《社会心理学》课程教学资源（文献资料）social cognition——Culture and cause - American and Chinese Attributions for Social and Physical Events

ATTITUDES AND SOCIAL COGNITION Culture and Cause:American and Chinese Attributions for Social and Physical Events Michael W.Morris and Kaiping Peng The autho ed the is tha pers were mor atmurdteandncountetfhctualjdg discussed. inerence is the ment of the u(Hume. of behavior).In sum,the precise mechanisms osited for this Psychologists traditi fundamental a nheedoih e decades.bu ing ner eptual or ju s.For intermaldispositionswa u ee a of its "levity")and social events (es n culture tlity )and a si is ak.Nis 985SbT974.1975S 195 oral oc and person tends toengulf the total field")and by Ross(1977)to raphi ions when asked tode hy such 1984 ngs do not dEonoo talked about their Ameri ty of Michi 3 TE. ds fin d E N rted b aus and to test ut the ures. Larrick.Haz nd Edwa r comn plinary appr nce in the research.an Psychological Approaches Perceptual Mechanisms 949

ATTITUDES AND SOCIAL COGNITION Culture and Cause: American and Chinese Attributions for Social and Physical Events Michael W. Morris and Kaiping Peng The authors argue that attribution patterns reflect implicit theories acquired from induction and socialization and hence differentially distributed across human cultures. In particular, the authors tested the hypothesis that dispositionalism in attribution for behavior reflects a theory of social behavior more widespread in individualist than collectivist cultures. Study 1 demonstrated that causal perceptions of social events but not physical events differed between American and Chinese students. Study 2 found English-language newspapers were more dispositional and Chinese-language newspapers were more situational in explanations of the same crimes. Study 3 found that Chinese survey respondents differed in weightings of personal dispositions and situational factors as causes of recent murders and in counterfactual judgments about how murders might have been averted by changed situations. Implications for issues in cognitive, social, and organizational psychology are discussed. If causal inference is the "cement of the universe" (Hume, 1739/1987), do cultures construct their models of the universe with different kinds of cement? Do principles of causal attribu￾tion vary across cultures? Psychologists traditionally assumed that they do not: Attribution patterns were explained in terms of underlying perceptual or judgmental processes. For example, a tendency to overemphasize internal dispositions was noted by Lewin (1935) in early scientific explanations for physical events (e.g., a log floats because of its "levity") and social events (e.g., a man kills because of his "hostility"), and a similar pattern has been experimentally documented by subsequent researchers in lay persons' attributions (see reviews by Holland, Holyoak, Nis￾bett, & Thagard, 1986;McCloskey, 1983; Ross, 1977). This pat￾tern was linked by Heider (1958) to perceptual gestalts (i.e., the person tends to "engulf the total field") and by Ross (1977) to Michael W. Morris, Graduate School of Business, Stanford Univer￾sity; Kaiping Peng, Department of Social Psychology, University of Michigan. This article draws on a dissertation completed by Michael W. Morris under the guidance of Richard E. Nisbett at the University of Michigan, which received the 1993 Society of Experimental Social Psychology Dissertation Award. This research was supported by a National Science Foundation (NSF) grant to Richard E. Nisbett and an NSF doctoral fellowship and a University of Michigan dissertation research grant to Michael W. Morris. Thanks to Richard E. Nisbett for sage advice and skilled assistance at every stage of this project. Thanks also to Lawrence Hirschfeld, Richard Larrick, Hazel Markus, and Edward Smith for commenting on earlier versions of the article, to Shaojian Chen, Chongkeng Cheng, Panfang Fu, Larry Quesada, and Marie Ting for assistance in the research, and to Amy Spade for assistance in preparing figures. Correspondence concerning this article should be addressed to Mi￾chael W. Morris, Graduate School of Business, Stanford University, Stanford, California 94305-5015. judgmental heuristics (i.e., personal dispositions have a higher "availability" and "representativeness" than situational causes of behavior). In sum, the precise mechanisms posited for this "fundamental attribution error" changed over the decades, but the assumption of cultural invariance did not change. By contrast, anthropologists traditionally reported that attri￾bution patterns vary, reflecting social structures or cultural symbol systems (Evans-Pritchard, 1937; Levy-Bruhl, 1910/ 1926). For example, ethnographers in non-Western cultures have long noted that behavior is explained with greater empha￾sis on the concrete situation, temporal occasion, and social context (Geertz, 1975; Hsu, 1953; Levy, 1973; Mauss, 1938/ 1985;Selby, 1974, 1975;Strauss, 1973). Consistent with ethno￾graphic claims are recent findings of cultural psychologists that Indians (compared with Americans) refer more to situational factors and less to dispositions when asked to describe a person they know (Miller, 1987; Shweder & Bourne, 1982) and when asked to explain a behavior by such a person (Miller, 1984). However, these findings do not necessarily indicate different processes of attribution because the objects of attribution also differed (Americans talked about their American acquain￾tances, and Indians about their Indian acquaintances). In other words, findings may merely reflect that cultures differ in the ac￾tual impact of personal versus situational causes on behavior (Argyle, Shimoda, & Little, 1978). Our studies were designed to close this evidential gap and to test hypotheses about the mech￾anism for dispositionalism that varies across cultures. Before presenting hypotheses in detail, we review traditional psycho￾logical and anthropological approaches and recent interdisci￾plinary approaches to causal attribution. Psychological Approaches Perceptual Mechanisms Early approaches to causal attribution were based on the Ge￾stalt theory principle that important abstract forms are per￾Joumal of Personality and Social Psychology, 1994, Vol. 67, No. 6, 949-971 Copyright 1994 by the American Psychological Association, Inc. 0022-3514/94/S3.00 949

950 MICHAEL W MORRIS AND KAIPING PENC mav be directly per fron xnerie ad B as Hume (1739/1987 gued. of displays tte conclu (Premack,1990.p.2). causality:nrn n whichA collides with B and Judgmental Heuristics n wh Others computationa rsky Kahne in the form "fo otion Ha man. 1973,1974 ments 2,1987)have rev tha such as pos collision ve ormto physical ostraints from highly s similar ies that 199). ct he istics o s of thei nigh avai dispositi onal force when object behavio ces.Ross concluded that"the ter rch o tion ofs ial causality bes with Heide and Simme 194 ate the role of dis expe ime ehavi s the "fund sion were similar to chotte's simultaneous mo nembe leads t mo d di d the nts witl Cognitive Structure nd s ons. "Just a ough the Finally,others have modeled attribution as "top-down"ap of pre-storc 1932:Gorman.1974:Min meaningful unit by re ank belso 1977).Re of p n terms of sonsistency in obi ct traiectories and have r d that th externa situational causes tends to son's imp is akin the early t tional dat he situation in social per cep ed thatd ional ca s on the ciser's 194 to in Gergen.1)or from a norm of social d ability(Jon 972ct I di nore by o perceptu ng th the ta t the (Bea cto annot s 1960 1057.1 &Cappellini.9:Gruber.Fink. iments monstrated th tion 1975)and lumination 98 'the actor (Ma ur Pos nd other proach and have demonstrated that subiccts.includingyoun uired kn

