JOURNAL OF EXPERIMENTAL SOCIAL PSYCHOLOGY 17, 80-95 (1981) Measurement and Interpretation of Situational and Dispositional Attributions FREDERICK D. MILLER ELIOT R. SMITH AND JAMES ULEMAN Received October IY, 1979 Four problems with the measurement of situational and dispositional causality are reviewed. These are: the assumption that dispositional and situational causality are inversely linked: the diversity of the causes considered within the situational and dispositional categories; the difficulties of differentiating between causes internal and external to the actor; and the low convergent validity of various closed-ended attribution measures. A study reaffirms the lack of convergence among closed-ended measures and between closed and open-ended measures as well. In a second study, subjects’ ratings of closed-ended attributions are taken as indicators that a freely chosen to not freely chosen dimension may better represent subjects’ attributional thought than the internal/external dimension does. Open-ended data from the convergence study are recoded using this scheme and achieve significantly better convergence with closed-ended data. The implications of this redefinition for solving the various measurement problems are discussed. Fritz Heider’s (19%) description of the naive analysis of action distinguishes personal forces from environmental forces, a distinction that has We wish to thank Nancy Greenberg and Jill Salberg of New York University for their assistance in coding open-ended materials as well as Shelley E. Taylor and Kay Deaux. Richard J. Harris. and Edward E. Jones for their comments on earlier drafts of this paper. Correspondence regarding this paper should be sent to Frederick D. Miller. IC-329C Bell Telephone Laboratories. Whippany. NJ 07981. 80 0022-1031/81/010080-16$02.00/O Copyright a 1981 by Academic Press. Inc. All rights of reproduction in any form resewed. MEASUREMENTOFATTRIBUTIONS 81 been explored in the many studies of causal attribution that ask subjects to rate the importance of situational and dispositional causes of events. In recent years, such measures have been widely criticized for their lack of convergent and construct validity, and for failing to adequately represent what people express in causal attributions. Yet little effort has been directed toward understanding what people actually do when filling out causal measures or toward providing a more adequate rationale for measurement strategies. Heider’s distinction reprises the two independent variables of Lewin’s classic B = f(P,E) equation and Murray‘s analysis of the interplay of needs and press. This distinction has had a great influence on attribution research. Consider, for example. Jones and Nisbett’s (1972) actorobserver hypothesis. They proposed that actors display a bias toward using situational explanations of behaviors that observers see as dispositionally caused. In tests of this hypothesis (Nisbett, Caputo, Legant, & Maracek, 1973; Storms, 1973) the researchers’ definitions of dispositional and situational causes have identified dispositional causes with stable characteristics of the person-traits, attitudes, and desires-that are transsituational. Situational causes have been identified with stimuli that elicit situation-specific responses. Similar definitions of situational and dispositional causality have been used in much of social psychology, for example, in research on self defensive biases (Miller & Ross, 1975) or on the styles of behavior associated with different strategies of self monitoring (Snyder, 1979). A variety of difficulties arising from the situational/dispositional distinction have become apparent, particularly its applicability to such specific issues as intrinsic motivation (Kruglanski, 1975) and the difficulty of reliably validating hypotheses using situational/dispositional measures (cf. Taylor & Koivumaki, 1976, on the actor-observer hypothesis). These problems have led some researchers to question whether causal attribution can be adequately measured (Ross, 1977). An answer to this question requires an assessment of the validity of the available measures and a determination of what those measures indeed do assess. First, however, the problems with the situational/dispositional distinction will be reviewed. Four separate problems can be identified. The problems are as fol1ows: I. The Hydraulic Assumption: Heider’s original formulation proposes that situational and dispositional causes are yoked together inversely such than an increment in one necessarily accompanies a decrement in the other. This assumption is operationalized by single-scale measures of causality with their endpoints labeled “situational causes” and “dispositional causes.” However, the predicted negative relationship between situational and dispositional causation has not been found in data from subjects who respond to separate scales of situational and dispositional 82 MILLER,SMITH,ANDLJLEMAN causality (Taylor & Koivumaki, 1976). Solomon (1978) reviewed the hydraulic assumption and concluded that researchers should not use bipolar situational-dispositional scales or difference scores based on combining separate situational and dispositional scales, as Storms (1973) did. Solomon urged researchers to measure situational and dispositional causality on separate scales. 