SURVIVAL CANNIBALISM OR SOCIOPOLITICAL INTIMIDATION? Explaining Perimortem Mutilation in the American Southwest
John Kantner
University of California, Santa Barbara
Over the past two decades, archaeologists and physical anthropologists investigating the prehistoric Anasazi culture have identified numerous cases of suspected cannibalism. Many scholars have suggested that starvation caused by environmental degradation induced people to eat one another, but the growing number of cases as well as their temporal and spatial distribution challenge this conclusion. At the same time, some scholars have questioned the validity of the osteoarchaeological indicators that are used to identify cannibalism in collections of mutilated human remains. To address these concerns, this study attempts to reconstruct the behaviors that produced the Anasazi skeletal trauma by first examining ethnographic, ethnohistoric, and archaeological material for analogues useful for interpreting mutilated human remains and then correlating these analogues with the evidence from the Southwest. The patterns suggest that different behaviors are responsible for the Anasazi skeletal mutilation seen in different time periods. To explain these differences, the study employs game theoretical models that examine how changing social and physical contexts altered the sociopolitical strategies that Anasazi groups would likely have employed. The results suggest that violent mutilation and perhaps cannibalism was an intentional sociopolitical strategy of intimidation used during Pueblo II (A.D. 900-1100), while environmental changes after this period promoted resource-based warfare and the incidental skeletal trauma associated with this behavior. Received February 27, 1998; accepted May 14, 1998.
Address all correspondence to John Kantner, Department of Anthropology, UC Santa Barbara, Santa Barbara, CA 93106. E-math
[email protected] Copyright 1999by Walter de Gruyter, Inc., New York Human Nature, Vol. 10, No. 1, pp. 1-50.
1045-6767/99/$1.00+.10
2
Human Nature, Vol. 10, No. 1, 1999 KEYWORDS" Anasazi; American Southwest; theory; Violence; Warfare.
Cannibalism;
Game
For several decades, researchers have been intrigued by evidence of skeletal mutilation among the Anasazi of the northern American Southwest. The suggestion that these remains may represent prehistoric cannibalistic activity has stimulated the public's imagination (e.g., Gibbons 1997). Only recently, however, have skeletal materials been intensively investigated with the intention of identifying the behaviors that produced the trauma (Malville 1989; Turner 1983, 1993; Turner and Turner 1990, 1992a, 1995; Turner et al. 1993; White 1992). Because many scholars believe that cannibalism is a likely explanation for the osteological patterns, a considerable amount of investigation has focused on identifying taphonomic indicators of this behavior. The result of this research is a clearly defined list of skeletal features that together are intended to provide a compelling argument that cannibalism occurred. However, the justifications for these taphonomic features are often unclearly explicated, and few studies (e.g., Bullock 1991; White 1992) have investigated the ethnographic record for examples with which to model and understand the wide range of behaviors that can produce perimortem mutilation. Even fewer studies have comprehensively tackled the important question of w h y the Anasazi patterns of mutilation and suspected cannibalism emerged w h e n they did, thereby failing to place the behaviors within the dynamic sociocultural contexts that shaped them. The following discussion contributes to the investigation and explanation of the unusual Anasazi perimortem trauma. First, 35 ethnographic and ethnohistoric cases of behaviors that result in perimortem skeletal mutilation are examined. This information is then used to develop osteoarchaeological correlates of these behaviors. Next, the data from 41 skeletal assemblages from the northern Southwest that show evidence of perimortem trauma are examined. Factor analysis is used to identify distinctive patterns in the data, which are then compared with the behavioral correlates derived from the ethnographic information. The result is a series of conclusions regarding the dynamic behaviors that likely generated the unusual osteological patterns seen in the Anasazi archaeological record. After behavioral interpretations are derived from the archaeological evidence, the discussion focuses on an explanation of w h y the patterns of perimortem mutilation emerged when they did. This is accomplished using game theory, an actor-based approach useful for investigating behavior in social contexts. Game theory has been an integral part of microeconomics for several decades, where it is currently enjoying a resurgence of popularity. A neo-Darwinian interpretation of game theory has
Survival Cannibalism or Sociopolitical Intimidation?
3
also been an important tool in evolutionary biology, and even economists agree that its most successful applications have been in this field (Binmore 1992:414; Parker and Hammerstein 1985). However, despite some initial interest (e.g., Buchler and Nutini 1969; Davenport 1960), game theory has never been widely used in anthropology. No doubt this reflects the general disappointment that many social scientists felt when early formulations of game theory did not immediately live up to their promise to resolve a variety of theoretical problems. More recently, however, applications of game theory to anthropological problems (e.g., Bonhage-Freund and Kurland 1994; Boyd 1988; Dunbar 1991; Hawkes 1992, 1993; Nettle and Dunbar 1997) have demonstrated the ability of this approach to balance the elegance and generality of a simple modeling technique with the capacity to directly accommodate social behaviors. This study illustrates the potential of game theory by applying it to the problem of explaining perimortem violence in the Southwest.
