on predicting some of the people some of the...

Psychological Review1974, Vol. 81, No. 6, pp. 506-520

ON PREDICTING SOME OF THE PEOPLESOME OF THE TIME:

THE SEARCH FOR CROSS-SITUATIONAL CONSISTENCIESIN BEHAVIOR

DARYL J. BEM l AND ANDREA ALLEN

Stanford University

The historically recurring controversy over the existence of cross-situationalconsistencies in behavior is sustained by the discrepancy between our intui-tions, which affirm their existence, and the research literature, which doesnot. It is argued that the nomothetic assumptions of the traditional re-search paradigm are incorrect and that by adopting some of the idiographicassumptions employed by our intuitions, higher cross-situational correlationcoefficients can be obtained. A study is reported which shows that it ispossible to identify on a priori grounds those individuals who will be cross-situationally consistent and those who will not, and it is concluded that notonly must personality assessment attend to situations—as has been recentlyurged—but to persons as well.

Our persistent belief in personality traits,the stubborn assumption that there are per-vasive cross-situational consistencies in anindividual's behavior, is, quite literally, oneof our most ancient convictions:

Penuriousness is economy carried beyond all mea-sure. A Penurious Man is one who goes to adebtor to ask for his half-obol interest before theend of the month. At a dinner where expensesare shared, he counts the number of cups eachperson drinks, and he makes a smaller libation toArtemis than anyone. . . . If his wife drops acopper, he moves furniture, beds, chests and huntsin the curtains. . . . [P]enurious men have haircut short and do not put on their shoes until mid-day; and when they take their cloak to the fullerthey urge him to use plenty of earth so that itwill not be spotted so soon [Theophrastus (372-287 B.C.), quoted in Allport, 1937, p. 57].

If this bit of historical personality theoriz-ing has a contemporary ring, it is, in part,because the same underlying assumption ofcross-situational consistency is still with us.It is most explicit in trait and type theoriesof personality, but some variant of it can bediscerned in nearly all contemporary for-mulations. Even psychodynamic theories,which are uniquely competent in dealingwith phenotypic inconsistencies in behavior,do so precisely by postulating an underlying

1 Requests for reprints should be sent to DarylJ. Bern, Department of Psychology, Stanford Uni-versity, Stanford, California 94305.

genotypic consistency in the personalitywhich rationalizes the apparent contradic-tions. Our intuitions are even more per-suaded. For them the assumption ofcross-situational consistency is virtually syn-onymous with the concept of personalityitself. There are few other beliefs abouthuman behavior which are as compellinglyself-evident.

But like many other assumptions, the con-sistency assumption did not fare well duringthe depression years, when three separatestudies with very similar methodologies be-gan to raise serious doubts about its validity.The earliest and best known challenge issuedfrom the extensive multivolume Studies inthe Nature of Character by Hartshorne andMay (1928, 1929; Hartshorne, May, &Shuttleworth, 1930), who found so littleconsistency among diverse measures of"moral character" in a group of children thatthey concluded that such traits as deception,helpfulness, cooperativeness, persistence, andself-control are "groups of specific habitsrather than general traits." Foreshadowingfindings which emerged from hundreds oflater studies on scores of personality traits,Hartshorne and May reported that theaverage intercorrelation of the 23 tests usedto construct a "total character score" was amodest +.30.

506

CROSS-SITUATIONAL CONSISTENCIES 507

During the same years as the Hartshorne-May inquiry, a less well known but equallytroublesome study on extroversion—introver-sion was published by Theodore Newcomb(1929), who explicitly set out to test theconsistency assumption. He kept daily be-havioral records on 51 boys at a summercamp for several weeks, recording behaviorsin 30 different situations. The behaviorswere conceptually organized into 10 separatetraits (e.g., volubility versus taciturnity,ascendancy versus submission, etc.) which inturn collectively defined the two personalitytypes of extrovert and introvert.

At the level of specific behaviors, New-comb found little or no consistency from onesituation to another. At the level of traitconsistency, the intercorrelations among be-haviors composing a given trait averaged only.14, almost identical to the figure obtainedfrom a randomly selected set of behaviors.And finally, there was only a slight tendencyfor traits to be related to one another asexpected by their extrovert-introvert classi-fication.

The third study is in some ways the mostdamaging of all since it investigated punctu-ality, a trait one would expect to be muchmore homogeneous than "moral character"or extroversion-introversion. In this study,Dudycha (1936) made 15,360 observationson over 300 college students, recording eachstudent's time of arrival at 8:00 a.m. classes,commons, appointments, extracurricular acti-vities, vesper services, and entertainments.The mean cross-situational correlationturned out to be +.19, with the highest cor-relation—between punctuality at entertain-ments and at commons—reaching .44.

These three studies are virtually uniquein that the investigators actually observed be-havior in vivo across several situations, adevotion to duty practically unknown intoday's literature on the same topic. Mean-while, the pencil-paper attempts to predictbehavior from a trait conception of person-ality were faring no better, leading Lehmannand Witty (1934) to conclude a review ofthe literature by saying that "over and over,a battery of tests designed to measure traitssuch as persistence, or aggressiveness, orhonesty, yields results so unreliable and un-

dependable . . . that one is led to questionthe actual existence of the general traits[p. 490]."

At the same time that the belief in cross-situational consistency was suffering theseempirical blows, stimulus-response be-haviorism was providing the theoretical argu-ment for the counter belief in the situationalspecificity of behavior. And with psycholo-gists like Gordon Allport (1937) and RossStagner (1937) willing to defend modifiedtrait conceptions of personality against thisonslaught, the controversy was a lively onefor nearly a decade before receding into thebackground just prior to World War II(Sanford, 1970).

All of this leads one to appreciate-thesense of deja vu that must currently beaffecting psychology's elder statesmen nowthat the "consistency problem" has suddenlybeen rediscovered (e.g., Alker, 1972; All-port, 1966; Argyle & Little, 1972; Averill,1973; D. Bern, 1972; Bowers, 1973;Campus, 1974; Endler, 1973a, 1973b; Endler& Hunt, 1968; Harre & Secord, 1972; Mis-chel, 1968, 1969, 1973a, 1973b; Moos, 1969;Peterson, 1968; Stagner, 1973; Vale & Vale,1969; Vernon, 1964; Wachtel, 1973; Wal-lach & Leggett, 1972).