950 MICHAEL W. MORRIS AND KAIPING PENG ceived with innate mechanisms that respond to patterns in the perceptual field (see Koffka, 1939). Michotte (1952) proposed that forms of physical causality may be directly perceived from trajectories of objects A and B, not necessarily derived from experience of succession of A and B, as Hume (1739/1987) had argued. After testing hundreds of displays, Michotte concluded that two evoke "universal" and "immediate" impressions of causality: entraining, in which A collides with stationary B and they move off together; and launching, in which A collides with stationary B and B alone moves off. Methodological flaws and failures to replicate cast doubt on Michotte's evidence for uni￾versality and immediacy,1 yet his thesis has been reincarnated in the form of proposed "modules" for causal perception. Ha￾bituation experiments (Leslie, 1982, 1987) have revealed that young children, and even infants, distinguish launching displays that conform to physical constraints from highly similar trajec￾tories that deviate from these constraints (e.g., object B begins to move just before A collides with it).2 Stewart (1984) found with similar displays that subjects perceive an object's behavior as caused by external, situational force when it conforms to physical constraints (essentially, Newton's laws of motion) but perceive an internal, dispositional force when object behavior deviates from these constraints. Research on perception of social causality began with Heider and Simmel's (1944) experiments involving animated displays of moving shapes. Trajectories that reliably evoked causal im￾pressions were similar to Michotte's: "simultaneous move￾ments with prolonged contact. . . successive movements with momentary contact . . . simultaneous movements without contact. . . successive movements without contact" (pp. 252- 255). Most striking in their results was the frequency with which subjects attributed behavior of shapes to internal per￾sonal dispositions, such as intentions, motives, and traits. Heider and Simmel offered a Gestalt account for this disposi￾tionalism: "Just as . . .a landscape seen through the window of a moving train can only be 'resolved,' or made to yield a meaningful unit, by reference to distant objects laid out in space, so acts of persons have to be viewed in terms of motives" (p. 258). Heider (1944, 1958) extended this to account for dis￾regard of external, situational causes: "Behavior . . . tends to engulf the total field, rather than be confined to its proper posi￾tion as a local stimulus whose interpretation requires the addi￾tional data of a surrounding field—the situation in social per￾ception" (1958, p. 54). Jones and Davis (1965) proposed that dispositional causes are most clearly perceived ("the heart is on the sleeve") when be￾havior deviates from expectations of a social role (Jones, Davis, & Gergen, 1961) or from a norm of social desirability (Jones & Harris, 1967). Jones and Nisbett (1972) extended the Gestalt account to explain why behavior is attributed to personal dis￾positions more by observers than by actors. In the perceptual field of an observer, the person is "figural" against the "ground" of the social situation. However, the actor cannot see himself as he acts; thus, in the perceptual field of the actor, it is the situa￾tion, and not the person, which is figural. Further experiments demonstrated that dispositionalism is affected by perceptual variables as mundane as perspective on the actor (S. E. Taylor & Fiske, 1975) and illumination of the actor (MacArthur & Post, 1977). Finally, some researchers have returned to Heider's ap￾proach and have demonstrated that subjects, including young children, distinguish the trajectories that indicate that behavior is caused by an intention (Bassili, 1976; Dasser, Ulbaek, & Pre￾mack, 1989). Modular theorists contend that "perception of in￾tention, like that of causality, is a hard-wired perception based not on repeated experience but on appropriate stimulation" (Premack, 1990, p. 2). Judgmental Heuristics Others approached attribution as a complex computational problem (Kelley, 1967), which people simplify by the use of heuristics (see Kahneman & Tversky, 1973; Tversky & Kahne￾man, 1973, 1974). Research on physical causality has revealed that perceptions of force in object collisions are based on sa￾lient, single dimensions of trajectory, such as postcollision ve￾locity, rather than on the correct multidimensional parameters (Gilden & Proffitt, 1989; Proffitt & Gilden, 1989). Patterns in attribution of social causality have been linked to heuristics of availability, representativeness, or consistency. Ross (1977) ex￾plained the bias toward personal dispositions in terms of their high availability (i.e., proximity of actor to act) and representa￾tiveness (i.e., similarity to the acts they are adduced to explain). After reviewing its consequences, Ross concluded that "the ten￾dency to underestimate the impact of situational factors and to overestimate the role of dispositional factors in controlling behavior" is the "fundamental attribution error" (p. 183). Pet￾tigrew (1979) proposed that the "consistency" of negative dis￾positions with stereotypes about outgroup members leads to heightened dispositionalism for deviant or undesirable behavior by an outgroup actor, which he designated the "ultimate attri￾bution error." Cognitive Structures Finally, others have modeled attribution as "top-down" ap￾plication of pre-stored knowledge in the form of an implicit theory, schema, or script (Bartlett, 1932; Goffman, 1974; Min￾sky, 1975; Schank & Abelson, 1977). Researchers of physical causality have identified a tendency to overpredict persistence or consistency in object trajectories and have proposed that the lay person's implicit theory is akin to the early scientific theory that the internal force of impetus drives a moving object (Kai￾ser, McCloskey, & Proffitt, 1986). The operation of an implicit theory in social causality attribution was suggested by Heider (1958) and Nisbett and Ross (1980). Both, in fact, drew on Ich￾heiser's (1943, 1949, 1970) description of the tendency "to in￾terpret in our everyday life the behavior of individuals in terms 1 Evidence for universality is weakened by the fact that, in many ex￾periments, he and his co-workers were the only subjects, and evidence for immediacy is weakened by the fact that displays were often shown repeatedly before recording the subject's response. In replications, as few as 50% of subjects have perceived causality immediately (Beasley, 1968; Boyle, 1960; Gemelli & Cappellini, 1958; Gruber, Fink, & Damm, 1957; Powesland, 1959). 2 Some have argued on this basis that perception of physical causality from object trajectories is an innate ability (Leslie & Keeble, 1987), whereas others have offered alternative explanations (see White, 1988), and others have demonstrated improvement in this ability through childhood and adolescence, which points to a substantial role for ac￾quired knowledge (Kaiser & Proffitt, 1984)

