Rock Varnish Dating, Surface Features and Archaeological Controversies in the North American Desert West

Abstract

Archaeological surface features on desert pavements, including geoglyphs, are notoriously difficult to assess. Lacking temporally diagnostic artifacts, they may be impossible to place chronologically, limiting their inferential utility. Not surprisingly, controversies have developed in the North American desert west over certain of these features. We describe methods for chronometrically constraining the ages of desert pavement features using three approaches to rock varnish dating: varnish lamination (VML), lead-profile dating, and the cation ratio (CR) as an additional tool. Each of these techniques may be applied to rock varnished cobbles that have been upthrust into areas previously cleared of the original pavement through cultural or natural processes. We use these methods to resolve two archaeological issues: the age of the intaglios (geoglyphs) along the lower Colorado River corridor and whether the Topock (or ‘Mystic’) Maze is the product of Precontact Indigenous or late-nineteenth-century railroad construction. Ethnographic analysis allows us to contextualize these features and to consider two additional issues: the antiquity of the Yuman speakers’ cultural pattern in the lower Colorado River region and the function of the Topock Maze.

1. Introduction

Surface archaeological features and sites are common if not ubiquitous in the North American desert west. Despite their pervasiveness, these types of archaeological remains—lithic scatters, petroglyphs, rock alignments and cairns, and geoglyphs, cleared circles, and trails in desert pavements—are problematic in the sense that their chronological placement is often unknown, limiting their inferential value. Indeed, in certain cases it is unclear whether specific features are human or natural in origin, complicating cultural heritage management (Caldwell et al. 2011; Bullard et al. 2012). In select cases, these features can be chronometrically dated, however, using rock varnish dating techniques. We describe and discuss these techniques and two issues in their application in the following sections. We then demonstrate the use of three techniques and the inferential implications of the resulting chronometric ages for two empirical case studies that help resolve long-existing controversies: the age of the intaglios (geoglyphs) along the lower Colorado River and thus the antiquity of the Yuman cultural pattern in this region and whether the so-called Topock (“Mystic”) Maze is a Precontact-era Indigenous sacred site or instead a by-product of late-nineteenth-century railroad construction.

We start with a brief discussion of rock varnish itself before turning to the methodological approaches and issues involved in its dating that we utilized. We next discuss the methodology of ethnographic analysis, including historical linguistic reconstructions, that we have followed before concluding with the results of our empirical studies and their implications for understanding the Precontact past of this region and the ritual importance of the Topock Maze.

Rock Varnish

Rock varnish is a paper-thin deposit that accumulates naturally on rock surfaces. It is composed of more than half clay minerals, Mn and Fe oxides that typically vary between a quarter to a half, and minor and major trace elements (Dorn 2024). Although it can form in any environment, rock varnish is most commonly recognized in arid regions (hence its non-technical label as ‘desert varnish’), where it typically appears in two forms: a chocolate brown to blue-black varnish on exposed rock surfaces and orange varnish found under ground-bands along the base of pavement rocks, as well as in cracks themselves (Dorn 1998). Rock varnish dating involves the first of these two rock coatings, and is the focus of the following discussion and analyses.

The clay minerals that are transfixed to lithic surfaces forming a rock varnish coating are derived from windblown dust and can vary based on regional geology (Dorn 2025). Despite this variability, Mn-oxides typically dominate, commonly resulting in varnish’s dark appearance. The clay particles, Mn- and Fe-oxides and trace elements that comprise this coating are fixed to rock surfaces by biotic and abiotic processes, including the effects of budding bacteria (Chaddha et al. 2024; Dorn 2024). Rock varnish is then an accumulated biogeochemical coating, not a patina formed from the chemical alteration of the underlying lithic surface, as it is sometimes incorrectly labeled.

Rock varnish’s development and its chemical constituents are also influenced by major paleoclimatic changes. During wetter periods, varnish is Mn-enriched; drier periods instead result in less Mn accumulation such that Fe dominates. Over time, as paleoclimatic conditions alternate, this results in a microstratigraphy of Mn- versus Fe-rich layers that is visible in ultrathin sections (Perry and Adams 1978; Liu et al. 2025). These changes appear as contrasting color differences using optical microscopy, with black (Mn-rich), orange (Mn-low), and yellow (Mn-very low) layers (Figure 1). Varnish develops first in small pockets or micro-basins on a rock surface, which, accordingly, typically preserve the most complete varnish layering sequence. This fact has important implications for rock varnish dating, discussed below.

It is important to note some additional caveats about varnish formation processes. First, multiple factors can impact or alter these processes and the preservation of complete layering sequences. Acid-producing lithobionts such as fungi (e.g., Burford et al. 2003; Favero-Longo et al. 2011; Favero-Longo and Viles 2020), for example, may develop as climatic conditions change or may be more prevalent in some environmental settings, and their acids can dissolve the Mn from rock varnish (Dorn 2019, 2024). Second, this (and other factors) may result in an irregular distribution of potential varnish sampling micro-basins useful for archaeological dating, which in turn may impede systematic sampling (Whitley 2013). Finally, there are other rock alterations that are Mn- and/or Fe-rich, including weathering rinds and case hardening that incorporate varnish components (Dorn 1998), and these may be visually mistaken for rock varnish without in situ technical analysis. All of these circumstances must be considered in archaeological applications of rock varnish dating.

2. Background and Methods: Rock Varnish Dating Techniques, Sampling Procedures, and Ethnographic Analyses

A series of different techniques have been developed and applied to rock varnish dating for archaeological purposes (Dorn 1994, 2001), not all of which have been successful. We discuss three here that have been frequently used in recent archaeological applications: varnish microlamination (VML), cation ratio (CR), and lead-profile dating.1

It is important to emphasize that, regardless of the chronometric approach employed, there is a time lag between the first exposure of a rock surface, such as the recently engraved groove of a petroglyph or the exposure of an upthrust cobble in an intaglio, and the initial development of a rock varnish coating. Although we provide new data relevant to this time lag below, here it is adequate to note that prior observations revealed that rock varnish starts to form in millimeter-sized rock surface depressions, or micro-basins, in hot deserts within a century. We observed this at Fort Paiute in the Mojave Desert (Dorn et al. 1986; Bamforth and Dorn 1988), on an intaglio made in the 20th century called the “Fisherman” (Dorn 1998), discussed below, and on a 20th-century engraving at South Mountain, Phoenix (Dorn et al. 2012). These prior findings do not indicate that an entire rock surface starts to revarnish within a century; it does not. The process starts in a few micro-basins initially and then slowly begins to cover other parts of a rock surface. As discussed below, this fact has implications for the use and efficacy of different varnish dating techniques and sampling strategies.2

2.1. VML Dating

Perry and Adams (1978) first observed alternating Mn- and Fe-rich microstratigraphic layers in rock varnish thin sections. These micron-scale laminations develop as a result of climatic changes (Liu and Broecker 2007; Liu et al. 2000, 2025). As they demonstrated, the resulting layered patterning visible in an ultrathin section can be used to determine the age of the sample when compared to a regional sequence developed using independently dated surfaces as calibration control points. VML dating is then conceptually analogous to tree ring dating, but it does not have the same chronological resolution as dendrochronology. When varnish grows slowly, as commonly occurs in hot deserts (Dorn 1998), the temporal resolution is low, and a sequence only records major climatic signals such as large glacial pulses. Where varnish grows more rapidly, as it does in moister microenvironments (Spilde et al. 2013; Dorn 2024) such as the north versus south face of some boulders, however, roughly millennial-scale climatic events may be recorded throughout the Holocene (Liu and Broecker 2007; Liu et al. 2025).

VML sampling requires the removal of a complete layering sequence from a rock varnished surface. This may be obtained mechanically, by carefully hand-prying off the varnish layer from a micro-basin using a tungsten needle, or by taking a hollow core sample, which preserves the varnish sequence. It is necessary to obtain multiple micro-basin samples from the surface being dated to ensure that the most complete layering sequence is obtained. Sampling petroglyphs requires the removal of intact micro-basins from within the pecked-out area that defines the image, whereas, for upthrust cobbles in cleared areas in desert pavements, they may be taken from anywhere on the exposed surface. We note that the area removed is typically only a few millimeters in diameter.

