Self-Schemata and Processing Information About the Self
Hazel MarkusThe University of Michigan
Attempts to organize, summarize, or explain one's own behavior in a particulardomain result in the formation of cognitive structures about the self or self-schemata. Self-schemata are cognitive generalizations about the self, derivedfrom past experience, that organize and guide the processing of the self-relatedinformation contained in an individual's social experience. The role of schematain processing information about the self is examined by linking self-schematato a number of specific empirical referents. Female students with schemata ina particular domain and those without schemata are selected and their per-formance on a variety of cognitive tasks is compared. The results indicate thatself-schemata facilitate the processing of information about the self (judgmentsand decisions about the self), contain easily retrievable behavioral evidence,provide a basis for the confident self-prediction of behavior on schema-relateddimensions, and make individuals resistant to counterschematic information.The relationship of self-schemata to cross-situational consistency in behaviorand the implications of self-schemata for attribution theory are discussed.
The quantity and variety of social stimu-lation available at any time is vastly greaterthan a person can process or even attend to.Therefore, individuals are necessarily selec-tive in what they notice, learn, remember, orinfer in any situation. These selective ten-dencies, of course, are not random but dependon some internal cognitive structures whichallow the individual to process the incominginformation with some degree of efficiency.Recently, these structures for encoding andrepresenting information have been calledframes (Minsky, 1975), scripts (Abelson,
The influence of cognitive structures on theselection and organization of information isprobably most apparent when we process in-formation about ourselves. A substantialamount of information processed by an indi-vidual (some might even argue a majority ofinformation) is information about the self,and a variety of cognitive structures are nec-essarily involved in processing this informa-tion. Yet in research on the self, in the per-
63
64 HAZEL MARKUS
sonality area for example, there has been anotable lack of attention to the structuresused in encoding one's own behavior and inthe processing of information about one'sown behavior. Research on self-perception(Bern, 1967, 1972) and research on self-monitoring (Snyder, 1974; Snyder & Mon-son, 1975) clearly suggests that the individualis an active, constructive information proces-sor, but no specific cognitive structures haveyet been implicated in this theorizing andresearch.
It is proposed here that attempts to orga-nize, summarize, or explain one's own be-havior in a particular domain will result inthe formation of cognitive structures aboutthe self or what might be called self-schemata.Self-schemata are cognitive generalizationsabout the selj, derived from past experience,that organize and guide the processing ofself-related information contained in the indi-vidual's social experiences. The main purposeof the present studies is to examine somefunctions of self-schemata in the processingof information about the self.
Self-schemata include cognitive representa-tions derived from specific events and situa-tions involving the individual (e.g., "I hesi-tated before speaking in yesterday's discus-sion because I wasn't sure I was right, onlyto hear someone else make the same point")as well as more general representations de-rived from the repeated categorization andsubsequent evaluation of the person's be-havior by himself and by others around him(e.g., "I am very talkative in groups ofthree or four, but shy in large gatherings,""I am generous," "I am creative," or "I amindependent").
Self-schemata are constructed from infor-mation processed by the individual in thepast and influence both input and output ofinformation related to the self. They repre-
This paper is based on a doctoral dissertationsubmitted to the Department of Psychology of TheUniversity of Michigan. I am especially grateful toRobert Zajonc, Richard Nisbett, and Keith Sendsfor advice and comments on an earlier version ofthis paper, and to Mary Cullen for preparation ofthe manuscript.
Requests for reprints should be sent to HazelMarkus, Institute for SSocial Research, Universityof Michigan, Ann Arbor, Michigan 48106.
sent the way the self has been differentiatedand articulated in memory. Once established,these schemata function as selective mecha-nisms which determine whether informationis attended to, how it is structured, how muchimportance is attached to it, and what hap-pens to it subsequently. As individuals ac-cumulate repeated experiences of a certaintype, their self-schemata become increasinglyresistant to inconsistent or contradictory in-formation, although they are never totally in-vulnerable to it.
Self-schemata can be viewed as a reflectionof the invariances people have discovered intheir own social behavior. They representpatterns of behavior that have been observedrepeatedly, to the point where a frameworkis generated that allows one to make infer-ences from scant information or to quicklystreamline and interpret complex sequencesof events. To the extent that our own be-havior exhibits some regularity or redun-dancy, self-schemata will be generated be-cause they are useful in understanding in-tentions and feelings and in identifying likelyor appropriate patterns of behavior. While aself-schema is an organization of the repre-sentations of past behavior, it is more than a"depository." It serves an important process-ing function and allows an individual to gobeyond the information currently available.The concept of self-schema implies that in-formation about the self in some area hasbeen categorized or organized and that theresult of this organization is a discerniblepattern which may be used as a basis forfuture judgments, decisions, inferences, orpredictions about the self.
There is substantial historical precedentfor the schema term and for schemalikeconcepts, and it would entail a very lengthydiscussion to trace the history of the term(cf. Bartlett, 1932; Kelley, 1972; Kelly,19SS; Piaget, 1951). In social psychology,schema-like concepts (e.g., causal schemata,scripts, implicit personality theories) havegenerally been vaguely denned heuristics withno real empirical moorings. Despite theirassumed cognitive consequences, they havebeen viewed primarily as epiphenomena, in-ferred on the basis of behavior or invoked invarious post hoc explanations. The investiga-
SELF-SCHEMATA 65
tion of self-schemata requires examining thehypothesized functions of schemata for theirparticular empirical implications. To datethis has not been done.
Recent work in the general area of cogni-tion suggests a number of ways of investigat-ing self-schemata. This work provides modelsof information processing (e.g., Anderson &Bower, 1973; Atkinson & Shiffrin, 1968;Erdelyi, 1974), indicates the possible func-tions of cognitive structures, and makes useof a variety of measures (recognition, recall,response latency, etc.) and techniques (signaldetection, chronometric descriptions of infor-mation flow, etc.) capable of empiricallyidentifying these functions. The experimentalwork in this area, however, has concentratedlargely on the processing of neutral or non-sense material. With the exception of somerecent work (Mischel, Ebbesen, & Zeiss,1976) there has been little empirical work onthe influence of cognitive structures on theselective processing of significant social in-formation (e.g., information about importantaspects of one's self).
The idea of self-schemata as cognitive gen-eralizations about the self has a number ofimplications for the empirical work on per-sonality and cross-situational consistency.For example, an endorsement of a trait ad-jective as self-descriptive or an endorsementof an item on a self-rating scale may reflectan underlying, well-articulated self-schema.It is equally possible, however, that the markon the self-rating scale is not the product ofa well-specified schema, but is instead theresult of the favorability of the trait term,the context of the situation, the necessityfor a response, or other experimental de-mands. Only when a self-description derivesjrom a well-articulated generalization aboutthe self can it be expected to converge andform a consistent pattern with the individu-al's other judgments, decisions, and actions.Thus, a person who does not really thinkabout herself as conscientious, yet would notobject to labeling herself as such, cannot beexpected to react to being late for an ap-pointment in the same way as one who ac-tively conceives of herself as conscientious,who can readily describe numerous displaysof conscientiousness in the past, and who
can enumerate the way she insures futureconscientious behavior on her part.