The major figure in this current roundof debate appears to be Walter -Mischel(1968), who, after reviewing both past andcurrent research, concludes that the pre-dictive utility of a trait-based approach topersonality still remains undemonstrated andthat situational specificity of behavior- ap-pears to be the rule rather than the excep-tion. Although other contemporary authorshave drawn similar conclusions (e.g., Peter-son, 1968; Vernon, 1964), it is Mischel whohas provoked the most controversy by argu-ing that the commonly observed + .30 ceilingon cross-situational correlation coefficientsprobably reflects true behavioral variabilityrather than imperfect methodology. Sincethis constitutes a fundamental conceptualchallenge, the controversy is once againfilling journal pages after a 30-year intermis-sion.

And the stubborn dilemma which sustainsthis conflict and accounts for its durabilitystill remains unresolved: The sharp dis-

508 DARYL J. BEM AND ANDREA ALLEN

crepancy between our intuitions, which tellus that individuals do in fact display per-vasive cross-situational consistencies in theirbehavior, and the vast empirical literature,which tells us that they do not. Intuitionsor research? One of them must be wrong.

ERRORS OF INTUITION AND THENOMOTHETIC FALLACY OF THE

RESEARCH PARADIGM

There are many persuasive reasons forbelieving that it is our intuitions which arein error (Jones & Nisbett, 1971; Mischel,1968). First, for example, we hold "im-plicit personality theories," preconceived no-tions of what traits and behaviors go withwhat other traits and behaviors. (SeeSchneider, 1973, for a recent review.) Thisleads us not only to generalize beyond ourobservations and fill in the missing datawith "consistent" data of our own manu-facture, (e.g., Passini & Norman, 1966) butalso to "see" positive correlations which are,in fact, not there (e.g., Chapman & Chap-man, 1969; Newcomb, 1929). Moreover, weare biased toward "primacy" effects (e.g.,Jones & Goethals, 1971); once we haveformed an initial impression of a person, wewill perceive inconsistent information abouthim as more consistent than it deserves tobe, assimilating it to our initial judgment.

Second, recent research in attributiontheory (e.g., Jones & Davis, 1965; Jones &Harris, 1967; Jones & Nisbett, 1971; Kelley,1967) has demonstrated that we tend tooverestimate the degree to which behavioris caused by traits of the individual andunderestimate the degree to which it iscaused by external factors. We are there-fore willing to generalize about his behavior,extrapolating it to other, unobserved settingsin which the situational forces might bequite different.

Third, the set of situations in which weobserve most individuals is probably morelimited than we realize, both in extent andrepresentativeness. For example, our ownpresence can frequently evoke a consistentmode of responding in another person (e.g.,Kelley & Stahelski, 1970). Accordingly,we are systematically excluded from observ-ing a whole host of situations across which

our acquaintances' behaviors are likely to bemore variable than they are across the situa-tions in which we do observe them.

Fourth, we probably misconstrue or over-generalize some of the consistencies whichare present. For example, Mischel's (1968)review reveals that the evidence for temporalconsistency in behavior is often quite re-spectable; an individual's behavior is oftenconsistent from one time to another if thesituations are similar. Our intuitions maywell go from this demonstrable temporalconsistency to an unwarranted cross-situa-tional consistency. Moreover, some featuresof behavior do show cross-situational con-sistency, such as intellectual ability, cogni-tive styles, expressive behaviors, and, ofcourse, simple physical appearance. To theextent that these behaviors or cues serve toanchor our inferences about other aspectsof behavior—via our implicit personalitytheories—we will again overgeneralize thedegree of cross-situational consistency actu-ally present.

Finally, our language entices us to thinkof human behavior in trait terms. As All-port and Odbert (1936) reported, there areabout 18,000 trait or traitlike terms in ourlanguage, nearly five percent of the entirelexicon. In contrast, we have an impov-erished and awkward vocabulary for labelingsituations.

These, then, are a sample of reasons forthinking that, in the matter of cross-situa-tional consistency, intuitions are wrong, andthe research is right. We, however, do notbelieve it. Despite the compelling impact ofthese arguments, we still believe that intui-tions capture reality more faithfully thandoes the research. In particular, we believethat there is a basic error in drawing infer-ences about cross-situational consistencyfrom the traditional research literature inpersonality, an error which was identifiednearly 40 years ago by Gordon Allport(1937). The fallacy resides in the fact thatthis entire research tradition is predicatedupon nomothetic rather than idiographic as-sumptions about the nature of individualdifferences. Thus nearly all of the researchis based on some variant of the nomotheticassumption that a particular trait dimension


or set of trait dimensions is universally ap-plicable to all persons and that individualdifferences are to be identified with differentlocations on those dimensions. For example,the Hartshorne-May study (1928) assumedthat an honesty-dishonesty dimension couldbe used to characterize all of the children inthe sample and that the differences amongthe children could be specified in terms oftheir degree of honesty. A more elaborateversion of the same nomothetic assumptioncan be found in factor-analytic formulationswhich assume that there is a universal factorstructure of personality and that individualdifferences are to be specified by differentpoints in the factor w-space.

In contrast, Airport's idiographic viewemphasized that individuals differ not onlyin the ways in which traits are related to oneanother in each person but that they differalso in terms of which traits are even rele-vant. Thus in commenting upon the factthat Hartshorne and May found lying andcheating to be essentially uncorrelated (r =.13), Allport noted that one child may liebecause he is afraid of hurting the feelingsof the teacher, whereas another may stealpennies in order to buy social acceptancefrom his peers. For neither of these twochildren do the behaviors of lying and cheat-ing constitute items on a scale called"honesty," a concept which exists in thehead of the investigator, not in the behaviorof the children. Accordingly, the low cor-relations "prove only that children are notconsistent in the same way, not that theyare inconsistent with themselves (1937, p.250)." To put the same objection in slightlydifferent terms, the research will yield theconclusion that a sample of individuals isinconsistent to the degree that their be-haviors do not sort into the equivalence classwhich the investigator imposes by his choiceof behaviors and situations to sample.

But there is more. Even if an entiresample of individuals does share the investi-gator's partitioning of behaviors into thesame equivalence class, there is a still morestringent requirement of consistency imposedby the traditional research paradigm: scal-

ability.2 That is, the sample of individualsmust all rank order the "difficulty levels" ofthe behaviors in the same way.

Consider, for example, the "friendliness"of the second author. She is very friendlyto undergraduates in her office, moderatelyoutgoing in a small seminar, and somewhatreserved before a large class. If we can agreethat all of these behaviors belong in a com-mon equivalence class labeled "friendliness,"she will be judged to be moderately friendlyon this trait dimension. She "passes" the"easy" item, has some difficulty with a"harder" item, and "flunks" the "most diffi-cult" item. Note that we do not judge herto be inconsistent any more than we' judgea student to be inconsistent when he solvesan addition problem but fails a calculus item.We do not do so because their behavior con-forms to our a priori ordering of the itemsin terms of their difficulty levels: their be-havior "scales" in the Guttman sense (Scott,1968; Stouffer, Guttman, Suchman, Lazars-feld, Star, & Clausen, 1950).