Liu and Broecker (2007, 2008a, 2008b) and Broecker and Liu (2001) identified and calibrated a Late Pleistocene—Holocene Mojave VML sequence (see Figure 1). This has been expanded into the lower Sonoran Desert with additional calibration points from radiocarbon-dated landslides (Dorn 2014) and is used in this paper. Liu and Broecker (2007) further noted—and Dorn (2014) verified—that when varnish grows fast enough, WH1 is often subdivided into three visible layers (e.g., Figure 1a) formed in the wetter phases of the Little Ice Age, about 300, 500, and 650 calendar years before the present. The chronological implications of this sequence have been described as follows:

“twelve wet event dark layers and thirteen dry event lighter layers … bracket the period from 300 yrs cal BP to 12,500 yrs cal BP. The lengths of the intervals between wet events vary from 250 to 1800 years, with an average of 970—roughly 1000 years. The resulting correlated VML ages are certainly not as precise as radiocarbon ages, but they are adequate for age assignment to the broad time periods comprising the regional cultural historical sequence”.

(Whitley 2013, p. 4)

VML, notably, has been subjected to, and verified by, blind testing using cosmogenically dated geomorphic surfaces (Liu 2003; Marston 2003). VML has also been used to date artifacts, petroglyphs, constructed surface features, and geomorphic surfaces in Africa, the Middle East, and North and South America (Carbonelli and Collantes 2019; Cerveny et al. 2006; Liu et al. 2021, 2025; Baied and Somonte 2013; Whitley et al. 2017), even though its primary use has been in geomorphology (Liu and Broecker 2008b, 2013). VML dating, for example, was used to predict that debris flows could be a major hazard in metropolitan Phoenix (Dorn 2010), a prediction that was confirmed in 2014 (Dorn 2016).

2.2. Lead-Profile Dating

The industrial revolution caused a significant increase in atmospheric pollutants, including lead (e.g., Candelone et al. 1995). One result is that rock varnish surfaces exhibit an observable increase, or spike, in lead (Pb), beyond the natural background levels in the rock coating. This lead spike only occurs in the very surface-most micrometers of rock varnish that accumulated in the 20th century. We identify the presence of lead in a VML sequence with an electron microprobe, using a focused micron-size beam for 200 s, which lowers the detection limit to 0.06% PbO (Dorn 1998). This 20th-century lead spike has subsequently and independently been verified in rock varnish as resulting from the manganese and iron scavenging of lead (Fleisher et al. 1999; Hodge et al. 2005; Hoar et al. 2011; Nowinski et al. 2013; Spilde et al. 2013; Sims et al. 2022).

Lead-profile dating was first used to determine if the “Fisherman” intaglio near Quartzsite, Arizona, noted above, was pre-20th century in age or a more recent potential fake. As discussed below, rock varnish coatings from upthrust cobbles within this geoglyph were sampled, all of which were exclusively characterized by the lead spike, indicating a recent/20th-century age for the creation of this surface feature (Dorn 1998). Note that this does not preclude the possibility that this feature may be Indigenous in origin since the ritual practice involved in the use of these geoglyphs has continued into the twenty-first century (Whitley 2014). That this intaglio is stylistically and thematically dissimilar from other examples, combined with its recent age, however, suggest that it may be Euro-American rather than Native American in origin.

Lead-profile dating has also been used to authenticate historical inscriptions (Dorn et al. 2012). An important use of this technique is the way we have employed it here: as an internal check, it verifies that the youngest Holocene rock varnish microlaminations at 300 (WH1a), 500 (WH1b), and 650 (WH1c) cal yr BP, correlated with three wet phases of the Little Ice Age (Liu and Broecker 2007; Liu et al. 2025), are likely correct, as discussed above. Repeating a point noted previously, rock varnish dating is best conducted using multiple techniques (Dorn 1994, 2001), as we employ lead-profile dating here.

2.3. CR Dating

The development of CR dating in the early 1980s (Dorn 1983; Dorn and Oberlander 1981) yielded the first calendrical age constraints on petroglyphs (rock engravings), resulting in the first chronometric rock art ages (Dorn and Whitley 1983, 1984). This technique is based on cation exchange processes in rock varnish that leach out mobile trace elements (K and Ca) more rapidly than stable elements (Ti) through capillary action (Dorn and Krinsley 1991). The resulting rate of change can be calibrated using independently dated control surfaces, such as river and lake terraces, and is used to provide minimum-limiting age estimates for a rock varnish coating (Dorn et al. 1990).

While CR dating has been independently replicated in a number of regions globally (e.g., Pineda et al. 1990; Zhang et al. 1990; Plakht et al. 2000; Sarmast et al. 2017; Ntokos 2021) and has successfully been subjected to blind tests in its archaeological use (Loendorf 1991; Bamforth 1997), it has certain analytical weaknesses. Varnish growth, as noted above, starts in micro-basins, with initial cation leaching likewise beginning at discrete points rather than continuously and uniformly across a rock surface. This discontinuity requires the removal (through mechanical abrasion of varnish within the micro-basins) and chemical analysis of a bulk sample in order to obtain an average value for the coating as a whole (Dorn 1994, p. 24). The resulting age estimates often have large error margins, and they can underestimate the true age of a varnish surface (see, e.g., Whitley 2013, Table 1), despite efforts to exclusively sample within micro-basins and thereby minimize contamination from the younger portions of a varnish coating. The most important weakness of the technique, however, is its basis in biogeochemical processes, which can potentially reverse over time. Despite these issues, a blind test matching CR to VML dating yielded comparable even if not fully equivalent results, demonstrating its value for broad-spectrum temporal analysis, such as identifying the age range of a corpus of petroglyphs or a surface artifact assemblage. But like obsidian hydration dating, CR is not appropriate for confident age assignments for individual specimens (Whitley 2013, p. 7). Given the considerable advantages of VML discussed above, we employ CR dating below as a subsidiary confirmation of the age assignment of a VML sequence.

2.4. Methods of Dating Surface Features in Desert Pavements

Dating surface features in desert pavements is based on the fact that the cobbles on the landforms in our study region are alluvial in origin, with the natural process of transporting these materials scouring off any prior rock coatings. Subsequently the wetting and drying of fines and the accumulation of silt push buried cobbles upwards, onto the ground surface, in areas that have been previously cleared of these accumulative ground coverings (Springer 1958; Haff and Werner 1996; Adelsberger and Smith 2009 Adelsberger et al. 2013; Haff 2014; Ugalde et al. 2020). When surface archaeological features involve the disturbance of the desert pavement (e.g., in the making of an intaglio or sleeping circle), VML, CR, and lead-profile dating can be used to provide a minimum-limiting age estimate on the subsequently upthrust lithic clasts and thus on the creation of the archaeological feature. Two technical matters involved in dating desert pavements using such an approach warrant discussion here. These are the time lag involved in the upthrust of cobbles onto the surface in a disturbed desert pavement area and the larger issue of the time lag in the initiation of varnish development on unvarnished rock surfaces of all kinds, noted briefly above.

2.4.1. Time Lag in Desert Pavement Regeneration

A variety of observers have claimed that desert pavement regeneration begins within a few months to a few years (Péwé 1978; Haff and Werner 1996; Dietze and Kleber 2012). We monitored the rate of pavement reformation in meter-scale disturbance plots created by Troy Péwé in the Sonoran Desert outside of Mesa and Quartzsite, Arizona (Péwé 1978; Bales and Pewe 1979), to test these estimates, for over four decades. We continued monitoring these pavements with Péwé’s permission, even after his passing in 1999.

It was important for our purposes to consider only those cobbles that emerged from the subsurface after Péwé cleared these sample plots and to ensure that any sampled cobbles were not remnants of the original pavement nor were inherited contaminants through disturbances of some kind from the surrounding areas. In addition to periodic monitoring, we focused only on those cobbles with inclusions of millimeter-scale fragments of pedogenic carbonate. Such carbonate coatings develop on buried clasts (Springer 1958; McAuliffe and McDonald 2006; Haff 2014), and precipitation would dissolve the carbonate once exposed on the ground surface after a few decades, allowing us to eliminate the possibility that our sampled cobbles were remnants of or derived from the original pavement surface. Note that thick calcium carbonate coatings can persist in surface contexts, potentially for thousands of years, as we have demonstrated previously (Cerveny et al. 2006). In the current case, the very small size of the carbonate fragments would be susceptible to weathering in a short period of time, guaranteeing recent exposure of the clasts.

We found that it took 14 years at the Mesa site and 22 years at the Quartzsite site for previously buried cobbles to emerge and start reforming a new desert pavement. The implication is that desert pavements begin to regenerate within a few decades of disturbances to them, indicating that rock varnish dating of subsequent upthrust clasts can provide an archaeologically reasonable minimum estimate of the age of manufacture of these types of surface features.