2. The Category Error: The categories defined by situational and dispositional causality are very broad. Dispositional causes-or causes internal to the person-may include stable traits and attitudes, unstable moods and feelings, and deliberate desires and choices. Situational causes encompass choices presented by the environment, coercion through social control, and stable environmental factors such as task difficulty. These two categories cut across dimensions that are of theoretical interest to social psychologists, such as responsibility and freedom. The use of such broad categories may confuse both subjects and researchers. as each tries to understand what sort of situational and dispositional cause the other had in mind. (See Cantor and Mischel(1979) for a discussion of the utility of categories of differing sizes.) The situational/dispositional distinction is attractive because it allows the use of ordinal scales of causes for which interval properties, and hence interval statistics, can be assumed. However, the categories of situational and dispositional causation may each be so broad that a single measure of dispositional or situational causes may be meaningless. This problem of category width has not been discussed before. Two solutions can be tentatively suggested. One is to ask subjects to choose causes from closed-ended categories with narrower definitions. Another is to gather open-ended attribution responses which researchers can then categorize in different nominal coding schemes as needed. 3. The Teleological Confusion: Kruglanski ( 1975) and Ross ( 1977) both identify a flaw in the distinction between situational and dispositional explanations of deliberate actions: the distinction breaks down whenever external causes are perceived and deliberately acted on by the person. The person is always the proximate cause of action, regardless of situational influences. Ross (1977) suggests that in such cases the distinction between dispositional and situational causes often rests only on the arbitrary phrasing of sentences. Kruglanski (1975) proposed that people understand deliberate actions as being caused endogenously (as ends in themselves) or exogenously (as means to further ends) rather than internally or externally. In a recent paper, Kruglanski. Hamel, Maides. and Schwartz (1978) proposed that each explanatory scheme will be applied when it will yield information that satisfies the person’s particular needs. A somewhat different position was advanced by Buss (1978) who argued that situational and dispositional “causal” attributions are distinct from “reason” attributions that explain the ends for which action occurs, and that the solution to the teleological confusion is to distinguish causal attributions from reason attributions. MEASUREMENT OF 83 ATTRIBUTIONS 4. The Convergent Validity Issue: A final methodological problem has been identified in studies that assess the convergent validity of various measures of situational and dispositional causality. In a recent paper, Herzberger and Clore (1979) used a multitrait-multimethod approach to show the lack of convergence among a variety of closed-ended attribution measures. This finding implies either that the scales are unreliable or that they are measuring different things. Herzberger and Clore gave qualified support to a two-scale measure. The present authors’ interest in attribution measurement was also stimulated by an initial interest in the issue of convergent validity among attribution measures. Below we will briefly report a study that demonstrates the poor convergent validity of a variety of attribution measures.’ Data from a further study will be presented to illuminate subjects’ understanding of situational and dispositional causality. A reanalysis of the convergent validity data will then be used to argue that situational and dispositional measures have meaning when they are understood in a fashion consistent with subjects’ naive psychology. STUDY 1: CONVERGENT VALIDITY OF EXPLANATIONS FOR HYPOTHETICAL EVENTS Uleman. Miller, Henken, Tsemberis, and Riley (in press) twice failed to demonstrate the actor-observer effect in between- and within-subjects versions of Storms’ (1973) study, using Storms’ two-scale attribution measure. During debriefing, many subjects reported that the explanations accompanying the dependent measures were confusing and that they felt uncertain about using the scales to