This article investigates the origins and antiquity of Native American (Indigenous North American) dice, games of chance, and gambling. For over a century, it has been well documented that the making of dice and the use of these objects for gambling and games of chance were widespread practices among historic Native American groups and that prehistoric artifacts resembling such dice are present in the North American archaeological record (Culin Reference Culin1907). Despite these facts, and continued finds of similar prehistoric artifacts in the intervening years (e.g., Figure 1), the antiquity of this phenomenon remains a mystery because no systematic attempt has been made to identify its earliest manifestations and track its persistence through prehistory.
Figure 1. Late Pleistocene (13,000–11,700 BP), Early Holocene (11,700–8000 BP), Middle Holocene (8000–2000 BP), and Late Holocene (2000–450 BP) diagnostic and probable prehistoric Native American dice: (a, d) Signal Butte, Nebraska (Middle Holocene), NMNH-A437076, NMNH-550791; (b) Agate Basin, Wyoming (Early Holocene), UW-11327; (c, f) Agate Basin, Wyoming (Late Pleistocene), UW-OA111, UW-OA448; (e, g) Lindenmeier, Colorado (Late Pleistocene), NMNH-A442165, NMNH-A440429; (h) Irvine, Wyoming (Late Holocene). (Figures 1a, d, e, and g courtesy of the Division of Anthropology, Smithsonian Institution, American Museum of Natural History. Figures 1b, c, f, and h courtesy of the Department of Anthropology, University of Wyoming. Photographs by the author.) (Color online)
This omission stems in large part from uncertainty about whether prehistoric artifacts in the North American archaeological record can be confidently identified as dice. As DeBoer (Reference DeBoer2001:237) observed, “Identification of [ancient Native American] dice is problematic.” This article attempts to address this problem and dispel some of this uncertainty through a two-step process. First, it develops an objective, morphological test for identifying prehistoric Native American dice based on diagnostic attributes shared among 293 sets of historic Native American dice from across North America documented in Stewart Culin’s (Reference Culin1907) ethnographic compendium Games of the North American Indians. Second, it applies this test to trace the origins and antiquity of these artifacts in the published archaeological record. This analysis has yielded two key findings with intriguing implications.
First, the evidence developed here suggests that Native American groups on the western Great Plains of North America were making two-sided dice (binary lots) and using them as randomizing agents in games of chance and for gambling by the closing centuries of the Pleistocene, no later than 12,000 years ago. This evidence comes from Folsom deposits at the Agate Basin, Lindenmeier, and Blackwater Draw sites in Wyoming, Colorado, and New Mexico, where carefully crafted artifacts exhibit the diagnostic attributes of historic Native American dice documented by Culin (Reference Culin1907). Remarkably, these Folsom artifacts predate the earliest currently known dice in the Old World by millennia (Depaulis Reference Depaulis2021). The significance of this finding is amplified by the fact that the invention of dice and games of chance are recognized by historians of mathematics and science as an important early milestone in humanity’s evolving recognition and understanding of the probabilistic nature of the physical universe and the observable regularities that underlie it (Acree Reference Acree2021:79–82; Bennett Reference Bennett1998:11–26; Bru and Bru Reference Bru and Bru2018; Burton Reference Burton2011:443; David Reference David1955, Reference David1962:21–22; Stigler Reference Stigler2014; Todhunter Reference Todhunter1865:1).
Second, the evidence developed here shows that artifacts exhibiting the diagnostic attributes of historic Native American dice appear in archaeological assemblages from diverse groups throughout all periods of North American prehistory—from the Late Pleistocene / Early Holocene (aka the Paleoindian period), around 13,000–8000 years before the present (BP); through the Middle Holocene (aka the Archaic period), around 8000–2000 BP; and into the Late Holocene (aka the Late Prehistoric period), around 2000–450 BP. This evidence suggests that Native American games of chance and gambling represent cultural practices of remarkable antiquity and persistence that have survived to and thrive in the present, now serving as the foundation of one of the most visible expressions of modern Native American sovereignty: tribal gaming. This notable persistence also suggests that ancient Native American games of chance and gambling were highly adaptive for the groups that engaged in them, acting as “social technologies of integration” (Weiner Reference Weiner2018:2) that allowed disparate groups with little or no preexisting relationships to interact; exchange goods, information, and mates; and forge new social bonds (DeBoer Reference DeBoer2001; Janetski Reference Janetski2002; Yanicki Reference Yanicki2021).
Finally, when the deep antiquity and persistence of Native American dice, games of chance, and gambling are considered together with their likely role as a tool of social integration, possibilities are created for new insights into open questions regarding Early Paleoindian social practices in the Great Plains. Specifically, the recovery of dice from Late Pleistocene deposits at the Agate Basin, Lindenmeier, and Blackwater Draw sites offers support for the unsettled proposition that Folsom bands on the Great Plains came together at periodic communal aggregations (Hofman Reference Hofman1994; LaBelle and Meyer Reference LaBelle and Meyer2021), and it suggests that dice might serve as an archaeological “signature of aggregation” (Hofman Reference Hofman1994:350) that could assist in identifying such locations in the archaeological record.
Background
Stewart Culin began the quest to understand the origins of Native American dice, games of chance, and gambling in his 1907 masterwork Games of the North American Indians. In this 809-page volume, Culin described and illustrated historic dice from across the continent and exhaustively documented the games of chance in which these dice were used. He noted that although rules varied, the essentials of the games remained generally consistent: (1) players took turns throwing sets of two-sided dice on a surface (or tossing them in a bowl or basket); (2) a score was assigned based on a count of the “up”-facing dice thrown by each player on their turn; (3) players’ cumulative scores were kept, often with counting sticks; and (4) the winner was the first player to reach a predesignated score (Culin Reference Culin1907:44–45). In the decades following Culin’s work, prehistoric artifacts resembling the historic Native American dice described by him were recovered in numerous North American archaeological excavations, and scholars noted this similarity (e.g., Roberts Reference Roberts1936:31).
Following Culin, scholars gathered evidence tracing Native American dice, games of chance, and gambling back in time from the historic period, though with limited success. DeBoer (Reference DeBoer2001) documented the manufacture of two-sided Native American dice and their use in games of chance and gambling through the colonial period, citing historical accounts from AD 1634 among the Massachusett, AD 1636 among the Huron, AD 1709 among the Carolina Siouans, and AD 1883 among Plains groups. DeBoer also compiled a list of various North American prehistoric artifacts identified in site reports as “gaming pieces” (Reference DeBoer2001:Tables IV–VI; see also Gabriel Reference Gabriel1996:Appendix III), but he drew no conclusions about how these datasets might illuminate the greater antiquity of Native American dice and games of chance. His reluctance to do so stemmed from his concern that there was little way of knowing whether the label of “gaming pieces” that was applied to these artifacts was anything more than a “guessed appellation” (DeBoer Reference DeBoer2001:237).
Other scholars have made advances on this front by connecting their finds to the ethnographic examples illustrated and described by Culin (Reference Culin1907). Morris and Burgh (Reference Morris and Burgh1954:63) noted that “gaming pieces” identified in early Ancestral Puebloan, Basketmaker II, and Pueblo I contexts closely resembled those documented by Culin, and they therefore argued that these artifacts were dice used in games of chance and gambling. More recently, Hall (Reference Hall2008), Janetski (Reference Janetski and Voorhies2017), Riggs (Reference Riggs2021), Yanicki and Ives (Reference Yanicki, Ives and Voorhies2017), and Weiner (Reference Weiner2018) have used comparisons to Culin’s ethnographic examples to identify two-sided Native American dice as far back as 2,000 years before the present among the prehistoric Fremont and Promontory peoples of Utah and the inhabitants of Chaco Canyon in northern New Mexico. Although these scholars’ contributions have advanced our understanding of Native American dice and gaming, no connection has been made between these relatively recent prehistoric appearances and earlier artifacts, and no attempt has been made to track this cultural practice farther back in time, leaving its origins unresolved.
The efforts of Culin (Reference Culin1907), Morris and Burgh (Reference Morris and Burgh1954), Gabriel (Reference Gabriel1996), DeBoer (Reference DeBoer2001), Hall (Reference Hall2008), Janetski (Reference Janetski and Voorhies2017), Yanicki and Ives (Reference Yanicki, Ives and Voorhies2017), Weiner (Reference Weiner2018), and Riggs (Reference Riggs2021) yield two important conclusions that serve as foundations for this study: (1) Native American groups have manufactured two-sided dice continuously from at least 2,000 years ago to the present; and (2) throughout this time, these dice have maintained a consistent, specific, and well-documented function as randomizing agents in games of chance and for gambling. In other words, these scholars have shown that the presence of two-sided Native American dice in archaeological assemblages strongly suggests the presence of games of chance and gambling.
Despite this important, foundational work, the antiquity of Native American dice, games of chance, and gambling beyond the last 2,000 years has remained unclear—though not from a lack of probative evidence. As DeBoer (Reference DeBoer2001:237) has noted, “Gaming pieces abound in archaeological site reports . . . [but] basic footwork needs to be done” because “there is no Culin-like survey of archaeological dice on a continental scale.” Moreover, without a well-accepted method for confidently identifying this artifact type, concern has remained that the identification of dice outside of well-documented contexts runs the risk of being little more than “an archaeological Rorschach test” (DeBoer Reference DeBoer2001:237).