We stress that pavement generation and regeneration is an ongoing process. The length of time it takes to form a completely intact pavement, measured by radiocarbon and optically stimulated luminescence, can range from about 5000–6000 years in the Mojave Desert to about 12,000 years in Death Valley, California, when cobble sizes are under approximately 10 cm (Seong et al. 2016). The samples we collected are thus from pavements that are still regenerating, and they tend to have even smaller cobble sizes. We emphasize that the varnish ages presented here are minimum estimates for the age of a pavement-disturbing event.

2.4.2. Time Lag in the Initiation of Revarnishing

It is also important to stress that, regardless of the chronometric approach employed, there is a time lag between the first exposure of a rock surface, such as the newly engraved groove of a petroglyph or the exposed surface of an upthrust cobble, and the initial development of a rock varnish coating—the start of its revarnishing for petroglyphs and the initiation of varnishing on upthrust cobbles. As noted above, prior observations demonstrated that rock varnish is present in millimeter-sized rock surface depressions, or micro-basins, in hot deserts within approximately a century, but the varnish only slowly begins to cover other parts of an exposed rock surface. Varnish age across a surface then varies from point to point, a fact which is critical in terms of effective chronometric sampling.

We used the opportunity afforded by the pavement regeneration study, discussed above, to improve the understanding of this time lag in the start of surface revarnishing. Although rock varnish was not detectable on most of the previously buried cobbles we observed on the ground surface at Péwé’s Mesa study locale, two had started to develop a varnish coating after 19 years of surface exposure. These coatings, however, were only visible using electron microscopy (Figure 2), not with the light microscopy used for VML and, as expected, were restricted to micro-depressions on the cobble surfaces. This finding illustrates the great variability in varnish accretion rates (cf. Spilde et al. 2013). This fact does not impact the reliability of VML, Pb-Profile, or CR dating using a sampling procedure focusing on multiple micro-basins, designed to account for this variability. But this invalidates methods that sample indiscriminately and rely on varnish accumulation rates.

We further assessed the revarnishing time lag using a natural petroglyph-like feature—a 0.5 cm wide scar on a rock surface, created by a landslide in the McDowell Mountains, Scottsdale, Arizona. A radiocarbon age of 1180–1290 cal year BP, obtained from a crushed paloverde tree under the landslide feature (Dorn 2014), provides a useful age constraint on the natural event that created this scar. This allowed us to assess the variability of rock varnish ages in different micro-basins within this feature, using VML dating, along with the rate and degree to which rock varnish spread across this groove, using X-ray mapping (Friel and Lyman 2006) to identify varnished micro-basins.

Only two of 10 VML cross-sections from micro-basins in the scar had the WH2 microlamination, thereby matching the radiocarbon age. The remainder provided younger estimates: four had the WH1 (Little Ice Age) microlamination at their base, and four were post Little Ice Age (younger than 300 cal yr BP). It may thus take roughly 1000 years for rock varnish to coat an exposed surface, with the ages of different specific micro-basins on this surface varying by that same number of years, illustrating that varnish development is not continuous across lithic exposure. We accordingly only use the oldest VML sequence for each feature in our age estimates, because the oldest VML provides the smallest time lag between cobble exposure and the onset of varnishing and is likely the closest to the true age of the feature. These methodological issues emphasize the importance of using a varnish dating technique like VML that can provide close minimum-limiting ages by sampling multiple micro-basins.

2.5. Methodological Principles of Ethnographic Analysis

We contextualized and interpreted our chronometric results with an analysis of existing ethnographic data collected by multiple anthropologists from Yuman-speaking tribal members, primarily the Mohave and Quechan. In addition to older published ethnographic data, we include recently collected information from contemporary tribal consultants and spokespersons. These accounts demonstrate continuity in the retention of traditional knowledge, given that the recent commentary correlates with the earlier recorded information, some of which is over a century old.

We followed the procedures and approaches to ethnographic analysis outlined by Whitley (2005, 2007, 2011) in our analyses. These emphasize four key methodological principles. The first is that ethnographic reports typically provide “raw data” that itself must be interpreted, rather than complete answers to specific questions. This is because the questions that anthropologists raise to their informants are rarely those that an archaeologist may be interested in, with archaeologically useful information often included in asides or seemingly tangential commentary. The second point is that ethnographic analysis is an iterative process that requires careful reading, and re-reading, the reports, after specific clues to meaning are successively identified.

Third, this iterative process is especially important when the interest is religious beliefs and symbolism. These topics can only be understood when the ethnographic data are contextualized in terms of the metaphysical beliefs—the ontological and epistemological commitments—of the informants. The failure to recognize this last fact caused archaeologists to claim for decades that Native Americans knew nothing about rock art, for example, because researchers attempted to interpret the ethnographic commentary from a Western, scientific perspective and were unable to do so. This further caused archaeologists to argue that rock art pre-dated the ethnographic tribes and that it was necessarily created by earlier peoples. What should have been obvious, however, is that religious beliefs of all types are matters of faith, not empirical science, and do not make logical or empirical sense in Western terms, regardless of the religion involved. Substantial ethnographic data on the making and meaning of rock art then exists, once it is understood that this can only be identified through careful reading and re-reading the ethnographic accounts and only understood in terms of Indigenous, not Western, ontological perspectives.

Fourth, as in most qualitative analyses, there are varying levels of reliability to or inferential strength of an ethnographic interpretation. Assessment factors may include the existence and numbers of confirming accounts; whether these were collected by different anthropologists, at different times, from different informants; and whether a particular interpretation fits a larger regional ethnographic pattern. The best results involve confirming results from multiple sources that correlate with regional beliefs and practices. We have followed these principles in our ethnographic analysis of the significance and meaning of the Topock Maze, as discussed below.

Linguistic and Ethnic Group Histories: Background and Methods

Our interpretive concern with respect to the ages of the intaglios involves a central problem in this part of the far west: the age of Yuman speakers’ (Mohave, Quechan and, historically, Halchidoma and other smaller tribal groups) arrival in the lower Colorado River region. The implications of the ages of the intaglio tradition to their origins have been recognized by von Werlhof (1995), and the data we present below is directly relevant to this issue. Existing interpretations of this topic, however, are only partly based on archaeological data, with significant—sometimes over-riding—weight given to historical linguistic reconstructions. Yet even a quick reading of the existing literature demonstrates that few archaeologists understand the methodological limitations of the different historical linguistic methods. This and a larger problem—what we mean by an ‘archaeological culture’—then warrant discussion here to better contextualize our results and our interpretation of their implications.

It is important to emphasize at the outset that there is no perfect correlation between language, culture, and physical/biological (“racial”) type, as the first half century of Boasian anthropological research demonstrated conclusively (Whitley 2020, 2021). But the parallel fact is that we cannot reduce a culture to an adaptive system or artifact assemblage. Cultures instead are defined in contemporary anthropological terms as cognitive systems of shared symbols and meanings, often tied to common ritual and belief systems, not sets of norms, artifact assemblages, or types, nor adaptive strategies, as many archaeologists still assume (Whitley 1992, 2020, 2024). As Sahlins (1985) has emphasized, furthermore, different aspects of human social life change at different rates, with technological and economic factors shifting quickly while belief and ideological systems—which is to say cultures—are much slower to change. Indeed, as Steward noted, a century of acculturation among the adjacent Numic (Paiute and Shoshone) speakers

“has not wiped out all Indian practices. Acculturation has consisted primarily of modifications of those patterns necessary to adjust to rural white culture … The Shoshoni retain, however, many practices and beliefs pertaining to kinship relations, child-rearing, shamanism, supernatural power and magic”.

(Steward 1955, p. 58)

Religion, a set of beliefs and practices related to supernatural agents and a supernatural realm, is often but not always a proxy for culture. It does not serve this role for large-scale societies, like our contemporary Western world with proselytizing religions (Keane 2003), but full concordance often exists between the cultures of small-scale societies, like those found in the North American desert west and their non-proselytizing religions (e.g., Harris 1940, p. 55; Opler 1940, p. 136; Siskin 1983, p. 11; Laird 1976, p. 24). Religious remains, including sacred objects and evidence of ritual practices, are then the most useful category of archaeological remains for studying cultural continuity and change, and intaglios fall within this category. Even if we cannot, with complete certainty, equate culture, as signaled by religious practices, with language and ethnicity alone, an assumed equivalence is reasonable in certain cases. A series of matters concerning the linguistic history of the Yuman language family then warrant discussion in light of our chronometric ages and their possible importance as an expression of an early cultural system for Yuman speakers along the lower Colorado River and because of the heavy reliance archaeologists have placed on linguistic reconstructions.