BEHAVIORAL INTERPRETATIONS OF ANASAZI PERIMORTEM VIOLENCE
Most scholars in the Southwest have viewed perimortem skeletal mutilation as evidence of some form of cannibalism, and most behavioral explanations have centered on this interpretation (e.g., Malville 1989; Turner 1993; Turner and Turner 1995; White 1992). To identify that cannibalism occurred, investigators have focused on a list of six criteria that are believed to be good indicators of cannibalism in skeletal assemblages: intentional bone breakage, cutmarks, burning, anvil abrasions, missing and crushed vertebrae, and "pot polish" caused by cooking (Turner 1983:233-234; Villa 1992; Villa et al. 1986; White 1992). Proponents of this list have noted that any single criterion is not a sufficient indicator of cannibalism, instead emphasizing the need for multiple criteria and a consideration of the totality of evidence (Turner and Turner 1992b). However, the reasoning behind each of the criteria has not always been explicitly outlined, and many seem to have been derived through the analysis of archaeological collections that are presupposed to represent cannibalistic activity (e.g., Turner 1993:421-422; Turner and Turner 1995:3-7). Some critiques have also pointed to ethnographic accounts suggesting that m a n y forms of extreme skeletal trauma can be attributed to behaviors unrelated to cannibalistic activity (e.g., Bullock 1991; White and Toth 1991). And even if cannibalism can be shown to have occurred, this still begs the question of w h y it was practiced. For example, was the behavior a way to contend with starvation, was it part of warfare, or was it pathological?
4
Human Nature, Vol. 10, No. 1, 1999
EthnographicAnalogues In order to develop behavioral models of perimortem skeletal mutilation, ethnographic, ethnohistoric, and archaeological information on 35 societies was assembled (Table 1). When possible, a description of the activity, its supposed function in the society, and the archaeological manifestations of the activity were noted for each society. All of the societies currently or at one time reportedly engaged in practices that would have inflicted severe trauma to a person either before or immediately after death, including violent mutilation, mortuary preparation, and cannibalism. The sample of societies is not random and should not be interpreted as representing the relative frequencies of the described behaviors that might be expected in all societies. The point here is to use descriptions of these behaviors to develop models that can be used to interpret osteoarchaeological patterns. Of course, in the following discussion, the accuracy of ethnographic accounts of cannibalism or related activities could be challenged (e.g., Arens 1979; Salmon 1995), but in general the cases examined in this study are those derived from recent sources in which the behaviors in question were carefully investigated and judiciously reported. Twenty-three of the 35 societies reportedly engaged in acts of cannibalism either of unrelated people or of people known by the aggressors. Sixteen of these 23 societies typically participated in cannibalism as a component of intergroup warfare, while 10 also used cannibalism as a form of social control over people or groups within their own society. This behavior was as simple as eating small amounts of flesh (e.g., Hopi, BiminKuskusmin) or as complex as the consumption of entire bodies (e.g., Aztecs, Maori, Rossel Island). In most examples, the threat of cannibalism was used to terrify enemies and political rivals or to control subordinates. The notions of intimidation or revenge played a role in almost all of these cases, with torture, corpse mutilation, and other indignities often accompanying the cannibalistic acts. For example, among the Busama of New Guinea, "the bodies of the slain were carried back and treated with great i n d i g n i t y . . , the underlying notion [of cannibalism] seems to have been not so much a desire to acquire the dead man's strength as a determination to humiliate him to the greatest possible degree" (Hogbin 1951:147). For the Bimin-Kuskusmin, "such defilement is believed to preclude proper funerals and ancestorhood for the enemy dead" (Poole 1983:15). On Rossel Island, when a chief was killed or died, the person believed to be responsible was captured and eaten during the funerary rituals (Armstrong 1928:106-108). In the Chinese county of Wuxuan, the Cultural Revolution of 1968 was associated with interfactional cannibalism, for "to chop up, cook, and masticate was a . . . complete way of offending bodily integrity, depriving the enemy of humanity by reducing him to the status of a comestible" (Sutton 1995:146; see also MacCormack 1983:59).
Survival Cannibalism or Sociopohtical Intimidation ?
• o~
o ~ ~ ~ o
,n.n ~
5
~ ~
~
~ ~
~
~~ :no~
~'~
~
,-~
~ ~"
~
.~ ~
"d
0
9
~
~
~
"~
0
0
o
G
%
Ox
~.~
0
f
G
~
~
~
~
o
2 d
o
rj
.~
;~
--.o r~
.~.
19
S u r v i v a l C a n m b a l i s m or Sociopolitical I n t i m i d a t i o n ?
I
r
I
I
I
I
I
ei
i
i
t6
r
trg
I
I
~ 9
~
~,
0
0
r
,-;
=5
r
td
~ 9
~
r
~
,.6
,.~
I
I
0
20
H u m a n N a t u r e , Vol. 10, No. 1, 1999
,..~.
~i c~
o o.
o
'~
~
,g
,>.,.
,~
o
,~
o.
o.