express their causal judgments. Though many of the explanations they offered for behavior during debriefing interviews could be categorized as either situational or dispositional by the researchers, the subjects themselves had trouble doing so using the available scales. This led to two questions. Did the closed-ended attribution measures accurately reflect subjects’ causal thought. and did different measures elicit similar thinking about causality? These questions were evaluated in a study in which subjects responded to attribution cues on several measures, thus allowing comparisons of responses made in different formats. Method Each subject read four single-sentence descriptions of behaviors and Koivumaki ( 1976) study. The sentences described one friendly by oneself and by a friend. After reading each sentence. the different measures ’ We have describing conducted two the studies causes of the of the convergent selected from the Taylor and one unfriendly action subjects completed three action. validity issue besides the one reported here, one using similar and one using entirely different materials and subjects. A full description of their results is deemed unnecessary. as they support the findings on nonconvergence that are already in the literature (Solomon. 1978: Herzberger & Clore. 1979). 84 MILLER. The subjects University. administered followed these order ended were 99 undergraduates The experiment booklet in which by the three SMITH, dependent AND drawn was conducted each behavior measures, ULEMAN from a psychology in class. description each class at New Each subject completed occupied a single page on a page of its own. Within York a selfand was each booklet. three dependent measures appeared in the same order after each description. This was counterbalanced acres\ booklets, though only the 12 orders in which the openmeasuresappeared first or last were used. because of the limited number of subjects.’ The words, Storms three dependent measures were: ( I) an o~p~n-t~&d measure asking. “In your why would you (your friend) behave this way’?” (2) a d@wwr measure taken (1973). consisting of definitions and separate 9-point scales of dispositional situational causality:’ dispositional tional causes” whose attributions; and “wholly difference difference measure. Responses to the open-ended situational definitions was then measure were coded to wholly dispositional causality. The of situational and dispositional causality sures. Traits. were classified taken .X8. moods. volitions, as dispositional Disagreements represent situational prior causes. experiences. Situational on a 5-point regarding codmg scale running and characteristic patterns causes included the actions were versus “wholly of terms $ltuaas the from wholly coding was done so as to mimic the provided with the closed-ended mea- istlcs of other people and characteristics of the setting.’ judges. Interjudge agreement on the full sample of storable p was to and (3) a 9-point hi/~olor scale with ends labeled dispositional causes” and the same definitions own from and resolved of behavior or character- All responses were coded by two responses was 77%. Spearman’\ through discussion. The results indicated low convergent validity among the various measures of dispositional and situational causality. Correlations were computed among the dependent measures following each sentence. These were then combined across sentences using Fisher’s r-to-z transformation. Most strikingly, responses to the open-ended measure did not correlate significantly with any other measures. The only significant correlation was between the bipolar and difference scales, though that correlation (r = .61) indicated only a modest equivalence of the two measures. In addition, the situational and dispositional scales of the difference measure showed a weak positive correlation (Y = .12. n.s.1. replicating J A fourth measure, actor‘s future behavior discussed here. ” These definitions asking subjects IO rate the certainty with which was also included among the counterbalanced were: Different factors can influence a person‘s they could measures. predict the but is not behavior. factora are: (a)prrsorlu/ ckuruct~risric~.s: These include a person’s personality, traits. ter. personal style, attitudes, and mood, and (bl churucterisrics of t/w situution: include the task a person is involved in. the actions of other people. and the physical Two such characThese setting. q In scoring open-ended responses we emphasized the semantic difference that troubled Ross ( 1977). In order to be scored a dispositional response. the subject’s answer had to refer speciiically to the actor’s traits. moods. or volition. Causal statements that only Implied the actor’s presence were scored as situational. Of358 open-ended responses. I IS were entirely dispositional. 76 entirely situational. 35 largely situational. 