This article undertakes to address this problem and dispel the uncertainty surrounding the identification of prehistoric Native American dice in two ways. First, it addresses the problem of accurately identifying prehistoric artifacts as dice by using the voluminous record of historic Native American dice compiled by Culin (Reference Culin1907) to derive an objective, morphological, attribute-based test for more confident identification of prehistoric examples. Second, it uses this morphological test to identify prehistoric Native American dice in the North American archaeological record, tracing their appearances over time and space. The combination of these two datasets enables new insights into the antiquity of Native American dice, games of chance, and gambling, and the implications of their appearance in the deep past.
Methods
Deriving a Morphological Definition of Native American Dice
The ethnographic dataset that forms the basis for the present study is Stewart Culin’s (Reference Culin1907) magnum opus Games of the North American Indians. This 809-page volume, containing 1,112 illustrations and 21 plates, catalogs the results of Culin’s nearly 14-year effort to compile a “classified and illustrated list of practically all the American Indian gaming implements in American and European museums, together with a more or less exhaustive summary of the entire literature of the subject” (Culin Reference Culin1907:30).
As described by Culin, his study began by examining the collection of the Field Columbian Museum of Chicago, where curator of anthropology George A. Dorsey had previously undertaken a “systematic collection of specimens of gaming implements of all the existing tribes” (Culin Reference Culin1907:29). Dorsey also provided Culin with voluminous field notes and manuscripts that he had prepared on the subject, and he served as a liaison for Culin with Native groups, posing questions and bringing back responses (Culin Reference Culin1907:29–30). In 1900, Culin traveled with Dorsey to a number of Native reservations, where they collected new materials and reports (Culin Reference Culin1907:30). Thereafter, from 1901 to 1905, Culin made yearly trips on his own to visit Native groups in furtherance of his study (Culin Reference Culin1907:30). He also consulted other scholars and collectors who provided him with additional information and access to their collections (Culin Reference Culin1907:30). In addition, Culin gathered descriptions of Native American games and gaming implements from explorers, missionaries, travelers, and traders (Voorhies Reference Voorhies and Voorhies2017:6).
His final report includes illustrations and descriptions of 293 unique sets of Native American dice from “130 tribes belonging to 30 linguistic stocks,” and it notes that “from no tribe [do dice] appear to have been absent” (Culin Reference Culin1907:48). In addition, Culin cites and quotes at length 149 ethnographic accounts of how these dice were used to power games of chance and for gambling. Based on this record, Culin suggested that “the wide distribution and range of variations in the dice games points to their high antiquity” (Reference Culin1907:48).
A careful examination of the historic Native American dice documented by Culin reveals common diagnostic attributes that can form the basis of a morphological definition that can be applied to prehistoric artifacts. As illustrated and described by Culin—and discussed in more detail below—historic Native American dice share four key diagnostic attributes (Figures 2–4). First, they are two-sided objects made of bone or wood. Second, their two sides are distinguished by applied color or markings. Third, their appearance in section is either flat, plano-convex, concave-convex, or convex-convex (with the latter in all cases being a peach or plum stone; Culin Reference Culin1907:45–46, 51). Fourth, they are of a size and shape such that two or more can be held in the hand and cast onto a playing surface. Each of these attributes is considered below.
Figure 2. Flat dice types illustrated by Culin: (left panel) “bone dice” (Reference Culin1907:Figures 21, 67, 189); (right panel) flat “stick dice” (Reference Culin1907:Figures 14, 30, 124).
The first diagnostic characteristic of historic Native American dice documented by Culin is that they are two-sided objects made of wood or bone (Figures 2–4). Of the 293 unique sets of historic dice illustrated and described by Culin, 97% (n = 284) are two-sided. As such, they are “binary lots,” random number generators that “when thrown and allowed to fall at random, would each come to rest with either of two distinct faces uppermost” (Parlett Reference Parlett2018:20–22). As described by Culin (Reference Culin1907:57), “The sides which fall uppermost express the count or the failure to count,” meaning the “‘marked’ and ‘unmarked’ sides are equivalent to the ‘1’ and ‘0’ of binary arithmetic” (Parlett Reference Parlett2018:22). Of the 284 sets of two-sided dice described by Culin, 96% (272 of 284) are made of wood or bone, with wood (including cane, peach and plum stones, and nut shells) accounting for 70% of the sets, and bone (including teeth and turtle plastron) accounting for 26%. The remaining 4% are made of postcontact or one-off materials (china, 2%; brass, <2%; shell, <1%; corn, <1%).
The second diagnostic characteristic of the historic two-sided Native American dice documented by Culin is that their two sides are readily distinguishable, typically because of the application of color or markings to one side, and in rare cases, simply based on shape (Reference Culin1907:45). All of the 284 sets of historic two-sided dice illustrated and described by Culin have their sides distinguished in one of these ways. In almost all cases (278 of 284, or 98%), one side is marked in some readily discernible way—that is, colored with paint in various hues; stained with pigment, such as hematite; or inscribed or burned with lines, notches, designs, or other markings (Figures 2–4). The rare exceptions with no markings (6 of 284, or 2%) have sides distinguished only by shape, with one side being convex and the other flat or concave (e.g., Culin Reference Culin1907:148).
The third diagnostic characteristic of historic two-sided Native American dice documented by Culin (Reference Culin1907) is their appearance in section. These dice occur in four shapes: (1) flat, (2) plano-convex, (3) convex-concave, and (4) convex-convex. Flat dice have two flat (planar) sides (Figure 2). They represent 35% (99 of 284) of the two-sided historic dice sets described by Culin (Reference Culin1907). Flat dice are categorized into two subtypes: “bone-dice” and “stick-dice” (Culin Reference Culin1907:46, 51). Bone-dice are primarily circular or oval in plan, although other shapes appear, and are mostly (but not exclusively) made of bone (Figure 2, left panel). Stick-dice are oblong ovals or rectangles in plan and are frequently (but not always) made of split sticks smoothed flat on both sides (Figure 2, right panel).
Plano-convex dice (Figure 3) have one flat side and one convex (rounded) side (e.g., Culin Reference Culin1907:159). They represent 43% (123 of 284) of the two-sided historic dice sets described by Culin (Reference Culin1907). Culin categorizes plano-convex dice into two subtypes: “stick-dice” and “wood dice” (Figure 3; Culin Reference Culin1907:46, 51). Plano-convex stick-dice are oblong-rectangular in plan, plano-convex in section, and typically made of wood—often a round stick cut to size and split lengthwise (Figure 3, left panel). Wood dice are rectangular in plan, plano-convex in section, and usually made of a thick, rounded piece of wood—often a thick stick or branch cut to size and split lengthwise (Figure 3, right panel).
Figure 3. Plano-convex dice types illustrated by Culin: (left panel) round “stick dice” (Reference Culin1907:Figures 80, 146, 280); (right panel) “wood dice” (Reference Culin1907:Figures 90, 97–98).
Concave-convex dice (Figure 4, left panel) have one rounded (convex) side and one concave side. When thrown, the concave side presents a “flat” surface by resting on the playing surface in the same manner as plano-convex dice when the convex side lands up. These represent 13% (38 of 284) of the two-sided historic dice sets described by Culin (Reference Culin1907). Most concave-convex dice are categorized by Culin as “cane-dice” because they are typically made from cane or similarly hollow material, cut to size, and split lengthwise (e.g., Culin Reference Culin1907:98). The final shape, convex-convex, has two rounded sides and presents a special case (Figure 4, right panel). In all instances documented by Culin, these dice are peach or plum stones with one side marked to distinguish it from the other. They represent 8% (24 of 284) of the two-sided historic dice sets described by Culin (Reference Culin1907).
Figure 4. Concave-convex and convex-convex dice types illustrated by Culin: (left panel) “cane dice” (Reference Culin1907:Figures 122–123, 100, 289); (right panel) peach- and plum-stone dice (Reference Culin1907:Figures 99, 118, 238).
The final diagnostic characteristic of historic two-sided Native American dice documented by Culin (Reference Culin1907) is that they are of a size and shape such that two or more can be held in the hand and cast on a playing surface. All of the historic two-sided dice sets identified by Culin share this characteristic, which not only describes their general morphology but also captures their function and use. As illustrated and described by Culin, long and narrow stick dice and cane dice (Figure 2, right panel; Figure 3, left panel; Figure 4, right panel) are held in bunches and grasped around their narrow center with the ends protruding from the top and bottom of the closed hand (Culin Reference Culin1907:Figure 89). Smaller bone dice (Figure 2, left panel), wood dice (Figure 3, right panel) and peach/plum stone dice (Figure 4, right panel) are held in the palm of the hand (or both hands clasped) and cast, or they are placed in a basket or bowl and tossed (Culin Reference Culin1907:77, 92, 195).