,,6
~~
q
,>-.
o
O0
cxl
o
I
oo
o5 ~o
~
~
~.
o~
~. tr~
O
o5 t~
8 8 r~
Survival Cannibalism or Sociopolitical Intimidation ?
--
o
,
-~.-~
~=
.~
~-~
cb"
o.,..
~
oo
~,~_~,_~.~,.
. . . . .
21 ~, ~.,
,~
EEE~EE~
o
I
I
I
I
I
c~
c~
c~
~
9
9
~"
~,~.~ EE
I
I
I
N
c~
oc~
~
c~
I
8-,'.,
~.~o, ~-~o, EE I
I
I
9
9
I
o
o
~.
~
~.
~
o0
~-,
~o
~o.
~
o
~
~
L~
~0
C~O
i 0
.~
o
~
'~
~'-
~,~.
~
.~o
c~
,,~
~.
"0 .~
o.
>.-,
p~,
o
p..,
~
~
p..,
~
o
~
~o~o~
22
Human Nature, Vol. 10, No. 1, 1999 I
".
..
2
9
Mancos
Chaco Small House
9 ", ~
9
~
Pefiasco Bltnco
LPH65030 9
1
Bluff 9
-
t ,, T~r N~ P~ 9
0
-
9 ~ H 37593
Tragedy House
9
LPH 37592 9 Sin Juan 9
Battle Cave
9
9
Ctmyon Butle Ruin 3
Ye~owJac~t 5M~ 9 t
9
Rattlesnake
Hansen Pueblo
9
--
Burnt Mesa
9~ ~
9 ~
Wu!~tlo I LG Bg20 9
Red Mesa
423-124, t
~ Cottonwood 9
-1
9
9 S~to
~ i
~Bg31 e~ O
"-
~
.s
,,
//t
Red Mesa423-131
/t
t
/
,
/
LG
BgS1
t i t
-2
LG Bg88B
9
*.
i
/
--
// t
I 9 Pucb|o[~s,te" Pueblolll ....
-3 -2
9 othcrd. . . .
nknc~n 9
LG Bg2 9
i .....
~dftomsmP ectoA l.amb
I
I
I
-1
0
1
...... I
Burning and fragmentation
Figure3. Plot of factor scores illustrating temporal clustering as well as indicating those sites where the skeletal remains were recovered from suspected burial contexts.
A plot of the t w o factor scores for the 29 sites appears to divide the sample into three different clusters and one isolated case (Figures 2 and 3). The first cluster, found on the left side of both figures, can generally be defined as those sites with relatively less skeletal mutilation and lower frequencies of burning and fragmentation. Several variables not u s e d in the factor analysis support the integrity of this group. Almost all of these sites date to Pueblo III times ( A . D . 1100-1300) ; n o n e of the skeletal collections exhibit pot polish or scalping (Figure 2); and no faunal remains were associated w i t h the h u m a n skeletal material. The few assemblages d o m i n a t e d b y males are all found in this Pueblo III group as well, and in general the proportions of males are higher. Only t w o of the 13 collections in this cluster
Survival Cannibalism or Sociopolitzcal Intimidation ?
23
were recovered from contexts suggestive of intentional burial (Figure 3); most of the remains were instead found scattered indiscriminately on the floors of pueblo rooms or pitstructures. Despite the strong similarities between most sites in this group, there are also three anomalous sites. The skeletal assemblages from Teec Nos Pos, Chaco Small House, and LA 37593 exhibit high frequencies of anvil abrasions a n d / o r cutmarks but little burning and fragmentation; an increase in the latter two variables would move these aberrant sites into the cluster to the right. The next cluster illustrated in Figures 2 and 3 consists of skeletal assemblages that generally exhibit a higher degree of intentional damage, fragmentation, and burning. The most unusual feature that m a n y of these sites have in common is the presence of scalping and pot polish; in fact, all but one of the sites in the entire analysis with these taphonomic features fall in this group, even though these variables were not considered in the factor analysis. Almost all cases where the remains appear to have been intentionally (if not ceremoniously) buried are also found in this cluster (Figure 3), and the demographic profiles are not overrepresented by a specific sex or age group. Most importantly, most of the sites in this cluster date to Pueblo II (A.D. 900-1100). The exceptions include Hansen Pueblo, Marshview Hamlet, Polacca Wash, and one of the Ram Mesa sites. However, the first two date to A.D. 1134 and 1150, respectively, which many archaeologists consider to be on the boundary between Pueblo II and Pueblo III. The pitstructure at Ram Mesa may date to the early Pueblo I period, but the fill from which the skeletal materials were recovered may actually be later since the site is surrounded by a Pueblo II community. Polacca Wash, which will be discussed shortly, likely dates to around A.D. 1600. In addition to showing the two large clusters, the plot of factor scores shows a small group of two sites and an outlying case. The small cluster consists of Bg2 and Bg51, both of which are characterized by an absence of cutmarks, anvil abrasions, or pot polish, but high frequencies of bone fragmentation and burning. These two sites both date to Pueblo III. All skeletal materials were found on the floors of structures and were not associated with n o n h u m a n faunal remains. In the upper right-hand corner of the plot is Mancos, a lone outlier that affects the factor analysis by defining the upper limits of each variable. This site, which probably dates to the middle A.D. 1100s, exhibits high frequencies of all the taphonomic features used in this analysis. Mancos is alleged to provide the best evidence for starvation cannibalism (White 1992), and its isolation in the plot is therefore an interesting result. The factor analysis seems to produce "real" groups. All of the groups have many features in common beyond those included in the factor analysis, even though the factor analysis accounted for only 70% of the variance
24
Human Nature, Vol. 10, No. 1, 1999
in the data. More consistently recorded and reported data might have produced clearer results and may have assisted in the interpretation of the aberrant cases. Nevertheless, the analysis produced patterns that appear to have a basis in prehistoric reality. At the very least, they suggest that the mutilated h u m a n remains found in the northern Southwest were the result of two or more different behaviors that produced distinctive patterns of skeletal trauma. The next task is to compare the results with the archaeological expectations derived from the survey of comparative material to identify the behaviors that created these Anasazi skeletal assemblages.