To demonstrate the construct validity ofthe concept of self-schemata, a number ofempirical referents can be specified. If self-schemata are built up from cognitive repre-sentations of past experiences, individual dif-ferences in self-schemata should be readilydiscovered because individuals clearly differin their past experiences. If a person has adeveloped self-schema, he should be readilyable to (a) process information about theself in the given domain (e.g., make judg-ments or decisions) with relative ease, (b)retrieve behavioral evidence from the domain,(c) predict his own future behavior in thedomain, and (d) resist counterschematic in-formation about himself. If a person has hadrelatively little experience in a given domainof social behavior or has not attended tobehavior in this domain, then it is unlikelythat he will have developed an articulatedself-schema.
Consistency in patterns of response on anumber of self-description tasks, as well asconvergence in results from a number of di-verse cognitive tasks involving self-judg-ments, should provide evidence for the ex-istence of an organization of knowledge aboutthe self on a particular dimension of behav-ior, or a self-schema. To the extent thatindividuals do not possess an articulated self-schema on a particular dimension of behavior,they will not exhibit consistency in response.Nor will they display the discrimination nec-essary for the efficient processing of informa-tion and the prediction of future behavioralong this dimension.
The procedure of the first study is to selecta dimension of behavior, to identify indi-viduals with schemata and those withoutschemata on this dimension, and then tocompare their performance on a variety ofcognitive tasks. Several tasks utilizing self-rating, self-description, and prediction of be-havior are combined to determine whetherthe processing of information about one's selfvaries systematically as a function of self-schemata. The second study investigates theselective influence of self-schemata on theinterpretation of information about one's ownbehavior. Individuals with articulated self-
66 HAZEL MARKUS
schemata (along a specific behavioral dimen-sion identified in the first study) and indi-viduals without such schemata are inducedto engage in behavior that is potentially diag-nostic of this dimension. The impact of thisinformation is evaluated for both groups.
Study 1
This study is concerned with the impact of self-schemata on the selection and processing of infor-mation about the self. Individuals with self-schemataalong a particular dimension of behavior are com-pared with individuals without such self-schemata.Also compared are individuals with different self-schemata along the same dimension of behaviorSpecifically, it is hypothesized that a self-schemawill determine the type of self-judgments that aremade and that these judgments will vary in latencydepending on the presence and content of self-sche-mata. Also, individuals with self-schemata shouldfind it easier to describe specific behavior that isrelated to their schema and should be relativelymore certain about prediction of their behavioralong this dimension than individuals without sche-mata.
Method
To gain a preliminary idea of each subject's self-schema on various dimensions, a number of self-rating scales were administered in introductory psy-chology classes. The most appropriate pattern ofvariation in self-ratings was found on the inde-pendence-dependence dimension and thus it wasselected as the dimension for further study. Fromamong the individuals completing this questionnaire,48 were selected to participate individually in thelaboratory sessions.
The first laboratory session consisted of threeseparate cognitive tasks designed to assess the in-fluence of self-schemata about independence on theprocessing of information about the self. These in-cluded:
1. Content and latency of self-description. Sub-jects were given a number of trait adjectives associ-ated with independence and dependence and wereasked to indicate for each whether it was self-descriptive or not. Response latency was recordedfor each judgment.
2. Supplying behavioral evidence for self-descrip-tion. Subjects were asked to select trait adjectivesthat were self-descriptive and then to cite instancesfrom their own past behavior to support their en-dorsement of a particular adjective as self-descrip-tive.
3. Predicting the likelihood of behavior. Subjectswere given a series of descriptions of independentand dependent behavior and were asked to judgehow likely it was they would behave in these ways.
Subjects
For the questionnaire phase of the experiment,subjects were 101 female students in introductory
psychology classes at a large university. Subjectsfor the first laboratory sessions were 48 studentsselected from this group. Only female students wereused in this study because the distribution of self-ratings on various dimensions appears to differ withsex. Using male and female students would haverequired selecting more dimensions.
Materials and Procedures
Initial questionnaire. Individuals in introductorypsychology classes were asked to rate themselves onthe Gough-Heilbrun Adjective Check List (Gough& Heilbrun, 196S) and on several semantic differen-tial scales describing a variety of behavioral do-mains. On the latter measure, subjects were alsoasked to rate the importance of each semanticdimension to their self-description. From these re-spondents, three groups of 16 subjects each wereselected to participate in the experimental sessions.
1. Independents Individuals who rated them-selves at the extreme end (points 8-11 on an 11-point scale) on at least two of the following se-mantic differential scales: Independent-Dependent,Individualist-Conformist, or Leader-Follower, andwho rated these dimensions as important (points8-11 on an 11-point scale), and who checked them-selves as "independent" on the adjective check listwere termed Independents.1
2. Dependents. Individuals who rated themselvesat the opposite end (points 1-4) on at least two ofthese scales, and who rated these dimensions as im-portant (points 8-11 on an 11-point scale), and whochecked themselves as "dependent" on the adjectivecheck list were termed Dependents.
3. Aschematics. Individuals who rated themselvesin the middle range (points S-7) on at least two ofthese three scales, and fell in the lower portion ofthe distribution on the importance scale, and didnot check themselves as either "independent" or"dependent" on the adjective check list were termedAschematics. The term aschematic is used here tomean without schema on this particular dimension.2
Invoking the importance criterion conjointly withthe extremity criterion made it possible to avoid
1 Although subjects were selected on the basis oftheir "extreme" scores on these self-rating scales,only 2 subjects of the 48 actually used the endpoints11 or 1 in their self ratings on the semantic differ-ential scales.
2 Another indicator of a subject's self-schemaabout independence was her score on the Autonomyscale of the Gough-Heilbrun Adjective Check List(ACL). The Autonomy scale is one of the ACL's 24empirically derived scales designed to correspond todimensions of the California Psychological Inven-tory and Murray's need-press system. Autonomy isdefined as the tendency to act independently ofothers or of social values and expectations. Subjectsselected as Independents in this study were amongthe 25 with highest scores on this measure, andthose selected as Dependents were among the 25with lowest scores. Aschematics scored in the middle
SELF-SCHEMATA 67
confusing Aschematics with persons who act (andthink of themselves) as independent in some classesof situations and as dependent in other classes ofsituations, and do so consistently. Making such finediscriminations would lead these individuals todevelop a fairly well-articulated conception of theindependence domain of social behavior, and thus itwould be incorrect to classify them as Aschematics.However, if these people had a well-articulatedconception of themselves as both dependent andindependent, they would no doubt be quite sensi-tive to social behavior in the domain of indepen-dence and would consider it to be a significant andimportant area. Hence, they would not be classifiedas Aschematics according to our criteria. Among theAschematics the average importance rating on thethree semantic differential scales was 6.4, whileamong the Schematics it was 9.5.
Three to four weeks after the questionnaire wasadministered, the 48 subjects were called individuallyto the laboratory and received identical treatment.They were not informed of a connection betweenthis session and the questionnaire, and it is unlikelythat they could have inferred such a connectionsince different experimenters were used.