These four diagnostic characteristics, distilled from Culin’s (Reference Culin1907) descriptions and illustrations, are used here to formulate an attribute-based morphological definition of Native American dice. However, two exclusions are made to prevent overinclusion. First, perforated objects—those with a drilled or pierced hole that extends through both sides—are excluded to prevent misidentification of beads or other decorative objects as dice. Only one of the 284 dice sets (<0.4%) illustrated and described by Culin (Reference Culin1907) includes a perforated die. Second, objects whose two sides are distinguished only by shape and not by markings are excluded to prevent misidentification of unmarked, two-sided, concave-convex, plano-convex, or convex-convex objects—such as split pieces of cane or stick—as prehistoric dice. Only six of Culin’s 284 dice sets (2%) are unmarked and distinguished only by shape. Although excluding objects that are perforated or whose sides are distinguished only by shape will prevent false positives, it should also be noted that, in some cases, these exclusions will necessarily produce false negatives. With these exclusions, the following four-part morphological definition is proposed: Native American dice are (1) two-sided, non-perforated objects made of wood or bone (2) with sides distinguished by applied color or markings (3) that appear in section as either (a) flat, (b) plano-convex, (c) concave-convex, or (d) convex-convex (and in this final case, are a peach or plum stone) and (4) are of a shape and size such that two or more can be held in the hand and cast on a playing surface.
This definition succinctly encapsulates the common morphological attributes of historic Native American dice. Any artifact meeting all four criteria would fit seamlessly and recognizably into this group. Moreover, this definition encompasses the functional nature of these artifacts. Two-sided Native American dice are binary lots, made and used to produce randomized outcomes that power games of chance (Culin Reference Culin1907; Parlett Reference Parlett2018:20). The definition captures this function because any artifact satisfying all four elements could have served as a binary random number generator suitable for use in a Native American game of chance. It not only allows for the identification of Native American dice in the existing archaeological record but also provides academic and cultural resource management (CRM) professionals with a succinct guide for identifying newly discovered artifacts as Native American dice.
Accordingly, based on the foregoing analysis of Culin’s (Reference Culin1907) illustrations and descriptions of 284 unique sets of historic Native American two-sided dice, his identification of these dice as implements for games of chance and gambling, and later scholars’ work connecting prehistoric dice of the last 2,000 years to games of chance and gambling (DeBoer Reference DeBoer2001; Janetski Reference Janetski and Voorhies2017; Morris and Burgh Reference Morris and Burgh1954; Weiner Reference Weiner2018; Yanicki and Ives Reference Yanicki, Ives and Voorhies2017), two working inferences are drawn here. The first is that prehistoric North American artifacts conforming to all four elements of the above definition can be identified with confidence as two-sided dice. The second is that these prehistoric dice likely served the same purpose as their historic counterparts: they were randomizing agents for games of chance and gambling. Proceeding from these inferences, the definition proposed above is used to identify prehistoric Native American dice, games of chance, and gambling in the archaeological record and trace their antiquity into the deep past.
On the Use of Ethnographic Analogy
In both instances, the above-stated inferences are based on ethnographic analogy—that is, inferring the function of an artifact type of unknown use based on formal similarities with an ethnographically documented cultural object of known use. Interpretation by ethnographic analogy has been a staple of North American archaeology since the early nineteenth century (Lyman and O’Brien Reference Lyman and Michael2001). It has been described as “not merely useful but essential to the interpretation of the archaeological record” (Randall and Hollenbach Reference Randall, Hollenbach, Walker and Driskell2007:223), and it is credited with the “ability to re-humanize the past and re-people ancient landscapes” (Eppich Reference Eppich2020:34). During the transition to processual and postprocessual archaeology, ethnographic analogy became “an object of uneasy mistrust among archaeologists” (Wylie Reference Wylie and Michael B.1985:63), based primarily on concerns that it failed to account adequately for cultural uniqueness and could therefore result in the uncritical projection of the present on to the past (McAnany and Woodfill Reference Metcalfe and Woodfill2020). More recently, however, “these debates have now mostly faded away” and ethnographic analogy “remains a preferred thread for coloring the past and making it recognizable to us and our audiences” (Lamoureux-St-Hilaire Reference Lamoureux-St-Hilaire2020:8).
Archaeologists attempting to identify the function of artifacts by reference to the ethnographic record typically employ two different types of analogies: direct historical analogies and general comparative (or cross-cultural) analogies (Eppich Reference Eppich2020; Randall and Hollenbach Reference Randall, Hollenbach, Walker and Driskell2007). In both cases, similarities and differences are noted between a source analog (the ethnographic artifact type of known use) and a subject analog (the archaeological artifact type of unknown use), and “if the resemblance in the form of the two artifact types is reasonably close, [the archaeologist] can infer that the archaeological type shares the technique, behavior, or other cultural activity which is usually associated with the ethnographic type” (Thompson Reference Thompson1958:5; see also Lyman and O’Brien Reference Lyman and Michael2001). In the case of direct historical analogies, cultural and geographic continuity exists between the source and subject analog, whereas in general comparative analogies, the source analog consists of repeated occurrences of the same artifact type, used in the same way, in a large number of otherwise unrelated cultural contexts (Deal Reference Deal2017; Lyman and O’Brien Reference Lyman and Michael2001).
Here, both of these forms of ethnographic analogy are arguably applicable and offer support for the identification of prehistoric Native American dice made here. For example, if the relevant cultural group and geographic area of the subject analogy are defined broadly as being Indigenous groups of North America, direct historical analogy would apply. This is so because the source analogy (the 284 sets of historic dice documented by Culin) would be seen as having a direct cultural and geographic connection with the subject analogy (the morphologically similar North American prehistoric artifacts identified here as dice). Conversely, if the relevant cultural groups and areas associated with the historic dice documented by Culin are defined more narrowly—as 130 separate groups operating in 130 different North American locations (Reference Culin1907:48)—this source analogy can be connected to the prehistoric artifacts identified here as dice by way of general comparative analogy, albeit one with unusually direct geographic and cultural connections.
Possible Alternative Uses
Finally, before moving forward with the definition of Native American dice presented above, it is important to note that the connection inferred here linking ancient dice with games of chance and gambling should not be interpreted as necessarily excluding the possibility that these dice were also used for other purposes, such as divination. Indeed, the use of dice (often binary lots) as a tool for divination has been documented ethnographically, historically, and archaeologically across the globe (e.g., Beerden [Reference Beerden2013], who discusses divination with dice in ancient Greece and Rome; Beeri and Ben-Yosef [Reference Beeri and Ben-Yosef2010], who discuss divination with dice in ancient Syria and Israel; Culin [Reference Culin1898], who discusses ethnographic examples of divination with dice in China, India, Japan, Korea, Liberia, Melanesia, New Zealand, Tibet, and Zimbabwe).
However, in North America, the evidence for a connection between dice and divination is sparser. Although Culin (Reference Culin1907) conducted an exhaustive study of the ethnographic record of Native American dice and their uses, he identified only one example of a reported connection with divination. He wrote, “I have no direct evidence of the employment of games in divination by the Indians apart from that afforded by Mr. Cushing’s assertion in regard to the Zuni sholiwe” (Culin Reference Culin1907:35). Moreover, the use of dice for divination in this one case appears to be less of an independent or alternative use for dice than an adaptation of a preexisting practice of using dice for games and chance and gambling. In the Zuni sholiwe, the act of divination was itself a game of chance in which priest-gamblers acting as proxies for the divine forces understood to be at play competed in an otherwise familiar dice game, with the winning player being seen as indicative of the forces then ascendent in—and therefore the probable outcome of—the matter under consideration (Culin Reference Culin1907:35, 210–225). Consequently, the record compiled by Culin provides only minimal support for the proposition that ancient Native American dice were used for divination as opposed to or in addition to for games of chance and gambling.
Searching for Prehistoric Native American Dice in the Archaeological Record
Having developed an attribute-based, morphological definition of Native American dice, the next step in the process undertaken here was to search for prehistoric artifacts in the archaeological record that satisfy its four-part test. As DeBoer (Reference DeBoer2001:237) has noted, artifacts identified as “‘gaming pieces’ abound in archaeological site reports.” Consequently, the search began with a review of compilations of references to gaming pieces compiled by previous scholars (DeBoer Reference DeBoer2001; Gabriel Reference Gabriel1996; Hall Reference Hall2008; Riggs Reference Riggs2021). The search was then expanded to include North American archaeological site reports contained in the EBSCO, eHRAF Archaeology, Google Scholar, JSTOR, and ProQuest electronic databases. These sources were searched for artifacts described as “gaming piece(s),” “dice,” “die,” or otherwise including the words “game” or “gaming.” In the course of these searches, references were found to similar artifacts reported in other publications not available in these electronic databases. Where possible, these additional sources were obtained from libraries, private collections, and other repositories. All of the searches described here were conducted by the author over a period of three years. The search did not include gray literature reports contained in nonpublic state historic preservation files. The search was intended to cover, to the extent possible, all of North America during all prehistoric periods of human occupation. Although the collections and subscriptions of the author’s home library (Colorado State University) are not unlimited and likely reflect a regional bias, the other electronic databases searched are not so limited and instead purport to include coverage for all regions of North America. Artifacts identified in these searches as possible matches were compared to the definition of Native American dice derived here, to the extent that information in the published reports was sufficient to do so. Based on these comparisons, the following two categories of relevant artifacts were identified:
1. “Diagnostic” prehistoric Native American dice: artifacts satisfying all four elements of the morphological test.