Reconstructing the Behaviors Responsible for Perimortem Mutilation The two clusters of Pueblo III sites appear to exhibit patterns of skeletal trauma consistent with ethnographic descriptions of warfare accompanied by torture or corpse mutilation. A good example of the evidence for this behavior can be found at Bg88B (Mackey and Green 1979:147), where the mostly articulated and primarily male individuals were found in a heap on the floor of a pithouse. These remains exhibited perimortem fractures and embedded arrow points, and rocks had been thrown d o w n on the bodies. Similar patterns characterize all 15 of the sites in these two clusters. Additionally, all but two of the assemblages were found unceremoniously scattered on the floors of pueblos or pithouses. The demographic evidence also supports the argument that the remains represent the victims of warfare, for adults, and particularly males, are generally found in higher frequencies. Overall, relatively little perimortem mutilation seems to have taken place, and there is no evidence decisively demonstrating that cannibalism occurred. None of the sites in the Pueblo III clusters exhibit compelling evidence of cannibalistic activity, despite the fact that this was a period of environmental degradation; warfare presumably related to conflicts over resources seems to have been the response to these conditions, not cannibalism, and indeed there is substantial evidence indicating that Pueblo III was an especially turbulent period (Haas 1990; LeBlanc 1997; Wilcox and Haas 1994). The cluster of Pueblo II sites is of particular interest, for some of these skeletal assemblages have been highlighted as providing the best evidence for cannibalism (Malville 1989; White 1992). In general, the remains in this group exhibit extensive mutilation, and starvation cannibalism has often been proposed as an explanation for the patterns (e.g., White 1992: 361-363). However, there are four reasons to suspect that this explanation is inadequate. First, although information from most of the sites is incomplete, none of the skeletal assemblages were in association with nonhuman faunal remains. Second, in those cases where demographic data are
Survival Cannibalism or Sociopolitical Intimidation?
25
reported, each site exhibits a normal demographic pattern, as if an entire nuclear or extended family had been killed in a single event. The weak, young, or old do not appear to have been differentially mutilated, and the numbers of people found in these assemblages are inconsistent with the attrition that characterizes historical examples of starvation cannibalism (Askenasy 1994; Hardesty 1997). Third, the site with the most compelling evidence for cannibalistic activity, Mancos, actually lies outside of the Pueblo II cluster. Finally, and perhaps most importantly, almost all of the sites in this group are from a period in which the environmental conditions on the Colorado Plateau were benign relative to previous and subsequent conditions (Plog et al. 1988:235). If starvation was not a motivating factor for the perimortem skeletal treatment found in the Pueblo II assemblages, do the patterns fit any of the other behavioral models? The recovery of some of the remains from pits is suggestive of typical burials, supporting a hypothesis that the trauma was associated with mortuary behaviors. However, as Turner (1993:431) notes, typical Anasazi burials are "usually singular, and often with grave goods. Such burials occur in patterned body orientations and predictable cemetery or interment locations." None of the skeletal assemblages discussed in this report nor the contexts from which they were recovered fit this description (Table 2). This evidence, as well as the overall scarcity of Anasazi skeletal assemblages exhibiting consistent patterns of perimortem trauma, suggests that the processing of the deceased is an unlikely explanation for the skeletal treatment seen at the sites analyzed in this study. Warfare or social control involving extremely violent perimortem mutilation and even cannibalism seems a more likely explanation for the patterns identified in the Pueblo II group (Turner and Turner 1995:13). This conclusion is supported by a variety of evidence. First, evidence of scalping is found at many of the sites. The degree of violence to the corpses is also especially striking, causing Turner and Turner (1992a:677) to observe that "the Canyon Butte 3 facial destruction does not register well with simple straightforward emergency cannibalism." At Polacca Wash, [The evidence] suggests that the faces were crushed while still covered with flesh . . . . With the exception of one adult female skull which had received only a single great shattering blow (probably inflicted with a club) . . . . all other skulls had received from three to as many as 20 or 25 blows (Turner and Morris 1970:32-33). Similar violence to the crania is reported at Yellow Jacket 5MT3 (Malville 1989:12), Burnt Mesa (Flinn et al. 1976:313), and Chaco Small House (Turner 1993:428). Most of the sites in the Pueblo I1 group exhibit some