CULTURE AND CAUSE 951 of specific personal qualities rather than in termsof specific sit- n e i as chin of beh 943,p.151 .H are tered,refl ting societies b du mechanism nderl nes. e be eoole attribute behav r to s ripted roles be cia did no ng dispositionalism and,he a1 to th hi of public dis Recently y Dweck and cole have traced individu d that beha vior is rstood pri urturshasnotS been investigated (for a review,see Dweck structures have been pr d to ex Am can ethn linked (mad 107 1973).Sel tural sys (G one of several sche mas f causal configurat s(e.g ople nderstand be rather tha atory r The be social relat r(1979)p d sp ecific sch re ted in a proverb see the face but do not wha d patterns in the di untine ofdi al caus r975.p.21.S orts that e rare and that ex are guided such as m rde were is& Schank Abelson.1977).Because their words t in their iude n for arti to do so fact tha els do thin years of hi convictio dispositionalism. olved after all the children( p.66) Anthropological Approaches Cognitive Structures Symbol Systems number of anthropol eists have shifted Ethnog phers ha long rece in causal expla e po tion that attributio on patterns reflect nbodie of dis -Bruhl(10/ and sm as r sition that net 28)des of traditional cult hey )to t in mode the pr ce miss structure has been spurred by sperb (1983,198 19 was cracked the hre,or a man was killed by and commun .Forexample a tight onOther Afr an ed patte Other connection models Turner,1975).An influential ount is that th ry-bas dee to count for American explanations of ess (St 155 公 t se the public tern sc the d on patterns of attribution that an ,1970) eral.such as the di terns in social x plana (1938/ guided has 0【h Cultural Psychology cia ositionalis concepts e received some support in recen

CULTURE AND CAUSE 951 of specific personal qualities rather than in terms of specific sit￾uations . . . based on the [ideological] presupposition of per￾sonal determination of behavior (as opposed to the situational or social determination of behavior)" (1943, p. 151). However, Heider and Nisbett and Ross did not posit that theories were the primary mechanisms underlying dispositionalism and, hence, did not limit their claims to the cultures they had researched nor suggest how attribution patterns might differ across cul￾tures. Recently, Dweck and colleagues have traced individual differences in dispositionalism among American subjects to an implicit social theory, but the distribution of this theory across cultures has not yet been investigated (for a review, see Dweck, Hong,&Chiu, 1993). More specific cognitive structures have been proposed to ex￾plain when dispositional attributions are made despite plausible situational attributions. Kelley (1972) proposed that attributors apply one of several schemas for causal configurations (e.g., multiple sufficient causes) that differ with regard to whether in￾ternal dispositional causes are discounted by external situa￾tional causes. Reeder and Brewer (1979) posited specific sche￾mas for types of dispositions (e.g., capacities) to account for finer grained patterns in the discounting of dispositional causes. Others have suggested that explanations are guided by scripts for particular routine behaviors (Abelson & Lalljee, 1988; Mor￾ris & Murphy, 1990; Schank & Abelson, 1977). Because the content of scripts varies across cultures, script-based models predict cultural differences in the explanations given for partic￾ular behaviors. However, these models do not yield predictions about cultural differences in general patterns such as dispositionalism. Anthropological Approaches Symbol Systems Ethnographers have long recorded patterns in causal expla￾nation and interpreted them as reflections of cultural systems of symbols or forms of discourse. Levy-Bruhl (1910/1926) and Fauconnet (1928) described tendencies of traditional cultures to attribute disruptive events (e.g., inclement weather or unsuc￾cessful hunts) to the presence of foreigners (e.g., missionaries or explorers). Evans-Pritchard (1937) interpreted a pattern of attributing an event to both a local, proximal cause (e.g., a pot was cracked by the fire, or a man was killed by a murderer) and simultaneously to a remote, ultimate cause (e.g., witchcraft) in terms of a Zandean metaphysics of dual causation. Other Afri￾canists have also described patterns in which disruptive events are attributed to ultimate causes in the social fabric (Marwick, 1982; Turner, 1975). An influential account is that theory-based attributions to unseen causes are found in both traditional Af￾rican cultures and modern Western cultures, but patterns differ because the public discourse comprising Western scientific the￾ory is open, whereas that comprising non-Western religious the￾ory is closed (Horton, 1970). Asian ethnographies were marshaled in support of Mauss's (1938/1985) thesis that the concept of a person (personne) guided by internal dispositions has evolved and replaced the concept of role or character (personage) only in modern West￾ern social conditions. Mauss's evolutionary argument has been discredited, but the relativity of social concepts has received considerable ethnographic support (Carrithers, Collins, & Lukes, 1985). Hsu (1953) argued that social conceptions of Americans are person-centered, whereas Chinese conceptions are situation-centered, reflecting societies based on individual￾ism versus interdependence. According to Geertz (1975), Bal￾inese people attribute behavior to scripted roles because social thinking occurs within forms of public discourse that direct at￾tention to roles, not dispositions. Scholars of Indian social thought also contend that behavior is understood primarily in terms of social relations, not individuals (Dumont, 1970). Dis￾tinctive patterns of causal attribution, such as those involving the notion of karma, have been traced to symbols in Indian philosophical and medical systems (O'Flaherty, 1980). Native American ethnographies have linked tendencies to￾ward situational explanations to cultural systems (Gearing, 1970; Selby, 1974; Strauss, 1973). Selby argued that the Zapo￾tec people understand behavior in "sociological, rather than psychological, concepts." The belief that internal traits have no "explanatory power for understanding social relations" is rep￾resented in a proverb—"We see the face but do not know what is in the heart"—which is not (as it would be to us) an expres￾sion of despair (1975, p. 21). Selby reports that even rare and deviant behaviors, such as murder, were explained in terms of the actor's social situation and context. Moreover, situation￾alism could be seen not only in their words but in their judg￾ments: A man who had murdered in one situation was not judged likely to do so in another, as evidenced by the fact that "within four years of his conviction for premeditated murder, he was holding a political post in the village and, ironically, it involved looking after all the children during fiestas" (1974, p. 66). Cognitive Structures In recent years, a number of anthropologists have shifted from the position that attribution patterns reflect disembodied symbol systems or social structures (sociocultural determin￾ism), just as psychologists have shifted from the position that they reflect innate, culturally invariant processes (psychobio￾logical determinism; Strauss, 1992a). Interest in models of cog￾nitive structure has been spurred by Sperber's (1983, 1985, 1991) critique of symbolism and call for descriptions of cultural representations that are consistent with how people store, re￾trieve, and communicate information. For example, a tightly structured "spirit attack script" has been proposed by Nuckolls (1991) to explain Jalari Indian attributions for sudden illness. Others have found connectionist models more appealing and have proposed loosely structured networks of semantic and ep￾isodic knowledge to account for American explanations of ro￾mance (Holland, 1992) and of success (Strauss, 1992b). In short, anthropologists have increasingly posited cognitive struc￾tures, but they have focused on patterns of attribution that are specific to particular events rather than on patterns that are more general, such as the dispositionalist patterns in social ex￾planations noted by previous ethnographers. Cultural Psychology Ethnographic reports that attribution is less dispositionalist in non-Western cultures have received some support in recent

