology as far as methodological matters are concerned.It can contemplate innumerable types ; these types are " lim ited ", and usuallycorrelated ; but if the facts are of this nature, then why should we vainlysearch for " pure ", uncorrelated types ? It does not seek to explain a type interm s o f a sim ple on e-to -o ne relatio nsh ip b etw ee n trait a nd u nd erlying te nd en cy ,but sees a vastly com plex connection betw een them .Previous w ork on introversion-extraversion has had a different standpoint.Ever since Stern and Thorndike (see Burt, 1938) refused to countenance"lim ited" types, psychologists have searched for underlying tendenciesbearing a simple one-to-one relation to observed reactions. This has led themfar astray. For example, empirical regard showed them that certain typesexisted ; and no one ever supposed for a moment that any such was " pure ".But instead of distinguishing these types, bringing them into focus for furtherregard, psychologists have searched for their explanation by looking forfundam ental tendencies. So indeed have they conducted m any post-m ortem so n im ag in ary b od ies.In conclusion two details deserve mention. On methodological groundsit is im portant to realize that out of a thousand persons only five m ight conformto a type ; all the others m ight be uncorrelated. Possibilities of this particularkind cannot be contemplated at all in ^-technique, where the general resultfor all persons governs the factor issues and the psychological explanations.Y et it m ight be just these five persons w ho are m ost significant for psychology.The other detail relates to the matter mentioned on p. 186 ; two personsmight be a type, but have different traits dominant in their personalities.The above technique takes due account of this ; for two persons may correlateonly 0'2O, and yet be of one and the same type. The small correlation isproof that the traits are of very divergent significance in the tw o personalities.


21/22

204 M ETHODOLOGICAL CONSIDERATION OF JUNG'S TYPOLOGY, [M ar.,

APPENDIX.From Jung's book (1923) I m ade a list of as m any traits and personalitydescriptions as I could without m aking a very exhaustive search for them .From this list, som e 500 in all, a random selection of 176 w as draw n, and thissample has been used in the example given in the above paper. Each of the176 traits or descriptions was written on a card (playing-card size) andnum bered. C ard No. 16, for exam ple, had on it the following :" S atisfa ctio n o f sim ple e leme nta ry b od y re qu ireme nts (c le an lin ess ,e tc .) n ot g iv en d ue con sid era tio n."No. 83 reads :" h ig hly c ommun ic ativ e ",No. 85 reads :" fits in to e xistin g c on ditio ns w ith re la tiv e e ase ".

A ll 1 76 w ere of this natu re, som e brief, an d som e am plified considerably.A ccording to the context into w hich Jung had placed these traits, the sam pleo f 1 76 c on siste d o f th e fo llowin g :N um ber o f tra itsPresumed type : of this class.1. General extravertive (E.) ..... 252. Unconscious ...... 93. T hinking extravertive (T.E.) ..... 274. Feeling (F.E.) 155. Sensory ,, (S.E.) ..... 136. Intuitive (U .E.) 127. General i ntrovertive (i.) ..... 68. Thinking (T.I.) 249. Feeling (F.I.) 2310. Sensory ,, (s.i.) ..... 811. Intuitive (u.i.) .... 14

Total 176Of the 176 traits, loi are thus considered by Jung to be extravertivere ac tio ns , a nd 7 5 a re in tro ve rtiv e.

REFERENCES.BECK, S.Introduction to the Rorschach M ethod, 1937. (Research M onograph No. i,Amer ic an Or th op sy ch ia tr ic A s so ci at io n. )BROWN , J. F . Ps yc ho lo gy a nd S oc ia l O rd er , 1 93 6. (M cG raw-H il l, L on do n.)B URT, C . I .. Brit. J ou rn . M ed . P sy ch nl., 1 93 8, x vii, p . 1 58 .Idem ." F actor Analysis by Sub-m atrices," Journ. Psychol., 19380, vi, p. 339.EAST, NORWOOD . Med ic alA sp ec ts o f C rim e, 1 93 6. (O xf or d Med ic al P ub lic at io ns .)


22/22

I939-] BY w- STEPHENSON, M.A. 205H EN DER SO N and ( IL LE SP IE. AText-book of P sychiatry, 192" . (O xford M edical P ublications.)HOLZIXCER, K. J., et al.-A Student M anual of 1 'actnr Analysis, 1937. (University ofChicago.)JUN G, C. G.Type Psychology, 1923. (Kegan Paul, London.)KELLEY , T. I,.C rossroads in the M ind of M an, 1928. (Stanford U niversity.)L EW IS , K . P rin cip le s o f T op olo gic al P sy ch olo gy , 1 93 6. (M cG raw-Hill, L on do n.)SPEA RMAN , C ." The Sub-structure of the M ind," Brit. Journ. Psychol., 1928, xviii, p. 249. d em . Th eA bilities o f M an , 1 92 7. (M acM illa n, Lo nd on .)Id em . Psyc ho lo gies o f 1 93 0, 1 93 0. (C la rk U niver sity P res s, U .S .A .)S TE PHEN SON. W . Psy ch om etrik a, 1 93 6. i, p . 1 95 .Idem." Two Contributions to the Theory of .Mental Testing," 1938. (To appear in Brit.

J nu in . P sy ch .)T HU RS TO NE , L . L .T he V ectors of the M ind, 193 5. (U niversity C hicago P ress.)W EB B, K." Character and Intelligence," B rit. Journ. Psych., 1919, M ono. Suppl., i, Xo. 3.

Stephenson Jung's Typology

Documents

Transcript of Stephenson Jung's Typology