But now consider the "friendliness" of thefirst author. He, too, passes one item andflunks one item. He is rather formal withundergraduates who appear in his office,moderately outgoing in a small seminar, andopen, personable, and friendly before a seaof 300 faces in introductory psychology. Butsomehow his behavior does not seem describ-able in terms of the same underlying dimen-sion. He appears not "moderately friendly,"but "blatantly inconsistent." And this isbecause his behaviors do not conform to' thea priori Guttman scale which we have im-plicitly imposed on this equivalence class ofbehaviors. He passes hard items but flunkseasy items.

Now reconsider the traditional researchstudy in which a sample of individuals isassessed on some trait across two or moresituations. To the extent that individualsin the sample scale the behaviors differentlyfrom one another—as the first and secondauthors do on "friendliness"—their rankingsrelative to one another will change from onesituation to another. The second author

2 We are indebted to Stanford colleague LeeRoss for bringing this point to our attention.


will rank first in friendliness in the officeencounter; the first author will rank first inthe large lecture hall. Under such circum-stances, the cross-situational correlation co-efficients will plummet toward zero. Onlyto the extent that all of the individuals inthe sample scale the behaviors in the sameway will the cross-situational correlations behigh.

In summary, then, the traditional trait-based research study will yield evidence ofcross-situational consistency only if the in-dividuals in the research sample agree withthe investigator's a priori claim that thesampled behaviors and situations belong ina common equivalence class and only if theindividuals agree among themselves on howto scale those behaviors and situations. Thefallacy to which Gordon Allport originallycalled attention thus becomes clear. Thetraditional verdict of inconsistency is in noway an inference about individuals; it is astatement about a disagreement between aninvestigator and a group of individuals and/or a disagreement among the individualswithin the group. This fallacy is a directconsequence of the traditional nomotheticassumptions about individual differences.(See a related line of argument by Baldwin,1946, and McClelland, 1951.)

In contrast to the empirical research, ourintuitions operate on idiographic rather thannomothetic assumptions. When we areasked to characterize a friend, we do notinvoke some a priori set of fixed dimensionswhich we apply to everyone. Rather, wepermit ourselves to select a small subset oftraits which strike us as pertinent and to dis-card as irrelevant the other 17,993 traitterms in the lexicon. Moreover, we try tocompose trait descriptions which conformto the individual's own partitioning of equi-valence classes. If John always does hisschool work early, is meticulous about hispersonal appearance, and is always punctual,it may well occur to us to describe him asconscientious. But if he is always con-scientious about his schoolwork, being negli-gent in these other areas, we may welldescribe him as a totally dedicated student,one who has time for little else. The im-

portant point is that we are not likely tocharacterize him as someone who is incon-sistently conscientious. That is, we do notfirst impose a trait term (e.g., conscientious)and then modify it by describing the in-stances which fail to fall into that equiva-lence class. Rather, we attempt first toorganize his behaviors into rational sets andonly then to label them.

We are, moreover, somewhat sensitive tothe scaling criterion. We will describe thesecond author as moderately friendly ratherthan inconsistent because we recognize theunderlying Guttman scale to which her be-havior conforms. But when we encounterthe first author, whose behavior does notscale according to the recognized "friendli-ness" dimension, we will typically try torepartition his behaviors before accepting averdict of inconsistency. Thus the firstauthor is, perhaps, a moderately aloof chapwho is a great stage performer. That is,we attempt to construct a new set of equiv-alence classes which better "capture" theindividual's personality. Note also that thisintuitive process automatically finessess theproblem of situational specificity by embrac-ing in a common equivalence class only thosebehaviors and situations which cohere forthe individual, thereby excluding a prioriany maverick behaviors and situations. Thetrait description is thus fractionated, ex-panded, contracted, and modified until a bestfit of greatest generality and parsimony isachieved. It is only when we fail to dis-cover a set of rationally-scaled equivalenceclasses which conform to the individual's be-havior that a judgment of "inconsistency" isfinally rendered. This is the essence of theidiographic approach to personality.

We are not here denying the well-docu-mented biases and illusions which plague ourintuitions, nor do we claim that the moreformalized idiographic procedures used byclinicians have a better track record in termsof predictive utility than nomothetic ones;they do not (Mischel, 1968). But in termsof the underlying logic and fidelity to reality,we believe that our intuitions are right; theresearch, wrong.


IDIOGRAPHIC ASSESSMENT ANDNOMOTHETIC SCIENCE

The problem with concluding that anidiographic approach represents the path totruth, however, has always been that one isnever sure what to do next. To the extentthat one accepts psychology's goal as theconstruction of general nomothetic principles,the idiographic approach appears a scientificdead end, a capitulation to the man-in-the-street view that a science of psychology isimpossible because "everybody is differentfrom everybody else." It is this pessimismwhich seems largely responsible for the factthat the field's respect and admiration forGordon Allport has never been translatedinto research programs based upon his con-ception of personality (cf. Sanford, 1970).Interestingly, the approach of behaviorismhas probably discouraged the study of per-sonality differences for much the same rea-son. To the extent that an individual's be-havioral repertoire faithfully reflects theidiosyncratic vagaries of his past reinforce-ment history, searching for some rationalnomothetic basis of personality organizationwould not appear to be a very hopeful enter-prise. As Mischel (1968) notes—not pes-simistically, however—the approach of socialbehavior theory does not

label the individual with generalized trait termsand stereotypes. . . . Behavioral assessment in-volves an exploration of the unique or idiographicaspects of the single case, perhaps to a greaterextent than any other approach. Social behaviortheory recognizes the individuality of each personand of each unique situation [p. 190].

But the impasse is not insurmountable.The use of idiographic assessment proce-dures did not appear to deter Freud fromformulating nomothetic principles of per-sonality organization. Similarly, albeit moremodestly, Mischel (1973b) has recently pro-posed a set of nomothetic principles withinthe idiographic assumptions of social be-havior theory. A third example is providedby George Kelly's (1955) psychology ofpersonal constructs and its associated idio-graphic assessment procedure, the Role Rep-ertory Test. In fact, it is Kelly's approachwhich best exemplifies the spirit behind the

present arguments. Thus Kelly permits theindividual to generate his own traitlike de-scriptors ("constructs") for characterizinghimself and his social world and to determinewhich behaviors and situations are to be em-braced by those descriptors, that is, to deter-mine what Kelly has termed the individual's"range of convenience" for the construct.Note that such an approach could reveal, forexample, that an individual who regardshimself as extremely conscientious might notconsider his casual attitude toward personalhygiene as pertinent to that trait. The factthat the investigator's concept or equivalenceclass of conscientiousness might include per-sonal hygiene within it is not relevant.