Task 1: Content and latency of self-description.Sixty-nine trait adjectives were prepared on 2 X 2inch (5X5 cm) slides, IS had been previouslyjudged (by another group of 50 subjects) to berelated to independence and nonconformity (inde-pendent words) and IS were judged to be relatedto dependence and conformity (dependent words).These 30 words were the critical schema-relatedstimuli.3 Thirty other words, included for compari-son with the schema-related adjectives, clusteredaround the notions of creativity and noncreativityand were used as control words. In each group of 30words, 10 were negatively rated, 10 were positivelyrated, and 10 neutral, according to Anderson's(1968) list of the hkableness of 555 trait adjectives.The words were either of high frequency or mod-erate frequency (according to the norms of Carroll,Davies, & Richman, 1971). The remaining 9 wordswere 3 practice adjectives, 3 adjectives which nearlyall subjects had indicated were self-descriptive onthe initial questionnaire (honest, intelligent,friendly), and 3 adjectives which nearly all sub-jects had indicated were not self-descriptive (rude,obnoxious, unscrupulous).
Each of these 69 adjectives was presented on thescreen for 2 seconds by a slide projector activatedby the experimenter. Following the presentation of
a word, the subject was required to respond bypushing a me button if the word was self-descrip-tive, or a not me button if the word was not self-descriptive. The response stopped an electronic clockwhich began with the presentation of the stimulus.The subject had to respond with one of the twobuttons before the next stimulus would appear. Foreach word the experimenter recorded both responselatency and the choice of me or not me. Subjectswere not aware that response latency was beingmeasured Four different randomly determined or-ders of presentations were used for the slides, with12 subjects in each order. In addition, for half ofthe subjects the me button was on the right side ofthe panel and for the remaining half on the left side.To insure that individuals were associating similartypes of behaviors to the trait adjectives, a particu-lar context was specified for the self-judgments. Theinstructions were
When you are making these decisions aboutyourself, try to imagine yourself in a typical groupsituation, one that might occur for example, in aclassroom, in the dorm lounge, or at a meeting ina friend's home. You are together to discuss animportant and controversial issue and to makesome decisions about it. Many of the people inthe group you know or are familiar to you, whileothers are not.
Task 2: Supplying behavioral evidence for self-descriptions. After the categorization task, eachsubject received a booklet containing 16 words (1on each page) from the set described in Task 1.Seven of these words were from the set of inde-pendent words and 7 were from the set of dependentwords. Two additional words were from the cre-ative/noncreative set. Of the 16 words, 4 were posi-tively rated for likableness, 4 were negatively rated,and 8 were neutral. The order of the adjectives ineach booklet was randomly determined. Subjectswere given written instructions to circle each adjec-tive they considered to be self-descriptive and werealso asked the following:
Immediately after you circle an adjective, listthe reasons you feel this adjective is self-descrip-tive. Give specific evidence from your own pastbehavior to indicate why you feel a particulartrait is self-descriptive. . . . List the first kinds ofbehaviors that come to your mind. Do not worryabout how other people might interpret a par-ticular behavior; use your own frame of reference.(Several examples were given.)
range on this measure It is important to note thatall of the subjects who would be labeled autono-mous on the basis of this measure (that is, theywere among the 2S highest scorers) also ratedthemselves extremely (points 8-11) on at least twoof the semantic differential scales. Those who wouldbe labeled nonautonomous or dependent on the basis°f this measure also rated themselves extremely(points 1-5) on the three semantic differential
Task 3: Predicting the likelihood of behavior.The third task utilized a series of specific behavioraldescriptions taken from a large number of descrip-tions that had been rated by a separate group of 40introductory psychology students as characterizingeither independence and nonconformity or depen-dence and conformity. This outside group of sub-jects was asked to decide how they would label orcategorize each act if they saw it or if they heardsomeone describe themselves in these terms. Thefinal list included 10 pairs of behavioral descriptionsmatched in content but differing in the way thebehavior would be categorized, for example, "Youhesitate before commenting, only to hear someoneelse make the point you had in mind" (rated de-pendent) and "You speak up as soon as you havesome comments on the issue being discussed" (ratedindependent). Several filler items also were included.A context for the behavioral descriptions similar tothe one in Task 1 was provided and then the sub-jects were given written instructions which read:
Listed below are a number of behaviors andreactions that might be true of you in a gather-ing like this. For each one, indicate how likely
Subjects
[^Dependents jp^ Aschematics ^ Independents
2 0
Dependent adjectives Independent adjectives
Figure 1. Top panel: Mean number of independentand dependent adjectives judged as self-descriptive.Bottom panel. Mean response latency for indepen-dent and dependent adjectives judged as self-descrip-tive and as not self-descriptive.
or how probable it is that you would behave orreact in this way. You may assign each item anynumber from 0 to 100. A 0 means that this couldnot be true of you, that it is extremely unlikelythat you would act or feel this way A 100 meansthat this could very well be true of you, that itis likely that you would act or feel this way.
Results
For the purpose of analysis, subjects weredivided into three groups labeled Indepen-dents, Dependents, and Aschematics, as de-scribed in the Procedure section.
Task 1: Content and latency of self-de-scription. As shown in the top panel of Fig-ure 1, the three groups of subjects clearlydiffered in the average number of the 15 de-pendent words judged as self-descriptive,^(2,45) = 14.89, p < .001. The three groupsalso differed in the average number of the15 independent adjectives judged as self-descriptive, ^(2,45) = 9.27, p < .001. Usingp < .05 as a criterion, Newman-Keuls com-parisons showed that Dependents judged sig-nificantly more dependent words as self-de-scriptive than did Independents, and con-versely, Independents judged significantlymore independent words as self-descriptivethan did Dependents.
The bottom panel of Figure 1 presents theaverage response latencies for self-descrip-tive judgments (me) and for not self-descrip-tive judgments (not me) for the independentand dependent adjectives. Dependent sub-jects were reliably faster at making me judg-ments for dependent words than for inde-pendent words, *(15) = 2.63, p < .01.* Con-gruently, Independent subjects were reliablyfaster at making me judgments for indepen-dent adjectives than for dependent adjectives,^(15) = 2.72, p<.01. The Aschematics,however, did not differ in response latency forindependent and dependent words.
When the top and bottom panels of Figure1 are considered together, a number of otherpoints about the self-categorization of thesethree groups of subjects can be made. A meresponse to a particular adjective may bethe result of an individual labeling her be-
• Except where specified, all t tests are two-tailed.
SELF-SCHEMATA 69
havior or reactions in this way or thinkingabout herself in these terms. But it may alsobe the result of several other considerations,such as the positivity or social desirability ofa particular adjective. Looking within groups,it can be seen from the top panel that De-pendent subjects responded me to signifi-cantly more dependent words than indepen-dent words; there is a clear differentiationhere, *(15) = 10.55, p < .001. Independentsubjects however, although responding me tomore independent words than either of theother two groups of subjects, found nearly asmany dependent adjectives to be self-descrip-tive, /(15) < 1. On the basis of these find-ings alone, one might conclude that thisgroup does not use independent or depen-dent words differentially or that independenceis not a meaningful dimension for these sub-jects. The bottom panel indicates that this isnot the case, however. Independent subjectsrespond much faster to the independent wordsthan they do to the dependent words. Thefaster processing times for the independentwords suggest that it is indeed easier for In-dependent subjects to think about themselvesin these terms or that they are used to think-ing about themselves in these terms.