952 MICHAEL W MORRIS AND KAIPING PENG eneral Chines people attrib ly leave groups,and they are sc ctalized to beha ationsh than did Americ Sh d bo e1982 The ant repr ntations in individualis ound 12 relationship. xtu in colle tivist Chines culture are root in Co ucia They nro with a"holisti w,”peT Thus,the centered theory that social b ms of stract disposi idualist culture cente ed theo y that 1 be (1984)extende arch to explanations for b is shape d by relatior tional conceptions in non-Western cultures may From this pro csize that ence erminar She aske r10u5 d i findings th d0c3gorzationanc of children in the two wer uch as phy cal(Keil,1986.1989 an mat (Ge like Yet with age (and presumabl acculturation)Ameri 1980 1982 Some evidence suggests that bour of the nedndGspti0m,9g2rson theories differ studies were d complem t the not held and s accordine to intentions bout 1976 98 virtues,such as protecting agains Thi would t f ty with m f th the hic rees s ence between Ame nd In terns d behavior And search indicat ce be ern and non- ern culture n the a In s this is Miller (1984) ted A 9781.0u studies were designed to co thi d sh bout a cu ted by s gave n Hypotheses -h d to fir hand info edsimplybecausemricnswereithscond-han about soc or that is r ore vide ad in individualis an st cu ial t as de ile(1840 1946.7a0 ). en vents in the nain and with public rep 19E3 vith cu 1953,197 collectivism dime n(Ba 1.1966) tures substa antial:Tr 988;Ho e B sion.,such as the Unit of s n&R的 or a re ds.1990

952 MICHAEL W. MORRIS AND KAIPING PENG cross-cultural psychological studies. Bond (1983) found that, al￾though American and Chinese attributions fell into the same general categories, Chinese people attributed more to circum￾stances of a social nature and to situations involving social re￾lationships than did Americans. Shweder and Bourne (1982) found that Hindu Indians gave more descriptions of an acquain￾tance's behavior as situated in a particular time, place, and so￾cial relationship, whereas Americans gave more decontextua￾lized descriptions in terms of general, cross-situational disposi￾tions. They proposed that, in cultures with a "holistic world view," persons and perhaps also physical objects are thought of in terms of specific occasions and concrete contexts rather than in terms of abstract dispositions. Miller (1984) extended this research to explanations for be￾havior by proposing that "individuals' acquisition of more rela￾tional conceptions of person in non-Western cultures may lead them to give less weight than Western attributers to general dis￾positions of the agent. . . [and more weight] to the contextual determinants of action" (p. 964). She asked subjects of various ages to explain a behavior of an acquaintance witnessed in ev￾eryday life. Explanations of children in the two cultures were alike. Yet with age (and presumably acculturation) Americans were increasingly dispositionalist and Indians were increasingly situationalist, particularly for deviant behaviors. This finding of cultural divergence was also obtained in descriptions of persons, both those known well and not known well (Miller, 1987). Our studies were designed to complement the evidence pro￾vided by previous studies. In previous studies, the object of ex￾planation or description was not held constant (Americans talked about their American acquaintances, and Indians talked about their Indian acquaintances). Although this design has many virtues, such as protecting against spurious cultural differences due to differential familiarity with stimuli, it has the drawback of confounding two possible sources of the effect: a difference between American and Indian subjects' attribution processes and an objective difference in the actual causes of their acquaintances' behavior.3 And research indicates such a difference between Western and non-Western cultures in the ac￾tual impact of personal versus situational causes on behavior (Argyle et al., 1978). Our studies were designed to close this evidential gap and to test hypotheses about a culturally variable mechanism fordispositionalism. Hypotheses We propose that dispositionalism in social attribution (the "fundamental attribution error") reflects an implicit theory about social behavior that is more widespread in individualist cultures than in collectivist cultures. We assume that an implicit theory about a domain is acquired from culturally bound expe￾rience with events in the domain and with public representa￾tions of the domain (e.g., folktales, sacred texts, laws, and works of art). Because the individualism-collectivism dimension cap￾tures substantial variation among national cultures in social ex￾periences and representations (Hofstede, 1980,1983,1991;Tri￾andis, 1990),4 we submit that the distribution of implicit social theories differs between cultures at opposite ends of this dimen￾sion. In highly individualist cultures, such as the United States, persons are primarily identified as individual units, they can leave groups at will, and they are socialized to behave according to personal preferences. In highly collectivist cultures, such as China, persons are primarily identified as group members, they cannot freely leave groups, and they are socialized to behave according to group norms, role constraints, and situational scripts. The dominant social representations in individualist American culture are rooted in the Judeo-Christian notion of the individual soul and the English legal tradition of free will. Those in collectivist Chinese culture are rooted in Confucian precepts about the primacy of social relations and the virtue of role-appropriate behavior (Hsu, 1981b; King & Bond, 1985). Thus, the person-centered theory that social behavior expresses stable, global, internal dispositions is more widespread in indi￾vidualist cultures; the situation-centered theory that social be￾havior is shaped by relationships, roles, and situational pres￾sures is more prevalent in collectivist cultures. From this proposal, we hypothesize that attributional differ￾ences between Americans and Chinese are broad in scope. The scope of implicit theories has been elucidated by developmental findings that categorization and inference rules are organized according to domains of things having the same kind of causal properties, such as physical (Keil, 1986, 1989), animate (Gel￾man, 1990; Gelman & Spelke, 1981), psychological (Shultz, 1980; Wellman & Gelman, 1992), and social kinds (Shultz, 1982). Some evidence suggests that boundaries of these do￾mains are culturally invariant (see Atran, 1989; Gelman, 1990; Jeyifous, 1985) even if content of domain theories differ. Do￾mains are marked by the way things move (animates can propel themselves, psychological creatures move on intentional paths, and social creatures move according to intentions about inten￾tions; see Bassili, 1976; Dennett, 1983, 1987; Premack, in press). Hence, the trajectory of motion in an event may trigger the implicit theory used to process it. This would account for cases when everyday perception is animistic (a leaf swirling in the wind seems animate) and anthropomorphic (trees swaying in the wind seem to be socially interacting). Attributional pat￾terns due to an implicit theory differ in scope from those due to 3 In an attempt to address this issue, Miller (1984) presented Ameri￾can subjects with narratives about behaviors that had been originally generated by Indian subjects, and she compared American explanations to the original explanations of the Indian subjects. She found, as pre￾dicted, that Americans gave more dispositional explanations. However, in this study, American culture is confounded with another factor that increases dispositional attribution: second-hand as opposed to first￾hand information about behavior (Gilovitch, 1987). The finding would be expected simply because Americans were working with second-hand information. 4 Although Hofstede's analysis of major dimensions of cultural varia￾tion provided the catalyst for psychological research on individualism￾collectivism, related constructs have been used previously by many so￾cial theorists such as de Tocqueville (1840/1946), Tawney (1926), We￾ber (1930), and Lukes (1973) as well as social scientists, including those concerned with culture (Hsu, 1953, 1971; Triandis, 1972), values (KJuckhorn & Strodtbeck, 1961), character (Riesman, 1950, 1954), so￾cial systems (Parsons & Shils, 1951), religion (Bakan, 1966), ecology (Berry, 1979), and so forth. The dimension has reliably emerged in sub￾sequent studies (Bond, 1988; Hofstede & Bond, 1984; Triandis et al., 1986) and has been found to predict free-riding in group tasks (Earley, 1989), frequency of social interactions (Wheeler, Reis, & Bond, 1989), favored types of verbal abuse (Semin & Rubini, 1990), and many other social psychological phenomena (for a review, see Triandis, 1990; for a critical view, see Kashima. 1987)