The basic point here is simply that thereis no inherent conflict between an idiographicapproach to assessment and a nomotheticscience of personality, whether one opts fora psychoanalytic orientation, a social learn-ing viewpoint, or a systematization of every-man's trait theory.

It should be clear, however, that idio-graphic assessment only permits one to pre-dict certain behaviors across certain situa-tions for certain people but not beyond that.Consequently, a conflict does arise if an in-vestigator refuses to relinquish the power todecide which behaviors of which people areto be studied in which situations; the logicof idiographic assessment requires that theindividual himself must be given this power,whereas the particular concerns of the in-vestigator may require these decisions to befixed parameters.

Consider, for example, the researcherwho wishes to study, say, need for achieve-ment in a particular setting in a particularpopulation. No matter how persuasive hefinds our arguments for the merits of idio-graphic assessment, he is simply not inter-ested in studying a different set of person-ality variables in each individual. But, onthe other hand, our arguments imply thatneed for achievement may not even be atrait dimension which usefully characterizesmany of the individuals in the sample. Ashis low validity coefficients will attest, thoseindividuals will contribute only noise to hisinvestigation. The dilemma is real. If ourarguments here are sound, one simply can-


not, in principle, ever do any better thanpredicting some of the people some of thetime. It is an idiographic fact of life.

Our advice to such an investigator, then,follows directly: Find those people. Separatethose individuals who are cross-situationallyconsistent on the trait dimension and throwthe others out, for by definition, only thebehavior of consistent individuals can bemeaningfully characterized by the investiga-tor's construct; only their behaviors can bepartitioned into the equivalence class underinvestigation. Perhaps a statistical meta-phor will make this proposal seem less illegi-timate : Unless an individual's variance ona particular trait dimension is small, it makesno sense to attach psychological significanceto his mean on that dimension.

We submit that even this token gesturetoward more idiographic assessment has itsrewards. First of all, one may obtain valu-able knowledge about the trait dimensionitself; it could be useful (as well as hum-bling) to discover why, which, and how manyindividuals fail to share the investigator'spartitioning of the world into his favoriteequivalence class. But perhaps even better,we believe that the rewards for this smallidiographic commitment can even be paid inthe sacred coin of the realm: bigger correla-tion coefficients! The following demonstra-tion illustrates the point.

A PRIORI ASSESSMENT OF CROSS-SlTUATIONAL CONSISTENCY 8

Our purpose in this study was to testwhether or not individuals can be divided onthe basis of self-report into those who are

8 We are grateful to many people for help inexecuting this study, particularly to Fred BartAstor who served as overall supervisor and re-search assistant and to Margaret Bond who trans-formed raw observations into meaningful data.For their multiple roles as trained observers, ex-perimental confederates, and general research as-sistants, we are grateful to John Backes, MargaretBond, Kathleen Chiappori, Tom Deremigio, GowenRoper, Jeremy Rosenblum, and Ann Scholey.Above all, however, we are grateful to our sub-jects, who were genuine partners in this researcheffort. In return for their cooperation, we at-tempted to keep them as fully informed of ourprocedures and rationale as the design wouldpermit.

cross-situationally consistent on a particulartrait and those who are not. Our hypothesisis straightforward: Individuals who identifythemselves as consistent on a particular traitdimension will in fact be more consistentcross-situationally than those who identifythemselves as highly variable. In popula-tion terms, the cross-situational correlationcoefficients of the self-identified low-vari-ability group should be significantly higherthan the coefficients of the high-variabilitygroup. In order to add a bit more persuasiveelegance to the study, we tested this hy-pothesis twice on the same population ofsubjects using two orthogonal personalitytraits, friendliness and conscientiousness.

MethodAs part of a questionnaire entitled the Cross-

Situation Behavior Survey (CSBS), all studentsin Stanford's introductory psychology course wereasked to assess themselves on several trait dimen-sions, including friendliness and conscientiousness.On each dimension, the individual was asked torate both his overall level and his variability. Forexample, on the friendliness dimension, he wasasked, "In general, how friendly and outgoing areyou?" and "How much do you vary from onesituation to another in how friendly and outgoingyou are?" Parallel pairs of questions were askedabout conscientiousness and other traits. Responseswere obtained on a seven-point scale which rangedfrom "not at all' to "extremely." It will benoted that these questions thus permit the in-dividual to employ his own concept of the traitdimension, to average across the situations he seesas pertinent and to ignore situations he sees asirrelevant. Accordingly, these global self-ratingswill be successful in predicting behavior only tothe extent that the individual's definition of a traitdimension coincides with the definition we willnecessarily be imposing by our selection of situa-tions to sample.

Using the same seven-point response scale, wealso obtained each individual's self-ratings on spe-cific behavior-situation items for each trait. Forexample, the CSBS included a 24-item scale whichassessed the trait of friendliness in specific situa-tions (e.g., "When in a store, how likely are youto strike up a conversation with a sales clerk?")and a 23-item conscientiousness scale (e.g., "Howcarefully do you double-check your term papersfor typing or spelling errors?"). Thus if theglobal self-ratings can be seen as reflecting theindividual's own definitions of the trait dimensions,then these CSBS scales can be viewed as reflect-ing the investigators' conception of these dimen-sions. The internal reliabilities (coefficient alpha)of the two trait scales were ,91 and .84 for friend-


liness and conscientiousness, respectively, and thecorrelation between them was +.13.

Cross-situational assessment. From the intro-ductory psychology course, 32 male and 32 femalestudents were recruited as subjects. In addition tothe initial testing session in which all studentsparticipated, the subjects were seen on three sepa-rate occasions, and they also signed release formsgiving us permission to obtain ratings on themfrom their parents and one of their close peers,usually a roommate. From these various sessions,the following six friendliness variables and sevenconscientiousness variables were derived.

Friendliness: (1) Self-report; (2) Mother's Re-port (3) Father's Report (4) Peer's Report:Each of these four judges provided us with anindependent assessment of the individual's friend-liness by rating him on the global friendliness itemand the 24-item CSBS friendliness scale. Foreach judge, these two measures were combined intoa single score. (5) Group Discussion: Each in-dividual was observed as he participated in agroup discussion with three other subjects of thesame sex. A measure of each individual's friendli-ness in the group was derived from the frequencyand duration of his vocalizations and the group'spostdiscussion rating of his friendliness. (6) Spon-taneous Friendliness: Each individual was ob-served as he waited in a waiting room with anexperimental confederate, and a measure of spon-taneous friendliness was derived from his "latency"in initiating conversation.