These start with the fact that many linguists assume that the Baja California peninsula was a linguistic dead-end in terms of the initial Late Pleistocene migration and colonization of the west coast, resulting in the development of an isolated speech community in this restricted area that eventually developed into the ancestral Yuman (and Cochimi) languages. This standard reconstruction posits that this initial colonization was followed by successive migrations northwards, out of this cul-de-sac homeland, creating a kind of linguistic stratigraphy, as it is sometimes described (Shaul 2020; cf. Massey 1961, 1966; Kowta 1984; von Werlhof 1995), yielding the relatively late arrival of Yuman speakers in the lower Colorado River region. But this reconstruction entails at least one significant confounding empirical problem. This is the 200 km separation between the Pai-speaking branch at the northern extreme of the Yuman languages, with the Iipay and Tiipay (“Diegueño”) in the coastal San Diego area and the Yavapai, Walapai, and Havasupai (“Upland Yumans”) in central- and north-western Arizona. This is typically, though not plausibly, explained by a very late, very rapid, hypothesized migration and thus split of Pai speakers in one direction or the other (e.g., Laylander 2015).

A more parsimonious reconstruction would suggest instead that there was an extra-peninsular Proto-Yuman heartland in southern California and/or the Colorado River Delta region, as Mixco (2006) argues, based on linguistic borrowing and reconstructed ecological terms. The separation over time into distinct speech communities and, ultimately, individual Yuman languages occurred at least partly (if not largely) in situ rather than primarily due to migrations. The geographical separation of the Pai resulted again in separation, in part or wholly, from the later movement of Takic speakers southwards into former Yuman territory, splitting the Pai apart, as Hinton (1991) first suggested, rather than exclusively due to late-dating Pai migrations. This last possibility accommodates the physical anthropological evidence suggesting that the Cupan branch (Gabrielino, Luiseño, Cupeño, and Cahuilla) of these Takic speakers were originally biologically Yuman or, alternatively, that a substantial proportion of Yuman speakers were absorbed into the in-moving Takic to account for the physical similarities (summarized by Sutton 2009), as well as the equivalences in their mythic corpora (cf. Kroeber 1906; Waterman 1909). It also fits with the genetic evidence (Monroe et al. 2013) and comparative trait studies (Jorgensen 1980) that suggest an earlier split for the Pai languages. Relatively greater antiquity, rather than recent arrival, then is a plausible even if unproven expectation for the Yuman speakers along the lower Colorado River.

The subtext of this first suggestion are estimates for the various divergences in the Yuman and surrounding language families, the second pertinent issue concerning the linguistic history of this region and its relevance to the origins of the lower Colorado River Yuman speakers. Three general approaches to the history of languages have been employed in the far west, an understanding of which is necessary for clarifying their value to archaeological interpretation, as well as for reconstructing the cultural history in this river region. The most common of these approaches involves linguists’ guesstimates commonly offered, insofar as we have been able to determine, with little or no detailed analytical basis but apparently based on very general analogies (e.g., Golla 2007). These guestimates appear tied to the following assumptions:

“The rule of thumb (derived from the study of languages with hundreds of years of written documentation…) is that after about 2000 years, language changes tend to have obscured the clear sorts of correspondences [for grouping language families and detecting borrowing, while] … After 5000 years there are few [such] correspondences”.

(Shaul 2014, p. 7)

The implications are that (1) all languages change at approximately the same speed, universally, regardless of population size and density, interaction networks, communication transmission and storage systems, demographic structure, or social complexity and that (2) the written texts of the Indo-European, Egyptian, and Semitic language families, the source of these estimates, are appropriate analogical models for the rate of linguistic changes, even to all manner of small-scale oral cultures and societies. Since language divergence and change is at least partly based on (if not heavily influenced by) the social structure of the speech communities using a given tongue, the nature of their interactions with speakers of surrounding languages (cf. Babel et al. 2013), and the differences in the various types of population movement and replacement that may occur (cf. Eshleman et al. 2004; Kemp et al. 2010), these assumptions are empirically implausible from the start (cf. Greenhill et al. 2023, p. 85).

But even more severe, from an archaeological perspective, is the temporal limit that such assumptions necessarily impose on the past, with the observable linguistic differences then required to have developed solely in the last few thousand years. Taken at face value and in aggregate, a synthesis of the far western Precontact past using these linguists’ guestimates would require numerous and frequent in- and out-migrations of different peoples, as if the far west were little more than a transit corridor for population displacements. Only occasionally is there a (right or wrong) logic suggested that would explain the cause for such movements (Bettinger and Baumhoff 1982, Diamond and Bellwood 2003, and Sutton 2009 are some notable exceptions). Even more rarely is there a plausible description offered concerning where those populations that were forced out or replaced moved, and the larger downstream demographic ripple effects of these geographical shifts. The effective result is models of linguistic change without people involved or social interactions considered. The similarities and close relationships between many far western Native American languages, instead, are more likely then a reflection of “a normal foraging adaptation in an arid region” (Shaul 2014, p. 66) resulting from a linguistic network rather than a dialectical chain, as opposed to numerous, necessarily recent population movements, as the guestimates require, despite the absence of archaeological data in their support.

A second, superficially more analytical approach to linguistic history is glottochronology. This has suggested that the core Yuman languages split up between 3500 and 1300 YBP (Shaul and Hill 1998). Note however that Shaul (2014) is no longer willing to assign specific dates to linguistic divergences, tacitly questioning his previously published estimate. This change in perception reflects a widespread rejection of glottochronology, the nature and problems of which are well described as follows:

“Glottochronology uses the percentage of shared ‘cognates’ between languages to calculate divergence times by assuming a constant rate of lexical replacement or ‘glottoclock’. Cognates are words inferred to have a common historical origin because of systematic sound correspondences and clear similarities in form and meaning. Despite some initial enthusiasm, the method has been heavily criticised and is now largely discredited. Criticisms of glottochronology, and distance-based [statistical] methods in general, tend to fall into four main categories: first, by summarizing cognate data into percentage scores, much of the information in the discrete character data is lost, greatly reducing the power of the method to reconstruct evolutionary history accurately; second, the clustering methods employed tend to produce inaccurate trees when lineages evolve at different rates, grouping together languages that evolve slowly rather than languages that share a recent common ancestor; third, substantial borrowing of lexical items between languages makes tree-based methods inappropriate; and fourth, the assumption of a strict glottoclock rarely holds, making date estimates unreliable. For these reasons historical linguists have generally abandoned efforts to estimate absolute ages”.

(Gray and Atkinson 2003, pp. 435–36)

The disqualifying problems inherent to glottochronological analysis are further emphasized by the fact that language changes sometimes involve punctuated bursts rather than gradual, progressive evolution (Atkinson et al. 2008), as glottochronology would require.

The third approach to linguistic history is the use of computational phylogenetic models derived from evolutionary biology. These overcome the four main problems with glottochronology, and they employ linguistic, genetic, and archaeological data to calculate dates for language change, with the archaeological and genetic data calibrating the linguistic rates. Although this approach has not yet been applied to the Yuman languages, to our knowledge, Greenhill et al. (2023) have provided an analysis of the neighboring Uto-Aztecan language family which has implications for the Yuman problem—the temporal relationship of the Yuman speakers to the neighboring Takic branch of Uto-Aztecan speakers and their residence in southern California. There is widespread agreement that the Yuman peoples were north of the Baja peninsula earlier than the Takic (e.g., Hinton 1991; Elliot 1994; Sutton 2009; Shaul 2014; Field 2018). Initially the arrival of the Takic in California was widely assumed to be linked to a southern origin and intrusive migration northwards of Uto-Aztecan speakers out of Mexico, perhaps associated with the spread of maize agriculture (e.g., Hill 2001). This necessarily placed Uto-Aztecan speakers in the region relatively late and, by implication, also accommodated a relatively more recent date for the presence of Yuman speakers. The alternative hypothesis was a northern origin for Uto-Aztecan speakers as a whole, including the Takic branch (Fowler 1983). Although this was initially a minority opinion, recent phylogenetically based reevaluations of Proto-Uto-Aztecan linguistic history provide significant support for the northern origin hypothesis, as well as deeper time depth for these people in the region (Merrill 2012; Merrill et al. 2009; Greenhill et al. 2023). The Proto-Uto-Aztecan language had differentiated into its own distinctive speech community about 4000 years ago at minimum, based on these latest reconstructions, in the central Mojave Desert adjacent to the Colorado River corridor. Assuming this reconstruction is correct, Yuman speakers would then necessarily have been in place outside of the Baja peninsula prior to about 4000 years ago.