The latency measure is also useful in inter-preting the results of the Aschematic group.From the top panel it can be seen that Asche-matics respond me to more dependent wordsthan independent words, f(15) = 2.42, p <.05. If the response latencies for these judg-ments are ignored, one might take this tomean that these subjects are similar to theDependents. It is evident, however, thatAschematic subjects do not really use thesetwo sets of words differentially in describingthemselves in the same way Dependent sub-jects do. There is no difference among Asche-matic subjects in processing time for the twosets of words. Even though they were con-strained to think of a specific social situa-tion, Aschematics appear to be equally at easelabeling their behavior with independent ordependent adjectives.6
Response latency for self-categorizationappears to be a sensitive measure which re-veals variations in judgments that ratingscales and check lists cannot. Endorsements
2 0 -
E 10
30
2 0
10
Subjects
[^Dependents Independents
16 16 16Honest
13 14 15Intelligent
Control odiectives
15 16 14Friendly
I I 13 14 12 14 16 12 14 14Independent Ambitious Individualistic
Schemo-reloted odiectives
Figure 2. Mean response latency for schema-relatedand control adjectives
which result from the positivity or desirabil-ity of a stimulus can potentially be sepa-rated from responses which reflect more validself-characterizations. This is clearly demon-strated in Figure 2. The top panel shows theresponses of three groups of subjects to thethree control words that were included inthe list of presented adjectives. The numberof subjects out of the total 16 that respondedme to each word is shown beneath the bar.
5 Overall, subjects find more dependent wordsthan independent words to be self-descriptive, de-spite the fact that the two sets of words wereinitially matched for positivity and frequency. Infact, across all subjects an average of 7.4 indepen-dent words were judged to be self-descriptive com-pared to an average of 10.9 dependent words. Thismay also explain the relatively longer response timesfor not me judgments of dependent words obtainedin all three groups of subjects. Across all subjectsthe average latency for a not me response to de-pendent words was 2.63 seconds compared with 2.22seconds for independent words. It is possible thatwithin the set of our 69 words (Anderson's subjectsrated a set of 555 adjectives), the dependent wordsappeared as more positive or desirable, and thus itwas difficult for subjects to respond not me to them.
70 HAZEL MARKUS
Not surprisingly, nearly all subjects viewedthemselves as honest, intelligent, and friendly.And there are no differences among the threegroups in the processing time for these adjec-tives. The overall average latency for thesethree words was much shorter (1.67 sec) thanthe overall average latency for all words (2.23sec). These adjectives appear to be synony-mous with general "goodness" and do notconvey unique information about an indi-vidual. The three groups of subjects also didnot differ in processing time for not me judg-ments to the negatively rated adjectives rude,obnoxious, and unscrupulous.
This was not true, however, for other ad-jectives which presumably are tied to morespecifically defined behaviors. The bottompanel of Figure 2 shows the responses of thethree groups of subjects to three schema-related words, that is, three words from theset of independent words shown to subjects.Here again, just on the basis of their overtresponses, it would appear that these subjectsdid not differ in their characterization ofthemselves, as the clear majority of subjectsalso responded me to these three schema-re-lated words. The latency measures exhibiteda much different pattern, however. Depen-dent subjects, for the most part, indicatedthat they were independent, ambitious, andindividualistic, but it took them significantlylonger to make this judgment than it did forIndependent subjects. Separate analyses ofvariance performed on the response latenciesfor these three words yielded significant dif-ferences among the three groups of subjects:ambitious, F(2,35) = 6.59, p < .01; inde-pendent, F(2,3S) = 6.59, p < .01; individu-alistic, F(2,37)-4.56, p < .01. It is prob-able that Dependent subjects would like tolabel themselves with these words and subse-quently do, but they experience some diffi-culty endorsing the words, a difficulty theydo not experience with the dependent words.This result cannot be explained by assumingthat Dependent subjects just take longer tomake self-judgments, for on the control ad-jectives the latencies of the Dependents didnot differ from those of the other groups ofsubjects. The faster processing times of theIndependent subjects on schema-related wordsmay be indicative of schemata which containinformation about independence and individu-
alism. Dependent subjects do not have infor-mation about themselves which might be rea-sonably labeled in this way, and their hesi-tation in making unsubstantiated judgmentsabout themselves is reflected in relativelylonger latencies.
The lack of differentiation in response la-tency to the schema-related adjectives shownby the Aschematics relative to the other twogroups of subjects has been interpreted asevidence for the absence of a schema on thisdimension. There is an alternative explana-tion, however. It may be that a clear self-definition in terms of one set of adjectives oranother is not a result of past behavior whichhas been categorized or labeled in this way,but rather a function of general cognitivedifferentiation or articulation. Independentsand Dependents might be individuals whogenerally prefer to have things compart-mentalized along a number of different di-mensions. Aschematics, in contrast, may haveglobal or undifferentiated cognitive styles. Adirect test of this possibility cannot be madegiven the present data. However, as an indi-rect test, an index which reflected the articu-lation of the schema for each subject wascalculated on the basis of the number of inde-pendent and dependent words judged me andnot me. The measure was the Kendall Tb, andit reflected a subject's departure from thestandard of responding me to all 15 indepen-dent words and not me to all 15 dependentwords. The closer the value to 1 or —1, themore clearly the subject defines herself onthis dimension. This measure was also calcu-lated for each subject on the basis of her meresponses to the 30 creative/noncreative ad-jectives which were included in the list pre-sented to subjects. A cognitive style expla-nation would predict that subjects with highTb values (either positive or negative) on theindependent/dependent adjectives would alsobe the subjects with high absolute values ofrb on the creative/noncreative words, reflect-ing a general tendency toward differentiationor articulation. In fact, there was no associa-tion between the sets of rb absolute values forIndependents (r = .00), Dependents (r =—.07), or Aschematics (r = —.14).
The fact that subjects with schemata onindependence-dependence do not necessarilyhave schemata on creativity-noncreativity
SELF-SCHEMATA 71
and vice versa indicates that differences inself-categorizations may be the result of anindividual's behavior and its subsequent cog-nitive interpretation rather than a reflectionof differences in the general complexity ofcognitive structure. It is also consistent withthe idea that individuals develop schemataon dimensions they choose to attend to anddo not develop schemata on others.
Task 2: Supplying behavioral evidence forself-description. In this task, it was hy-pothesized that if one has a schema which isa reflection of past behavior, then one shouldbe readily able to provide specific behavioralevidence related to it; that is, to generatespecific instances of behavior which werelabeled or categorized by oneself or others ina particular way.
This task appeared to be generally mean-ingful to subjects and they performed it withlittle difficulty. For example, the dependentadjective conforming elicited responses suchas "I didn't go to any of the rallies aboutthe teaching-fellow strike because my friendsdidn't" or "I watched a television show Icouldn't stand last night just to save ahassle with my roommates" or "I pierced myears because all my friends did."
Independent subjects wrote more behav-ioral descriptions for independent words thandid either of the other two groups of subjects.The differences in the mean number of be-havioral examples written for each indepen-dent adjective for the three groups were sig-nificant, F(2,45) = 4 . 9 1 , p< .005. Acrossall of the seven independent adjectives, theIndependent subjects supplied almost onespecific example of behavior for each word(.93 behavioral descriptions per adjective)compared to .56 for the Aschematics and .36for the Dependent subjects. A significantopposite pattern occurred for the dependentwords, ^(2,45) = 3.59, p < .05.
