Rock Art and Hunter–Gatherer Landscapes: Iconography, Cosmology and Topography in Southern Africa

Abstract

Landscape studies of hunter–gatherer rock art often suffer from logical flaws. Some of these failings stem from the founding question that researchers ask: “Why do some places have images while others do not?” This question is misleading and not particularly helpful in some—but not all—contexts where there is no direct ethnographic evidence to provide an answer. Instead, we suggest that a better question from which to begin is: “How are rock art images related to landscape?”. To answer this question, we examine the relationship between iconography, cosmology and topography in two areas of southern African San rock painting. We argue that cosmology guided iconography and that the imagery, in turn, manipulated topography into landscape for the San. In this view, we do not need to rely on cognitive templates that invest topography a priori with significance that then determines the choice of locale for art. Instead, landscape for the San was socially and symbolically constructed through the placement of imagery.

1. Introduction

Landscape is a recurring feature of archaeology. At the turn of the twentieth century, phenomenological approaches, coming out of post-processual archaeology, provided new impetus to the topic (for a review see Johnson 2012). More recently, ontological approaches coming out of New Animism, have also shown limited interest in rock art and landscape (e.g., Challis 2019). Although phenomenology and ontology are different spheres of philosophy, being concerned with human experience and interpretation of phenomena and with the classification and relation of things respectively, in archaeological landscape studies, there is often an overlap and blurring of these fields. Thus, much of what we say about phenomenological approaches (but not all) applies to ontological studies as well. Phenomenological approaches emphasize the human experience of space and although they have produced a wide array of novel ways of looking at archaeological landscapes, they lack a clearly defined methodology (Barrett and Ko 2009; Knapp and Ashmore 1999; Smith and Blundell 2004). At best, a phenomenological method for doing landscape studies is one in which scholars think their way back into the bodies of past peoples who once moved through the same space (see, for example, Tilley 1994). In doing so, researchers experience the present topography in—assumptively—the same way as those past peoples. Such approaches ignore cultural differences that manifest spatially and temporally; moreover, they project modern perceptions onto archaeological landscapes made by very different people in a distant time.

As part of the phenomenological influence at the turn of the twentieth century, global rock art research also engaged with landscape studies in new ways (e.g., Frederick 2000; Chippindale and Nash 2004; Tilley 2008). Broadly speaking, rock art studies on landscape can be divided into those that use ethnography and those that do not, either because it is not available or its relationship to the imagery is elusive. Studies that use ethnographic evidence tend to be more successful. In North America, for example, there is an abundance and diversity of material that shows complex and well-developed beliefs about topographical features and the relationship of cardinal points to a “sacred” geography (e.g., Sundstrom 1996). Topographical features, such as volcanic rocks, are typically regarded as sources of supernatural power and Native Americans would visit these areas as part of vision quests; their visions were typically engraved or painted on nearby rock surfaces (Whitley 2024). Where ethnographic material is not available, research on the rock art landscape is less convincing because of assertive generalizations based on the author’s own experiences (see Smith and Blundell 2004).

Nowhere is the power of ethnographic evidence to elucidate ancient iconography more evident than in southern Africa, where the ‘ethnographic turn’ in the 1970s revolutionized San hunter–gatherer rock art research (Lewis-Williams and Dowson 1994; Blundell 2004, p. 53ff). Until recently, research on rock art and landscape in the region has concentrated on engravings rather than the better-understood paintings (e.g., J. Deacon 1988; Ouzman 1995, 1998, 2001; Parkington and Rusch 2010). The historical progression of San rock art scholarship over the last four decades, with its emphasis on ethnohistorical sources, is responsible for this apparent anomaly (see, for example, Lewis-Williams 1995; Blundell 2004; Witelson 2024). While those sources show that the art concerns the religious beliefs and ritual practices of San peoples, there is little direct evidence on how those are related to topography, with the notable exception of the /Xam San material on engraved Karoo landscapes (J. Deacon 1986). The /Xam ethnography is concerned with terrain in which there are engraved images and very few so-called ‘classic San fine-line’ painted images (for an overview of southern African engravings, see Dowson 1992). The ethnographic-based research on /Xam engravings and more recent work on rock paintings use ethnography to understand landscape (e.g., J. Deacon 1986, 1988; Ouzman 1998; Lewis-Williams 2010b; Morris 2010; Challis 2019; Stewart and Challis 2023). We argue that, while these ethnographic approaches have made inroads in understanding San imagery and landscape over studies without ethnography, they nevertheless have logical weaknesses. These flaws are common to many rock art and landscape studies throughout the world and by considering the southern African example, we may possibly find ways to circumvent them. We propose analysing the inter-relationship between iconography, cosmology and topography as a starting point to overcome some of the criticisms that we raise here. Understanding this inter-relationship allows us to discern past hunter–gatherer landscapes in some areas and at some times at least.