26
Human Nature, Vol. 10, No. 1, 1999
level of trauma to the facial region, and most of the assemblages have also been intentionally burned. There is some evidence indicating that the mutilation in the Pueblo II group was the result of violent social control rather than warfare. If these assemblages were the result of warfare over critical resources, the inhabitants of the sites would likely have been driven away and the sites abandoned (Ferguson 1990:32-33; Wilcox and Haas 1994:215, 234--236), as appears to have happened to several sites in the Pueblo III group. However, many of the Pueblo II remains were actually buried in pits and the sites continued to be occupied. This pattern may be more characteristic of violence stimulated by revenge, punishment, or intimidation (Table 1), and it is consistent with other indications that Pueblo II, although characterized by sociopolitical complexity, was a period that has not been associated with much evidence of warfare (Haas 1990; LeBlanc 1997; Wilcox and Haas 1994). An additional intriguing bit of anecdotal evidence supporting the argument for violent social control is provided by the inclusion of Polacca Wash in the Pueblo II cluster. There is considerable evidence indicating that the 30 individuals found in the pit at Polacca Wash were the victims of the legendary A.D. 1700 massacre in which the other Hopi villages attacked, captured, tortured, killed, and perhaps cannibalized people living in Awatobi village owing to their alliance with the Spanish (Turner 1993:435; Turner and Turner 1992a:676-677). Historical evidence suggests that the event was precipitated by an individual interested in promoting his own political position and supported by village leaders whose traditional authority was being threatened (Rushforth and Upham 1992:103-106). Similarly, Nass and Bellantoni (1982:268) speculate that the Monument Valley skeletal remains may be the result of the killing and mutilation of a family accused of witchcraft. If the Pueblo II assemblages analyzed in this study were the result of violent social control, were the remains also cannibalized, perhaps in an attempt to cause the highest insult to the deceased, humiliate kin, and terrify others? This remains a contentious issue (e.g., Bullock 1991, 1992; Turner and Turner 1992b). Turner and Turner (1992b:189) believe that the totality of evidence for the Pueblo II sites they have examined shows that cannibalism did occur. The higher degree of mutilation exhibited by the Pueblo II group does suggest behavior beyond mere torture or corpse mutilation. The presence of pot polish in this group may be the most convincing evidence of cannibalism (White 1992), for there w o u l d seem to be few other activities that could produce this osteological attribute. However, pot polish is a relatively recent feature of taphonomic studies that has not yet been extensively examined in other burial populations. As Turner notes (1993:427), "polishing is difficult to identify because it usually occurs on a very small surface area, usually less than 0.25 mm 2." The possibility of post-
Survival Cannibalism or Sociopolitical Intimidation?
27
depositional processes producing similar results has also not been ruled out. The evidence therefore suggests that cannibalism is a likely explanation for at least some of the Pueblo II sites, but severe perimortem mutilation without cannibalism could still conceivably explain these patterns. To summarize, the evidence suggests that two different sets of behaviors resulted in the perimortem skeletal mutilation found at the unique Anasazi sites examined in this study. The first behaviors, exhibite~l primarily by Pueblo II skeletal assemblages, consisted of acts of extreme violence possibly accompanied by cannibalism. The resulting osteoarchaeological patterns appear to be most similar to ethnographic examples of extremely violent social control that is often associated with factional competition or the maintenance of sociopolitical asymmetries. The second set of behaviors, exhibited primarily by Pueblo III assemblages, is associated with relatively minimal evidence of mutilation and probably no cannibalistic activity. The patterns are instead suggestive of trauma sustained during warfare that was likely precipitated by resource conflicts. However, identifying the different behaviors that produced mutilated skeletal remains in Pueblo II and Pueblo III is only half the battle. What contextual factors led to the emergence of extreme violence in Pueblo II? Why does this behavior appear to change in Pueblo III times?