2. “Probable” prehistoric Native American dice: artifacts that satisfy some (and likely all) of the elements of the morphological test for which the evidence of one of the elements is unclear. Specifically, these are two-sided, non-perforated artifacts that are marked on at least one side and of the appropriate shape and size, but evidence regarding either the distinctness of the two sides or the shape of the artifact in section is either (a) suggestive but inconclusive or (b) not available due to limitations in the source material.
For both categories, data were collected (where available) on site name, location, dating, period (Late Pleistocene / Early Holocene, 13,000–8000 BP; Middle Holocene, 8000–2000 BP; and Late Holocene, 2000–450 BP), associated cultural complex, and morphological characteristics.
In addition, for the Late Pleistocene artifacts identified in the search as diagnostic or probable dice, an attempt was made to physically examine and photograph each object. This effort resulted in visits by the author to the Smithsonian Institution National Museum of Natural History, the University of Wyoming Archaeological Repository, and the Denver Museum of Nature and Science to view diagnostic and probable dice from the Lindenmeier (Colorado) and Agate Basin (Wyoming) sites.Footnote 1
Before considering the results produced by these efforts, it is important to keep in mind both the purpose of the search conducted for this study and the inherent limitations of the dataset produced by it. As discussed above, the goal of developing and applying a morphological test based on well-documented ethnographic examples is to overcome the pervasive uncertainty that has hampered past efforts to trace prehistoric Native American dice into the deep past by producing a sample that can be identified with confidence based on objective criteria.
However, in considering this sample, it should also be understood that it is both incomplete and an underrepresentation of the true corpus of prehistoric Native American dice in the North American archaeological record. This is so because the results reported in this article are based on a review of a published record that is highly fragmentary, frequently difficult to access, and characterized by large numbers of site reports that are inaccessible through online services, which are unavailable in hard copy from most libraries or through interlibrary loan or which exist only in the form of privately produced, “gray literature” CRM and compliance reports. Consequently, it is certain that numerous published reports of artifacts matching the morphological test derived here were not captured by this study. In addition, the sample described here is also underrepresentative (by design) because it excludes artifacts that are described in site reports as “gaming pieces” if they are unaccompanied by a description or image sufficient to apply the four-part morphological test derived here. Given these limitations, the results produced in this article should be thought of as a reliable indicator of where and when some Native American dice appeared in the ancient past, rather than as a definitive statement as to where all such dice appeared, or where dice were necessarily absent.
Despite these limitations, the sample produced in this study has significant value for tracing the origins and antiquity of prehistoric Native American dice. By focusing only on diagnostic and probable dice that can be identified with confidence based on their morphology and physical characteristics, this dataset allows for a consideration of this artifact class in a way that is unburdened by much of the uncertainty that has hampered past inquiries. With these caveats, the temporal, geographic, cultural, and subsistence diversity of the groups that made and used the diagnostic and probable dice identified are discussed below.
Results
The results of the search for prehistoric North American artifacts matching the four-part morphological test for Native American dice discussed above are reported in Table 1 and Figures 5–8. The findings include the following identifications:
1. 565 “diagnostic” Native American dice from 45 prehistoric North American archaeological sites
Table 1. Diagnostic and Probable Prehistoric Native American Dice.
Note: Dates stated in years before present (BP) are derived by (1) converting BC/AD reported dates to years BP or (2) calibrating reported radiocarbon dates using the OxCal 4.4 software (Bronk Ramsey Reference Bronk Ramsey2009) and the IntCal20 calibration curve (Reimer et al. Reference Reimer, Austin, Bard, Bayliss, Blackwell, Ramsey and Butzin2020) at a 95.4% probability.
a Dating information is from the primary reference cited or from the secondary reference designated “(D).”
2. 94 “probable” dice from 20 prehistoric archaeological sites (including 13 unique sites not in the diagnostic group)
In total, 659 diagnostic and probable Native American dice from 57 archaeological sites in 12 states, spanning more than 2,500,000 km2, were identified in site reports documenting more than a century of archaeological investigation of North American prehistory (Table 1; Figure 5). At a majority of these sites (40 of 55), more than one diagnostic or probable die was present in the archaeological assemblage (Table 1). At eight of these sites, diagnostic or probable dice were present in multiple, temporally distinct, stratified layers spanning thousands of years of successive occupations at the same location (Table 1).
Figure 5. Late Pleistocene through Late Holocene (13,000–450 BP) sites with diagnostic and probable Native American dice. (Dice illustrations by D’arcy N. R. Madden.)
Temporal Distribution and Continuity
The chronological distribution of the diagnostic and probable prehistoric Native American dice identified here is significant. They appear in deposits independently dated to all precontact periods (Late Pleistocene / Early Holocene, Middle Holocene, and Late Holocene), spanning roughly 13,000 years of North American prehistory (Table 1).Footnote 2
As previously noted, prior research has convincingly documented Native American dice, games of chance, and gambling from historic times (Culin Reference Culin1907), through the colonial period (DeBoer Reference DeBoer2001), and back as far as 2,000 years into prehistory among Basketmaker II, Fremont, Puebloan (including Chacoan; Janetski Reference Janetski and Voorhies2017; Morris and Burgh Reference Morris and Burgh1954; Weiner Reference Weiner2018), and Promontory cultures (Yanicki and Ives Reference Yanicki, Ives and Voorhies2017). The results obtained here confirm and extend our understanding of prehistoric Native American dice from the Late Holocene period (2000–450 BP), identifying 520 diagnostic dice and 71 probable dice from deposits at 46 archaeological sites dated to this period (Table 1; Figure 6).
Figure 6. Late Holocene (2000–450 BP) sites with diagnostic and probable Native American dice. (Die illustration by D’arcy N. R. Madden.)
Of even greater significance is evidence indicating the presence of dice—and by inference, games of chance and gambling—extending far deeper into the past. From the Middle Holocene (8000–2000 BP), 31 diagnostic and 12 probable dice were identified originating from 11 archaeological sites (Table 1; Figure 7). From the Late Pleistocene / Early Holocene (13,000– 8000 BP), 14 diagnostic and 11 probable dice were identified originating from six archaeological sites (Table 1; Figure 8). The Late Pleistocene Folsom record is particularly robust and will be discussed in more detail below.
Figure 7. Middle Holocene (8000–2000 BP) sites with diagnostic and probable Native American dice. (Die illustration by D’arcy N. R. Madden.)
Figure 8. Late Pleistocene / Early Holocene (13,000–8000 BP) sites with diagnostic and probable Native American dice. (Die illustration by D’arcy N. R. Madden.)
This remarkable temporal continuity is reflected in the appearance of Native American dice not only at a multiplicity of different archaeological sites over time but also in multiple strata at the same sites dated to different periods. For example, at the Agate Basin site in Wyoming, four diagnostic and probable dice appear in 12,000-year-old Folsom deposits, and a fifth appears in later deposits dating to around 10,000 years ago, associated with the more recent Agate Basin cultural complex (Table 1). Similarly, temporally distinct, stratified deposits containing Native American dice appear at (1) the Cowboy Cave site in Utah (six dice in two levels spanning more than 6,000 years), (2) the Sudden Shelter site in Utah (six dice in four levels, spanning more than 4,000 years), (3) the Pictograph Cave site in Montana (five dice in two levels, spanning more than 3,000 years), and (4) the Tularosa Cave site in New Mexico (12 dice in four levels, spanning more than 700 years; Table 1).
These results significantly revise our understanding of the antiquity of Native American dice, games of chance, and gambling. They demonstrate what appears to be a deep and recurring record of the manufacture and use of two-sided dice by Native American groups stretching back more than 12,000 years (Table 1; Figure 5).
Geographic Distribution
The great majority of sites with diagnostic and probable prehistoric Native American dice are concentrated in a relatively narrow corridor in western North America, approximately 1,200 km wide, covering more than 2,000,000 km2 (Figure 5). These sites—and those on the periphery of this corridor—extend from the western Great Plains to the Colorado Plateau and westward into the Great Basin region, and they are found in parts of Arizona, California, Colorado, Idaho, Montana, Nebraska, New Mexico, North Dakota, Texas, Washington, Wyoming, and Utah.
A striking feature of this distribution is that most of these sites cluster around the north–south axis of the six Late Pleistocene / Early Holocene sites containing the earliest examples of Native American dice (compare Figures 5 and 8). This geographic pattern, and its progression over time, suggest that dice, games of chance, and gambling may have originated along this axis in the Late Pleistocene and in the millennia thereafter spread to surrounding areas. As shown in Figure 8, the earliest Late Pleistocene / Early Holocene sites are clustered within a 750 km east–west and 1,300 km north–south corridor straddling the Rocky Mountains and extending through the states of Arizona, New Mexico, Utah, Colorado, and Wyoming. Thereafter, as shown in Figure 7, the Middle Holocene sites continue this geographic pattern, filling in and extending the corridor north and northwest into Idaho and Montana. Finally, as shown in Figure 6, by the Late Holocene, sites with Native American dice continue to be concentrated in this western corridor, with more distant examples found in Texas, North Dakota, Washington, and California. This concentration of sites in this core zone, beginning in the Late Pleistocene and continuing through the Late Holocene, supports the view that Native American dice, games of chance, and gambling originated in this western core over 12,000 years ago and then spread throughout the immediate region and reached into some western peripheries over the intervening millennia.