Note, however, that the archaeological data that Greenhill et al. used in their analyses were from reports that partly based their chronologies on the discredited Lamb (1958) glottochronological study of the Numic branch of Uto-Aztecan, potentially biasing the phylogenetic results towards a younger estimated date. This was pointed out in personal correspondence with Simon Greenhill, who kindly re-ran his analyses with those data removed. This yielded no appreciable difference in the results and thus his estimated 4000 YBP date for the development of Proto-Uto-Aztecan in the Mojave Desert (email communication to Whitley, 18 July 2024). Unanswered, however, is a related question: what effect would the inclusion of earlier archaeological ages based on evidence suggesting much longer-term Numic cultural continuity, extending back into the Terminal Pleistocene (Whitley 2013; Whitley et al. 1999), have on the calculated estimate for Proto-Uto-Aztecan in the region?

Regardless of the answer to that specific question, this review of historical linguistical methods indicates that there is no reason to infer that Yuman speakers were relatively late arrivals in the region. Instead, and in fact, it suggests that they may have been living along the lower Colorado River region much earlier than previously suggested. The rock varnish minimum ages for the intaglios, discussed below, support this inference.

3. Age of the Colorado River Intaglios

Substantial, popular interest has developed over the last half century in the intaglios (geoglyphs or earth figures) found on the alluvial terraces along the lower Colorado River in California and Arizona. These commonly include very large (tens of meters in their maximum dimension) stick-like human figures, rarer depictions of animals (mountain lions, lizards, snakes), and geometric patterns like spirals and zigzags, often associated with cleared paths and circles (Figure 3). Like the more famous Nazca Line intaglios of Peru, much of this popular appeal involves spurious claims about putative associations with ancient aliens when, in fact, there is Native American ethnography on the meaning and ritual use of these surface features (Bourke 1889; B. Johnson 1985; von Werlhof 2004; Whitley 2014, 2023). This demonstrates that the sites are specific locations of mythic events for the River Yuman-speaking Quechan and Mojave; that the motifs portray the actors and events that occurred at each location; and that the sites were—and at least until two decades ago continued to be—used for ritual pilgrimages undertaken to re-experience the creation of the world. They are somewhat then akin to the commemorative and ceremonial purposes of the Catholic Stations of the Cross, on a landscape scale.

While there is widespread professional archaeological acknowledgement of the Indigenous ritual origin of these surface features, their age has been uncertain, as has the potential antiquity of the associated pilgrimage and ritual belief systems tied to them. This is important in part because the history of Yuman-speaking peoples in this region is uncertain, as discussed above. There is archaeological consensus that Yuman speakers were present along the river corridor by at least 700 CE (Schroeder 1979), based on the development of the Hakatayan/Patayan archaeological “culture,” but their initial arrival in this region is unresolved. Many archaeologists suggest a relatively late River Yuman arrival (e.g., von Werlhof 1995; Laylander 2001, 2010, 2015). Linguistic evidence, however, discussed above, alternatively suggests that they may have been present much earlier, especially if the Colorado River Delta and/or Colorado Desert area is the homeland for the Proto-Yuman language, as Mixco (2006) and others (Law 1961) have hypothesized. Our varnish dates were partly intended to help resolve that question.

3.1. VML and CR Ages on Intaglios

At the request of the Bureau of Land Management El Centro, Needles, and Yuma field offices, we collected upthrust cobbles from the culturally cleared areas within intaglios at a series of sites, accompanied by members of the Fort Yuma Quechan Tribe Historic Preservation Office and the Fort Mojave AhaMakav Cultural Society.

Table 1. Minimum rock varnish ages for intaglios obtained in this and prior research, as explained in the text, using three different rock methods. The CR ages without matching VML ages have been calibrated from raw data presented in von Werlhof et al. (1995), using the procedures in Bamforth and Dorn (1988).

Intaglio, Site	VML Age cal yr BP	CR Age cal yr BP	Lead Profile
Zigzag/Snake, Quien Sabe	650	500 ± 300	Pre-20th century
Stick Figure, Winterhaven	n/a *	750 ± 300	n/a
Anthropomorph, Pilot Knob	n/a	850 ± 350	n/a
Anthropomorph 1, Blythe Giant	n/a	1000 ± 350	n/a
Quadruped, Blythe Giant	n/a	1100 ± 400	n/a
Anthropomorph 2, Blythe Giant	n/a	1100 ± 400	n/a
Anthropomorph, Ripley Complex	between 1100–1400	1200 ± 450	Pre-20th century
Cross, Ripley Complex	between 1100–1400	1250 ± 450	Pre-20th century
Largest Anthropomorph, Quartszite	n/a	1350 ± 350	n/a
Amorphous Form, Quartszite Airport	n/a	1400 ± 400	n/a
Anthropomorph, Quien Sabe	1400	1500 ± 450	n/a
Lizard Figure, Ripley	n/a	1600 ± 500	n/a
‘Snake’ Head, Singer Complex	n/a	1900 ± 550	n/a
‘Snake’, Museum complex near Ocotillo	n/a	2900 ± 750	n/a
Schneider Dance Circle, Yuha Mesa	n/a	3200 ± 750	n/a
Anthropomorph, Quien Sabe	5900	6100 ± 1200	n/a

* n/a—not available.

presents a compilation of rock varnish minimum-limiting ages for these intaglios, found on both sides of the lower Colorado River, including new VML, CR, and lead-profile ages. Included in the table are newly calibrated CR ages from a prior study (von Werlhof et al. 1995)3 in addition to the new chronometric estimates. The VML ages represent the oldest microstratigraphic sequence observed from 10 cobbles analyzed from inside each of these features. Control cobbles from the unaltered desert pavement surrounding each feature were also analyzed. These “controls” were all Late Pleistocene in age, precluding the possibility that our sampled cobbles in the archaeological features were contaminated by the surrounding intact desert pavement.

The minimum-limiting age estimates range from 650 to 5900 cal yr BP, or from the Late Precontact period to the middle Holocene. Figure 4 presents the oldest of the observed VML sequences for two intaglios from the Ripley Complex, Arizona, perhaps the largest intaglio site along the Colorado River. An orange layer underneath the black WH2 layer is shown in Figure 4A, indicating that its varnish began to develop between WH2 and WH3, or from 1100 to 1400 cal yr BP. Figure 4B displays only a very small section of the orange laminae, with the black WH2 layer as its basal layer. Since varnishing always post-dates exposure of an alluvial pavement clast, and since upthrust cobble exposure itself takes time, an appropriate minimum age assignment would be 1100 to 1400 cal yr BP for both intaglios.

Note that an area labeled “bacteria cast” in Figure 4A is a cluster of budding bacteria, with their cellular casts encrusted in Mn and Fe. These casts are normally dissolved and remobilized as varnish grows (cf. Dorn 2024), but some preserved areas may occur where varnish is particularly fast growing. This is consistent with preservation of the three Little Ice Age wet events, present in both profiles and indicative of relatively quick varnish development. Lead-profile dating is also consistent with these analyses, as are the CR ages.

That both Ripley intaglio VML ages fall within the same time range suggests that the two geoglyphs may have been created in the same event, or at least during the same general time period. This possibility is enhanced by the CR ages from the famous Blythe Giants site, all three of which are internally consistent, with average dates of 1000–1100 cal yr BP. This perhaps indicates that at least some of these sites were planned compositions rather than accretions of motifs added over time. Note further that there seems to be a cluster of minimum ages roughly occurring in the first millennium of the Common Era, when the CR dates on samples without accompanying VML ages are included. This may represent a period of intensified ritual activity.

The Quien Sabe Point site, however, presents a very different circumstance. Figure 5 shows three previously undated geoglyphs from this locale. Ages indicated by the oldest VML sequence from each of the sampled intaglios are presented in Figure 6. These range from 650 to 5900 cal yr BP, which is corroborated by the CR estimates, suggesting that new intaglios were added to this site over a very long period. This demonstrates both the long-term continuity in the use of a specific location for rituals, as has been documented in other Native American cases (e.g., Sundstrom 1996; Whitley et al. 1999), as well as change over time in the sense that new motifs were successively added to the markers, signaling these locations as sacred. This last fact counters any claim that the intaglio ritual tradition was somehow ahistorical—that is, unchanging over time—given that it involved the addition of new symbolic elements and, as indicated by the results from the Blythe Giants and Ripley sites, additional sites/ritual locations over time.