92 evenly mixed . 24 largely dispositional. and 13 uncodable. In the subsequeni reanalysis of these data 342 open-ended responses mislaid. are reported because the raw data from four subjects (Ih responses) had been MEASUREMENT CORRELATIONS BETWEEN TABLE I CLOSED-ENDED ATTRIBUTIONS ATRIBUTIONS IN STUDY I Self/ friendly Open by bipolar Open by situational Open by dispositional Open by difference -.I3 scale scale score 85 OF ATTRIBUTIONS Self/ unfriendly Other/ friendly .I3 .I3 .oo .Ol .I? .09 - .?8 AND OPEN-ENDED Other/ unfriendly .05 -.O’ .I5 -.I .I8 p.14 r - c .05 .oo .23 -.24 .I4 -.I5 Taylor and Koivumaki’s findings of nonsupport for the hydraulic assumption. The correlations between open and closed-ended measures are summarized in Table 1. These weak effects are not due to characteristics of the distributions of scores on the various measures, all of which were reasonably normal in shape and had adequate variances. A second striking aspect of the responses to the different dependent measures was that they elicited differing patterns of attributions in response to the same cues. Two-way repeated measures analyses of variance using origin of the behavior (self/friend) and positivify of the behavior (friendly/unfriendly) as within-subject classifying factors were performed on each of the attribution measures. The open-ended responses provide support for the actor-observer effect, though examination of the cell means for this analysis indicates that this effect was due largely to a difference in explaining positive behaviors of the self and others. This produced a significant interaction between origin and positivity. The Storms difference scale also yielded a significant main effect supporting the Jones and Nisbett hypothesis, but no interaction. The bipolar scale yielded a significant positivity effect as reported by Taylor and Koivumaki. This confusion of effects is summarized in Table 2, where it can be seen that the dependent measures appear to produce different patterns of effects. though they were all completed by the same subjects responding to the same stimuli. While there is support in these data for both the actor-observer and positivity hypotheses, the support varies as a function of measurement strategy.5 5 Because the dependent analytic procedure testing Harris. 1977. p. XI). measures in these analyses were for a difference among patterns interval scales, no multivariate of means was appropriate (cf. 86 MILLER,SMITH. ULEMAN AND TABLE? EFFECTS OF WITHIN-SUBJECT ACTOR-OBSERVER INTERACTIONS Degrees ON THE of freedom” POSITIVE-NEGATIVE SCALES, (1.72) Difference (1.91) FACTORS AND 1 Positivity F = 33.180, p < .OOl F = 9.946. p = ,003 n.5. ! 1.86) STUDY Actor/ observer Open-ended Bipolar AND THREE F = Interaction n.s. F = 10.387, p = .002 n.s. n.s. 11.705. n.s. p < ,001 (i Degrees of freedom apply to all F statistics in the corresponding row of the table Discussion The results of this study support what is already widely suspected: that responses to a variety of measures of situational and dispositional causality display little convergence. The lack of relationship between closedended attributions and open-ended attributions scored according to attribution theorists’ definitions of situational and dispositional causality indicates that many subjects do not think about situational and dispositional causality as theorists expect. The important question that follows from this is what, then, are subjects doing in responding to attribution questions? It is conceivable that they are answering at random, but then one would not expect reliable findings in closed-ended measures, such as the positivity effect. Alternatively, subjects may share a definition of situational and dispositional causality that is different from researchers’ definitions of these concepts. Study 2 examines subjects’ understanding of dispositional and situational causality in the hope of illuminating the strategies they use to answer attributional questions. STUDY 2: SUBJECTS’ EVALUATIONS OF SITUATIONAL DISPOSITIONAL ATTRIBUTIONS AND Method Subjects closed-ended received a booklet containing attribution questions which four sentences, the subjects were each followed told had been by completed answered by a fellow student. The sentences described either a positive or a negative act by oneself and by a friend, with each of two positive and two negative sentence stems paired either with oneself and one’s friend in different booklets. Each sentence was followed by the usual definitions of situational and dispositional causality and the 9-point attribution measures. filled out either with a 9 on the situational scale and a 1 on the dispositional scale (high situational). a I on the situational scale and a 9 on the dispositional scale (high dispositional). or a 5 on both scales (mixed). Attribution responses were randomly assigned to sentences in