CULTURE AND CAUSE 953 perce aplicit theories would and social vents involved a gr swimming in a al dis sition vould not be so bre as to ex Our studic d the eadth hypoth ith empha ntern Amer udies inese people would com the 'funda ttr sical events.we hvpothesized that Am an e.we test Chine nhuman c hat torsand accords motio that deviates from these co nts ith refe encetodin ted tha ofrest (ega statior vimming in front of a gr oup. American night at rest).A traint inv rather Runeson's(97)pro rial ob nal fo n the group). here ed that trajectories w ole mieht see a grou ng an indi orsocial events,we exp cer h s can t dy er s to exteral forces (e.kicks)b namics would he i interpr ted sociall (anth hiz more by Chin ubjects and akin to A di due and colle rom other ects in the ing of be pomorph ulting in a harmon nding the beh ior (D ck Hong Chiu.1993) those rest ant di vision (Bond Hwane 1986:Hsu,1981a.In dict th ang it oa group. e expe of Erdley 93) in ion individual and enta itivi impact of future 19 y to the Sullivan 85 986 arkus Kitayama,19 expected Chinese subi cts to be more e likely to anthropomor cognition dA more likely to ing whether merican and chir and group cohesion (Redding Wong.196). ed w Method dy ?)t d Chinese oole repr High School Sample they Subjects.We sampled 100 Chinese secondary so s fro olsttcnts,hal we d ring their judgments )and ha Stu The Chinese exp nenter (a psycho y graduate nt a ugh t vith its nti cpeldinth d

CULTURE AND CAUSE 953 other proposed mechanisms: Any event that triggers a social theory would be processed differently by Americans and Chi￾nese people. Cultural differences due to implicit theories would extend across types of social events (unlike differences due to scripts) and across types of social actors (unlike differences due to stereotypes). Yet differences would not be so broad as to ex￾tend across domains (unlike differences due to world views). Our studies investigated the breadth hypothesis with empha￾sis on types of actors, which had not been varied in previous studies. The events that subjects explained in previous studies were behaviors of acquaintances. Our studies focused on behav￾iors of strangers and of outgroup members. We predicted that Chinese people would commit neither the "fundamental attri￾bution error" nor the "ultimate attribution error." Further￾more, we tested whether cultural differences extend even to nonhuman events that are interpreted as social events. Because Chinese people and Americans would interpret animal behavior with reference to different social theories, we predicted that they make different attributions. Imagine, for example, one fish swimming in front of a group. Americans might attribute the fish's behavior to an internal disposition (e.g., leadership abil￾ity), whereas Chinese people might attribute it to an external, situational force (e.g., pressure from the group). In short, where Americans might see an individual leading a group, Chinese people might see a group chasing an individual. This cultural difference, however, would not extend to physical events, as these would not trigger social theories. Imagine, for example, a soccer ball bouncing down a soccer field. Its movements can be causally attributed to internal properties (e.g., its elasticity) or to external forces (e.g., kicks) but one would not expect Ameri￾cans to emphasize the former and Chinese people the latter. Cultural differences due to implicit theories would also be cognitively deep. Dweck and colleagues have explored how sub￾jects who hold the social theory predictive of dispositionalism differ from other subjects in their processing of behavior. They attend to and encode different features into the representation of the behavior (Dweck, Hong, & Chiu, 1993), which makes them more likely to infer dispositions based on limited evidence (Gervey, Chiu, & Dweck, 1992) and to overpredict the consis￾tency of future behavior (Erdley & Dweck, 1993). Processing may result in a more decontextualized representation of behav￾ior that hinders judgmental sensitivity to the impact of future contexts (Semin & Fiedler, 1991). In sum, because theories shape encoding, representation, and inferences drawn from be￾havior, cultural differences would be found not only at the level of verbal discourse about behavior but also in other modes of causal cognition. We tested predictions about differential encoding by investi￾gating whether Americans and Chinese people differ in their on￾line visual perceptions of causality (Study 1). We tested whether different kinds of attributions are generated by comparing the explanations of American and Chinese newspaper reporters for the same events (Study 2). We investigated whether American and Chinese people represent events differently by comparing how they evaluate various kinds of explanations (Study 3). We also investigated whether they draw different inferences from their representations of events by comparing their judgments about counterfactual situations (Study 3). Study 1 In Study 1, American and Chinese subjects watched cartoon displays of physical and social events and reported their causal perceptions. The contexts chosen were familiar in both cul￾tures: Physical events involved an object moving across a soccer field, and social events involved a group offish swimming in a lake. For physical events, we predicted no cultural differences: An object's trajectory will be attributed to internal dispositions to the extent that it deviates from certain trajectory constraints. For social events, we predicted cultural differences: A fish's be￾havior that deviates from others will be attributed more to its internal dispositions by Americans and more to its external sit￾uation by Chinese subjects. All subjects were shown many events of each kind to investigate additional hypotheses about perception of dynamics from trajectories. For physical events, we hypothesized that Americans and Chinese people have the same implicit theory that accords ob￾ject motion within certain trajectory constraints to external fac￾tors and accords motion that deviates from these constraints to internal factors. One proposed constraint (Stewart, 1984) that we investigated is conservation of rest (e.g., a stationary soccer ball stays at rest). Another constraint investigated involved ve￾locity. The constraint was not conservation of velocity, but rather Runeson's (1974) proposed constraint that terrestrial ob￾jects gradually lose velocity or decelerate, (e.g., a rolling soccer ball gradually decelerates). For social events, we expected that trajectories would be more ambiguous as to internal versus external causation. On the basis of Heider and Simmers finding that trajectories that resemble familiar social dynamics cue perception of social relations, we investigated whether fish trajectories akin to Chinese social dy￾namics would be interpreted socially (anthropomorphized) more by Chinese subjects and those akin to American social dynamics more by Americans. We attempted to vary this tra￾jectory factor within three sets of social displays. In displays of fish compelling others to move through compliance or confor￾mity, we expected Chinese people to be more likely to anthro￾pomorphize those resulting in a harmonious bonding, whereas we expected Americans to anthropomorphize those resulting in a discordant division (Bond & Hwang, 1986; Hsu, 1981a). In displays of a fish changing its relation to a group, we expected Chinese subjects to be more likely to anthropomorphize events resulting in connection of individual and group, whereas Americans would be more likely to anthropomorphize events resulting in separation (Bellah, Madsen, Sullivan, Swidler, & Tipton, 1985; Hoosain, 1986; Markus & Kitayama, 1991). In displays of a group's response to the presence of an outsider, we expected Chinese subjects to be more likely to anthropomor￾phize dispersion and Americans more likely to anthropomor￾phize convergence because of different dynamics of leadership and group cohesion (Redding & Wong, 1986). Method High School Sample Subjects. We sampled 100 Chinese secondary school students, half of whom we drew from a school serving a university neighborhood (Beijing University High School) and half from a school serving a work￾er's neighborhood in an eastern coastal city (Qingdao City High School). The Chinese experimenter (a psychology graduate student at Beijing University) accessed classrooms in the former school through psychology department channels and in the latter school through family connections with its principal. We sampled 95 American secondary