Across all the adjectives, the subjects withschemata (the Dependents and the Indepen-dents) and the Aschematics did not differ inthe average number of words that werejudged as self-descriptive, but they did differin the average number of behavioral descrip-tions that were written for each word, t(A6)= 1.78, p < .05,° with the Aschematics sup-Plying somewhat fewer examples of behaviorthan ^rfyects with schemata. This result is
consistent with the expectation that individu-als without schemata on this dimension prob-ably have not used many independent or de-pendent adjectives to label their behavior,and thus it should be more difficult for themto supply specific behavioral descriptions.
A more detailed analysis of the individualwords revealed that the smaller number ofbehavioral examples supplied by the Asche-matics was primarily the result of these sub-jects producing fewer examples for the fournegatively rated adjectives. It is interestingin this respect that for the Aschematics therewas a substantial relationship between thepercentage of individuals judging a word asself-descriptive and the positivity of the word(r = .53, p < .05). This relationship was notevident for the other two groups of subjects(Dependents, r = .11; Independents, r —.21). As a group, then, the Aschematics ap-pear to be relatively more affected by thepositivity of the adjective and may use thisattribute to decide whether a particular wordis self-descriptive. Independent and Depen-dent subjects are relatively more willing touse negative labels for their behavior. It maybe that an individual with a schema abouther behavior on a particular dimension isaware of both the positive and negative as-pects of it and has evidence for both.
This task employed 14 of the same adjec-tives used in Task 1 and thus it is possibleto evaluate the consistency of self-descriptiveresponses for the three groups of subjects.For the Independents, the correlation be-tween me responses on the two tasks was .64(p < .01), for the Dependents .56 (p < .05),and for the Aschematics it was .20. Of the 14adjectives employed in this task, independentsubjects exhibited the shortest processingtime for the words independent, self-confi-dent, and cooperative in Task 1. In the pres-ent task, these were the 3 words that weremost likely to be judged self-descriptive byIndependent subjects, and in addition, thesewere the 3 words that elicited the largestnumber of behavioral descriptions. Dependentsubjects responded fastest to the adjectives
6 One-tailed test of the hypothesis that individualswith schemata are able to provide more specificbehavioral evidence for their self-judgments thanAschematics.
72 HAZEL MARKUS
cooperative, cautious, and moderate in Task1. Again, in this task these were the wordsthat were the most likely to be judged self-descriptive by the Dependent subjects andalso the words for which they gave thegreatest number of behavioral descriptions.This pattern, however, was not found for theAschematic subjects; the adjectives requiringthe least processing times were not thosemost likely to be judged self-descriptive. Thelack of congruence between these two judg-ments suggests either that Aschematics werefluctuating from Task 1 to Task 2 on whichadjectives they considered to be self-descrip-tive or that different types of considerationswere mediating the two judgments. For theIndependents and the Dependents, the twojudgments appear to be mediated by a simi-lar type of consideration, presumably whetheror not they have previously characterized orlabeled their own behavior in this way.
Task 3: Predicting the likelihood of be-havior. In this task, it was expected thatpeople with self-schemata on the independ-ence-dependence dimension would assigneither relatively higher or lower probabilitiesto independent and dependent behaviors thanindividuals who do not have a self-schemaon this dimension of behavior. It was assumedthat subjects with self-schemata would berelatively more aware or more certain of whatbehaviors would be elicited from them in thesesituations and could make more confident pre-dictions of their behavior.
Dependent subjects assigned a significantlyhigher likelihood to dependent behaviors thanto independent behaviors, t(l5) — 3.30, p <.01. In contrast, Independent subjects as-signed a reliably higher likelihood to inde-pendent behaviors than they did to dependentbehaviors, /(IS) = 3.31, p < .01. For theAschematics, however, there was no differ-ence between the likelihood assigned to inde-pendent behaviors and the likelihood assigneddependent behaviors, t(l5) < 1.
It was evident that the Independent andthe Dependent subjects differ in the actionsthey think likely of themselves, although someitems were better than others in differentiat-ing among the groups. The average subjectivelikelihood assigned to the dependent behaviorswas 35.7 for Independent subjects, 45.9 for
the Aschematic subjects, and 54.4 for De-pendent subjects, F(2, 45) = 5.57, p < .01.The average likelihood assigned to the inde-pendent behaviors was 53.8 for Independentsubjects, 45.7 for Aschematic subjects, and37.1 for Dependent subjects, F(2, 45) = 7.40,p < .001.
Overall, the Independent and the Depend-ent subjects differ markedly from the Asche-matic subjects. The former two groups arerelatively polarized in their estimations of theprobabilities of these behaviors occurring, in-dicating that they are more certain aboutwhat types of behavior might be character-istic of them in particular settings. For theindependent behaviors, the average deviationfrom the mean likelihood rating was 7.60 forIndependents, and —9.12 for Dependents. Forthe dependent behaviors the average deviationfrom the mean likelihood was —9.63 for theIndependents and 8.99 for the Dependents.Aschematics, however, do not show this polar-ization in their judgments. For the independ-ent behaviors their average deviation was .45,for dependent behaviors it was —.05. Forthese individuals there appears to be littledifference in the subjective likelihood of in-dependent and dependent behaviors; they areequally likely to occur or not to occur. Recallagain in this respect that the subjects wereconstrained to think of a fairly specific con-text. These data suggest, therefore, that theAschematics have no articulated cognitivegeneralizations or self-schemata along thedimension of independence-dependence.
One possible alternative explanation for thelack of difference in the mean ratings ofindependent and dependent behaviors forAschematics is that Aschematics are really aheterogeneous collection composed of approx-imately equal numbers of subjects respondinglike Independents and Dependents. However,inspection of the distributions of likelihoodratings yielded no evidence to support thispossibility. The distributions of likelihood rat-ings for all three groups of subjects with re-spect to both independent and dependent be-haviors are unimodal and fairly symmetricabout the mean.
The consistent pattern of responses ob-served for the Independents and Dependentsacross these diverse tasks argues for the exist-
SELF-SCHEMATA 73
ence of a prevailing self-schema which facil-itates the processing of social information.Individuals assumed to have schemata clearlyperformed differently on these tasks than didindividuals assumed not to have schemata.
Aschematics did not discriminate among theindependent and dependent stimuli on any ofthe tasks. It seems that for these individuals,independence-dependence was not a meaning-ful dimension of behavior; that is, they didnot categorize or make distinctions on thebasis of the independence or dependence oftheir actions.
The pattern of findings describing theAschematics clearly indicates why self-charac-terizations such as adjective self-descriptionsmay often be misleading as indicators of fu-ture behavior. For those individuals who hada self-schema about their independence or de-pendence, the responses to the self-categoriza-tion task were highly consistent with their re-sponses to the other two tasks and would mostlikely be consistent with behavior along thisdimension. The responses of the Aschematicsto the self-categorization task, however, couldnot be generalized even to the other self-description tasks. It would be surprising,therefore, if these responses were consistentwith observable behavior.