To show the heuristic potential of this three-pronged strategy, we consider two areas of San rock paintings, each with an unusual motif. The first is the Matobo Hills in Zimbabwe, where unusual ‘formling’ motifs cluster and the second is the Cape Fold Mountains in South Africa, where bird-tailed therianthropes occur (Figure 1). Both motifs—formlings and bird-tailed therianthropes—have long-intrigued researchers and recent, independent, interpretative advances allow a consideration of their relationship to landscape. Combining this interpretative work with Michael Taussig’s (1993) discussion on mimesis and Alfred Gell’s (1998) ideas of secondary agency allows for a compelling argument for landscape as it was conceptualized, for at least some of the San artists in both areas. We begin with a consideration of landscape archaeology and, specifically, within rock art studies before discussing interpretations of the two motifs, and we then end with a discussion on how the motifs link to landscape.

2. Landscape: Problems and Possibilities

Leaving aside the philosophical issue of whether there is a world external to our embodied sensory apparatus, we define topography as the forms and features of land surfaces external to our bodies that we experience. Landscape, on the other hand, we regard as the construction or manipulation of those forms and features into symbolically and socially meaningful things. We focus in this paper on what is traditionally regarded as landscape study in rock art in that we consider the role of the locales of imagery within a world of meaning for the peoples who made them. We do not consider the issue of potential landscape imagery in the iconography itself (on this, see Solomon 1997). We accept that there is no way of perceiving topography that is not socially and symbolically loaded and that words such as ‘topography’, ‘landscape’, ‘space’, and ‘place’ have historically received and culturally specific values (Cosgrove 1985; Casey 2001). Nevertheless, the distinction between topography and landscape is analytically useful, particularly in archaeological contexts. If we understand it as symbolically and socially constructed, then it follows that landscape is also contested and that different people within the same community might have different landscapes. It also follows that landscape would be spatially and temporally contingent; different cultures will have different landscapes and landscapes could change through time, even if the topography remained relatively constant.

If landscape is the social and symbolic construction or manipulation of topography, then we can postulate that hunter–gatherer image-making played a role in such constructions or manipulations. Indeed, this is not a contentious statement, and it is the founding assumption for much of global hunter–gatherer rock art research on landscape. It is, however, where researchers go from this starting assumption, particularly in the absence of relevant ethnography, that creates logical difficulties. A common problem, for example, is the claim that there is a relationship between prominent topographical features, such as mountains or rivers, to a rock art site or sites (see Smith and Blundell 2004). The assertion here is that the locale might have been ‘chosen’ because of these features (Taçon and Faulstich 1993). In discussing prominent topographic features, researchers often talk about the view from sites and the viewshed between different sites, as if what a person chooses to see from a specific place is a universally shared value (Robinson 2010; Acevedo et al. 2019). The development of Geographical Information Systems (GIS) software has probably enabled the growth of such arguments, while simultaneously allowing for the elision of critical scrutiny of assumption.

In southern Africa, for example, Grant McCall (2010) uses multivariate statistical and spatial analysis (GIS) to identify what he labels as large residential, large public ritual, and small isolated private ritual sites in the Didima Gorge of the Drakensberg Mountains. This analysis, however, is based on a subjective reading of a place that he never visited (for a critique and response, see Challis et al. 2012; McCall 2012). Consequently, his desktop-GIS study degrades into an interpretation loaded with the terminology of a Western gaze that includes ideas such as “remote”, “large”, “small”, “public”, and “private”, without any effort to demonstrate those concepts within a San worldview. Moreover, McCall uses surface artefacts to support his argument. As Aron Mazel (2023, p. 37) points out, he “misinterprets the relationship between surface artefacts and painted assemblages by erroneously believing that there was a proportionate ‘relationship between the number of artefacts present at sites and the number of rock art elements’”. In another remote sensing (GIS) example from the Drakensberg region, James Pugin (2016) attempts to develop a predictive model for the location of rock art sites. He concludes that GIS-based studies have predictive value but only if they are based on large datasets. Those large datasets—of course—are based on rock art sites previously located without the aid of predictive GIS models and, to date, there are no convincing demonstrations of the putative prognostic capabilities of such GIS-based models in southern Africa.