booklets. After reading each pair of attribution measures, the subject was asked to fill out I I nine-point scales describing the attribution. The first five asked for ratings of the MEASUREMENT OF ATTRIBUTIONS 87 attributer-as honest, fair. intelligent, sympathetic. and likeable-qualities chosen because they seemed likely to depend upon the attributions themselves. The next five scales asked the subject what the attributor meant by the attribution: did the attributor regard the actor as responsible, free, having chosen the act. having enjoyed it. and likely to repeat it. Finally, the subject was asked whether the attribution was correct. Sixty undergaduate volunteers at New York University completed the questionnaire during class time. Because of the incomplete within-subjects design, the data were analyzed separately for each of the four types of sentence (self doing friendly act, self doing unfriendly act. etc.). Separate multivariate analyses of variance were run on the five scales describing the attributor and on the five scales describing the attribution, and a univariate analysis of variance was run on the accuracy judgment. all using a 3 x 2 (attribution by sentence stem) design. Results First, there were no significant effects on subjects’ estimates of the accuracy of the attributions. Different attributions were not seen as more or less accurate explanations of events. Similarly, the MANOVA results showed that different attributions were not seen as diagnostic regarding rater characteristics such as fairness or intelligence. Across 4 multivariate and 20 univariate F’s, no significant @ < .05) main effects or interactions with attribution were found. In contrast, all four multivariate F’s for the main effect of attribution on ratings of the attribution were significant. Fourteen of the twenty univariate F‘s were significant, and the direction of the means was the same in all 20. For these subjects, dispositional attributions implied that the rater believed that the actor chose the action, did so freely, enjoyed it, was responsible for it, and was likely to do it again. Subjects rated mixed and situational attributions as implying less freedom, choice, enjoyment, responsibility, and repetition, but did not differentiate clearly between these two attributions. These analyses are summarized in Table 3. Discussion These results imply that subjects define dispositional causality as denoting acts chosen freely by the actor, and situational causality as denoting acts for which choice and responsibility are limited, at least in this subset of behaviors where the subject is considering the source of an action and not seeking to attribute a particular ability or attitude. The behaviors under consideration are all behaviors that an actor can perform. The task of the attributor shifts to assessing try, in Heider’s terms. This interpretation is consistent with the definitions of internal and external causality used in intrinsic motivation studies (Lepper et al., 1973), where internal causation implies choosing the action, but is inconsistent with the definition in actor-observer research, such as Storms (1973), where dispositional causality can be identified with stable traits that determine behavior across situations. It should be recalled that one of Jones and Nisbett’s original explanations of the actor-observer effect was a motivational 88 .ANDULEMAN Y P MEASUREMENT OF ATTRIBUTIONS 89 desire to believe one’s own behavior to be free (and hence responsive to situational causes) while seeing other people as predictably guided by stable dispositions that allow little variation. The present data indicate that subjects believe the opposite. A checkmark on the dispositional end of a closed-ended measure appears to signify a freely chosen act, not the operation of a personality trait that rigidly determines behavior. This suggests an explanation for the poor convergence of open- and closed-ended measures discussed above. Our scoring of the open-ended measures, though consistent with the definitions in the instructions for the closed-ended scales, may have been inconsistent with the meaning of subjects’ closed-ended ratings. The relationship between open-ended and closed-ended scales should be stronger if the open-ended data were coded in a fashion consistent with subjects’ definitions. To test whether subjects identify dispositional causality with freely chosen action and situational causality with actions that are situationally constrained, the open-ended attributions from Study 1 were recoded using a new scoring system. Reanalysis of Study 1 The new coding scheme used three nominal categories: internal acts were those deliberately chosen by the actor because the actor enjoyed them or wished to attain some desirable effect; external acts were those resulting from coercive aspects of the situation, or causes which were not