Study 2
This study focuses on how self-schematamay produce differences in the selection andinterpretation of information about the self.If self-schemata produce differences in judg-ments about the self, in description of pastbehavior, and in prediction of future behavior,it is reasonable that they should also producedifferences in interpreting new informationabout the self. Thus, it is hypothesized that aself-schema along a particular dimension ofbehavior will make an individual resistant tocounterschematic information about his or herbehavior. This study employed the same in-dividuals who took part in Study 1, providedthem with information about themselves (arigged score on a test of suggestibility) whichwas incongruent with their assumed schema,and then evaluated the impact of this in-formation. Subjects with definite schematashould be unwilling to accept or believe the
incongruent information produced by thismeasure, while Aschematic subjects shouldfind the test results relatively more credible.Following the suggestibility test and a subse-quent questionnaire, content and latencies ofself-descriptions (Task 1 from Study 1) wereagain assessed.
Method
Subjects
Forty-seven of the 48 subjects who participated inExperiment 1 also participated in this experiment.
Materials and Procedures
Three weeks after completion of the laboratorysession in Study 1, subjects were scheduled (or asecond experiment. They were given the QPAT Sug-gestibility Test, a fictitious test prepared especially sothat the experimenter could provide feedback thatwas incongruent with the subject's self-schema. Inde-pendents were given information that stated theywere not independent at all but rather very sug-gestible and good followers. Dependents were giveninformation indicating that they were very inde-pendent and not at all suggestible The Aschematicswere randomly assigned to one feedback conditionor another.
Upon arriving at the laboratory, subjects weretold:
Before beginning today's experiment, I wonder ifyou would like to take part in a Psychology De-partment project. They are in the process ofvalidating a new test . . . that involves measuringphysiological changes by attaching these [elec-trodes] to your fingers.
After subjects agreed to take part, electrodes wereattached to two of their fingers. Standard galvanicskin response apparatus was in clear view. Subjectswere then given the QPAT Suggestibility Test In thistask subjects were asked to relax and then to performa number of simple movements such as standing upand squeezing hands together In addition, they wereasked to imagine a number of specific scenes andwere asked questions about these images. The testtook about 10 minutes. The experimenter then re-moved the electrodes and ostensibly calculated thesubject's score. The subject was told:
This is a test of suggestibility in everyday life.It has been found to be a quick and reasonablyaccurate way to measure suggestibility. Yourphysiological data, the responses you made, as wellas the time it took you to answer, indicate thatyou are highly suggestible (not at all suggestible).. . . This sheet will tell you what this score
74 HAZEL MARKUS
Subjects were then handed one of the two descrip-tions, depending on their experimental condition. In-dependents and the Aschematics assigned to thiscondition were given the following description:
For individuals who score SS-6S: Individualswho receive a score in this range are generallyquite susceptible to social influence. They are quitelikely to be influenced by others Because they areopen-minded and receptive to the suggestions ofothers . . . they are generally good team mem-bers. . . Individuals in this range are not usuallyconcerned with making their own point or standingup for a particular issue . . . This score also sug-gests that one is quite sensitive and able to seethings from the other person's point of view.
Dependents and the Aschematics assigned to thiscondition were given a description that was verysimilar in form, but written to describe an individualwho is not suggestible and not likely to be influencedby the ideas of others.7
None of the words used in these descriptions werethe same as words used in any of the tasks in Study1, although suggestibility is clearly related to thegeneral independence-dependence dimension. Aftersubjects finished reading their description, acceptanceof the incongruent information was assessed bymeans of a brief questionnaire. Following the com-pletion of this questionnaire, subjects were asked toput their questionnaire in a campus mail envelopeaddressed to the Psychology Department and weregiven a name and number to call in case they wantedto find out more about the tests
Following this procedure, subjects were told thatthe experiment they were called for would begin.Subjects were given Task 1 (content and latency ofself-description) from Study 1 again. The task wasidentical using the same 69 adjectives and requiringa subject to respond either me or not me Each sub-ject was randomly assigned to one of the four ordersof presentation, with the exception that no subjectreceived the same order she received in Study 1.
Results
In the analysis of these results individualsassumed to have schemata on the independ-ence-dependence dimension were comparedwith individuals assumed not to have schemataon this dimension.
Suggestibility test. Reactions to the sug-gestibility test support the expectation thatindividuals with schemata are less willing thanAschematics to accept incongruent or counter-schematic information as self-diagnostic.When asked, "How accurately does this testdescribe you?" Aschematics felt that the sug-gestibility test described them more accuratelythan did the other two groups of subjects,
*(45) = 2.11, p < .05. In addition, a largerproportion of Independents and Dependentsindicated some disagreement or disbeliefabout their score than did Aschematics, t(45)= 2.11, p < .05. More of the Aschematicswere also willing to take the test again, al-though this difference was not reliable.
The degree to which subjects accepted testfeedback is indicated by how distant from theneutral point of the 11-point scale of sug-gestibility they placed themselves followingthe manipulation. The Independents and theDependents placed themselves on the average2.17 points away from the neutral point in thedirection indicated by the suggestibility in-formation they received. The Aschematics,however, placed themselves 2.83 points awayfrom the neutral point in the direction indi-cated by the suggestibility information, £(45)= 1.86, p < .05.8
Self-description task. For each of thethree groups, there were no significant differ-ences between number of endorsements attimei and endorsements at time2 (which im-mediately followed the suggestibility test) onany of the four judgments (we-dependentwords, not we-dependent words, we-inde-pendent words, not w»e-independent words).Aschematics, however, exhibited relativelymore inconsistency in their pattern of me/notme responses than did the Independents andDependents. A correlation between the Ken-dall Tb for each subject at timei and time2 forthe pooled Independents and Dependents was
7 The two descriptions were written with the in-tention of making the independent and the dependentfeedback convey different types of information, butgive equally favorable overall impressions. To testfor this, the two descriptions were given to femalestudents in several introductory psychology classesnot participating in this experiment. These studentswere asked to rate the two descriptions for "Howmuch would you like an individual who fits thisdescription?" on a 7-point Like-Dislike scale. Halfof the respondents received one description first andthe other half received the alternative description.The Dependent, or suggestible description, receivedan average rating of 4.55 and the Independent, ornot suggestible description, received an average rat-ing of 4.45, a nonsignificant difference.
8 One-tailed test of the hypothesis that Aschemat-ics are relatively more accepting of the informationprovided by the suggestibility test.
SELF-SCHEMATA 75
.82 compared to .41 for the Aschematics.While the overall number of independent anddependent words that were endorsed by theAschematics as a group did not change, therewas substantially less stability in their re-sponses over time. This type of inconsistencywas also noted in the Aschematics between thetwo self-characterizations tasks of Study 1.
As for processing times, the pattern is verysimilar to that shown in Figure 1, with De-pendents again responding faster to dependentwords than to independent words, 2(15) =3.53, p < .01, and Independents again re-sponding faster to independent words thanto dependent words, t(l5) = 2.61, p < .05.Aschematics again showed no differences inprocessing times between the two types ofjudgments. Subjects with schemata (Inde-pendents and Dependents) however, showedsignificantly longer average latencies for theseself-judgments on this second task than onthe first, t(3l) - 2.05, p < .05. The Asche-matics, however, did not evidence longerprocessing times for these judgments on thesecond task, t(lS) < 1.