E X P L A I N I N G THE E M E R G E N C E OF PERIMORTEM VIOLENCE
Explaining w h y these violent behaviors emerged among the Anasazi and w h y they changed over time is a difficult task. Scholars researching mutilation and cannibalism in the Southwest have generally addressed this question on a cursory basis, probably because much of their effort has been focused on identifying convincing evidence of cannibalism. In line with the influence of cultural ecology on Southwestern archaeology, the tendency has been to blame environmental degradation for alleged cannibalistic activity. For example, in White's brief discussion of the causes of the Mancos skeletal evidence (1992:360-363), he tentatively concludes that starvation was the motivating force behind the cannibalism that he believes occurred there (see also Flinn et al. 1976; Nickens 1975:291-292; Turner 1988:82). In contrast, Turner has moved from the view that the behaviors were caused by isolated cases of social pathology to the opinion that cannibalism was a form of institutionalized violence practiced by some segments of Anasazi society (Turner 1993:431-434). Turner (1997; Turner and Turner 1995:13-14; 1999) has recently proposed that the introduction of a Mesoamerican cult into the Southwest promoted cannibalism in certain populations. This is in accordance with his view that institu-
28
Human Nature, Vol. 10, No. 1, 1999
tionalized violence required that "the Anasazi had evolved, borrowed, or had imposed upon them from Mesoamerica, complex social organization with ranked individuals and specialized organizations, including quasimilitary war societies" (Turner 1993:434). Even if these scenarios are based on convincing reconstructions of the behaviors that created the skeletal assemblages, few adequately explain w h y mutilation and perhaps cannibalism should have emerged when they did. After all, both environmental degradation and contact with Mesoamerica had occurred many times prior to Pueblo II and since Pueblo III, yet evidence for these violent behaviors is temporally clustered. Furthermore, none of the proposed scenarios has recognized the intriguing differences between Pueblo II and Pueblo III patterns of mutilation. Perhaps most important, no approaches have explicitly recognized and modeled these violent activities as social behaviors. Killing and mutilating someone else occurs within a social context in addition to the physical environmental context that most theories emphasize. An explanation of the Anasazi patterns might therefore benefit from an approach that explicitly models social interaction while also considering temporal changes within the social and physical environment. As the Anasazi interacted with one another in their attempts to acquire both physical and sociopolitical resources, they would have made decisions not only in response to the physical environment but also according to the expected behaviors and reactions of other individuals and groups around them. For this reason, game theory is a particularly appropriate tool for explaining w h y perimortem violence appeared in Pueblo II and Pueblo III communities. To augment our understanding of the interactions represented by the patterns of mutilation, in this section game theoretical models are developed that evaluate different social strategies that can arise in contexts specific to Anasazi prehistory. The games are not as detailed as they could be, and they do not employ actual values derived from specific situations in the prehistoric Southwest. Rather, consistent with other applications of game theory to anthropological problems, general games are developed and their implications for the violent behaviors in question are discussed. This is an appropriate procedure for modeling prehistoric behavior since the parameters needed for developing detailed models would be difficult to specify. Another advantage is that the simple games developed in this study can be applied to similar prehistoric or contemporary situations. Note that in the following discussions, game solutions are derived using both the concept of evolutionarily stable strategies used in evolutionary game theory (e.g., Maynard Smith 1982) as well as the concepts of dominance and Nash equilibria prominent in noncooperative classical game theory (e.g., von N e u m a n n and Morgenstern 1944). In most cases, the results of the two approaches are the same (Cressman
Survival Cannibalism or Soczopolitical Intimidation ?
29
1996). After the models are developed and compared with the perimortem violence data from the Southwest, the utility of these game theoretical approaches in the analysis of h u m a n sociocultural behavior is discussed. Pre-Pueblo II Behavior