We emphasize, again, that the oldest VML and associated CR ages for each intaglio are the most appropriate estimates because all varnish dates are minimum-limiting ages. This is due to time lags in pavement reformation and in varnishing. The 5900 cal yr BP age for one of the Quien Sabe anthropomorphs relative to the other two dates from this site, however, raises an obvious question of reliability. This early date is supported by the VML age distribution of the other nine sampled cobbles from this specific intaglio. Two of these have the same WH6 sequence (5900 cal yr BP) age; five have the WH3 sequence (1400 cal yr BP); and two have WH1c at their base (650 cal yr BP). The VML age distribution for these ten sampled cobbles from a single intaglio could reflect ongoing pavement reformation processes, ongoing lag in varnishing, or some combination of these processes. The age distribution might also reflect imperfect “upkeep” (that is, periodic clearing of newly upthrust cobbles in the intaglio), since we know that, in a relevant parallel case, ritual initiates were required to maintain the Trail of Dreams by tamping down this trail used in their puberty rites with logs (Whitley 2014). Since the VML age of the adjacent pavement control cobbles is at least 36 ka, we are confident that WH6 (5900 cal yr BP) is an accurate minimum age assignment for the oldest of the intaglios and not a result of contamination from the surrounding pavement.

The results from the Singer Complex (Figure 7) warrant brief discussion. This site is located in an area where armored infantry troops trained during World War II, west of the river corridor, and pavement disturbance from tanks is visible in the immediate site area. This has led to questions about the authenticity of this site. Are the geoglyphs Indigenous in origin or instead the result of military activities of some kind (cf. Casey 1992)? This last possibility might seem supported by the site’s location, some distance from where most other intaglio sites occur, and due to the relatively unique nature of its intaglios—very long, snake-like forms. The 1900 cal yr BP resolves that question, demonstrating that these intaglios long pre-date Euro-American arrival in North America.

3.2. Implications of the Intaglio Ages

There are two issues concerning our intaglio rock varnish ages, ranging from 650 to almost 6000 years ago, that warrant discussion. The first is methodological. It concerns the fact that dating desert pavements is predicated on the processes that re-generate these surfaces over time, as new cobbles are upthrust into a cleared surface. The important question then is how an intaglio can persist for as long as 6000 years, given that much younger archaeological features may sometimes be buried under desert pavements as they re-form (e.g., Ahlstrom and Roberts 2001). The likely explanation, as noted above, is that these geoglyphs were actively even if not perfectly maintained, keeping them (mostly) clear of newly upthrust cobbles over the millennia. Ethnographic evidence indicates that the cleared surface of the so-called Trail of Dreams, which runs through desert pavement and was used for the puberty initiation run of Quechan boys, was tamped down periodically by the boys, using logs, so that it would be cobble-free (Whitley 2014). It is likely that a similar practice existed for the intaglios themselves and that this accounts for their persistence.

There is then a plausible even if not conclusive argument suggesting that the 5900-year antiquity of the intaglios reflects an early and, ultimately, enduring River Yuman religious and ritual system given the localization of the intaglios along the river corridor (von Werlhof et al. 1995, p. 259; Billo et al. 2013, p. 1298; Wright et al. 2024, p. 83) where these peoples resided. This would comprise a third long-lived North American rock art tradition, joining the Great Basin petroglyphs (Whitley 2013, 2019; Whitley et al. 1999) and the Pecos River pictographs (Steelman et al. 2025), each of which also lasted for thousands of years. The three traditions provide a stark contrast to the changes in subsistence strategies and artifact assemblages over the same time periods in these three respective regions. This points, on the one hand, to the poverty of adaptive patterns and tool-making practices for understanding the cognitive and religious lives in Precontact Native America (Whitley 2013). But it does not, on the other, imply that these traditions were necessarily timeless and ahistorical, as is sometimes alleged. Such criticisms are based on confusions about the nature of religious change, in that they assume that historical changes necessarily require catastrophic breaks or massive shifts in practice and belief. As Bloch (1986, 1992) has illustrated, religious change over time often instead may involve a persistent core of ritual practice that alternately is elaborated and embellished, followed by a retrenchment to the basic pattern. Such a pattern of change, in fact, can be cited for the desert west, where the historical Ghost Dance developed out of traditional Paiute and Shoshone beliefs. With the failure of its prophecies, however, the Basin religion returned to its core of traditional beliefs and practices (Jorgensen 1986). There is justification in assuming that similar processes of religious expansion and retraction, in other words, may have occurred along the Colorado River, contributing to persistence yet also variability over time.

4. Age, Origin, and Meaning of the Topock Maze

Few surface rock features are visually more impressive, and more controversial, than the Topock (or ‘Mystic’) Maze geoglyph site (CA-SBR-219). Located on adjacent Colorado River terraces near Needles, California, the site consists of an extensive series of low, linear gravel rows that have been scraped into the desert pavement, creating numerous parallel cleared paths, each about 1.3 m wide (Figure 8 and Figure 9); despite the name, these cleared paths do not form an actual maze. The site area has been heavily impacted by over a century of development, starting with the Atlantic and Pacific Railroad right-of-way through the site in 1893. The subsequent construction of Interstate-40 and the railroad grade now divide the locality into northern and southern areas. The southern section is bounded by a PG&E pipeline compression station and county wastewater treatment plant on the east and west, respectfully, with steep terrain to the south, while the northern portion has been impacted by a variety of projects, including the construction of a marina along the river. Extant sections of the gravel alignments today cover about 13 hectares (34 acres), while the remnants suggest that it may have originally been, very roughly, 100+ hectares (247 acres) in size. The site was listed on the National Register of Historic Places (NRHP) in 1978, signaling its importance as a heritage resource.

Controversy over the origin and function of the site has long existed (see Haenszel 1978; Musser-Lopez 2011), however, despite its NRHP listing, with some contending that the site resulted from the collection of gravel for railroad track ballast. This argument has been putatively supported by an early article describing construction techniques for the 1893 Red Rock Bridge, which crosses the Colorado River at this location. This states in part that, for construction purposes,

The process of gathering [gravel] was to rake these fragments of [surface] stone into windrows and haul them by wagon to a pile where convenient to load into a car when needed … Indian labor was used very successfully for this as well as for labor about the caisson.

(Rowe et al. 1891, pp. 692–93)

Ethnographic and archaeological evidence, in contrast, supports an Indigenous religious origin and use of the site that pre-dates the railroad and bridge. Edward S. Curtis, writing at the turn of the century for example, published an ethnographic account of the Topock Maze, stating the following:

The Mohave Indians near-by have utilized the area so marked, in recent years, as a maze into which to lure and escape evil spirits, for it is believed that by running in and out through one of the immense labyrinths one haunted with a dread may bewilder the spirits occasioning it, and thus elude them.

(Curtis 1908, p. 55)

It has been further reported that this ritual was specifically followed by warriors returning from travel outside Mohave territory who were concerned with ensuring that they would not be contaminated by their exposure to foreign dangers manifest in ghostly spirits (Haenszel 1978).

The unpublished ethnographic notes of John Peabody Harrington (n.d.), also written early in the twentieth century, further support a Mohave origin for the site. Harrington’s notes include an account of a discussion he had with Arthur Woodward, an early California archaeologist, about the site. According to Harrington, Woodward was aware that bridge construction had resulted in the creation of gravel windrows on the Arizona side of the Colorado River, but Woodward was certain that the much larger “maze” on the California side was Indigenous. Supporting this conclusion, Woodward noted that it originally included a large anthropomorphic intaglio, similar to the giant earth figures that are documented in numerous places along the river. What may be a small segment of this motif is still present at the site (Haenszel 1978). The intaglios depict mythic actors, were placed at the locations of mythic events, and were used in a ritual pilgrimage that celebrated the creation of the world (Whitley 2014), as described above.

Malcolm Rogers was the first archaeologist to publish information on the site, although his written discussion was brief. But his unpublished notes, written prior to 1939, concur with and elaborate on Woodward’s observations to Harrington. Rogers states in part that

Early settlers claim that when the Santa Fe R.R. [originally the Atlantic and Pacific] was built (1893) several acres of the lower end were gathered up for ballast for the RR tracks and that a large shrine at the lower [i.e., northern] end on the River Trail was at the same time destroyed. This shrine contained potsherds and artifacts. In the assemblage was [sic] stone axes and some turquoise jewelry. The informant F.M. Kelley of Needles said that near the base toward the river there was previous to this a large intaglio human figure similar to those at [another site]. Mohaves in early days disclaim having built the maze and that it had always been there but that in the old days they used it for ceremonial purposes occasionally.