Another striking geographic pattern that emerges here is that no prehistoric (pre–450 BP) Native American dice were identified from sites in the eastern half of the North American continent (Table 1; Figure 5). This absence is unexpected given the historic and ethnographic record of the use of dice for games of chance and gambling in eastern North America in the period after European contact. Indeed, an analysis by DeBoer (Reference DeBoer2001) of the geographic distribution of ethnographically documented postcontact Native American dice games found them present among 18 Tribes in 11 eastern states (Alabama, Delaware, Illinois, Louisiana, Maine, Massachusetts, Michigan, New York, North Carolina, Ohio, and Pennsylvania) and three eastern Canadian provinces (New Brunswick, Ontario, and Quebec; DeBoer Reference DeBoer2001:217–226, Figure 2, Appendix).
These results raise the question of why no prehistoric dice from eastern North America were identified. One possible answer is that there are none to be found because dice were a western North American tradition that did not extend east of the Mississippi River until after European contact. The other possibility is that Native American dice were used in eastern North America before European contact, but some bias in this study or in the archaeological record has caused them to escape detection.
On the question of bias, as previously noted, the list of prehistoric sites containing Native American dice identified here cannot be considered to be a complete accounting—meaning the failure of the present study to identify prehistoric dice in the eastern North American archaeological record is not a determination that there were none. That said, the search conducted here was intended to encompass all of North America and included numerous databases (EBSCO, eHRAF Archaeology, Google Scholar, JSTOR, ProQuest) with coverage of the east of the continent, which means that if prehistoric dice were present there in any number, at least some of them should presumably have been identified here. It is also possible that the present failure to identify prehistoric dice in eastern North America is the result of one or more biasing factors in the discovery, preservation, or recording of such artifacts in the region, but no such bias is readily apparent.
Alternatively, if the geographic distribution found here is accepted as at least generally accurate, it suggests that prehistoric Native American dice were primarily—and perhaps exclusively—a western North American tradition that did not cross over into the east of the continent until after the period of European contact. It may be that the social turmoil caused by colonization—environmental and demographic change, disease, dispossession, and forced migration (Farrell et al. Reference Farrell, Burow, McConnell, Bayham, Whyte and Koss2021; Silliman Reference Silliman, Pauketat and Loren2005)—played a part in this apparent postcontact eastern expansion, bringing previously disconnected groups together and creating new possibilities for cultural diffusion. In these conditions, dice, games of chance, and gambling might have served an adaptive integrative function, creating a social context in which otherwise unfamiliar groups could interact and exchange with each other (as discussed in more detail in a later section of this article).
Cultural and Subsistence Diversity
The 57 sites containing diagnostic and probable prehistoric Native American dice are associated with diverse cultural groups with varied subsistence patterns. These sites are associated with 22 distinct cultural complexes (Clovis; Folsom; Agate Basin; Desert; McKean; Hueco; Mogollon; Fremont; Avonlea; Basketmaker II and III; Woodland; Pueblo I, II, III, and IV; Fremont; Uinta Fremont; Promontory; Apishapa; Mandan; and Plateau Interaction Sphere), as well as numerous unclassified prehistoric groups (Table 1). These groups range from mobile hunter-gatherers (Clovis, Folsom, Desert, McKean, Avonlea), to semisedentary groups (Basketmaker II and III, Mogollon), and to sedentary agriculturalists (Fremont; Pueblo I, II, III, IV; Mandan). This pattern suggests that the making and using of dice for games of chance and gambling were a cross-cultural phenomenon not tied to any particular language or ethnic group or to any social or subsistence adaptation.
Late Pleistocene Folsom Diagnostic and Probable Dice
The record of diagnostic and probable Native American dice associated with the Folsom cultural complex—a Late Pleistocene archaeological tradition dated between 12,845 and 12,255 BP (Buchanan et al. Reference Buchanan, David Kilby, Hamilton, LaBelle, Meyer, Holland-Lulewicz and Andrews2021)—is particularly robust, with examples appearing at three different sites in the western Great Plains: the Agate Basin site in eastern Wyoming, the Lindenmeier site in northern Colorado, and the Blackwater Draw Locality No. 1 site in eastern New Mexico. (Table 1; Figure 8). Thirteen diagnostic and seven probable dice were recovered in Folsom deposits at these sites (Figure 9; Table 2). These diagnostic and probable dice include examples of three of the dice types—flat, concave-convex, and plano-convex—that comprise 92% of Culin’s (Reference Culin1907) historic examples.
Figure 9. Folsom diagnostic and probable Native American dice. (Figure 9a, b, d, and g: Agate Basin, Wyoming, UW-OA005, UW-OA109, UW-OA111, UW-OA448, courtesy of the Department of Anthropology, University of Wyoming. Figure 9c: Lindenmeier, Colorado, DMNS-A900.179, courtesy of the Denver Museum of Nature and Science. Figure 9e–f, h–i, k–p, r: Lindenmeier, Colorado, NMNH-A443046, NMNH-A442165, NMNH-A44890, NMNH-A441178, NMNH-A440429, NMNH-A441841; NMNH-A442122, NMNH-A443755, NMNH-A443850, NMNH-A443658, NMNH-A441839, courtesy of the Division of Anthropology, Smithsonian Institution, American Museum of Natural History. Figure 9j: Lindenmeier, Colorado, CSU-7805-6, courtesy of the Department of Anthropology, Colorado State University. Figure 9q: Blackwater Draw, New Mexico; drawing by D’arcy N. R. Madden after Hester [Reference Hester, Lundelius and Fryxell1972:Figure 9b, by Phyllis Hughes]. All photographs, except [j], are by the author.) (Color online)
Table 2. Late Pleistocene Folsom (12,845–12,255 BP) Diagnostic and Probable Dice Attributes.
All 20 of the Folsom dice are two-sided, non-perforated objects, 19 of which are made of bone (Table 2; Figure 8). The sole outlier is a steatite (soapstone) disk with incised edges from the Lindenmeier site (Figure 8j). Because it is not made of wood or bone, this steatite disk does not technically meet the definition of Native American dice derived here. Nonetheless, it is included as a probable die because of its obvious similarity (in size and markings) to eight other edge-incised, disk-shaped dice recovered in the same Folsom deposits at the Lindenmeier site (compare Figure 8j with Figures 8c, f, k–l, n, q).
All 20 of the Folsom dice are also marked on at least one side. For 16 of them, these markings clearly distinguish the sides. For the remaining four, information is currently available for only the marked side of the artifact, which means that it is not possible to definitively establish that the sides are distinguished. For this reason, these four artifacts are identified as probable—rather than diagnostic—dice. All of the Folsom dice are also similar to the historic dice illustrated and described by Culin (Reference Culin1907) in the ways in which their two sides are marked and distinguished. For instance, all are marked with carved incisions that distinguish their sides (Table 2): in some cases, edge-ticking (Figure 9c, e–f, h, j–l, p–r), and in others, parallel or perpendicular lines (Figure 9a–b, d, g–i, m–n). This same technique appears on 28% (70 of 284) of the historic dice sets described by Culin (Reference Culin1907). Another common method for differentiating the sides of historic dice, appearing on 36% of the sets recorded by Culin (Reference Culin1907), is the application of red pigment to one of the sides. Similarly, three of the Folsom dice show traces of what appears to be a red pigment or stain on one side (Figure 10g–i).
Figure 10. Detail of Late Pleistocene Folsom dice: (left panel) shape in section of concave-convex and moderate to minimal plano-convex diagnostic and probable dice; (right panel) red coloration traces on diagnostic dice, with color enhancement using DStretch (YDT colorspace), a decorrelation stretch digital imaging tool created for the enhancement of rock art images (Harman Reference Harman2015). (Photographs by the author.) (Color online)
In addition, when viewed in section, all of the Folsom dice are of the appropriate shape called for by the definition of Native American dice derived here. Eleven of them are flat on each side (Table 2; Figure 8a, c, f, j–l, n–q), a shape matching 35% (99 of 284) of the two-sided, historic dice sets described by Culin (Reference Culin1907). Flat dice appear at all three Folsom locations (Table 2). Of these 11 flat dice (some whole, others fragmentary), nine are disc-shaped (Table 2; Figure 8c, f, j–l, n, q). Disc-shaped dice make up 37% (37 of 99) of the flat dice types described by Culin (Reference Culin1907). Some of the more complete Folsom discs closely resemble the flat “bone-dice” that appear repeatedly in Culin (Reference Culin1907; compare Figure 2, left panel, with Figure 9c, f, k). These flat, disc-shaped dice appear at both the Lindenmeier site in Colorado (n = 8) and the Blackwater Draw site in New Mexico (n = 1; Tables 2; Figure 8c, f, j–n, q).