During the Basketmaker (before A.D. 750) and Pueblo I (A.D. 750-900) periods in the northern Southwest, most Anasazi groups were relatively small and widely dispersed in pifion-juniper woodlands over a large area. Communities appear to have consisted primarily of closely related individuals, and many were occupied on a seasonal basis, especially prior to the A.D. 700s (Gilman 1987; Wills 1991; Wills and Windes 1989). The values of basic resources such as arable land were probably low since populations were relatively small and mobile (Dean et al. 1994; Gumerman and GellMann 1994), and the archaeological evidence suggests that personal property was not rigorously delineated until very late in Pueblo I (e.g., Kohler 1992; Sebastian 1992:27-29). Sociopolitical differentiation appears to have been very ephemeral, with little undisputed evidence for leadership or unequal access to resources (e.g., Lightfoot and Feinman 1982; Mathien 1993:43-43; Sebastian 1992:103-105). In this context, we might expect that behaviors concerned with competition for resources would be consistent with the widely applicable strategies commonly modeled in the basic Hawk-Dove game. The Hawk-Dove game has been described in detail in m a n y contexts (e.g., Binmore 1992:416; Maynard Smith 1982; Rasmusen 1989:112-114). This game models the decisions that two individuals with equal power would make when contesting a resource that they both want, a situation that is clearly relevant to a large number of h u m a n interactions. The basic game is a symmetric, non-zero-sum model with two potential strategies. The Hawk-Dove model is abstracted in Figure 4, in which V indicates the value of the resource and C is the injury received for fighting over it. For this general model, the payoff (V) indicates the increases or decreases in units of the resource, but if we assume that the resource is necessary for somatic or reproductive functions, the payoff can also be a proxy for fitness (Smith and Winterhalder 1992:49). While the basic Hawk-Dove game is an excellent way to model numerous types of h u m a n interactions, the game benefits from minor augmentation. Specifically, two additions can be made that enhance the realism of the game without requiring additional assumptions or sacrificing its general applicability. First, the cost of fighting (C) can be divided into two costs: c is the cost to the individual for playing the Hawk strategy, while C is now the injury to the opponent of a Hawk. The second change to the Hawk-Dove game is the addition of a cost for playing Dove when meeting
30
Human Nature, Vol. 10, No. 1, 1999 Hawk
Dove
(V-C)/Z Hawk
(v-c)/2
V V
Dove
V/2 V/2
Figure 4. The basic Hawk-Dove Game in abstracted form. V is the value of the resource to the player, and C is the cost of fighting for the resource. a Hawk. Dove no longer gets away with no cost when it meets a Hawk, but rather pays a small cost that is a proportion of C. Cushing (1995) demonstrates that this is likely a more realistic result of Hawk-Dove interactions, particularly in social situations. The significance of these two alterations to the basic Hawk-Dove game will become apparent later in this discussion. The enhanced game, which we will call Costly Hawk-Dove, is illustrated in Figure 5. Depending on the specific relationship between values and costs, this game will have different equilibria. When the value of the resource is greater than the cost of injury from a Hawk (V > C) and also greater than the cost of playing Hawk (V > c), the best strategy for both players is to choose the Hawk strategy. However, when the value of the resource is less than or equal to the cost of playing Hawk (V < c), the best strategy for both players is to choose the Dove strategy. Finally, when the value of the resource is less than the damage you receive for fighting for it (V < C), there is no clearly dominant strategy, but instead the best solution is to play H a w k part of the time (with a probability of [V - 2c]/C) and Dove the rest of the time. In short, the best strategy in the Costly HawkDove game is to play Hawk until either the cost of playing H a w k (c) or the potential injury (C) gets too high. Note that the result of this game is the same whether one relies on the rational decision-making of classical game theory or the selection processes of evolutionary game theory. On the surface, the solution to the Costly Hawk-Dove game seems clear. However, although Hawk is the dominant strategy, the game is also a Prisoner's Dilemma in which the players receive less than they would if they could somehow agree to both play Dove. Furthermore, the payoff to this game is negative until the value of the resource is much larger than the costs (V > 2[C + c]). In those situations where the payoffs are negative, but the players still need the resource, the players will likely focus on conditional asymmetries to determine who wins (e.g., Hines 1987:217-219;
31
Survival Cannibalism or SoclopohticaI Intimtdation 7
Hawk
Dove
V/2-C-c
Hawk
V/2-C-c
-C/2
V-'c V-c
Dove
-C/2
V/2 V/2
The Costly Hawk-Dove Game, in which the cost of fighting for the resource is divided into two costs: c, the cost to play Hawk, and C, the injury that Hawk causes the other player.
Figure 5.
Packer and Pusey 1985). The most famous conditional strategy, Bourgeois, was first described by Maynard Smith (1982:22-23, 94-97) as a frequently encountered asymmetry that could be used to influence the outcome of Hawk-Dove competitions. The Bourgeois strategy is to play H a w k if the player is the first to find a resource and thus become its owner, but to play Dove if the resource is already owned by another player. In its usual incarnation, the Bourgeois strategy is calculated as if half the time Hawk is played and half the time Dove is played (Maynard Smith 1982:95). This game is illustrated in Figure 6. The Costly Hawk-Dove-Bourgeois game has different equilibria depending on the relationship between values and costs. When the value of the resource is greater than the injury caused by Hawks (V > C) and also greater than the cost of playing H a w k (V > c), there are two possible solu-
Hawk
Hawk
Dove
V/2-C-c
-C12
V/2-C-c
V-c
-C/2
V/2 V/2
3V/4-Cl2-c Bourgeois
Figure 6.
V/4-3C/4-c/2
V14-3CI4-c12 3V/4-C/2-c
V-c Dove
Bourgems
3V14-c/2 V/4-C/4
V/4-C/4 3V/4-c/2
The Costly Hawk-Dove-Bourgeois Game.