(M. Rogers n.d.)

Rogers’ unpublished account suggests that, while the site is in fact Indigenous in origin, it was created in Precontact times, was used ritually, and a portion of it was destroyed for railroad track ballast. This reconciles the contrasting non-Indigenous versus Indigenous accounts of its origin.

Widespread public perception has also acknowledged the Indigenous origin of the Topock Maze from the turn of the century onward, pointing to its ritual importance. A Fred Harvey Company postcard (Figure 10) dating from the early 1900s, for example, is a painting of the site showing the gravel rows, a rock shrine, and two Native Americans with horses. These postcards were sold at restaurants (“Harvey Houses”) at rail stations and were thus widely distributed across the West. Popular accounts, similarly, have associated it with the land of the dead and/or the Mohave mourning ceremony, though somewhat ambiguously (Haenszel 1978; Meloy 2003).

A third interpretation of the origin of the Topock Maze developed during the 1970s when Robert F. Heizer and C. William Clewlow, Jr. mapped and sampled the site to test the hypothesis that it resulted from Indigenous farming practices (personal communication from Clewlow to Whitley 1977). They took soil samples for pollen analysis and, finding none that could be the result of Indigenous farming practices, concluded that the site instead resulted from railroad gravel operations. Although they did not publish their study, Clewlow provided a useful summary included in Musser-Lopez’s (2011) discussion of the site.

More recently, Musser-Lopez (2013) has revived the railroad track ballast hypothesis, in part by arguing that the gravel rows were created partly using horse-drawn Fresno or Buck scrapers. As is clear, this would require a recent—less than 150 years—age for the site. It would further discount the Indigenous knowledge that has supported a ritual use for the locale, including the rationale for recent tribal efforts to protect the site.

4.1. VML and Lead-Profile Dating of the Topock Maze

Ten cobbles were collected from within the cleared pathways between the cobble rows from the extant southern section of the Topock Maze, along with three control cobbles from an adjacent, unaltered desert pavement area as controls. Four cobbles from within the archaeological feature had varnish coatings, with a VML sequence that started to form when microlamination WH2 (Wet Holocene 2) was deposited, about ~900–1100 cal yr BP, providing the oldest age estimate for the surface feature. The corresponding CR ages for these for cobbles are 1200 ± 400; 1150 ± 400; 950 ± 400; and 850 ± 350 cal yr BP.

The other six cobbles had younger VML and CR ages. It is instructive to provide here the range of younger ages for the Topock Maze, because they are all consistent with the finding that this feature is prehistoric. The microlamination sequence for five other cobbles started at the WH1c (Wet Holocene 1c) layer, deposited about 650 cal yr BP (Figure 11, while one cobble had a microlamination sequence falling between 650 and 900 cal yr BP. The CR ages for the other six cobbles are 700 ± 350; 700 ± 350; 700 ± 350; 600 ± 250; 550 ± 250; and 400 ± 250 cal yr BP.

Lead-profile dating was also employed, yielding findings that are consistent with the VML and CR ages in two ways. First, all ten cobbles had surface lead spikes with background levels of lead underneath, confirming that the Topock Maze is pre-20th century in age. Second, the lead values are measured every micron; this provides overlap in measurements, but it also gives a stratigraphic perspective on where the 20th century (high PbO values in Figure 11) spike ended in relationship to when the Little Ice Age ended, at about 300 cal yr BP (WH1a in Figure 11).

These results likely reflect one or more different possibilities that are not mutually exclusive: the process of upthrusting cobbles into the cleared paths was ongoing; younger results reflect a time lag in the revarnishing process; and/or cleared pathways were imperfectly maintained. This uncertainty demonstrates the importance of multiple samples for age determinations. We conclude, accordingly, that the Topock Maze is about 900 years old at minimum.

The natural pavement control cobbles collected outside the Topock Maze had much older VML sequences that were late Pleistocene age. There was one exception (Figure 12). A cobble collected about 0.5 m from the maze showed an angular unconformity produced by varnish erosion followed by a new sequence of VML deposited on top of the eroded surface. The new varnishing started during WH6 (about 5900 cal yr BP). The cause of this unconformity is unknown, although it is not due to wind abrasion. We speculate that it may be evidence of a mid-Holocene construction phase at this feature, and we include this information partly to emphasize that our varnish ages are minimum-limiting dates.

4.2. Function, Meaning and Symbolism of the Topock Maze

VML, CR, and lead profile results are all consistent with the interpretation that the Topock Maze is at least 900 years old and thus represents a Precontact feature, not a recent railroad-related activity. It is important to emphasize, however, that the upthrust sampled cobbles from within the surface feature were exclusively derived from the extant portion of the site—a relatively small area at its southern extreme. The large majority of the site to the north had been destroyed, including an intaglio that was once associated with the gravel rows. As with the intaglios, discussed above, it is possible (if not likely) that this surface feature was ritually maintained over time, to prevent the regeneration of a complete pavement surface. It is also unlikely that the feature was created during a short period of time, given its large size, the labor required to create a construction this massive, and the small-scale society that occupied the river corridor in Precontact times.

We accordingly do not suggest that 900 years necessarily represents the maximum antiquity of this feature but instead that it is simply the currently identifiable minimum age of the southernmost section of the site. It is plausible, if not likely, that the northern destroyed sections of the feature were built at different times, perhaps much earlier. For example, the unconformity seen in the varnish in Figure 12, dating about 5900 cal yr BP, could potentially reflect earlier construction activity, as noted above. This emphasizes again that our derived chronometric ages are minimum-limiting constraints and not necessarily estimates of the overall antiquity of this feature and its associated ritual practice.

But the Pre-Contact rock varnish age on and the cryptic ethnographic and popular descriptions of the Topock Maze provided above beg an important question: are there additional ethnographic data that support and augment our understanding of its symbolic meaning, ritual function, and use? Perhaps not surprisingly, there is, in fact, significant but heretofore overlooked evidence important to fully understand this cultural landscape feature. It starts with interviews obtained in the last few decades which support the earlier reported ethnographic accounts. But what is also included is overlooked information from older ethnographic studies about traditional fears of travel outside of Mohave territory and its association with ghosts and thus the need to conduct rituals to nullify this ‘foreign contamination.’ Further information is provided, more generally, about the association of Topock with beliefs about the spirits of the dead and thus its symbolic importance as a religious place along the Colorado River.

Recent accounts provided by Mohave tribal members about the Topock Maze, first, have been summarized as follows:

“interviewees suggested that stories or songs telling of its construction were present in the Mojave culture, but these stories are only told in some family lines and are not known by everyone … Other interviews in the 20th Century suggested that the Mojave would use the Maze to purify themselves by running through the Maze or by navigating through the Maze without walking over a windrow, leaving evil spirits or ghosts in the Maze, or that the purpose of the Maze is to help the deceased atone for their life before fully passing to the afterlife”.

(AECOM 2010, Section 4.4, p. 34)

This corroborates the earlier accounts claiming that the Topock Maze was used for ritual purification involving the shedding of ghosts obtained while travelling outside of Mohave lands.

Second, according to Devereux (1961, p. 129), helping to explain this perceived need, a “’xenophobia of fear’ is a major theme in [traditional] Mohave culture,” causing them to “refrain from all close contact with other tribes, and even more with intimate connections with alien races.” Alien contamination illness, ahwe: hahnok, was in fact one of three categories of disease linked to ghosts (Devereux 1961, p. 21), and contact with foreigners was thought the chief cause of insanity (ibid, p. 32); “the Mohave experience one and the same type of dread in connections with aliens, magical powers, and the ghosts of the deceased” (ibid, p. 133). Their fear of aliens and thus ghosts was linked to their beliefs that “the soul of the dead person draws them to the land of the shadows” (Devereux 1937a, p. 411; our translation), that is, the Land of the Dead.

The paradox is that the Mohave were great travelers and warriors, with trading, raiding, and scalping all putatively exposing them to this contaminating illness. Scalpers were especially subject to the “nefarious influence of scalps, prisoners and aliens,” which resulted in fainting, hollering at night, and more general aberrant behavior (Devereux 1937a, p. 43). According to Fathauer:

“The scalper, being a shaman, has power over this disease and can cure people afflicted with it. The scalps … bring beneficial power to the tribe after they have been tamed. The scalper, then, contributes to tribal welfare by his power to tame the scalps and to cure the “enemy sickness.” He also directs one of the most important Mohave ceremonies [the Mourning Ceremony]. Scalper is one of the most important religious statuses”.