The longer response latencies at time2 forthe self-judgments of individuals with sche-mata may have been the result of these sub-jects trying to be consistent from timei totime2. This, however, should also have beenthe case for the control adjectives, the creativ-ity/noncreativity words, but these judgmentsdid not evidence longer latencies in the secondself-categorization task; in fact, they gen-erally became shorter. It appears, then, thatthe longer processing times for the inde-pendent/dependent words may well have beendue to the counterschematic information pro-vided by the suggestibility test. Subjects withschemata appear to realize that they have re-ceived information about themselves that doesnot fit with their current self-conception onthis dimension. While this realization is notsufficient to warrant a change in self-charac-terization (and thus no change in adjectiveendorsement is observed), it probably causedthese subjects to reflect slightly longer tocheck this information against their schematabefore making a judgment.
Like the Independents and the Dependents,the Aschematics did not change their self-characterizations in the direction of the
feedback. However, unlike subjects withschemata, their processing times for the inde-pendent/dependent words remained unchangedfollowing the suggestibility information. Themost plausible explanation for this finding isthat individuals without schemata on a par-ticular dimension of behavior may indeed notsee the relationship or the convergence amongthe suggestibility information and independ-ent/dependent adjectives used in the self-description task. For Aschematics the sug-gestibility test score may have been just anisolated fact about a noncentral dimension. Itshould be recalled here that the suggestibilitytest feedback was carefully worded to avoidany of the words used in this self-categoriza-tion task. Thus, telling someone that she isvery suggestible may not imply anything atall about whether or not she is conformingor dependent, if she does not categorize herbehavior along an independence-dependencedimension. It should, and in this study did,influence her assessment of her relative sug-gestibility, however. Generalizations acrossvarious chunks of information about the selfmay only occur if individuals possess schematarelating them. Because the Aschematics didnot have an integrated picture of themselveson the independence-dependence dimension,the suggestibility information was not per-ceived as relevant to the judgments beingmade. The self-judgments and latencies forthese judgments were, therefore, not affected.
General Discussion
These studies provide converging evidencefor the concept of self-schemata, or cognitivegeneralizations about the self, which organize,summarize, and explain behavior along a par-ticular dimension. Systematic diSerences wereobserved among the three groups of subjectsin the selection and processing of informationabout the self. Thus, the group of individualswho thought of themselves as "independent"endorsed significantly more adjectives asso-ciated with the concept of independence thandid individuals who did not characterize them-selves this way. Additionally, these people ie-quired shorter processing times for me judg-ments to words concerned with independencethan to other types of words, were able to
76 HAZEL MARKUS
supply jaflafiwrify imore specific examples ofindependent Ibdhavior, thought they werelikely to **"*g»ffp iin future independent be-havior., and \wme iBeaistant to the acceptanceof infanmaiifln ifhat implied they were not in-dependent. A (parallel pattern of results wasfound TOfli ytojaenflenl stimuli for those indi-viduals who tfihoHght of themselves as "de-pendent*1 JWHiJEflfi.
In dear .™mf-.raBt to the Independents andDependents went rtflie Aschematics who didnot differ in tihek pnocessing times for inde-pendent and depemlBErt words, had relativelygreater dS&oalJty im pomading behavioral evi-dence of "MflfpmaflprtTy and dependence,thought they were as likely to engage in inde-pendent as dependent behavior, and wererelatively accepting of information aboutthemselves on rt»fc dimension. Aschematicsdid not appear to view themselves along anindependence-dependence dimension at all.
There are, of course, a number of alternativeexplanations for some of these findings whichare taken as evidence for the self-schema con-cept. One of the major findings of these studieswas that the group of Aschematics did notrespond differentially to the independent anddependent stimuli. This lack of selectivity wasinterpreted as indicating that these individ-uals do not categorize their behavior accord-ing to its independence or dependence andthat this is not a meaningful dimension forthem. It could be argued, however, that thesepeople are not really aschematic with regardto independence and dependence, but insteadhave a schema about themselves which en-compasses elements of both independent anddependent behavior. Aschematics may simplyhave monitored their behavior (Snyder, 1974)on the independence-dependence dimensionand have found a mixed pattern. Some of theAschematics, for example, may have felt thatthey were independent from social influencebut not at all aggressive or assertive. It mightbe argued, then, that there are no substantialdifferences between subjects with schemataand those without schemata in the types ofcognitive structures that mediate informationabout the self. Both groups may simply beoffering accurate reports of their behavior inthe independence-dependence domain.
A number of results, however, argue againstthis alternative. First, if the underlying cogni-tive structures were equally articulated amongthe groups, the same consistency among thetasks should be found for the Aschematicsthat was found for the Independents and De-pendents. Yet, considerable inconsistency wasexhibited by the Aschematics between the twoself-description tasks of Study 1, as well asbetween the self-categorization task in Study1 and its identical counterpart in Study 2.Second, if there were no differences in thearticulation and precision of the underlyingcognitive structures between Schematic andAschematic subjects, there should have beenno differences in their readiness to make pre-dictions about their future behavior in thedomain of independence-dependence. All sub-jects would be expected, under these circum-stances, to be equally polarized in the likeli-hood they assigned to future behaviors, eventhough they may have differed in the specificbehaviors they considered likely. For example,a subject who monitors her past behavior andrecalls that she has always been quite activein group discussions and shows no hesitationin giving advice and opinions, but generallystops short of starting an argument to defenda point, may see herself in the middle of theindependence-dependence dimension. Never-theless, this subject, when asked, should bequite ready to predict the manner of her fu-ture participation in group discussion. Thepresent results on behavioral predictions, how-ever, clearly indicate that Aschematics do notassign extreme likelihoods to any of the be-haviors and report that they are as likely toengage in dependent as in independent be-havior. It would appear, therefore, that sub-jects who have been categorized as Asche-matics do not have clear and precise cognitivestructures about the self in the domain of in-dependence, for if they did, these structureswould allow them to generate relatively un-ambiguous judgments about their futurebehavior.
There is, of course, the more remote pos-sibility that Aschematics are individuals whoare truly inconsistent in their behavior, suchthat one day they may be timid and shy ina discussion and the next day surly and ag-
SELF-SCHEMATA 77
gressive. In this case, inconsistency in self-description from one time to another might beexpected if individuals think about themselvesas highly variable or inconsistent in a par-ticular domain. But one should also expectlonger latencies for self-judgments, reflectinguncertainty or conflict over which trait labelsare most appropriate, or at least more vari-ance in judgments. None of these results wasfound, however.
The results of these studies have a numberof implications for research on personality andthe search for cross-situational consistency inbehavior. It seems likely that those indi-viduals who have schemata about themselveson a particular behavioral dimension are thosemost likely to display a correspondence be-tween self-description and behavior and toexhibit cross-situational consistency on thatdimension. In contrast, those individuals whohave no clear schema about themselves areunlikely to exhibit such consistency in be-havior. For example, Bern and Allen (1974)were able to identify a group of individualswho displayed substantial cross-situationalconsistency in behavior. Individuals whoidentified themselves as consistent on a par-ticular trait dimension (claiming that theydid not vary on this dimension across situa-tions) exhibited substantial correspondencebetween self-description and behavior andwere also cross-situationally consistent in theirbehavior. The subjects in Bern and Allen'sstudies who claimed they were consistent ona particular dimension may have been ac-knowledging a self-schema on this dimension.And the cross-situational consistency that wassubsequently noted for these consistent indi-viduals may have resulted from these indi-viduals engaging in behavior that was mo-tivated from a desire to be consistent withtheir self-schemata. Differences in the waysindividuals generalize and interpret their ownbehavior may thus be a significant source ofindividual differences, and such differencesmay actually predict differential behavior.