Five of the Folsom dice are concave-convex in section and are found only at the Lindenmeier site (Table 2; Figure 8e, h–i, r), although a concave-convex die does appear in post-Folsom, Early Holocene deposits at the Agate Basin site dated to around 10,600–9500 BP (Figure 1b; Table 1). This shape represents 13% (38 of 284) of the two-sided historic dice sets described by Culin (Reference Culin1907; Figure 4, left panel). Three of the Lindenmeier concave-convex dice share the key characteristics of Culin’s “cane-dice” type in that they are made from a hollow material (a small, cylindrical bone) that is cut to size and split lengthwise to create the distinctive concave-convex shape in section (Figure 8e, r; Wilmsen and Roberts Reference Wilmsen and Roberts1978:Figure 128k). Two other dice from the Lindenmeier site are made from larger pieces of split bone shaft, with the cancellous material removed to create a concave-convex appearance in section (Figures 9h–I and 10d–e). One additional die from the Lindenmeier site has a shape that is suggestive of a concave-convex section, but it does not appear to be made from a split bone shaft. Instead, it seems to be a fragment of a discoidal object that, in section, appears to display a concave-convex shape (Figures 9m and 10i). Because of the ambiguity of the shape in section of this latter artifact, it is identified here as a probable rather than diagnostic die.
The remaining three Folsom dice are cut and incised rib sections from the Agate Basin site, similar to each other in outline and in the longitudinal lines incised on their faces, with shapes in section that vary from moderately to minimally plano-convex (Figure 9b, d, g; Figure 10a–c). Forty-three percent (123 of 284) of the two-sided historic dice sets described by Culin (Reference Culin1907) are plano-convex in section; Figure 3). The first of the three displays a clear convex curvature on one side but a slightly irregular flat surface on the other side (Figure 10a). The next has one side that is generally convex, whereas the other is somewhat flatter, although the flatness is not well defined (Figure 10b). The final artifact lacks a well-defined flat surface, and the curvature on both sides is generally symmetrical (Figure 10c). This shape is more similar in section to the convex-convex peach- and plum-stone dice that make up 8% of the two-sided ones described by Culin (Reference Culin1907; Figure 4, right panel). Because of the ambiguity of the shape in section of these latter two artifacts, they are labeled here as probable rather than diagnostic dice (Table 2).
Finally, all of the Folsom dice meet the fourth element of the definition derived here in that they are all of a shape and size that would permit two or more to be held in the hand and cast on a playing surface. Indeed, the size and overall shape of the Folsom dice have close analogs in the historic dice described by Culin (Reference Culin1907), with nine of them resembling historic disk-shaped bone dice (compare left panel of Figure 2 with Figure 9c, f, j–n, q); five resembling rectangular wood dice (compare right panel of Figure 3 with Figure 9b, d, g–i); three resembling ovoid, rectangular historic bone dice (compare left panel of Figure 2 with Figure 9a, h, o–p); and two resembling concave/convex cane dice (compare left panel of Figure 4 with Figure 9e, r).
Consequently, the evidence for identifying the 20 Folsom artifacts discussed here as Native American dice is robust. Their shape and markings closely match Culin’s (Reference Culin1907) historic dice types. Their number and presence at multiple Folsom sites suggest that they represent a distinct artifact class with a specific purpose. Most significantly, they are not temporal outliers separated from Culin’s historic dice by a dozen millennia. Instead, they connect to these well-documented historic examples through a near-continuous record of prehistoric Native American groups manufacturing and using two-sided dice over the intervening 12,000 years in the same geographic region (Table 1; Figure 5).Footnote 3
Implications
Probability in the Pleistocene
A significant implication of the findings reported here is that the earliest Native American dice appear to predate their earliest known Old World counterparts by millennia. As DePaulis (Reference Depaulis2021) has recently summarized, “The earliest objects we can understand with certainty as ‘dice’” appear from and after around 3500 BC (5450 BP) in Mesopotamia, the Indus civilization, and the Bronze-Age Maikop culture of the western Caucasus (DePaulis Reference Depaulis2021:11; see also Dales Reference Dales1968; Finkel Reference Finkel and Finkel2007; Piccione Reference Piccione1990:20, Reference Piccione and Finkel2007). Remarkably, these earliest Old World dice postdate the 12,000-year-old Native American Folsom dice discussed here by more than 6,000 years. Moreover, later Native American examples discussed here also predate the earliest-known Old World dice, including the 10,000-year-old die at the Agate Basin site in Wyoming (Figure 1c) and the six 7,000- to 8,000-year-old dice at the Sudden Shelter and Cowboy Cave sites in Utah (Table 1). Therefore, the findings presented here place prehistoric Native American groups at the forefront of the invention of dice, games of chance, and gambling.
This finding is all the more significant because historians of mathematics frequently identify the invention of dice and games of chance as a crucial early step in humanity’s evolving discovery and understanding of randomness and the probabilistic nature of the universe (Acree Reference Acree2021:79–82; Bennett Reference Bennett1998:11–26; Bru and Bru Reference Bru and Bru2018; Burton Reference Burton2011:443; David Reference David1955, Reference David1962:21–22; Hacking Reference Hacking2006:3–8; Stigler Reference Stigler2014; Todhunter Reference Todhunter1865:1). This is so because dice throws produce “characteristic random or patternless sequences of outcomes” to which “the calculus of probability applies straightforwardly” (Franklin Reference Franklin, Hájek and Hitchcock2016:1). Consequently, the making and using of dice represent humans’ first known efforts to intentionally generate, observe, and record streams of controlled, random events (David Reference David1962:ix) and harness the probabilistic regularities that appear in them (such as the so-called law of large numbers) to power games of chance. Indeed, the role of dice as a means of exploiting and exploring the nature of randomness is nowhere better displayed than in the fact that our modern understanding and formulation of probability theory originated with sixteenth- and seventeenth-century mathematicians, such as Cardano, Galileo, Fermat, and Pascal, who were studying the principles governing the outcomes of dice rolls in games of chance (Debnath and Basu Reference Debnath and Basu2015). As Laplace (Reference Laplace1902:195), a pioneer of probability theory, noted in 1814, “It is remarkable that a science which commenced with the consideration of games of chance, should be elevated to the rank of the most important subjects of human knowledge.”
The findings presented here suggest that some of the earliest steps on this intellectual journey were taken not by complex societies in the Near East and Eastern Europe around 5,500 years ago but rather by Native American hunter-gatherers in western North America in the waning centuries of the Pleistocene, no later than 12,000 years ago. This would represent an uncommon case in which the initial steps toward an intellectual advance central to modern civilization—the understanding and use of probability—originated in the New World.
The connection made here between ancient Native American dice and probability is premised on the objective behavior of these groups in making and using dice as observed in the archaeological record, informed by the recognition that “some appreciation of ‘chance’ is inseparable from the use of dice” (Erasmus Reference Erasmus1950:376). Although the depth and sophistication of this appreciation is impossible to know, there is some basis for expecting that it may have been nontrivial.
In a landmark article, foundational to the field of behavioral economics, Tversky and Kahneman (Reference Tversky and Kahneman1974:1130) argued that humans do not infer the statistical regularities embedded in everyday experience because they “are not coded appropriately”—meaning that the quantitative features inherent in these experiences are not isolated, noted, and organized in ways that reveal probabilistic patterns that are usually obscured by the noise of other incoming experience. Intriguingly, Native American dice games appear to perform such a “coding” function. They produce a simplified stream of random events that are carefully observed and recorded at multiple levels: in the scoring of individual dice throws, in the keeping of cumulative scores in single matches, and in tallying wins and losses in multiple matches over time as recorded by the giving or receiving of goods. Therefore, by observing and recording the patterns appearing in these outcomes, ancient Native American dice players repeatedly presented themselves with the very type of “coded” experiences that Tversky and Kahneman (Reference Tversky and Kahneman1974) argued would allow humans to observe and infer the presence of underlying probabilistic regularities.
As to the more subjective question of how the Native American groups that made and used dice for games of chance and gambling might have thought about or conceptualized chance and the random outcomes produced by dice, the evidence is opaque, although some clues can be gleaned from descriptions of the beliefs and practices of historic groups that carried forward this ancient practice. For example, numerous ethnographic accounts of Native American traditions depict dice playing as a sacred activity that was inherently pleasing to the gods and celestial powers (who were themselves dice players), with ceremonial and secular dice games being played at festivals and seasonal events “with the object of securing fertility, causing rain, giving and prolonging life, expelling demons or curing sickness” (Culin [Reference Culin1907:34], Zuni; see also Culin [Reference Culin1907:116, 119], Seneca, Wyandot; Dye [Reference Dye and Voorhies2017], Pawnee, Menominee).