Of significance here is that arguments for the importance of viewshed often require the researcher to turn their back on the images, in painted shelters for example. The rock art becomes a backdrop, a simple marker of the significance of ‘site’—an idea imposed by contemporary scholars rather than being an established ‘fact’ of the way past peoples thought of the locations of their art (see Witelson 2024). This is the sentiment behind claims that rock art was created to mark the landscape (e.g., Bradley 1997; Parkington and Rusch 2010). Moreover, arguments for the importance of view and viewshed typically rely on large topographical features. Some scholars, for example, argue that rock art was made on or near large and prominent topographical features to demarcate boundaries for past peoples (e.g., Lee and Hyder 1991; Bray 2002; Munson 2002, on the use of style and boundary, see Brady and Bradley 2014). Such arguments are often underpinned by ideas of resource stress, which purportedly drives the need for boundary marking, and they have functionalist tendencies. The emphasis on macro-topography and boundary-marking is, in the light of some ethnographic evidence, misleading. For many hunter–gatherers, small micro-topographical aspects often have greater significance than larger elements, and hunter–gatherers rarely show the same obsession with boundaries as modern people living in the relatively recent phenomena of nation-states (Smith and Blundell 2004). As Alfred Kroeber (1925, p. 939), the renowned Californian ethnographer, noted of Native American rock art:

It has sometimes been conjectured that the symbols served as boundary marks, direction signs, or for some analogous practical purpose. Yet this interpretation fits neither their character, their location, nor the habits of native life. The Indian knew the limits of his territory and his way around it; and as for strangers, his impulse would have been to obscure their path rather than blazon it.

A recent development that seemingly presents an alternative to view and viewshed is that of archaeoacoustics and the study of “soundscapes”. Archaeoacoustics in rock art research involves analysing the properties of painted or engraved sites for echoes, resonance, reverberation, intensity, and so forth (e.g., Waller 2002). Archaeoacoustics differs from the study of viewshed in that the researcher’s equipment is pointed at the actual imagery or area of imagery; they do not turn their backs on the iconography. Nevertheless, there is at least one attempt to marry GIS to acoustic research to show “soundsheds” in topography (Primeau 2022). The underlying assumption of archaeoacoustics is that differences in sound qualities played a significant role in the choice of locale at which to make imagery. In southern Africa, for example, Riaan Rifkin (2009) and Neil Rusch (2024) have undertaken archaeoacoustic research on rock engravings and Aron Mazel (2011) has promoted such studies for rock paintings.

Rifkin (2009, p. 587) argues that an engraved site in the north-west South Africa was an “engraved soundscape within a cultural landscape”. Rifkin deftly intertwines San ethnography with acoustic measurements; he shows how “percussive sound was an integral part of the performance of trance rituals” at a single site in the Korrannaberg and that the sound “relates pertinently to the engraved depictions” (Rifkin 2009, p. 593). This pertinent relationship finds expression in the placement of engraved imagery at the “exact points from which echoes emanate” (Rifkin 2009, p. 594). Mazel (2011, p. 293), on the other hand, working on San rock paintings in the same area where McCall undertook his GIS analysis, suggests that “the richness of Didima Gorge’s rock art and its acoustic properties…were connected in establishing this locale as a special place for the hunter-gatherers”. Although Mazel does not deploy any equipment to analyse the acoustic properties of shelters in the Didima, he points to the meaning of the word “Didima” in Zulu, which means “reverberating one” (Mazel 2011, p. 291ff; see also Pager 1971, p. 4). The word has a Xhosa origin and possibly derives from a /Xam San word that means “to roar” (Mazel 2011, p. 291ff). The reverberating probably refers to the sound of thunder so common in the Drakensberg Mountains during summer storms. Mazel (2011, p. 291) proposes that “Didima Gorge’s notable acoustic characteristics were central to the hunter-gatherer creation of its exceptional corpus of rock art.” Most recently, Neil Rusch (2024), inspired by Mazel’s work, has used archaeoacoustic equipment at the site of Kurukop, in the Northern Cape. Here he finds correlations between echoes and densities of engraved images and concludes that “sound plays a role that influences experiences of transformation, mutability and liminality” (Rusch 2024, p. 169).

