World-system and regional linkages as causally implicated in local level conflicts at the ethnographic horizon

Critique of dirposiiianal explanarions for local level conflict shows anomalies in Ross's theor? of violence and in ethnographic accounrr using "culture ofviolence" explanarions of conflict. This paperparallels inrernaiionai relations theories (e. g. Waltz 1959) in examining hypothcrer about regional md -or idryarem linkages as causally implicated in local-level wariare. In apilor riudy usinga fraction of rhe srindaid sample, two variables are found to predicr external conllict: Frcquenc? of inrersocictal contact and relacation forced bvoowers in thelareer world-rvsrcm.Thecorrelarion between external and internalrontlicr is , . poriri\,e in morepcriphcr~l zones of rhe world-system, bur negative in rhc moreceniralzoncs. Ercrrnal war combined with theevrenr of stare level organization predicts the rtrengh of fraternal inrerert groups, oncof the dir~osirional ~rcdictois ofinrrrnzl violence. Such evidence mieht ruooorr arcenario for r e~e i s ine the di" .. recrion of causation in Rorr'stheory of violencesuch that inremairiolenceandrocieral features -di$oringm 10 inrernal violence (including features of socialization) may rerulr from conflicrual pressures at the worldsystem, regional linkages and state levels. Cultural anthropology in recent decades has shifted away from interest in purely indigenous aspects of social systems to exploration of the linkages between loc31 communities and the larger systems in which they are embedded. The focus of this paper is on conflict, partly as a local boundary phenomenon, operating at different lei-els. often related to ethnocentrism, but also h a a result of pressures generated in thc l ~ r s c r licld in which societies interact. After considering some leading theories about wartarc, the paper tests provisional hypotheses about the role of regional and ~orld-system linkages in warfare at the local level. A general theory of violence, incorporating multiple hypotheses from previous studies, was recently formulated and tested by Ross (1985, 1986). It was intended to predict the extent of internal conflict and external warfare from structural and dispositional features of societies in the Murdock and White (1969) standard sample. His general argument consists of two parts. The dispositional pan of his argument is that child training practices (harsh socialization, nnaffectionate socialization) and male identity conflicts predict higher amounts of violence, including both internal conflict and esternal war. A strong form of this dispositional argument would imply that generalized "cultures of violence" exist tovarying degrees in different societies. The structural part of his argument is that there are social structural features that mitigate against violence by providiwgmeans for conflict resolution, and can account for variation in the type of UuugIar K.White: World-syrrrm and rrgion~l inkages 113 violence (e.g., internal conflict versus esternal warfare). Ross hypothesized that crosscutting ties and community intermarriage would predict lower internal but higher external conflict, while fraternal interest y o u p s would predict higher internal conflict only. 'The present paper reviews various weaknesses in the theory of violence presented by Ross, and explores the possibility that the external environment is one of the major causes of conflict, both external and internal. Ross'svariables will be used to test hypotheses where relevant, supplemented by variables drawn from the cumulative standard sample database', and in particular several new but provisional variables on world-system linkages of societies in the standard sample (see the Appendix for definitions and sources of variables). Ross coded half (90 societies) of the Standard Cross-Cultural Sample (Murdock and Whire 1969), and the hypotheses tested here will also use his subsample. Ross's scales for external and internal conflict will he replicated using a single-factor scoring procedure (White 1990); other scales for measuring a single concept with multiple measures will be derived by the same procedure2. Weaknesses in the argument for a generalized disposition to conflict After testing predictions of internal and external conflict from dispositional and structural variables, Ross (1985) shows that adding internal conflict as an independent variable helps to predict external conflict. and that adding external conflict as an indepen' This srudy depends on the curnularive data seu irom diffeicnr and independent codingproiecrr rhat have been assembled an a common rct of racicrier and pinpointed rimer and ethnographic foci in the Srandard Cross-Cultural Sample (Murdock and Whire 1969. Barry and Schlrgel 1980). The cumulative database is