954 MICHAEL W.MORRIS AND KAIPING PENG an and an-A resultin vith the Mar cer ball and almost entirely (Thi nts scem influenced by bled those of Michotte (in his Ex 229,3nd ctvely).The cir When subjects had read the in Figure 1).Ac ing re s onsofishswimmieinlhky ents i dngations ting fish." Sub circle a and tra 一2e大hcma传h (see di rele in cdsieht5f v frictio au do the other is'smovements tion that n th s not to invite a the elikecoil the bl n. the fo theorderofdisplayswithineachetwasTeerse swim sepa t of Graduate School Sample on d tedineirher nd th In th Figure4.omeeiheriencrjoinedtheouhereaiovariCwhhDa PR. g K qual r paths and the bl either the center or the 2 (o gA tota d neid in the Ins was tran he qu ed as fo Always try di the four phase

954 MICHAEL W. MORRIS AND KAIPING PENG school students, whom we drew from a school serving the mixed college and factory town of Ypsilanti, Michigan (Willow Run High School). Classrooms were accessed through psychology department channels. The American sample comprised roughly equal numbers of European￾American and African-American students plus a smaller number of Asian-American students. In each country, we selected several classes and recruited the participation of all students in these classes, resulting in a sample evenly divided by gender. Materials. We produced a series of animated displays with the Mac￾romind Director program on a Macintosh Ilci computer and then trans￾ferred them to videotape so that they could be displayed on 20-inch television screens. Physical displays featured a round object moving across a soccer field. All displays featured a solid black circle (5 cm diameter) and some also featured a yellow rectangle (10 cm X 2 cm). We produced two sets of five displays: 1. Collision displays resembled those of Michotte (in his Experiments 2, 1, 29, 31, and 3, respectively). The circle always began at rest, center￾screen, then moved rightward; what varied was the degree to which its movement coincided with being struck by the rectangle (see diagrams in Figure 1). Across five displays, the circle increasingly deviated from conservation of rest: (a) It moved while in continuous contact with the moving rectangle (entraining); (b) it moved immediately on impact of the rectangle (launching); (c) it moved 500 ms after impact (time gap); (d) it moved before impact, when the approaching rectangle was still 2 cm away (space gap); and (e) it moved without any rectangle present (starting). 2. Acceleration displays resembled those of Stewart and Runeson. The circle always appeared from beyond the left edge and traversed the screen; what varied was its change in velocity as it crossed the screen (see diagrams in Figure 2). Across five displays, the circle increasingly deviated from the slight deceleration characteristic of inanimate objects in the terrestrial ecology: (a) It decelerated slightly as if by friction (de￾celerating); (b) it held a constant speed (constant); (d) it stopped sud￾denly (stop); (d) it accelerated slightly (accelerating); (e) it suddenly stopped, started, stopped, and started again (stopping and starting). Social displays portrayed events involving a group of fish. All fish were identical in size (6 cm from gill to tail), features, and manner of swimming, but each was a different color. The blue fish (on which ques￾tions were focused) swam on a trajectory that deviated from that of others. The animations were designed to resemble a film of actual fish as much as technically possible so as not to invite anthropomorphiza￾tion any more than actual fish do.5 We produced three sets of displays: 1. Compulsion displays were like collisions between the blue fish and the group. In these displays (see Figure 3), one party approached the other and compelled it to move. This compelling took the form of either harmonious entraining (parties swim together and in synchrony after contact) or discordant launching (parties swim separately and out of synchrony after contact). We varied which party compelled the other. 2. Connection displays showed movements that resulted in either con￾necting or separating the blue fish and the group. In these displays (see Figure 4), one either left or joined the other. We also varied which party made the move. 3. Collection displays showed the group either collecting or dispersing in the presence of the blue fish. In these displays, group members moved on radial paths and the blue fish swam in either the center or the periphery. Groups of 5-10 subjects sat at desks in classrooms in front of a large TV and were given questionnaires in English or Chinese. The English version was translated into Chinese and back-translated successfully. The questionnaire cover stated as follows: "In this experiment you will answer questions in this booklet while watching cartoons on the televi￾sion. Always try to answer quickly, based on your first impression." Instructions for Part 1 described cartoons set on a soccer field in the evening, which feature a dark circular shape moving across the field that could be "either a soccer ball or an animal." Subjects were instructed to observe "how much its movement is influenced by internalfactors, such as the pressurized air inside the soccer ball that makes it bounce, or the muscles of an animal that allow it to run" and "how much the move￾ment is influenced by external factors, such as a person kicking a ball or scaring an animal." Subjects were instructed to familiarize themselves with the following questions: 1. "Does the dark thing look like a soccer ball or like an animal?" answered on a 5-point scale labeled looks like soccer ball, more like soccer ball than animal, could be either, more like animal than soc￾cer ball, and looks like animal. 2. "To what extent do the thing's movements seem influenced by internal factors?" answered on a 5-point magnitude scale labeled hardly at all, slightly, moderately, greatly, and almost entirely. (This scale was used with all influence questions.) 3. "To what extent do the thing's movements seem influenced by external factors?" When subjects had read the instructions, the videotape was started by the experimenter. The videotape showed each display three times, each time preceded by an attention-getting beep and followed by a ques￾tion then a 30-s pause. During the pause, subjects responded by marking a scale in their questionnaire. The entire running time of Part 1 was slightly over 10 min. Instructions for Part 2 described cartoons offish swimming in a lake. Subjects were instructed to observe the blue fish's relationship to the group and the influences on fish movements, including internal factors, "such as when a fish is hungry and swims to look for food," and external factors, "such as when one fish . . . follows another fish." Subjects were instructed to familiarize themselves with the following questions: 1. "Does the blue fish seem to be an important member of the group?" answered on a 5-point scale labeled not at all, less than average, average, more than average, and most important member. 2. "To what extent do the blue fish's movements seem influenced by internal factors?" 3. "To what extent do the blue fish's movements seem influenced by the other fish?" 4. "To what extent do the other fish's movements seem influenced by the blue fish?" Each of the 12 cartoons was shown four times, each time preceded by a beep and followed by a question then a 30-s pause. The entire running time of Part 2 was just over 15 min. Half of subjects saw the forward￾order videotape and half saw the reverse-order videotape. In the forward tape, the five physical displays in each set were ordered by increasing deviation from natural trajectories (as in Figures 1 and 2), and the four social displays in each set were ordered randomly. In the reverse tape, the order of displays within each set was reversed. Graduate School Sample Subjects. Subjects were students in the mechanical engineering and chemistry graduate programs at the University of Michigan. We con￾tacted all Chinese students (citizens of the People's Republic of China [P.R.C.], Hong Kong, or the Republic of China [R.O.C.]) in the 2nd and 3rd year cohorts and an equal number of randomly selected American students (U.S. citizens). They were offered $ 10 for participating. A total of 22 (of 28) Chinese and 22 (of 29) American students participated. Materials and procedure. The same materials were used as in Study 1. Sessions were held in the Institute for Social Research laboratory. American and Chinese natives were both tested in English. For each 5 We used a sequence of computer graphic images showing a fish in four phases of its swimming motion, each with different fin and tail positions. In the 100 frames of each animation, we cycled through this sequence (a film loop) as we translated the fish along its trajectory