While there is much value to what Rifkin, Mazel, Rusch and other scholars mentioned argue, there are at least three problems with archaeoacoustic arguments. First, they confuse correlation with causation. Rifkin (2009, p. 594), for example, prefaces his observations about the linkages between images and echoes by stating, “It is becoming increasingly evident that sound reflection may have been an essential motivating influence in the production and placement of rock art”. There is no independent evidence for such claims, external to a correlation between the visual imagery and the audio qualities of location. It is quite conceivable that another reason—inconceivable to a modern observer—led to the choice of that location for image-making and that the presence of acoustic qualities at the site was a subsequent, opportune discovery by the artists. Second, claimants for acoustic causality in the choice of image-making locale rarely test their hypothesis against the potential places for rock art that do not have imagery (but, see Diaz-Andreu et al. 2023). Do such potential sites not have any audio qualities? If they do have such audio qualities, how do they differ from the audio qualities of actual engraved or painted sites? In rare cases, this has been performed, as in Steven Waller’s (2005) research in Utah, where archaeoacoustic tests for echoes revealed a locale of rock art images hidden behind vegetation; however, such studies remain the exception rather than the norm. It is more than a little ironic, then, that claims about the non-visual aspects of engraved or painted sites rely on visual iconography to identify the locale as significant in the first instance. Third, claims for a correlation between acoustic qualities and location choice rely on large datasets to be inductively compelling, if not deductively conclusive. Rarely do researchers provide such substantive data, instead relying on a single site or a handful of sites to argue their point (for an exception to this, see Kleinitz 2004). In some recent studies, where substantive data are provided, there is no correlation between the acoustic properties of locales and the presence of imagery. For example, in the Altai region of Russia, investigators have stated that “we have to conclude that there is no evidence for acoustics to be considered an essential element for the selection of the places to be carved” (Diaz-Andreu et al. 2023, p. 14). In this study, only seven sites were tested—two ‘major’ sites, three ‘minor’ sites, and two areas without imagery (Diaz-Andreu et al. 2023, p. 5). It is thus hardly a substantive dataset; however, already in such a small sample, there is no correlation. In a more substantive study, this time from South Africa, 27 shelters in the Drakensberg Mountains were analysed for acoustic properties; here again, the authors conclude that “in our study area, shelters with paintings do not possess acoustic properties capable of significantly increasing the sensory impact of sonorous cultural practices conducted in them, and that concentrations of images are not related to the presence of particular acoustic effects” (Santos da Rosa et al. 2025). These two studies alone raise scepticism about any claims for correlation between the acoustic properties of locales and the presence of rock art that relies on a single site or a handful of sites in a small area. The debilitating issues that plague GIS-based studies on vistas and viewsheds, then, also hobble archaeoacoustic research on soundscapes and soundsheds.

In the absence of specific ethnographic evidence about precise places, arguments that rock art sites mark boundaries and those that suggest the view, viewshed, acoustics, or soundshed were significant in the choice of location for image-making tend to treat the imagery as passively subordinate to terrain rather than considering the way in which iconography interconnects with topography to construct landscape. In McCall’s case discussed earlier, for example, there is no effort to link the iconography specifically to topography at Didima; instead, the numbers of images are used to talk about site size as ‘large’ or ‘small’ (McCall 2010). In the case of the archaeoacoustic research that we mentioned, studies also focus on the number of images rather than the iconography; denser concentrations of imagery are thought to be located at places where the acoustic qualities are most pronounced (e.g., Rusch 2024). The emphasis on the view, vista, viewshed, acoustics, soundshed, and concomitant ideas of boundary-marking probably says more about the modern archaeologist’s gaze in the present than they do about the landscape for archaeological hunter–gatherers. This type of ‘gaze and guess’ approach has long been eschewed in a southern African context in favour of ethnographic interpretation for the rock art imagery (Lewis-Williams and Dowson 1999, p. vi).

The use of ethnography, often extended with the near-universal human experiences in altered states of consciousness (a prominent element of San ritual practice), has produced significant insight into the meanings and motivations of San rock painting (for essays on ethnography and San rock art, see Blundell et al. 2010). Herein, then, lies the reason for the relative paucity of research on the landscapes of San rock paintings; there is very little ethnographic evidence available on how San rock painters perceived the topography in which they made paintings (but, see Orpen 1874). As mentioned earlier, the ethnographic material on the San comes predominantly from the semi-arid and largely flat interior Karoo region in which engraving, not painting, was the dominant mode of representation or from the Kalahari Desert, another flat area where there are few mountains and few paintings (see Campbell et al. 1994). Consequently, studies of rock art and landscape have focused on rock engravings.