in succerrivc installments of the World Culiurer elccrronic iournal. These dataover 1,250 variables from 50 different coding proiectrhave been assembled for elecrronicacccrs using rhr MAPTAB software package. MAPTAB way urcd in rhe prcscnr paper to relecr md recode relevaor variables, rodncioss. * , pasite scaler (see noteZ),and the Am-CORprogram (Reitz,Dow and Whire 1989) to terr multiple regression models and re-esrimarc final models gircn rhe hirrorical non-independence of cases. ? Validity and reliability are major issues in testing hypothercr from empirical measurer in comparativeresearch. Single measures of concepts provide n o means of assesring reliability or porrihle sources of bias in measurement. IWhilc intercoder reliabilitv is reooned in some srudics. estimates of aerecment rmone cad~~ ~ , , . ers workingon thcsameprojeet isinsufficient i ~ r m a n ~ p u r p o r c s i n srudpingreiiabiliry and bias: 1)thcy typically do nor apply to the entire sample, 2) researchers do nor publish "duplicate" codes from differcnrcod. ~ .. . err, but only summary or consensus codes. 3) rirhour multiple indrpcndenr coder, the independent codci contriburionr clnnar bccombined into summa. scales across praiccrs.4) arrerrmentsofcoding biarcr cannot be made]. An earlier paper (White 1989) on ~orld-system impacrsaf war explored the issuer of reliability, rneasuremenr bias, and validity in considerable derail. Wherever possible, this study user multiple measurer of each concept, chosen from independent studies. This allows exolicit asrer~ments of reliability (Whire 1990) and validity (Campbell and Fiske 19591, and uses . . . . . the reliability Bnd validity results to arrive at more accuiate measurement of concepts. This type of work cannot be done virh the results of aringle coding ~rojecr , since the best assessments of reliability and validiry aredone from strictly independenr measuraafrhesameorrimilar conceprs,suchar bydifferentcoding projects. dent variable helps to predict internal conflict. H e argues that this result supports the argument for a generalized disposition to conflict. One flaw in this argument is the fnilure to consider that structural conflicts leading to violence may also be generalized: dispositional factors are not the only explanation of generalii.cd (internal and external) conflict. Regional pressures and competition, for example, may act to generalize local level conflicts both externally and internally. Ross (1985: 553-554) states that "As a number of modern nations haveshown ..., teaching a citizenry to fight outside enemies often produces more fighters inside as well. Similarly, if violence is a mechanism for dealing with internal opponents, whv would we not expectit to he used with outsiders too?" The first proposition, however, does not necessarily imply a psychological disposition, and applies equal]!well to the case where structural conditions might effect a transfer of warfare skills from external to internal conflicts. The second statement, for which no evidence is adduced, exposes another possible flaw in Ross's dispositional argument, discussed below. There is a striking anomaly in Ross's data. Fourteen of his 90 cases (Ganda, Dahomey, Bambara, Basseri, Burusho, Saulteaux, Pawnee, Huron, Chiricahua. Papago, Aztec, Miskito, Mundurucu, and Timbira) are extremely high in external conflict, but I extremely low in internal conflict. Yet there are no comparable cases that are extremely high in internal conflict, and low in external conflict. The latter should occur, according to the theory (Ross 1986), with fraternal inteiest groups and patrilocal residence, the absence of cross-cutting ties and communiry intermarriage in decentralized societies, and the lack of socialization patterns that create the psychological conditions for the generalization of violence. According to his theory, one should find internal conflict in the absence of external conflict. But this does not fit the pattern of empirical observations. Thus, it would seem that Ross's structural theory is flawed when it comes to explaining the lack of cases of high internal conflict with low external conflict. Specifically, his theory (1) lacks structural causes of external conflict, and (2) posits structural causes of internal violence which predict a class of cases (high internal but l r ~ ~ v esternal conflict) which tends not to occur empirically. Thus, the anomaly in Ross's data and structural argument poses a serious problem for his dispositional interpretation. That is, if his dispositional argument for violence were valid, one would expect that cases would arise where the "disposition' to conflict was expressed within the population as internal conflict, but had not yet generalized outwards, either due to lag, opportunity, or isolation (e.g., island societies with "disposition to conflict" resultant from child training would haveinternal but not external conflict). But there are very few cases (Ifugao, Kapauku, Goaiiro, Trobriandcrs, Copper Eskimo) where internal conflicts outweigh external conflicts to a significant degree (and even then, the exceptions are marginal). Even in island societies in this sample (there are other well known exceptions), where there are no external enemies there tends to he little internal conflict. A dispositional model in which violence is endemic to some societies, beginning with internal and expanding to external violence, finds little support. 