(Fathauer 1951a, p. 275)

A ritual, conducted by the scalper, or “Ghost Doctor,” was then required to “tame” the scalps, that is, to nullify the dangerous power they otherwise contained and that resulted in alien contamination sickness. Subsequently,

“When the scalper returns with the war party he turns the scalps over to the kwaxot, or custodian of the scalps, who is the principal Mohave religious leader. The custodian of the scalps prepares a great celebration in honor of the returning warriors … After the feast the custodian of the scalps places them in large pottery ollas for safekeeping”.

(Fathauer 1951a, p. 275; cf. Kroeber 1925, p. 752)

The Mohave were also renowned for capturing girls and young women, but not men, on their raids. Because of these females’ potential for introducing foreign illness, “a ceremony had to be made over them else they would bring sickness into the land; and even after purification they seem more generally not to have been married” (Fathauer 1951a, p. 275). Fathauer notes, however, that

The female captives are given by the custodian of the scalps to some of the old men who need wives. Young men are afraid to take these women because of the ‘enemy sickness’ but the old men are glad to have them since they have lived a long time and do not have long to live under any circumstances.

(Fathauer 1951a, p. 275; cf. Devereux 1961, p. 42)

Mohave ethnography then supports the implications of this limited commentary on the ritual use of the Topock Maze in the sense that the Mohave considered themselves spiritually contaminated, by ghosts, while traveling outside their lands and from interactions with foreigners, despite the fact that this was a relatively common experience, especially for traders and warriors conducting raids. It also demonstrates that specific rituals were conducted to prevent the ghost sickness that could result from some of these activities. It follows that a ritual would be necessary to ensure good health after foreign travel more generally, and thus the Topock Maze may have served that purpose, as suggested by an account by Curtis (1908) and a recent account.

Third, there is, furthermore, important ethnographic evidence that explains the logic of the ritual use of the Topock Maze as a place to “lose” a foreign ghost after traveling outside of Mohave territory, based on its association with spirits and the Land of the Dead. Although there is minor variation (or confusion) in the specific geographical details, three different accounts all clearly refer to the Topock area. (Note that the general toponym ‘Topock’ strictly refers to a small, census-designated community in Arizona, immediately across the Colorado River from the ‘maze’ in California, possibly explaining the confusion.) These start with Bourke’s early publication, which stated,

“That other sharp, high mountain, down there near the Needles, in Arizona, was also a spirit mountain; that was where the Mojaves went when they died. (It was the Mojave Elysium)”.

(Bourke 1889, p. 172)

Devereux augmented Bourke’s account, despite some degree of uncertainty over the meaning of his informant’s comments. He noted that

The entrance to the “land of the dead” (cilia’yt) is somewhere near Needles, California, almost by the Colorado River on the Arizona side. There is something that looks like a big invisible “wash” containing a big invisible shed [i.e., ritual ramada] near a place called Ahatcku-pi’lyk, which is but a few feet [sic] from the land of the dead.

(Devereux 1937b, p. 419)

The Mohave in fact held that the afterworld was at or near the confluence of the Colorado River and the Bill Williams Fork. It most likely was in the sandy Chemehuevi Valley on the western, California side of the Colorado across from the confluence of the two rivers, roughly 20 miles southwest of Topock, with its name variously transcribed as cilia’yt (ibid.), calya: at (Devereux 1961, p. 146), sil’aid (Wallace 1947, p. 256), salya:yt (Stewart 1973, p. 316), and saly’at, “the Happy Hunting Ground,” from selye’aya, ‘sand’ (Munro et al. 1992, p. 160; cf. Kroeber 1948), although some instead claimed that it was under the Colorado River or in sand dunes on the Arizona side of the river (see Bourke 1889, p. 174; Harrington 1910, p. 333; Kroeber 1902, p. 280; 1925, p. 727; Devereux 1935, p. 114; Devereux 1961, p. 146; Drucker 1941, p. 148; Laird 1976, p. 134). Regardless, Fathauer clarified the earlier comments, observing that

“Following cremation, the soul remained near the site of the [funeral] pyre four days. At the end of this time the soul changed into a ghost which was then able to see the road to the afterworld. This started at Topock and ran south into the desert in the neighborhood of the Bill Williams River”.

(Fathauer 1951b, p. 605; emphasis added)

Fathauer’s account is confirmed by recent ethnographic data which state that “The Topock Maze … is the passageway to the next dimension, to the land of those who have passed on” (BLM 2012, p. 50). The Topock Maze locale was then the portal to the trail to the land of the dead, which, following the widespread belief of Yuman speakers, ran to the south. Regardless of the exact location of the afterworld, the Topock Maze was the place from which ghosts departed towards that destination and was, symbolically, the entrance to the Land of the Dead. It was thus the appropriate location for a ritual intended to send off ghosts to a location where they would be harmless to living people.

That the maze was the start of the path to the Land of the Dead suggests a potential additional ceremonial use of this locale, beyond its employment by those returning to their home needing ritual purification. This is implied by a comment by Devereux who stated that, “when a man dies his relatives often request a shaman to visit the land of the dead to check up if he reached it” (Devereux 1937b, p. 419).

“When rumors spread that a person has died as a result of evil practices, the members of their family send a bona fide sorcerer to visit the land of the dead to see if the deceased’s soul has arrived there. If the messenger does not encounter this soul, it is ipso facto proof that the soul is being held captive somewhere by the sorcerer”.

(Devereux 1937a, p. 410)

Sometimes a shaman would even transport an individual to this ghostly realm to visit the deceased.

The ghost doctor also could take people to the spirit world, although he did not encourage this because it was dangerous. He warned the person who wished to see a dead relative: ‘Be careful. If our hands slip apart, I’ll have to look for you all night. If I don’t find you before morning we will both be stuck here.’ The shaman and the person who was to accompany him dressed in their best clothes and painted themselves. About twilight they built a small brush shelter and then lay down to sleep with their hands clasped. In less than an hour they were transported to the afterworld. The ghost doctor knew exactly where the person’s family was, so they went directly there

(Fathauer 1951b, p. 605)

It is possible, although far from certain, that these rituals to enter the Land of the Dead occurred at the Topock Maze location.

There is, finally, one other implication of the Topock Maze that warrants brief mention. This concerns the implications of its original size of roughly 250 acres. This surface feature is, by any measure, a monumental construction, a trait not previously thought associated with the small-scale, low-population-density horticultural societies like the Mohave. It seems unlikely that a surface architectural feature of this size could have been created in a single episode, given these constraints. The most likely scenario would be its construction over multiple generations, perhaps with additions to its size occurring whenever it was used ritually, much as male puberty initiates were tasked with maintaining the Trail of Dreams used in their ritual run noted above. The angular unconformity evident in the varnish coating of a control cobble (Figure 12) would be consistent with ongoing construction over a much longer period of time, with the varnish erosion event occurring at WH6 or 5900 cal yr BP.

5. Conclusions

The use of three different rock varnish dating techniques, all of which have been replicated and blind-tested, has allowed us to resolve two longstanding controversies concerning surface features on desert pavements in the North American desert west. The intaglios along the lower Colorado River, first, represent an example of ritual activity in this region. Rock varnish dates on these extend back to 5900 years B.P., demonstrating the existence of a lengthy religious tradition and artistic practice. Although only occasionally is there a perfect correlation between culture, language, and physical/biological type, we believe the mid-Holocene age of these geoglyphs may signal the presence of Yuman-speaking peoples in this area by that early date. This in turn would suggest that existing archaeological models and linguistic reconstructions arguing for a much more recent arrival of Yuman speakers are wrong. Instead, our evidence supports other reconstructions suggesting that the lower Colorado River and southernmost California region may have been the original homeland of these peoples.

Second, the Topock Maze has long been a source of popular fascination and professional debate. Some have argued that it is a by-product of late-nineteenth-century railroad construction, while others, including the Mohave tribe, have contended that it is both Indigenous in origin and that it was used for ritual purposes. Our chronometric ages indicate that the extant portion of this large construction, originally roughly 250 acres in size, is at least 900 years old. Other, now destroyed, sections could have been older or younger. Our review of pertinent ethnographic data, furthermore, indicates that it represented the start of the path to the Land of the Dead; hence, it was understandably associated with ghosts and spirits and is correspondingly a place of great religious importance.