Beyond the propitiation of higher powers, ethnographic reports of Native American traditions regarding dice games and gambling suggest something more fundamental: a conception of randomness and chance as powerful and essential forces that control the fates of men and constrain the will of the gods—both on earth and in the sacred realms. For example, in a Zuni story in which the Twin War Gods are engaged in a game of dice, their playing surface is described as being akin to the surface of the earth, the place where “the gods themselves count up the score of their game” as they “play games of hazard, wagering the fates of whole nations in mere pastime” (Culin Reference Culin1907:33; Cushing Reference Cushing1901:387). Likewise, the Zuni divination ceremony, based on dice and gambling, is described as an attempt to gain insight into “the natural powers by which humanity was assumed to be dominated” (Culin Reference Culin1907:35). Similarly, Crow tradition describes the fate of humans as being controlled by the random outcomes of celestial dice games: “everything that happened to him depended on the fortune of his guardian spirit in a stick dice game . . . as the dice went, so went a man’s career.” (Eyman Reference Eyman1965:43).
This view that the random outcomes observed in dice games are manifestations of a pervasive and powerful natural force can also be seen in the creation stories of some historic Native American groups. In the Pawnee creation tradition, the bowl-and-dice game (in which dice are shaken in a slightly concave bowl or basket rather than tossed by hand) is described as representing the celestial, earthly, and watery/underworld realms, with the casting of dice deciding the outcomes of the relationships between them (Dye Reference Dye and Voorhies2017). Similarly, in the traditions of the Maricopa, Navajo, and Zuni Tribes, dice games are associated with creation stories, particularly those involving the emergence of humans (Culin Reference Culin1907:203–204, 480; Gabriel Reference Gabriel1996:88, 139; Stevenson Reference Stevenson1903:480). For example, in the Zuni tradition, the emergence of the first humans from the underworld is dependent on the cultural hero Poseyemu prevailing over his father, the Sun, in a dice game (Gabriel Reference Gabriel1996:139).
A common thread in these traditional accounts is the suggestion of a probabilistic worldview in which randomness—such as is produced and observed in the use of dice in games of chance—is conceived of as a powerful and sacred force that determines the fates of humans, nations, and the world. Although the gods play their part in deciding these outcomes, they do so not as all-powerful beings imposing their will from above but rather as players in a celestial game whose outcome is dictated by chance—a power seemingly beyond even their control. In this respect, these Native American traditions suggest an ontology that might be familiar to Western science: that “the true logic of this world is to be found in the theory of probability” (Debnath and Basu [Reference Debnath and Basu2015:1] quoting James Clerk Maxwell). Whether the ancient Native American groups that made and used the prehistoric dice identified in this article possessed such a probabilisitic worldview is impossible to say with any certainty on this record. What can be said with more certainty is that these groups appear to have been purposefully creating, operationalizing, and observing the properties of randomness and chance that underlie probabilistic ontologies.
Persistence, Adaptive Fitness, and Social Function
An additional implication of the findings presented here emerges from the remarkable temporal continuity of dice, games of chance, and gambling among Native American groups. In recent years, the persistence of Indigenous culture has been a focus of archaeological study in North America. In this respect, persistence is examined by considering cultural continuity in material culture and practices over time, including a focus on the effects of colonialism on Native American culture (Lightfoot et al. Reference Lightfoot, Martinez and Schiff1998; Panich Reference Panich2013; Pezzarossi and Sheptak Reference Pezzarossi and Sheptak (editors)2019). It has been argued that by examining the persistence of aspects of Native American culture, “archaeologists can provide important insights into the interconnected nature of continuity and change in indigenous polities and identities” (Panich Reference Panich2013:105).
More broadly, evolutionary archaeologists view cultural traits as reflected in material culture as “units of transmission that permit diffusion and create traditions—patterned ways of doing things that exist in identifiable form over extended periods of time” (O’Brien et al. Reference O’Brien, Lyman, Mesoudi and VanPool2010:3797). Generally speaking, these patterned ways of doing things are seen as persisting, changing, or passing away based on their adaptive fitness (Prentiss Reference Prentiss2021). Under this framework, the persistence of specific forms of material culture over very long periods are seen as the result of highly stable cultural transmission processes operating under the influence of strong cultural norms, creating what have been referred to as a “meta-traditions” (Araujo et al. Reference Araujo, Pugliese, Dos Santos and Okumura2018:2515–2516).
The findings presented here suggest that Native American dice, games of chance, and gambling are such meta-traditions—cultural traits of tremendous persistence—stretching back in time over at least 12,000 years, continuing through all periods of prehistory, surviving colonialism, and remaining a vital cultural force in the present in the form of the modern tribal gaming industry (Gabriel Reference Gabriel1996:26; Weiner Reference Weiner2018). Currently, over 250 Tribes operate more than 500 gaming locations in the United States, producing over $40 billion in annual revenue (National Indian Gaming Commission 2025). The importance of this adaptation of an ancient cultural practice to current conditions goes beyond the economic benefits it provides to modern Tribes. Tribal gaming has also enhanced, empowered, and challenged tribal sovereignty in new and transformative ways. As Cattelino (Reference Cattelino2005:197, Reference Cattelino2008) has documented, “sovereignty is not only a backdrop to gaming: sovereignty also is enacted through gaming operations, challenged by outside attacks upon casinos, and, most of all, realized in governmental activities enabled by gaming.”
Clues to the adaptive fitness of Native American dice, games of chance, and gambling—cultural forms and practices with a 12,000-year-old pedigree—can be gleaned from studies of the ethnographic and archaeological records of North American and beyond. For example, after a review of prehistoric and ethnographic Native American gambling, Weiner (Reference Weiner2018:2) concluded that it functioned as a “mechanism for transcending social distance” between otherwise disconnected groups, allowing them to interact economically and socially for their mutual benefit.
Native American gambling was able to perform this distance-spanning function because it is a system of exchange that operates outside the confines of preexisting relationships—a constraint that is endemic in traditional societies. There, as described by Sahlins (Reference Sahlins1972:168), exchange “is usually a momentary episode in a continuous social relation” in which the “decisive function” is not economic but rather the maintenance and creation of long-term reciprocal social relations, leaving little room or social context for exchange with rarely encountered outsiders. With gambling however, the context for exchange is created not by preexisting social relationships but rather by what Yanicki (Reference Yanicki2021:123) has termed “a shared fluency of gambling games that transcends barriers of language and ethnicity.”
Central to this shared fluency was not only a knowledge and acceptance of how the games were to be played (Yanicki and Ives Reference Yanicki, Ives and Voorhies2017) but also a working understanding of how probability shaped their outcomes, allowing players to be confident that they had an equal chance of winning and could therefore interact and exchange with other players on an equitable basis. This necessity of a shared understanding of fairness was well summarized in the sixteenth century by Gerolamo Cardano, an early pioneer in the study of probability (and himself a gambler), when he observed that “the most fundamental principle of all in gambling is simply equal conditions . . . to the extent to which you depart from that equality, if it is in your opponent’s favor, you are a fool, and if in your own, you are unjust” (Ore Reference Ore1953:189). Consequently, by ensuring both the appearance and reality of fairness in a mutually understood and transparent way, Native American dice, games of chance, and gambling created a social context for exchange with outsiders similar to that described by Woodburn (Reference Woodburn1982:443) as prevailing among Hadza hunter-gatherers of Tanzania, in which “the transactions are neutralized and depersonalized by being passed through the game.”
Given these features, it is not surprising that scholars of Native American gambling have found that it acted as an “in-between or liminal activity” that “brought together people who were neither close friends nor complete strangers” (DeBoer Reference DeBoer2001:235). Consistent with this view, scholars of Native American dice and gambling in the Late Holocene (2000–450 BP) have suggested that they functioned as an important means of social integration that acted to “facilitate interactions among diverse individuals, allowing them to come together to exchange information, goods, and marriage partners and to foster a larger group identity” (Weiner Reference Weiner2018:35). Likewise, DeBoer (Reference DeBoer2001) and Yanicki (Reference Yanicki2021) have argued that gambling in ancient North America was an outward-directed, intergroup activity reserved for outsiders that took place on territorial frontiers and at large intertribal gatherings, making gambling an important mechanism for cultural transmission and knowledge exchange. Similarly, Riggs (Reference Riggs2021) has argued that gambling in the ancient Southwest played an integrative role in societies that were aggregating or coming into frequent contact with outside groups.
Moreover, Native American gambling not only expanded the range of individuals and groups with whom social interaction and integration could occur but also created a system of exchange with these peripheral groups that incorporated efficiency-enhancing features otherwise incompatible with traditional systems of continuing reciprocal relationships. For example, in systems of exchange based on general reciprocity, the receiving of goods creates only a diffuse obligation to reciprocate at some unspecified time in some unspecified way that is not tied to the value of the goods received (Sahlins Reference Sahlins1972:175–176). In contrast, exchange by gambling is direct and immediate, with goods trading hands at the close of each game. In addition, because players must agree in advance on the stakes to be wagered, they must engage in direct price negotiation—another efficiency-enhancing feature—to reach agreement on the value, quality, or quantity of the goods to be given or received depending on the outcome of the game. Furthermore, although it may seem that a zero-sum transaction is hardly efficient for the losing party, this pattern of one-sided benefits holds true only in the context of a single match. However, in a series of contests over time in which players have an equal chance of success, the tally of wins and losses will, on average, tend toward an equal (50/50) distribution by operation of the law of large numbers (Dekking et al. Reference Dekking, Kraaikamp, Lopuhaä and Meester2005:185).