Conscientiousness: (1) Self-Report; (2)Mother's Report; (3) Father's Report; (4)Peer's Report: As with friendliness, each of thesefour judges provided us with an independent as-sessment of the individual's conscientiousness byrating him on the global conscientiousness item andthe 23-item CSBS conscientiousness scale. (5)Returning Evaluations: During the quarter, eachindividual received four evaluation forms by mailas part of an ongoing assessment of the intro-ductory psychology course. Each form asked himto evaluate one of the course lectures and to re-turn the form anonymously prior to the subsequentclass period. (The forms for our subjects werenumerically coded.) Measures of his promptnessin returning each of the four forms were combinedinto a single index. (6) Course Readings: Eachof the course evaluation forms described aboveasked the individual to check off the course read-ings he had been able to complete up to that time,providing us with four separate reports of his con-scientiousness in school work. (7) Neatness: Theneatness and cleanliness of each individual's hairand clothing were rated on two separate occasionsby three independent judges and his living quarterswere rated on nine aspects of neatness during asurprise visit paid during the last week of thequarter. These several observations were combinedinto an overall neatness score.

Finally, it should be noted that experimentalassistants and observers were all blind with respect

to the individual's scores on the trait dimensions,and no observer made more than one observationfor each trait on the same individual.

Results

The first step in the analysis of resultswas to classify each individual on a priorigrounds as a low-variability or a high-vari-ability subject in a way that would not beconfounded with his actual position on thetrait dimension. Accordingly, for each trait,a subject was first classified into one ofseven subgroups on the basis of his responseto the question "In general, how friendlyand outgoing [conscientious] are you?".Then, on the basis of his response to thequestion "How much do you vary from onesituation to another in how friendly and out-going [conscientious] you are?", he wasdesignated as a low-variability or a high-variability subject, respectively, dependingupon whether he was below or above themedian among the same-sex subjects at thesame point on the trait scale. Thus low andhigh variability were redefined at each of theseven points on the global trait scale inorder to partial out any relationship betweenan individual's self-rated variability and hisself-rated position on the trait dimension.

In order to assess each individual's cross-situational consistency for each trait, we con-verted each of the 13 variables to a standardT score with a mean of SO and a standarddeviation of 10 across the 64 subjects. Wethen calculated each individual's standarddeviation across the six friendliness variablesand across the seven conscientiousness vari-ables. These two standard deviations thusreflect the individual's cross-situational vari-ability on friendliness and conscientiousness,respectively; the larger the standard devia-tion, the more variable he is across situa-tions.

Friendliness. With respect to the friendli-ness dimension, our hypothesis was con-firmed. Individuals who indicate that theydo not vary much from one situation toanother do, in fact, display significantly lessvariability across situations than do thosewho say they do vary (6.42 versus 7.90; t= 2.34, p < .02, one-tailed test). Moreover,an individual's self-rated friendliness per se


TABLE 1

INTKRCORRELATIONS AMONG THK Six FRIKNDLINESS VARIABLES FOR Low- AND HIGH-VARIABILITY SUBJECTS

Low All variables

6. SpontaneousFriendliness

is not related to his cross-situational vari-ability ; in particular, individuals in the low-est, middle, and highest thirds of the distri-bution on the self-rated friendliness scaledid not differ from one another in theircross-situational variability, F (2, 61) =1.10, ns.

Table 1 shows how the differential cross-situational consistency of low- and high-vari-ability individuals translates into cross-situa-tional predictability in correlational terms.The intercorrelations of the six variables forthe 32 low-variability subjects are shownabove the diagonal; intercorrelations for the32 high variability subjects, below the di-agonal. The bottom row of Table 1 servesto summarize the matrix by showing themean correlation between the column vari-able and the remaining five variables for thetwo groups separately.*

As Table 1 shows, 13 of the 15 intercor-relations are higher for low-variability sub-jects than they are for high-variability sub-jects, six of them significantly so (p < .05,one-tailed test). The mean intercorrelationamong all the variables is +.57 for the low-

*A11 calculations and analyses involving cor-relation coefficients are actually performed on their^-transforms.

variability group and +.27 for the high-variability group. Note also that the pre-dicted effect is quite general across differentpairs of situations. For example, Mother'sReport and Father's Report, two measure-ments that one would expect to be highlysimilar and "contaminated" by one another,show a correlation of +.75 for the low-variability group but only +.28 for the high-variability group (p < .005, one tailed test) ;similarly, Group Discussion and SpontaneousFriendliness, the two methodologically in-dependent behavioral observations, show acorrelation of +.73 for the low-variabilitygroup but only +.30 for the high-variabilitygroup (/> < .009, one-tailed test). Thus, notonly have our expectations been confirmed,but the magic +.30 barrier appears to havebeen penetrated.

It will be noted that the "moderating vari-able" in this analysis, the variable whichseparates the population into groups whichare differentially predictable, is the indi-vidual's response to the single question"How much do you vary from one situationto another in how friendly and outgoing youare?" To see whether such an item couldenhance the utility of a standard personalityinventory as well, we computed the cor-

CROSS-SITUATIONAL CONSISTENCIES SIS

relations between the six friendliness vari-ables and the extraversion-introversionscale of the Eysenck Personality Inventory(Eysenck & Eysenck, 1968), which all ofour subjects had taken at the initial testingsession. Using the same criterion for lowand high variability as before, Table 2 dis-plays the correlations for the two groupsseparately. It is seen that the extraversion-introversion scale of the Eysenck PersonalityInventory does indeed have greater predictiveutility for self-identified low-variability sub-jects than for high-variability subjects (t =3.96, p < .01). Moreover, the effect remainssignificant even if the methodologically simi-lar Self-Report variable is removed from theanalysis (t = 3.22, p < .025).

Conscientiousness. As we noted earlier,the global self-rating items used to classifyour subjects permit the individual to utilizehis own definition of the trait, to implicitlyaverage across behaviors he regards aspertinent and to ignore all others. As wepointed out, this seemed likely to yield ourpredicted results only to the extent that thesubjects' trait definitions coincided with ourown. For the trait of friendliness, an ade-quate commonality of definition was achieved.For example, the correlation between theindividual's global self-rating of his friendli-ness and his mean score on our CSBSfriendliness items was +.84. The corre-sponding correlation for the conscientious-ness trait, however, was significantly lower,r = .62; Zdlffei.ence = 2.74, p < .006, two-tailed test, implying that the trait term "con-scientiousness" is more likely to denote dif-ferent equivalence classes of behaviors fordifferent individuals than is the trait term"friendliness." Not unexpectedly, then, wewere not able to replicate the friendlinessfindings for the conscientiousness trait whenwe employed our subjects' global self-ratingsas the classification variables.