These studies give empirical substance tothe idea that not all people have a subjectiveposition on every dimension of behavior. It isentirely possible that systematic effects insocial behavior depend less on people havingsome amount of a particular substantive at-
tribute, such as independence or dependence,and more on the readiness or ability to cate-gorize behavior along certaini dimensions.This, of course, was the notion behind Kelly's(1955) methodology of allowing the indi-vidual to generate his own constructs- for cate-gorizing himself and the social environmentand is pertinent to Bern and Allen's- (1974)call for an idiographic approach to personal-ity. If the dimensions under study are notthe ones an individual characteristically at-tends to, he cannot be expected to. make cor-responding conceptual and behavioral dis-criminations along that dimension.
The concept of self-schemata that functionas selective mechanisms resulting in differen-tial attention and processing of informationabout the self also has implications for self-perception and attribution theory. In thestudy of social attribution, there are a numberof well-known studies (e.g., Davison & Valins,1969; Schachter & Wheeler, 1962; Valins &Ray, 1967; Weick, 1967; Zimbardo, 1969)that reveal behavior changes in the absenceof subsequent change in self-categorizations(how one says he thinks, feels, or is). Thesestudies are discrepant with the model of self-attribution that assumes that inferences aboutinternal states, dispositions, or attitudes followfrom behavior and thus have been somethingof a puzzle for attribution theorists. Valinsand Ray (1967), for example, gave snakephobic subjects false feedback about theirfear of snakes. These subjects were able to ap-proach a snake more closely than controlswithout false feedback. They did not, how-ever, report themselves as any less fearfulas a result of their experience. In view of thepresent results, this finding is perhaps notsurprising. A correspondence between self-categorization and overt behavior depends onthe mediating self-schemata. For example,snake phobic individuals probably have afairly well-articulated self-schema about howthey feel toward snakes. The elements of thisgeneralization include cognitive representa-tions of specific encounters with snakes and thesubsequent evaluation of these encounters.If such an individual is induced to handlesnakes, it does not follow that he will im-mediately perceive himself as no longer afraidof snakes. This type of dispositional attribu-
78 HAZEL MARKUS
tions would be (a) based on only one isolatedexperience and (b) contradict a well-estab-lished schema. However, individuals with noparticular generalizations about their attitudestowards snakes may readily interpret theirpositive experience with a snake as unam-biguously self-diagnostic—as indicative thatthey don't mind snakes and are not afraid ofthem. If this argument is correct, it impliesthat individuals will use information abouttheir own behavior to make an inference abouttheir own internal state, disposition, or atti-tude only when the behavior appears to theindividual to be related to the self-charac-terization to be made and when it does notrun counter to a prevailing generalizationabout the self. Within an attribution theoryframework, self-schemata can be viewed asimplicit theories used by individuals to makesense of their own past behavior and to directthe course of future behavior. And a givenchunk of behavioral information has decidedlydifferent self-attribution consequences for anindividual with an implicit theory or self-schema than it does for one without such aschema.
Reference Note
1. Abelson, R. P. Script processing in attitude forma-tion and decision making. Paper presented at theEleventh Carnegie-Mellon Symposium on Cogni-tion, April 24-26, 197S.
References
Anderson, J. R., & Bower, G. H Human associativememory. Washington: V. H. Winston, 1973.
Anderson, N. H. Likeableness ratings of 555 per-sonality-trait words. Journal of Personality andSocial Psychology, 1968, P, 272-279.
Atkinson, R. C, & Shiffrin, R. M. Human memory:A proposed system and its control processes. In K.W. Spence & J. T. Spence (Eds ), Advances in thepsychology of learning and motivation research andtheory (Vol. 2). New York: Academic Press, 1968.
Bartlett, F. C. Remembering. London: CambridgeUniversity Press, 1932.
Bern, D. J. Self-perception. An alternative interpreta-tion of cognitive dissonance phenomena. Psycho-logical Review, 1967, 74, 183-200.
Bern, D. J. Self-perception theory. In L. Berkowitz(Ed.), Advances in experimental social psychology(Vol. 6). New York: Academic Press, 1972.
Bern, D. J., & Allen, A. On predicting some of thepeople some of the time: The search for cross-situational consistencies in behavior. PsychologicalReview, 1974, SI, 506-520.
Bobrow, D. G., & Norman, D. A. Some principles ofmemory schemata In D. G. Bobrow & A Collins(Eds.), Representation and understanding Studiesin cognitive science. New York: Academic Press,1975.
Carroll, J B., Davies, P., & Richman, B. Word fre-quency book. New York: American Heritage, 1971.
Davison, G. C, & Valins, S. Maintenance of self-attributed and drug-attributed behavior change.Journal of Personality and Social Psychology, 1969,11, 25-33.
Erdelyi, M H. A new look at the New Look: Per-ceptual defense and vigilance. Psychological Re-view, 1974, SI, 1-25.
Gough, H. G & Heilbrun, A B. The Adjective CheckList Manual. Palo Alto, Calif : Consulting Psy-chologists Press, 1965.
Kelley, H. H. Causal schemata and the attributionprocess. In E. E. Jones, D. E Kanouse, H. H.Kelley, R. E Nisbett, S Valins, & B. Weiner(Eds) , Attribution. Perceiving the causes of be-havior. New York: General Learning Press, 1972.
Kelly, G A. The psychology of personal constructs.New York: Norton, 1955.
Minsky, M A framework for representing knowledge.In P Winston (Ed ), The psychology of computervision. New York: McGraw-Hill, 1975.
Mischel, W., Ebbesen, E. B , & Zeiss, A. M. Deter-minants of selective memory about the self. Jour-nal of Consulting and Clinical Psychology, 1976,44, 92-103.
Piaget, J. The child's conception of the world. NewYork- Humanities Press, 1951.
Schachter, S , & Wheeler, L. Epinephrine, chlorpro-mazine, and amusement. Journal of Abnormal andSocial Psychology, 1962, 65, 212-218.
Snyder, M. Self-monitonng of expressive behavior.Journal of Personality and Social Psychology, 1974,JO, 526-537.
Snyder, M., & Monson, T. C. Persons, situations,and the control of social behavior. Journal of Per-sonality and Social Psychology, 1975, 32, 637-644.
Stotland, E., & Canon, L. K. Social psychology: Acognitive approach. Philadelphia: Saunders, 1972.
Tesser, A , & Conlee, M. C. Some effects of time andthought on attitude polarization. Journal of Per-sonality and Social Psychology, 1975, 31, 262-270.
Valins, S., & Ray, A. A Effects of cognitive de-sensitization of avoidance behavior. Journal ofPersonality and Social Psychology, 1967, 7, 345-350.
Weick, K. E. Dissonance and task enhancement: Aproblem for compensation theory? OrganizationalBehavior and Human Performance, 1967, 2, 175-216.
Zimbardo, P. G. (Ed.), The cognitive control ofmotivation. Glenview, HI.: Scott, Foresman, 1969.