intentional or not the result of the immediate situation, such as actor’s mood. The remainder of the open-ended attributions were treated as unclear regarding whether an act was internally or externally caused. Out of 342 open-ended attributions in Study 1, 156 were “internal,” 119 were “external,” and 67 were not diagnostic. Interjudge agreement was found for 84% of a sample of ratings. The recoded open-ended attributions were correlated with the closedended bipolar, difference and individual situational and dispositional measures. Examination of the crosstabs of open- and closed-ended measures supported the notion that open-ended responses that were unclear about whether the act was free or not could be placed between internal/free and external/not free attributions to form a three-level attributional measure for which interval statistics could be used, as is common practice. This new coding was virtually orthogonal to the original open-ended coding (see Table 4). Coded according to this free-not free scheme, the openended attributions were strongly related to scores on the bipolar scale and significantly related to closed-ended attributions on the other three measures. All the individual correlations and Fisher’s r to I tests that estimate the combined summary correlations are included in Table 4. The correlations shown in Table 4, ranging from .25 to .60 (the summary column) are to be compared with those in Table 1. where using the original codes, the open-ended measure’s correlations with other scales 90 MILLER.SMITH.ANDULEMAN MEASUREMENT OF ATTRIBUTIONS 91 ranged from .OOto .15. In every instance, the new coding is more strongly related to the open-ended measures, 10 of the 16 differences being significant at the .05 level. Clearly the new coding scheme is much better at capturing the aspect of the subjects’ responses that is common to the closed-ended measures. When the recoded open-ended scores were used as dependent measures in a repeated measures analysis of variance examining the effects of origin and positivity on attributions, both actor-observer (F(1,65) = 10.82, p = .002) and positivity (F(1,65) = 27,59, p < .OOl) main effects emerged. Friendly actions were seen as more likely to be freely chosen and actions by oneself were seen as less constrained then actions by others. This latter effect is consistent with the notion that observers want to view others’ behavior as predictable, or at least as more predictable than their own. The two independent variables did not interact in this analysis. DISCUSSION The data here should encourage researchers who are interested in pursuing the measurement of causal attributions, if they attend to the caveat that subjects’ answers must be treated as describing what subjects mean and not what researchers would wish them to mean. Subjects’ use of closed-ended categories may be more influenced by the subjects’ understanding of the categories they denote than by researchers’ brief and, from the subjects’ viewpoint, sometimes muddled instructions regarding those categories. In this regard, we researchers find ourselves in a position analogous to that of trial judges who sometimes despair of jurors. unwillingness to abandon their own interpretations of law in the face of extensive instructions from the bench (Kalven & Zeisel, 1966). Yet, by attending to the actual content of subjects’ naive psychology, some order can be restored to our understanding of causal responses. How then to measure causal attributions? The data provided by Study 2 and by the reanalysis of the open-ended data from Study 1 support the notion that subjects’ responses to both bipolar and difference measures of causality are based in part on beliefs that dispositional causality implies that action was voluntary and deliberate while situational causality denotes action taken under conditions of less choice and greater constraint. This suggests that subjects’ responses to situational and dispositional causality scales be regarded as reflecting an underlying chosen-not chosen dimension. It also suggests that in future studies of causal attribution researchers may want to explicitly measure the degree to which acts are deliberately chosen by the actor, instead of due to internal or external causes. Closed-ended choice-no choice measures could be devised easily. 