CULTURE AND CAUSE 955 COLLISION ENTRAINING LAUNCHING LAUNCHING W/TIME GA oves after delay LAUNCHING W/SPACE GAP STARTING there were no cultural differences in perc ption tched a fo or extern s.Ame can and Chi nese per sample of subjects.The lack of cultural differences in the phys Results Physical Events urietcoicenlbaperCCpdootfiaiemlforce,extemalforsg ion from na al tra

CULTURE AND CAUSE 955 COLLISION ENTRAINING LAUNCHING LAUNCHING w/ TIME GAP (moves after delay) LAUNCHING w/ SPACE GAP STARTING Figure 1. Diagrams showing trajectories of objects in collision displays. Arrowed lines show direction of movement; vertical lines show where the object stopped, if it stopped. session, one American and one Chinese graduate student were seated near opposite ends of a long table, each in front of a TV and VCR (vid￾eocassette recorder). One of them watched a forward-order videotape and the other a reverse-order tape. They controlled the pace of the ex￾periment by pausing the VCR with a remote control as needed. Results Physical Events We predicted that causal perceptions of physical events by American and Chinese subjects would not differ. As predicted, there were no cultural differences in perception of internal force or external force. As Table 1 shows, American and Chinese per￾ceptions did not differ with either set of displays or with either sample of subjects. The lack of cultural differences in the phys￾ical domain suggests that we successfully translated the ques￾tionnaire into Chinese and supports interpretation of culture effects in the social domain as differences in perception, not fail￾ures of communication (Campbell, 1964). We predicted that perception of internal force, external force, and animacy would be cued by an object's trajectory. Specifi￾cally, we tested the prediction that deviation from natural tra-

956 MICHAEL W.MORRIS AND KAIPING PENG ACCELERATION DECELERATING CONSTANT SUDDEN STOP ACCELERATING SUDDEN STOPS AND STARTS jectory constraints increases rception of internal force,de increasingly deviated of rest,s acy(se Figure 5).Across the five acceleratic displays,in whi e en g d d as ph cts like ral p a perceptu to the ce more re

956 MICHAEL W. MORRIS AND KAIPING PENG ACCELERATION DECELERATING CONSTANT SUDDEN STOP ACCELERATING SUDDEN STOPS AND STARTS Figure 2. Diagrams showing trajectories of objects in acceleration displays. Dashes of constant length represent constant velocity; dashes of increasing length indicate acceleration; dashes of decreasing length indicate deceleration; vertical lines show where the object stopped, if it stopped. jectory constraints increases perception of internal force, de￾creases perception of external force, and increases perception of animacy. Across the five collision displays, in which the circle increasingly deviated from conservation of rest, subjects per￾ceived more internal force, less external force, and more anim￾acy (see Figure 5). Across the five acceleration displays, in which the circle increasingly deviated from gradual deceleration, sub￾jects showed the same predicted pattern (see Figure 6).6 This general pattern might also be predicted from a perceptual mod￾ule, but the fact that science graduate students were relatively more responsive to trajectory cues inclines the evidence in favor 6 Slopes of the functions in Figures 5 and 6 (estimated by linear con￾trasts) differed from zero in the predicted direction and did not differ by culture. We excluded the entraining display from analyses on learning that it is perceived as physical causation only with square objects like Michotte's, not with round objects like ours (Beasley, 1968). Because they are tangential to the central thesis of the article, details of slope computations and comparisons are omitted here and can be obtained by writing to Michael W. Morris

CULTURE AND CAUSE 957 HARMONIOUS DISCORDANT GROUP ENTRAINS GROUP LAUNCHES 纱 INDIVIDUAL ENTRAINS INDIVIDUAL LAUNCHES signif Social Events be rans n artifacts be use th American and Chine se subjec would differ in emph asis on in wereentireyabsentinourgradlatcdcS pulsion cor We ale nection,and collection display sets,an effect highly significant CONNECTING SEPARATING 一 GROUP JOINS GROUP LEAVES INDIVIDUAL JOINS INDIVIDUAL LEAVES