114 . Zeirrchiifr fiir Erhnolo~iv 115 (1990) Ross (personal communication) notes that differences between internal and external conflict might account for this anomaly: "Internal conflict is much more complex in the ways it creates division and cohesion among people who (by definition) share long standing social links. Internal conflicts much more easily become a feature of daily life which involve people in much more complex ways than external warfare does." Ross argues that a society with high internal but low external conflict is probably uncommon because such a society is unstable ("all the forces promote division and few produce cohesion", particularly since external conflict is lacking). "Internal cooperation and effective conflict management, because they are crucial for long term survival, are going to select against the high internalllow external conflict in societies in the long run. Hence their less then expected occurrence in the sample." In addition, although high external conflict may promote internal peace through ingroup cohesion, Ross (personal communication) argues that there is no reason for high internal conflict to promote external peace. If Ross's arguments explaining the anomaly are valid, however, they would require a modification of his structural hypothesis. Factors promoting internal conflicts would have to readily promote external conflicts in order to he easily generalized so as to account for the anomalous lack of high internal hut low external societies. H o wever, factors which predict internal conflict behavior, such as fraternal interest groups o r patrilocal residence, d o not appear to predict external conflict. O n e problem is that Ross's theory ignores the possibility that patterns of violence might originate in the field of external intersocietal relationship (i.e., structurally induced from the outside), and spread from the outside in, as in his example of nations training a military in skills that are then used in internal conflicts. Is the anomalous asymmetry in the distribution of internal and external conflicts statistically valid? Table I shows a cross-tabulation of Ross's inrernal and external violence scalesJ. There are very few cases in the lower left triangle of this table (high internal, low external), hut many cases in theupper right (high external, low internal). In fact, if one removed the lower left cells from the contingency table analysis, the rcsidual correlation between internal and external conflict would drop to zero, thus providing no support for the dispositional generalization argument. ' There scales for internal and external conflict, the main dependent variables employed in this study, are composed of seven and three items. rerpecrivcly, ar defined by Rorr (1985: 574) and coded by Rorn (1983). ?hey are combined into summan, indices bv the rinele-factor rcorine orocedure (White 19901 similar to ,. . Rorr's factur-5conng mcrhod. A; with ~os; 'r rc>ler,'rhe cxturnal ~onflicr r;ale loads highly bn warfare vsrnblcs, but theinternal conllicr scde lords hlghlv on moregencra.ired \iolence vsriablc5. Cronhlrh ~lplra reliaollotncr for rhrre rcrlrr i r e .YC md 8 8 , r r ~ p c c r t \ ~ c h For pumowr of ;rurr-trbulatlon. the factor <core . . values are reasrigred inreger values between I d d 9, w'here each scale has a mean of 5 and a standard deviation of 2. Douglas K. Whire: World-svsrem znd regional linkages 115 ! Table 1. Cross-Tabulation of external and internal violence scales External war Hifih internal conflict Law (dacr rescalcd from Ross 1983) Hich 3 3.84 2.01 1 4 1 1 7 5 5 4 7 1 0 0 1 11_1--33 I 8 5.67 1.41 Low 9 7.00 .OO 0 0 0 0 0 0 1 4 I Table 2 shows the same anomaly for Ross's (1983: 179-183) variables coding the frequency of war. Here, the frequency of external war tends to exceed that of internal war. This is again contrary to the view of internal war as the origin of external war. If internal war were (causally) a necessary condition of external war, especiallv in "pristine" conditions, external war should occur less often on its own than internal war. This version of the dispositional view predicts the opposite of what we find empiri-