92 MILLER.SMITH.ANDULEMAN The association ofdispositional causality with choice has been explored in some recent attributional research on perceived freedom. Kruglanski and Cohen (1973) demonstrated that observers attribute more freedom to an actor whose behavior is congruent rather than incongruent with internal dispositions. Trope and Burnstein (1977) and Trope (1978) have also shown that ratings of freedom are dependent on perceptions of congruence with internal dispositions. The use of such a distinction sheds some new light on each of the four problems with situational and dispositional measures described above. Heider’s original distinction of forces internal and external to the person was subject to the teleological confirsion pointed out by Ross (1977)-that many situational influences are internally processed and hence neither truly internal nor external. This problem is avoided by a chosen-not chosen distinction. Internal causes are now identified with the deliberate and desired choices of a person rather than with the presence of a person as actor, which had the shortcoming of implying that all actions have a “dispositional” component. This suggests that subjects’ criteria for an internal cause appear somewhat more relaxed than Buss’ criteria for a reason attribution. The present criteria define any freely chosen act as internal regardless of whether a deliberate intent can be identified. Additionally, our data indicate that subjects make such reason/cause distinctions in responding to the dispositional/situational scales. This indicates that these latter are not purely causal scales, as Buss describes them. Subjects’ internal/choice category can encompass both endogenous and exogenous attributions (Kruglanski, 1975) as being freely chosen, which differs from Kruglanski’s suggestion that only endogenous actions are regarded as free. To argue that exogenous action-for example, delay of gratification-cannot be taken freely implies an extremely alienated concept of choice and volition. The substitution of a chosen-not chosen dimension also addresses the category width and hydruulic assrrmption problems. The newly proposed categories are narrower, especially dispositional (chosen) causality, which no longer encompasses both rigid traits and volition. The external (not chosen) dimension may require further specification as it still unites such disparate influences as coercion and moods. The absence of a negative relationship between dispositional and situational measures may be due to their imperfect realization of an underlying dimension (chosen-not chosen) that does operate hydraulically. Chosen and not chosen are opposites, but situational and dispositional causes can operate either in unison or opposition in ordinary language, and hence are unlikely always to be placed in opposition by subjects. The possibility that bipolar and difference measures are each imperfect measures of an underlying dimension may explain their only modest MEASUREMENTOFATTRIBUTIONS 93 convergence. Their convergent validity will remain a problem as long as they are used to stand for notions of actions originating inside or outside the person, which are inconsistent with subjects’ thought. The correlation between open-ended responses and the bipolar scale indicates that a single scale may be the best way to operationalize the chosen-not chosen dimension. The weaker relationship with the difference score may reflect the fact that many subjects checked the same scale point (often an endpoint) on both measures of the difference score. Such responses can reflect an artifactual confusion, as subjects who wish to respond that both causes were important can do so by checking the same point on each scale regardless of its scale value. Though the choice-no choice dimension has utility, it must be remembered that it has emerged here from research on a limited set of behaviors in college student samples. It is best to consider such a dimension as a worthwhile avenue to be explored in further research across a wider variety of behaviors. However, if we assume that the choice-no choice dimension has meaning for a majority of our subjects, when judging the causes of actions, what does this indicate concerning findings using situational-dispositional measures? Studies using situationaldispositional measures can be readily reinterpreted in the light of our results. The reliable positivity effect first reported by Taylor and Koivumaki (1976) indicates that subjects identify desirable acts as chosen by people, while identifying undesirable acts as less deliberate. This seems to hold for acts by oneself and by others. Such findings may reflect cultural norms. If college students hold moderately optimistic views such that they expect themselves and their friends to behave in a generally nice fashion (cf. Kanouse & Hansen, 1972). then it would not be surprising that they regard positive actions as freely chosen and unfriendly actions as somewhat less a matter of choice. Similarly, the fundamental attribution error described by Ross (1977)-a tendency to cite dispositions more than situations as causes of events-may reflect cultural beliefs and values that stress that individuals choose and are responsible for their actions, even in many instances where the choice is in fact limited. By citing dispositions, observers may be affirming the belief that people usually choose their own behavior, a widespread