CULTURE AND CAUSE 957 HARMONIOUS GROUP ENTRAINS INDIVIDUAL ENTRAINS DISCORDANT GROUP LAUNCHES INDIVIDUAL LAUNCHES Figure 3. Diagrams showing trajectories offish in compulsion displays. Vertical lines show where the fish stopped, if it stopped. The blue fishes have the darkest arrows in these diagrams. of an implicit theory shaped by experiences and public representations. Social Events We predicted that causal perception of social behavior by American and Chinese subjects would differ in emphasis on in￾ternal versus external causes. Our predictions were borne out for the high school sample: Americans perceived more influ￾ence of internal factors and Chinese perceived more external influence on the blue fish's motion. Americans attributed more to internal force than did Chinese subjects in compulsion, con￾nection, and collection display sets, an effect highly significant for the first of these (see Table 2, top panel). Chinese subjects attributed more to external force in compulsion, connection, and collection display sets, an effect significant for all three dis￾play sets (see Table 2, top panel). These effects are not likely to be translation artifacts because the analogous task produced no cultural difference in the physical domain (compare Table 2 to Table 1). We expected that these main effects of culture might be attenuated among graduate students, who have more in com￾mon culturally than high school students. We found, however, that cultural effects were entirely absent in our graduate student sample (see Table 2, bottom panel). We also predicted interactions of culture and trajectory in that perception of social relations among fish (anthropomorphi￾CONNECTING GROUP JOINS INDIVIDUAL JOINS SEPARATING GROUP LEAVES INDIVIDUAL LEAVES Figure 4. Diagrams showing trajectories offish in connection displays. Vertical lines show where the fish stopped, if it stopped. The blue fishes have the darkest arrows in these diagrams

958 MICHAEL W MORRIS AND KAIPING PENG esosabhtatrbuieaemareeoarea and Chinese Subiccls es becaus Display set Collision Perception M F M F P erer and Chine reporte nteaore 28182830 .4650 83535 .9633 Graduate school subjects ns rather than situaou member.However, ” 2363 35823 63 Method Chinese 835220 88 35 Materials d=1,193.d=1,42 Both and cia nthropo 199 by Ch a Chin Thi lost hi and high antly)and with graduate stu usine s do d d206u iandoncabou ce in the pre ence of nurder h high sch Table 3).The validity of fish displays for a oent of Procedure familiar social dynamics. Study2 able attrib ude.lo (e.Lau&Russel.1980:Schoeneman&Rubanowitz,1985)

958 MICHAEL W. MORRIS AND KAIPING PENG Table 1 Causal Perception of Physical Events by American and Chinese Subjects Display set Collision Acceleration Perception M M High school subjects8 Internal force American Chinese External force American Chinese 2.56 2.70 3.40 3.11 1.82 .18 .35 3.04 3.11 2.74 2.85 .46 .96 .50 .33 Graduate school subjectsb Internal force American Chinese External force American Chinese 2.80 2.71 3.77 3.61 .23 .63 .35 3.58 3.77 2.40 2.21 .23 .63 .35 *df= 1,193. bdf= 1,42. zation) would be greater for trajectories that resemble a social dynamic from the perceiver's culture. We measured anthropo￾morphization with an index of ratings 1, 3, and 4 (blue fish's relationship to the others, influence by the others, and influence on the others). In compulsion displays, we manipulated the har￾mony of trajectories, predicting that Chinese subjects would be more likely to anthropomorphize harmonious events and that Americans would anthropomorphize discordant events. This pattern was obtained with both high school students and grad￾uate students (see Table 3). In connection displays, we manipu￾lated the outcome of trajectories, predicting that Chinese sub￾jects would be more likely to anthropomorphize events resulting in connection and that Americans would anthropomorphize events resulting in separation. This pattern was obtained with high school students (nonsignificantly) and with graduate stu￾dents (significantly; see Table 3). In collection displays, we ma￾nipulated the flow of the group offish, predicting that Chinese subjects would be more likely to anthropomorphize group di￾vergence in the presence of an outsider and that Americans would anthropomorphize group convergence in the presence of an outsider. This pattern was obtained with high school students (significantly) and with graduate students (nonsignificantly; see Table 3). The validity of fish displays for assessment of social perception is supported by these general tendencies of subjects to perceive social relations when trajectories resembled those in familiar social dynamics. Study 2 In Study 2, we compared attributions for mass murders in newspapers serving American and Chinese communities. The reason for studying attributions generated spontaneously in this natural context rather than in response to a researcher's request (e.g., Lau & Russel, 1980; Schoeneman & Rubanowitz, 1985) was to establish that attributional patterns are not artifactual responses to research tasks. This is a crucial step in studying how patterns of attributions differ across cultures because spu￾rious cultural differences can arise from the differential famil￾iarity of a research task. First, we predicted that Americans show a stronger tendency toward the "fundamental attribution error." That is, for a given murder, American reporters would make more attributions to personal dispositions of the mur￾derer and Chinese reporters more attributions to situational pressures. Second, we predicted that culture of the reporter would interact with culture of the murderer, such that Ameri￾cans but not Chinese reporters would show the pattern of the "ultimate attribution error." In other words, we expected that American reporters would attribute murder to personal dispo￾sitions rather than situational pressures to a greater extent when the murderer is an outgroup member. However, we expected that Chinese reporters would attribute primarily to situational factors for ingroup and outgroup murderers alike. Method Materials We compared articles about murders in the leading English-language (The New York Times) and Chinese-language (World Journal) U.S. newspapers. Both newspapers are based in New York and circulated worldwide. We focused on attributions for mass murder because unu￾sual, negative behaviors evoke the most spontaneous attribution (Weiner, 1985). We selected two comparable crimes committed in au￾tumn 1991 by Chinese and American murderers, respectively: 1. Gang Lu was a Chinese physics student who had recently lost an award competition, unsuccessfully appealed it, and subsequently failed to get an academic job. On October 31, he entered the University of Iowa Physics Department and shot his advisor, the person who handled his appeal, several fellow students and bystanders, and then himself. 2. Thomas Mcllvane was an Irish-American postal worker who had recently lost his job, unsuccessfully appealed the decision with his union, and had failed to find a full-time replacement job. On November 14, he entered the Royal Oak, Michigan, Post Office and shot his super￾visor, the person who handled his appeal, several fellow workers and bystanders, and then himself. From each paper, we photocopied every article about the murders that appeared between November 1,1991, and December 31, 1991. For coding purposes, we segmented articles into units corresponding to each clause, using Miller's (1984) method. The unitizing was done for both English and Chinese articles by a bilingual psychologist: Three English articles about the Lu murder totaled 206 units, and one about the Mcllvane murder totaled 84 units; 17 Chinese articles about the Lu murder totaled 1,099 units, and one about the Mcllvane murder totaled 41 units. We gave coders 22-page booklets consisting of an enlarged copy of each article with units demarcated and numbered as well as coding sheets with corresponding numbers. Procedure We constructed a scheme to enable exhaustive coding of units. A unit was coded as an attribution to a personal disposition of the murderer, an attribution to a situational factor, or neither (nonattribution or unclas￾sifiable attribution). Coders were instructed that a personal disposition is a property that the murderer carries across time, place, and social context (such as a personality trait, temperament, stable value or atti￾tude, long-standing goal, habit, chronic pathology, general capability, physical characteristic, character flaw, etc.). We instructed coders that a situationa! factor is tied to a particular time (such as an emotional crisis