V/2 V/2
32
Human Nature, Vol. 10, No. 1, 1999
tions: to both play Hawk or to both play Bourgeois. However, note that the Hawk payoff is always worse than the Bourgeois payoff, suggesting that rational decision-making would tend to favor the Bourgeois strategy. When the value of the resource is less than the cost of playing H a w k (V < c), the game has two equally good solutions that yield the same payoff: play Dove or play Bourgeois. Finally, when the cost of injury is greater than the value of the resource (C > V), the game has one solution: to both play Bourgeois. In short, the Bourgeois strategy is effective in this game, for it always does better than H a w k and it does at least as well as Dove. As in the Costly Hawk-Dove game, these solutions are the same whether one relies on the rational decision-making of classical game theory or the selection processes of evolutionary game theory. The Costly Hawk-Dove-Bourgeois model is likely an accurate representation of the behavioral choices available to Basketmaker and Pueblo I social entities as they engaged in contests over important resources. These resources, such as arable land or good hunting territories, would have had comparatively low values since populations were relatively small and distributed over a large area. Individuals or groups competing for resources therefore would have had relatively little incentive to engage in aggressive Hawklike behaviors, particularly if the cost of playing Hawk was high. Instead, a basic Bourgeois conditional asymmetry would have likely dominated contests over resources since it would have yielded the best results for everyone involved. The Bourgeois solution indicates that evidence of Hawklike behavior in the archaeological record for the pre-Pueblo II Anasazi, such as warfare-related trauma, should be relatively uncommon. The skeletal evidence discussed earlier supports this conclusion, for very few examples of perimortem trauma used in this study date to these early periods in Anasazi prehistory. Nevertheless, specific contexts similar in effect to the prevailing conditions described for Pueblo II and Pueblo III may have promoted violent behaviors for which evidence is lacking in the mutilated skeletal remains analyzed in this study (e.g., Farmer 1997; LeBlanc 1997; Wilcox and Haas 1994:226-230). Pueblo II Behavior
The transition from Pueblo I to Pueblo II (A.D. 900-1100) is associated with changing physical environmental conditions and sociopolitical contexts in the northern Southwest. First was a proliferation of sedentary Anasazi communities (Dean et al. 1994), with some scholars suggesting that new groups immigrated into the region (Berry 1982; Judge 1989: 220-221). Associated with the growing populations was a reduction in the availability of necessary resources. Arable land would have been an especially valuable commodity, and this may explain the appearance of field-
Survival Cannibalism or Sociopolitzcal Intimzdation ?
33
houses as people became increasingly concerned with protecting their fields (Kohler 1992). Pueblo II mortuary practices (Akins 1986), architecture (Kantner 1996), access to prestige items (Mathien 1993), and even skeletal indicators of health (Akins 1986; Martin 1994:102) indicate that levels of sociopolitical and economic inequality were increasing, perhaps owing in part to disparities in the productivity of resources that people held (e.g., Kantner 1996). This was also a time when multivillage sociopolitical systems such as the Chaco Anasazi and Mimbres were emerging (e.g., Cordell et al. 1994; Gumerman and Gell-Mann 1994; Sebastian 1992). In general, the overall context was one in which competition for resources was increasing at the same time that sociopolitical differentiation was developing. The game theoretical perspective indicates that Bourgeois strategies would cease to be successful in the context of these Pueblo II conditions. This is because the Bourgeois strategy encounters problems when there is a decreasing quantity of a nonrenewable resource. As fewer and fewer unclaimed units of the desired resource are available, the Bourgeois player seeking additional units will fail to obtain any; since Bourgeois owners act like Hawks, the resource-seeking Bourgeois player is forced to act repeatedly as a Dove and retreat after suffering some injury (C/2). When this situation develops, the Bourgeois strategy is no longer viable and the game reverts back to the original Costly Hawk-Dove model. Subsequently, as the value of the resource begins to increase, the H a w k strategy will be selected by those individuals or groups desperate to obtain resources. In this context, players would seek other asymmetries to use to their advantage in contests over resources. Those individuals and groups benefiting from the unequal sociopolitical positions that appear to have emerged during Pueblo II would clearly have had an advantage in these interactions. Specifically, the class of game theoretical strategies known as "threat" strategies is particularly relevant to the Pueblo II situation and the development of perimortem violence in the Southwest. Although the "threat" strategies are most often modeled as complex dynamic games (e.g., Rasmusen 1989:96-101), they can also be modeled in other ways (e.g., Brains 1990:139-154; Maynard Smith 1982:151). One solution is to incorporate the concept of the "threat" with the symmetric Costly Hawk-Dove game by adding a new strategy that is a modification of the original H a w k position. Recall that in the Costly Hawk-Dove game, when the injury caused by a H a w k and the cost of playing H a w k are less than the value of the resource (V > C and V > c), H a w k is a dominant strategy, whereas the opposite conditions lead to either a Dove or a mixed strategy. During Pueblo II, as V increased and the Bourgeois strategy ceased to be viable, everyone would end up playing H a w k and paying high costs when competing for resources. However, for the self-interested social entity, one w a y
34
Human Nature, Vol. 10, No. 1, 1999
around this predicament would be to make the cost of playing Hawk (c) as low as possible while raising the injury caused to your opponents (C) as high as possible so that, in effect, they are forced to play Dove. This introduces a new option that might be called the "Intimidator" strategy. The Intimidator threatens to cause extreme injury (I) if the opponent does not play Dove and retreat from the contested resource. Because this is only a threat, the Intimidator initially pays a low cost (i), although in theory they do have to make the threat credible by actually following through on occasion. Note that when two Intimidators meet, they both play Hawk, under the assumption that there is little point in trying to threaten an Intimidaton The Hawk-Dove-Intimidator game is outlined in Figure 7. The Intimidator strategy is extremely effective, for it clearly dominates the Hawk and Dove strategies regardless of the value of C. As long as the cost of playing Hawk is greater than the cost of being an Intimidator (c > i) or as long as the injury caused by an Intimidator is greater than that caused by a Hawk (I > C), the Intimidator strategy is dominant. When the value of the resource is less than the cost of playing Hawk (V _