Therefore, the adaptive fitness underwriting the remarkable persistence of Native American dice, games of chance, and gambling is not difficult to discern. They likely functioned as a social lubricant that allowed for interaction, integration, and efficient exchange outside of normal social channels for groups coming into contact only infrequently. As DeBoer (Reference DeBoer2001:235) has observed, “It would be hard to devise a better suite of conditions for facilitating the rapid circulation of goods, bodies, and information over large areas.”
New Insights into Early Paleoindian Social Complexity and Organization on the Great Plains
An additional implication of the findings presented here is that they offer the possibility for new insights into the social practices of Early Paleoindian groups on the Great Plains, such as those that made and used the diagnostic and probable dice recovered from Folsom deposits at the Agate Basin and Lindenmeier sites. The presence of multiple dice at both of these sites not only suggests an unusually high level of social complexity but also offers support for the unsettled proposition that Folsom bands on the Great Plains came together at communal aggregation events such as those described above.
“Social complexity” is an umbrella term for a variety of cultural processes and characteristics, including inequality, large-scale networks of cooperation, institutionalized leadership, and hierarchical forms of governance (Feinman Reference Feinman, Chacon and Mendoza2017:459). It is typically associated with—and considered a hallmark of—so-called complex societies (Daems Reference Daems2021:2). However, if we infer that Folsom groups used dice for the same purposes as their Late Holocene and historic ethnographic descendants, it would appear that seemingly “simple” hunter-gatherers in the Late Pleistocene were engaged in highly structured, socially complex behavior—specifically, the creation and maintenance of large-scale networks of cooperation and exchange outside the confines of traditional reciprocal relationships based on a novel technology (dice and gambling) that provided a social context for such interactions.
Taken a step further, the findings made here also shed light on the decades-long debate as to whether Early Paleoindian bands on the Great Plains gathered together for planned communal aggregations. Although the periodic aggregation of small hunter-gatherer bands into larger groups is commonly observed ethnographically (Kelly Reference Kelly2013:174), a seminal article by Hofman (Reference Hofman1994) questioned whether this was necessarily so on the Great Plains in the Late Pleistocene. As Hofman framed the issue, although many common assumptions are proffered for the supposed necessity of aggregation among hunter-gatherers—such as high intergroup competition, uneven or seasonally unreliable resources, mobility limited by competitors, and exogamous marriage patterns—“Early Paleoindians may not fit any of these assumptions” (Reference Hofman1994:348). Hofman also warned that “we would probably not recognize Paleoindian aggregation sites even if we found them,” because we lack a sufficient understanding of the social structures and regional and site-specific assemblages associated with these groups necessary to identify archaeological “signatures of aggregation” (Reference Hofman1994:349–350).
Following Hofman, debate and uncertainty has continued as to whether Early Paleoindian groups on the Great Plains gathered together in communal aggregations (Andrews et al. Reference Andrews, LaBelle and Seebach2008; Bamforth Reference Bamforth2011; LaBelle and Meyer Reference LaBelle and Meyer2021), largely because “the archaeological signature of aggregation in a Paleoindian context is still poorly understood” (Guarino Reference Guarino2017:239). For example, communal bison kill sites and large camp sites with large quantities of materials have been suggested as possible Great Plains Folsom aggregation locations (Frison and Stanford Reference Frison and Stanford2014:362–363; Hofman Reference Hofman1994; Wilmsen and Roberts Reference Wilmsen and Roberts1978:179), but significant uncertainty remains as to whether these large sites represent contemporaneous aggregations or accumulations of artifacts over time (palimpsests) resulting from the reoccupations of favored locales by consecutive Folsom groups (LaBelle and Meyer Reference LaBelle and Meyer2021).
The findings presented here offer new evidence and suggest a new approach that can shed light on these long-standing questions. As discussed above, scholars of Late Holocene prehistoric Native American gambling have argued that it is best understood as an outsider-directed activity occurring primarily at intergroup gatherings and aggregations of normally dispersed groups (DeBoer Reference DeBoer2001; Riggs Reference Riggs2021; Weiner Reference Weiner2018; Yanicki Reference Yanicki2021). If we assume that this same connection between gambling and aggregation was operative in the Late Pleistocene on the Great Plains, then Native American dice could constitute what Hofman (Reference Hofman1994:349–350, 358) referred to as a “signature of aggregation,” providing a much needed (and currently absent) “means of recognizing such sites in the archaeological record.”
Following this approach, the presence of multiple diagnostic and probable dice in Folsom deposits at the Agate Basin (n = 4) and Lindenmeier (n = 15) sites marks both as prime candidates for Great Plains aggregation locations. Consistent with this view, both sites have been suggested as possible Folsom aggregation locations based on their spatial extent and large assemblages (Frison and Stanford Reference Frison and Stanford2014:362–363; Hofman Reference Hofman1994; LaBelle and Meyer Reference LaBelle and Meyer2021; Wilmsen and Roberts Reference Wilmsen and Roberts1978:179). Consequently, the identification of Native American dice in site assemblages (whether Paleoindian or later), particularly where other site characteristics indicate the possibility of large-scale communal activities, suggests that such sites may have served as aggregation locales where usually disconnected groups gathered together for interaction and exchange facilitated by games of chance and gambling as a means of social negotiation.
Conclusion
This article has attempted to dispel the prevailing uncertainty regarding the origins and antiquity of Native American dice, games of chance, and gambling by deriving an objective morphological test for identifying dice in the archaeological record based on Stewart Culin’s (Reference Culin1907) illustrations and descriptions of 293 unique sets of ethnographically documented Indigenous North American dice, set forth in his Games of the North American Indians. The results of this effort suggest that dice, games of chance, and gambling have been a persistent feature of Native American culture—one that served a critical role in social integration—for at least the last 12,000 years and continues into the present. Although some of the implications of these findings are discussed, they mark only the beginning of what further study of dice, games of chance, and gambling—as intellectual and social technologies—may reveal about ancient Native American complexity, social organization, exchange, religion, and worldview now that their roots can confidently be traced into the deep past.
For example, in recent decades, questions related to gender roles have become a frequent topic of discussion in North American archaeology (e.g., Gero and Conkey Reference Gero and Conkey1991), and there is evidence to suggest that ancient Native American dice and dice games may have had a gendered component. Specifically, an analysis by DeBoer (Reference DeBoer2001:226, Table I) of ethnographic accounts of such games found that out of 131 instances where the gender of the participants was noted, 81% were played exclusively by women, with 12% being played by both sexes, and only 7% played by men only. Additional research could shed light on whether this historical pattern extends into the prehistoric past, suggesting the possibility that women may have been leaders in the social and intellectual innovations associated with ancient Native American dice, games of chance, and gambling discussed in this article.
Another question raised here that merits further study is whether Native American dice are in fact an exclusively postcontact phenomenon in eastern North America. A review of museum and private collections, SHPO files, and other sources might reveal prehistoric dice not identified here. In addition, a study of the ethnographic record focusing on postcontact eastern North American groups that made and used dice could serve as a starting point for identifying culturally associated precontact assemblages whose contents might be expected to include Native American dice.
Additionally, the findings made here suggest that further study of the geographic distribution and change in stylistic types of Native American dice might reveal regional variations in game play and shed light on questions of the diffusion and adoption of games of chance and gambling in ancient North America over time and space. Likewise, the identification here of Native American dice as a possible marker of aggregation suggests that a reexamination of site assemblages where these artifacts are found in number—such as the Early and Middle Holocene sites of Sudden Shelter, Cowboy Cave, and Hogup Cave in Utah, Signal Butte in Nebraska, Pictograph Cave in Montana, Dipper Gap in Colorado, and Marsh Pass Cave in Arizona—could reveal patterns of social interaction and integration that have not been previously considered at these locations.
These and other questions are open for inquiry now that the origins and antiquity of Native American dice, games of chance, and gambling have been traced through prehistory and their roles as intellectual accomplishments and technologies of social integration are recognized. Let the games begin.
Acknowledgments
I thank Jason LaBelle for his advice and encouragement, and for his comments on a previous draft of this article. I thank Jared Orsi for his review and comments on a previous version of this article. I also thank the reviewers of this article, whose comments and constructive criticism significantly improved the final product. Thanks are also extended to the following institutions and archivists for allowing access to their repositories: (1) the Smithsonian Institution, National Museum of Natural History, and museum specialist James Krakker; (2) the University of Wyoming and Marcel Kornfeld; and (3) the Denver Museum of Nature and Science and collections manager Angela Rueda. Finally, I thank D’arcy N. R. Madden for the dice illustrations that accompany the maps included in this article.
Funding Statement
This research received no specific grant funding form any funding agency, commercial, or not-for-profit sectors.
Data Availability Statement
The data supporting Table 1 are in the published sources referenced in it. The artifacts appearing in Figures 1, 9, and 10 are held in the repositories of the National Museum of Natural History, the University of Wyoming, and the Denver Museum of Nature and Science.
Competing Interests
The author declares none.
Open access