Accordingly, we turned to our definitionof conscientiousness as the basis for subjectclassification, designating each individual aslow or high variability as a function of hisvariance on the CSBS conscientiousnessscale. In particular, we calculated each in-dividual's variance across the 23 conscienti-ousness items and divided it by his variance

TABLE 2

CORRELATIONS BETWEEN EYSENCK's EXTRAVERSION-INTROVERSION SCALE AND Six FRIENDLINESS

VARIABLES AS A FUNCTION OFSELF-RATED VARIABILITY

Extraversion versus:

Self-ReportMother's ReportFather's ReportPeer's ReportGroup DiscussionSpontaneous Friendliness

Mean correlation"Mean correlation

(omitting self-report)11

Self-ratedvariability

Low

.77

.54

.26

.71

.34

.25

.51

.44

High

.65

.37

.24

.41

.18-.12

.31

.22

» Correlated I = 3.96, p < .01, one-tailed test.b Correlated t = 3.22, p < .025, one-tailed test.

across all 86 items of the questionnaire. This"ipsatized" variance index not only correctsfor the individual's tendency to respond con-sistently or inconsistently to CSBS itemsirrespective of their content, but it also has amore conceptual interpretation as well. Itreflects the degree to which an individual"extracts" the particular trait-scale itemsfrom the total pool of items and "clusters"them into an equivalence class. Statistically,the ipsatized variance is like an inverted Fratio, representing the ratio of two varianceswhich assumes a value of zero if the indi-vidual responds identically to each item onthe trait scale and a value of one if he doesnot "cluster" the items on a trait scale at all.5

Subjects were first formed into matchedpairs on the basis of their CSBS con-scientiousness scores, and then each indi-vidual was classified as low or high vari-ability, respectively, depending upon whetherhis ipsatized variance was lower or higherthan that of his matched partner. It willbe recognized that this again serves to par-tial out any relationship between the indi-vidual's variability and his location on thetrait dimension. As before, the individual's

6 In 1961, Berdie used an intraindividual variancemeasure as a moderating variable for predictingmathematical aptitude. (See also Campus, 1974,and Fiske and Rice, 1955.)


TABLE 3

INTERCORRELATIONS AMONG THE SEVEN CONSCIENTIOUSNESS VARIABLES FORLow- AND HIGH-VARIABILITY SUBJECTS

LowHigh Self-report Mother's

reportFather'sreport report

Returningevaluations

Coursereadings Neatness All

variables

1. Self Report

2. Mother's Report.27

,31 .833. Father's Report

.23 .46

.444. Peer's Report

.25

,38

.49

5. ReturningEvaluations

.43 .60 .42 .49

-.12 .03 .10 .11

6. Course Readings.40 .20 .31

-.31 -.24

-.01

.18

.31 137. Neatness

.55 .39

Meancorrela-tions

.40

.20

,36

.12

Mean correla-tions (omittingNeatness)

.41

.11

.40

.06

.45

.09

cross-situational variability is assessed byhis standard deviation across the severalsituations.

With this method of classification, the re-sults confirm our hypothesis, parallelingexactly our findings for the friendlinessdimension. Thus low-variability subjectswere significantly less variable across situa-tions than high-variability subjects (7.46versus 8.89, correlated t = 2.80, p < .005,one-tailed test). And again, an individual'sstanding on the trait itself was not relatedto his cross-situational variability; indi-viduals in different thirds of the distributionon the CSBS conscientiousness scale did not

differ from one another in their cross-situational variability, F(2,61)<1, ns.*

Table 3 translates the differential cross-situational consistency of low- and high-vari-

6 We believe that, like conscientiousness, mosttraits would not attain the degree of definitionalconsensus between subject and investigator dis-played by the friendliness trait. Accordingly, webelieve that the ipsatized variance index, ratherthan the single item self-rating of variability, willprove to be the more promising candidate for themoderating variable in any future work. More-over, the ipsatized variance index can be calculatedfor any set of questionnaire items the investigatorchooses even when the layman has no trait termfor labeling the potential equivalence class so de-nned.

CROSS-SlTUATIONAL CONSISTENCIES 517

ability individuals into correlational terms.As before, correlations for the 32-low vari-ability subjects are found above the diagonal,the 32 high-variability subjects, below thediagonal.

It is seen in Table 3 that 15 of the 21intercorrelations are higher for low-vari-ability subjects than they are for high-vari-ability subjects, with 9 of them significantlyso (p < .05, one-tailed test). As the sum-mary rows of the table reveal, only theNeatness variable fails to conform to ourhypothesis accounting for all but two of thereversals in the matrix. The bottom rowof the table shows the mean intercorrela-tions obtained when this variable is omitted.

But it is precisely the Neatness variablewhich illustrates the main point of thisarticle. As an inspection of the correlationmatrix reveals, it is only we, the investi-gators, who think that school-related con-scientiousness (Returning Evaluations andCourse Readings) and personal neatnessought to belong in the same equivalenceclass. Our subjects do not. Our low-vari-ability subjects find them to be orthogonal( — .01 and —.11), and our high-variabilitysubjects apparently have time to do theirschoolwork or to keep things neat and clean,but not both (r = -.61).

The judgments by self, parents, and peersare also interesting in this regard. An in-spection of the correlations in the matrixreveals that judges of low-variability subjectsappear to be responding to both kinds ofconscientiousness almost equally, with somebias toward greater attention to school-re-lated conscientiousness. But judges of high-variability subjects seem to be ignoringschool-related conscientiousness and respond-ing primarily to the inversely-related per-sonal neatness. It is this latter patternwhich causes the Neatness variable to violateour hypothesis that low-variability subjectswould have the higher correlations.

The moral of all this, of course, is that weneed to move even further toward idio-graphic assessment. In this demonstration,we have relinquished the presumption thatall traits are relevant to all people but stub-bornly retained the right to dictate whichbehaviors and situations shall constitute the

trait itself. When an investigator is willingto release this degree of freedom as well, hisvalidity coefficients will reward him with anappropriate increment in magnitude. In thepresent study, for example, we might haveasked the subjects to rate the several CSBSitems for their relevance to the various traitdimensions. In this way, we might havediscovered a priori those subjects who donot share our personal delusion that the con-scientious person does his schoolwork andattends to his personal neatness.