belief in our culture. CONCLUSION Our data suggest that while notions of internal and external causation as used by attribution theorists may have to be modified, they may have more utility and validity than Ross (1977) or Buss (1978) suggest. Social psychologists are often tremendously clever in constructing independent variables, but more attention needs to be paid to obtaining reliable and valid measures as well. A first step for any naive psychology must be to establish agreement on terms between professional and naive psycholo- 94 MILLER.SMLTH.AND ULEMAN gists. Closed-ended measures have to accurately reflect subjects’ phenomenology. In the early stages of research, open-ended measures may be preferable to establish the meaning of closed-ended measures, the latter being preferable for their ease of use once their validity is established. But using closed-ended measures without first validating them can lead to the sorts of problems Ross ( 1977) and Solomon (1978) insightfully pointed to. We hope that this paper has been a step forward in clarifying problems of measurement in one area of attribution, and a goad toward greater concern with dependent measures more generally. REFERENCES BUSS, A. R. Causes Personality and Cantor. N., and reasons Social & Mischel. in attribution Psychology, W. Prototypes A conceptual critique, 131 l-1321, perception. In L. Berkowitz theory: 1978. 36, in person Journu/ of‘ (Ed.), Ad- vances Harris, Heider. in experimenfal sucialpsychology (Vol. 12). New York: Academic Press. 1979. R. J. A primer of multivariate statistics. New York: Academic Press, 1977. F. The psychology of‘ inferpersonal relations. New York: Wiley 1958. Herzberger, S. multimethod Jones, D., & matrix. Clore. E. E.. & Nisbett. causes of behavior. behavior. Kalven. Kanouse. R. E. The Morristown. D. E.. Attribution: L. Actor and of Research observer A. N. J.: General & Hanson. Learning Press. 1972. Boston: Little, Brown, jury. L. R. Negativity in evaluations. In of behavior. Morristown, the causes The A. endogenous-exogenous W.. Personality & A. W., special Cohen, and Social Kruglanski. partition theory. in attribution case directions M. Attributed Psychology, Psychological Hamel, 1. Z., Maides, of lay epistemology. in attribution freedom Social D. T.. Psychological Psychology, & Ross, 1973. M. (Vol. 2). N. 1915. I.. The intuitive psychologist process. IN L. Berkowitz (Ed.), 28, ~f theory as a (Eds.), Nens interest with 1978. intrinsic Journal of Personality 129- 137. and and Social his Advances 1971. and Fact as seen by the actor 1913.27, Distortions in experimental situational of causality: Psychology, shortcomings: In L. Berkowitz York: Academic dispositional Bulletin, Storms, M. Videotape of view. Journal J.: Erlbaum, biases in the attribution 82, 213-225. P.. & Maracek, J. Behavior New York: Academic Press, Snyder, M. Self monitoring processes. social psychology (Vol. 12). New S. Measuring Journa/ & R. F. Kidd children’s hypothesis. Self serving Bulletin, causality. J. M. Attribution W. Ickes, Hillsdale. Nisbett. R. E., Caputo. C., Legant, seen by the observer. Journal ofPersonality Psychology personal S. A.. & Schwartz, In J. H. Harvey, theory and 1913. 26, 245-250. Lepper. M. R., Greene, D.. & Nisbett. R. E. Undermining extrinsic reward: A test of the “Overjustification” Solomon. 1966. E. E. Jones et al. (Eds.). N, J .: General Learning 1975. 82, 387-406. Kruglanski. Ross, in a multitraitl-15. 13, actor and the observer: Divergent perceptions of the et al. (Eds.). Attribution; Perceiving the causes qf The American H. Perceiling Reviel*s, and 1979. 1972. Press, Krnglanski, Miller, attributions in Personality, In E. E. Jones & Zeisel, H., G. Journal in the socialpsychology or fiction’? and as 154-165. attribution (Vol. IO). Adrlances in experimental 1919. attributions. Persona/i@ and Social (Ed.), Press. 1978, 4, 589-594. and the attribution of Personality and process: Social Reversing Psychology. actors’ 1973. and observers‘ 27, 165-175. points MEASUREMENT 95 OF ATTRIBUTIONS Taylor, S. E.. & Koivvmaki, J. H. The perception of self and others: Acquaintanceship, affect, and actor-observer differences. Journal of Personality and Social Psychology, 1976, ceived Trope. Y. ceived Trope. Y., Journal 33, 403-408. freedom. Journal Extrinsic rewards. freedom. J-ournal & Burnstein. E. A of Experimental of‘ Personality und Social Psychology. congruence between dispositions of’ Personality und disposition-behavior Sociul Psychology. Social Psychology. 1978, 36, and behaviors, 588-597. and per- 1978. 36, 588-597. congruity model of perceived freedom. 1977. 13. 357-368. Uleman, J.. Miller, F. D.. Henken, V.. Tsemberis. S., & Riley, E. Visual perspective or social perspective’? Two failures to replicate Storms’ videotape reversal and support for Monson and Snyder. Replications in Sociul Psychology. in press.
© Copyright 2026 Paperzz