PREDICTIVE UTILITY OF TRAITSAND SITUATIONS

We have argued in this article that it isnot possible, in principle, to do any betterthan predicting some of the people some ofthe time. Furthermore, our argumentswould seem to imply that an investigatormust simply abandon the highly variableindividual since the trait under investigationhas no predictive utility for him. But thisis not always true. As Mischel (1968,1973b) has persuasively argued, variabilityis not synonymous with either capriciousnessor unpredictability. Indeed, an individual'scross-situational variability may well be themark of a highly refined "discriminativefacility" (Mischel, 1973b), the ability to re-spond1 appropriately to subtle changes insituational contingencies. Although such anindividual cannot be predicted from aknowledge of his standing on a personalitytrait, he may be precisely the individual whois most predictable from a knowledge of thesituation. In short, if some of the peoplecan be predicted some of the time from per-sonality traits, then some of the people canbe predicted some of the time from situa-tional variables.

This point is nicely illustrated in recentwork on sex roles by S. Bern (1974, inpress). Whereas previous research in thisarea has been concerned either with the sex-typed masculine males and feminine femalesor, occasionally, with the sex-reversed femi-nine males and masculine females, S. Bernhas constructed a sex-role inventory whichpermits her to identify "androgynous" indi-viduals as well, individuals who attributeboth masculine and feminine characteristics


to themselves in about equal amounts with-out regard for their sex-role connotations.Thus, with respect to either of the two traitterms masculinity or femininity, androgynousindividuals are the "high variability" sub-jects; neither trait has predictive utility forthem. And as hypothesized, Bern finds thatandrogynous individuals of both sexes varytheir behavior cross-situationally so that theyare able to "do well" at both masculine andfeminine tasks and behaviors, whereas sex-typed individuals do not do well in situationsor settings which call for behaviors incon-gruent with their self-described sex roles.For example, only the androgynous subjectsshowed both "masculine" independence in anAsch conformity experiment as well as"feminine" nurturance when given the op-portunity to play with a baby kitten (S.Bern, in press). There are, in short, someoccasions when one can predict some of thehigh-variability people some of the time.

It should be clear from this discussionthat the position we have argued in thisarticle cannot be characterized as opposingeither side of the debate between those whobelieve that behavior is consistent acrosssituations and those who believe that be-havior is situationally specific. The shift toidiographic assumptions about the nature ofindividual differences dissolves this falsedichotomy and permits one to believe inboth propositions. As noted early in thisarticle, the actual cleavage is between no-mothetic and idiographic criteria for con-sistency and inconsistency.

Similarly, it should be clear that we havenot been arguing an "anti-Mischel" position.Both Mischel and we agree that the nomo-thetic assumptions of the traditional ap-proach in personality virtually guarantee a+ .30 ceiling on validity coefficients and,hence, that trait-based approaches predicatedon such assumptions will continue to fail thetest of predictive utility. Mischel and weagree that only an idiographic approach canbreak through this predictive barrier. Mis-chel and we agree that the classification ofsituations must be an integral part of any as-sessment procedure; moreover, we agree thatsuch classification will have to be in terms

of the individual's own phenomenology, notthe investigator's (Mischel, 1973b), a sug-gestion that is bound to increase further thedeja vu of any psychologist old enough toremember Kurt Lewin (1935). It is truethat Mischel and we have chosen differentconceptual languages in which to expressthese points, and future divergences willsurely emerge as a consequence, but the twoformulations are far more similar in theirbasic assumptions than their formal appear-ances would suggest.

The failure of traditional assessment pro-cedures and the belief that person-situationinteractions will account for most of thepsychologically interesting variance in be-havior have led several recent writers toemphasize that personality assessment mustbegin to attend seriously to situations. Weagree. We have merely chosen to emphasizethe perfectly symmetric, but perhaps moresubtle, point that personality assessmentmust also begin to attend seriously topersons.

REFERENCES

Alker, H. A. Is personality situationally specificor intrapsychically consistent? Journal of Per-sonality, 1972, 40, 1-16.

Allport, G. W. Personality: A psychological in-terpretation. New York: Holt, 1937.

Allport, G. W. Traits revisited. American Psy-chologist, 1966, 21, 1-10.

Allport, G. W., & Odbert, H. S. Trait-names: Apsycholexical study. Psychological Monographs,1936, 47(1, Whole No. 211).

Argyle, M., & Little, B. R. Do personality traitsapply to social behavior ? Journal for the Theoryof Social Behaviour, 1972, 2, 1-35.

Averill, J. R. The dis-position of psychologicaldispositions. Journal of Experimental Researchin Personality, 1973, 6, 275-282.

Baldwin, A. L. The study of individual personalityby means of the intraindividual correlation. Jour-nal of Personality, 1946, 14, 151-168.

Bern, D. J. Constructing cross-situational con-sistencies in behavior : Some thoughts on Alker'scritique of Mischel. Journal of Personality,1972, 40, 17-26.

Bern, S. L. The measurement of psychologicalandrogyny. Journal of Consulting and ClinicalPsychology, 1974, 42, 155-162.

Bern, S. L. Sex-role adaptability: One conse-quence of psychological androgyny. Journal ofPersonality and Social Psychology, in press.


Berdie, R. F. Intra-individual variability and pre-dictability. Educational and Psychological Mea-surement, 1961, 21, 663-676.

Bowers, K. S. Situationism in psychology: Ananalysis and a critique. Psychological Revieiv,1973, 80, 307-336.

Campus, N. Transituational consistency as adimension of personality. Journal of Personalityand Social Psychology, 1974, 29, 593-600.

Chapman, L. J., & Chapman, J. P. Illusory cor-relations as an obstacle to the use of validpsycho-diagnostic signs. Journal of AbnormalPsychology, 1969, 74, 271-280.

Dudycha, G. J. An objective study of punctualityin relation to personality and achievement. Ar-chives of Psychology, 1936, 204, 1-319.

Endler, N. S. The case for person-situation inter-actions. Paper presented at the meeting of theAmerican Psychological Association, Montreal,August 1973. (a)

Endler, N. S. The person versus the situation—A pseudo issue? A response to Alker. Journalof Personality, 1973, 41, 287-303. (b)

Endler, N. S., & Hunt, J. McV. S-R inventoriesof hostility and comparisons of the proportionsof variance from persons, responses, and situa-tions for hostility and anxiousness. Journal ofPersonality and Social Psychology, 1968, 9, 309-315.

Eysenck, H. J., & Eysenck, S. B. G. Manual forthe Eysenck Personality Inventory. San Diego:Educational and Industrial Testing Service, 1968.

Fiske, D. W., & Rice, L. Intra-individual responsevariability. Psychological Bulletin, 1955, 52,217-250.

Harre, R., & Secord, P. F. The explanation ofsocial behavior. Oxford: Basil, Blackwell, &Mott, 1972.

Hartshorne, H., & May, M. A. Studies in thenature of character. Vol. 1. Studies in deceit.New York: Macmillan, 1928.

Hartshorne, H., & May, M. A. Studies in thenature of character. Vol. 2. Studies in serviceand self-control. New York: Macmillan, 1929.

Hartshorne, H., May, M. A., & Shuttleworth, F.K. Studies in the nature of character. Vol. 3.Studies in the organisation of character. NewYork: Macmillan, 1930.

Jones, E. E., & Davis, K. E. From acts to dis-positions : The attribution process in personperception. In L. Berkowitz (Ed.), Advancesin experimental social psychology. Vol. 2. NewYork: Academic Press, 1965.

Jones, E. E., & Goethals, G. R. Order effects inimpression formation: Attribution context andthe nature of the entity. New York: GeneralLearning Press, 1971.

Jones, E. E., & Harris, V. A. The attribution ofattitudes. Journal of Experimental Social Psy-chology, 1967, 3, 1-24.

Jones, E. K, & Nisbett, R. E. The actor and ob-server: Divergent perceptions of the causes 0}

behavior. New York: General Learning Press,1971.

Kelley, H. H. Attribution theory in social psy-chology. In D. Levine (Ed.), Nebraska Sym-posium on Motivation: 1967. Lincoln: Uni-versity of Nebraska Press, 1967.

Kelley, H. H., & Stahelski, A. J. Social inter-action basis of cooperators' and competitors' be-liefs about others. Journal of Personality andSocial Psychology, 1970, 16, 66-91.

Kelly, G. A. The psychology of personal con-structs. New York: Norton, 1955, 2 vols.

Lehmann, H. C, & Witty, P. A. Faculty psy-chology and personality traits. American Jour-nal of Psychology, 1934, 44, 490.

Lewin, K. A dynamic theory of personality. NewYork: McGraw-Hill, 1935.

McClelland, D. C. Personality. New York:William Sloane Associates, 1951.

Mischel, W. Personality and assessment. NewYork: Wiley, 1968.

Mischel, W., Continuity and change in personality.American Psychologist, 1969, 24, 1012-1018.

Mischel, W. On the empirical dilemmas of psy-chodynamic approaches: Issues and alternatives.Journal of Abnormal Psychology, 1973, 82, 335-344. (a)

Mischel, W. Toward a cognitive social learningreconceptualization of personality. Psychologi-cal Review, 1973, 80, 252-283. (b)

Moos, R. H. Sources of variance in responses toquestionnaires and in behavior. Journal of Ab-normal Psychology, 1969, 74, 405-412.

Newcomb, T. M. Consistency of certain extro-vert-introvert behavior patterns in 51 problemboys. New York: Columbia University, Teach-ers College, Bureau of Publications, 1929.

Passini, F. T., & Norman, W. T. A universalconception of personality structure? Journal ofPersonality and Social Psychology, 1966, 4, 44-49.

Peterson, D. R. The clinical study of social be-havior. New York: Appleton-Century-Crofts,1968.

Sanford, N. Issues in personality theory. SanFrancisco: Jossey-Bass, 1970.

Schneider, D. J. Implicit personality theory: Areview. Psychological Bulletin, 1973, 79, 294-309.

Scott, W. A. Attitude measurement. In G. Lind-zey & E. Aronson (Eds.), The handbook of so-cial psychology. Vol. 2. Menlo Park, Calif.:Addison-Wesley, 1968.

Stagner, R. Psychology of personality. NewYork: McGraw-Hill, 1937.

Stagner, R. Traits are relevant—logical and em-pirical analysis. Paper presented at the meetingof the American Psychological Association,Montreal, August 1973.

Stouffer, S. A., Guttman, L. Suchman, E. A.,Lazarsfeld, P. F., Star, S. A., & Clausen, J. A.Studies in social psychology in World War II.


Vol. 4. Measurement and prediction. Prince- and the implacable experimenter: An inquiryton: Princeton University Press, 1950. into the consistency of personality. Journal of

Vale, J. R., & Vale, G. R. Individual difference Abnormal Psychology, 1973, 82, 324-334.and general laws in psychology: A reconcilia- Wallach, M. A., & Leggett, M. I. Testing thetion. American Psychologist, 1969, 24, 1093- hypothesis that a person will be consistent:1108. Stylistic consistency versus situational specificity

Vernon, P. E. Personality assessment: A critical in size of children's drawing. Journal of Per-survey. New York: Wiley, 1964. somlity, 1972, 40, 309-330.

Wachtel, P. Psychodynamics, behavior therapy, (Received January 21, 1974)

NOTICE TO APA AUTHORS

The American Psychological Association announces publication of the secondedition of the Publication Manual in August 1974. This new edition, whichsupersedes the 1967 Manual, will be adopted by the 14 APA journals in 1975.

The new Manual is more comprehensive than the previous edition. It updatesAPA publication policies and procedures and incorporates changes in editorialpractice since 1967. For instance, APA now sends many authors their copy-edited manuscripts for review before they are set into type, and some editorsnow use blind review procedures. The second edition also includes up-to-datestatements of the coverage of each APA journal including the Journal oj Ex-perimental Psychology which will be published in four separate sections in 1975.

The new Publication Manual initiates several changes in APA style.These changes are announced in the August 1974 American Psychologistand will be introduced in the APA journals in January 1975. During the periodof transition to the new style, authors should note that (a) all manuscriptspublished in 1974 will be copy-edited according to the 1967 Manual, (b)manuscripts accepted in 1974 and published in 1975 will be copy-edited to con-form to the new Manual. Starting in 1975, accepted manuscripts that departsignificantly from the Manual will be returned to authors for correction.

Authors will be encouraged by the changes in the second edition. The newAPA style simplifies reference forms; eliminates unnecessary underlines,brackets, and other devices; supports appropriate use of "I" and "we"; andgenerally clarifies typing requirements. Material is arranged for maximum con-venience to authors and typists, and all sections are cross-referenced and indexed.

The new Publication Manual is available after August 1 for $3; Send ordersto APA Publication Sales, 1200 Seventeenth Street, N.W., Washington, D. C.20036. Orders of $15 or less must include payment unless they are submittedon institutional purchase order forms.

on predicting some of the people some of the...

Documents