The most notable work in this regard is that of J. Deacon (1988), whose pioneering effort was on the Northern Cape Karoo homeland of the /Xam San, with whom Wilhelm Bleek and Lucy Lloyd worked in the 1870s (Lewis-Williams 2000). Using the Bleek and Lloyd ethnography, Deacon was able to identify the actual places where the /Xam San lived and where their immediate ancestors made rock engravings. Apart from being a remarkable piece of detective work, Deacon made poignant observations about the relationship of engraved imagery to topography. She pointed out, for example, that at the sites she worked on, engraved images tended to cluster around known waterholes, on northern hillslopes from whence rain would come in that part of the Karoo or at places “that have legendary connotations for rain” (J. Deacon 1988, p. 136). Deacon concluded that the engraved imagery at these places “reinforced the power already present there because the animals and other themes depicted carried metaphorical significance that inspired the rainmakers and their assistants” (J. Deacon 1988, pp. 136–37). Later, she argues,

If the power lay in the place and its legendary significance as well, then it would attract rock art and localize it to a greater extent than if the art alone provided the inspiration and this might explain why rock art often appears as localized concentrations whereas nearby places remain without art. Local variations in themes and engraving techniques would be the result of both the individual trance hallucinatory experiences of the artists and of the metaphorical meaning attached to the place through the shape or prominence of landscape.

(J. Deacon 1988, p. 138)

Deacon’s seminal work was the first to recognize the significance of topography in southern African rock art. Since then, other scholars, including Sven Ouzman (1995), David Morris (2002), and Jeremy Hollmann (2007), have built on her work. Ouzman’s (2001) work on engravings, in particular, has forced consideration of the ‘non-visual’ elements of rock engraving sites. To varying degrees, each of the scholars mentioned above talks about the “power of place” when discussing the relationship of rock engravings to topography. Yet, rarely do they articulate what they mean by “power”, and it is unclear whether they think that the San thought of rock art locales as infused with supernatural energy or as a source of such energy or if they simply mean that the place must have been special to the San in some way.

Importantly, in emphasizing the power of place, scholars often give significance to the locale independently of the iconography. David Morris (2002, p. 188), for example, suggests that at the engraved site of Driekopseiland,

…the glaciated rock which came to be exposed in the bed of the river, was pregnant with symbolic possibilities….That the site was so richly festooned with engravings is strongly suggestive of the view that the place and the rock surfaces themselves came, indeed, to be construed as meaningful.

It is, of course, the case that all rock must have been “pregnant with symbolic possibilities” for the San. The issue, however, is how we determine that certain places were more significant than others if the possibilities were not actualized through the making of images there. Similarly to Morris, J. Deacon (1988, p. 131) herself states:

The long-lasting quality of the engravings and their placement on dolerite boulders at selected spots can be seen as a deliberate attempt to mark the landscape. The fact that there is a wide choice of suitable sites, yet relatively few have been selected, would suggest that the choice of a site for engravings was purposeful and therefore held some meaning.

Hollmann (2007, p. 61), citing J. Deacon (1988, p. 138), states the argument succinctly: “The markings and motifs are a consequence of the qualities of the place, not the prime reason for its importance”.

The problem with these arguments that talk about the locales of rock art as powerful places is that such claims are made in the absence of any explicit ethnographic support for the specific places discussed. There are no statements in the ethnographic record, as far as we are aware, where San describe specific rock art sites as “powerful places”, even in the case of the /Xam San. In the absence of such independent statements, arguments tend towards circularity. Places where rock art is found are said to be powerful: What makes those places powerful? The fact that rock art is found there. The scholars mentioned here are, of course, different individuals, and our summary does not do justice to the various nuances of each person’s argument. In particular, Deacon’s original ground-breaking and subsequent work is nuanced and more circumspect than many other claims for rock art locales as “places of power” or for assertions that iconography “marked” the landscape. In later work (J. Deacon 2001), for example, she posed the issue of what came first—the art or the place—as a conundrum. She is circumspect about her answer to this conundrum, and she offers two contradictory solutions. First, she answers it by pointing out that, “while both the art and the meaning of the place may not have been ‘born’ simultaneously, they enhance and support each other like religious beliefs, myth and reality did in /Xam thought” (J. Deacon 2001, p. 248). Second, she states that the “underlying structural principle that answers the conundrum is that a place must first be perceived as having potency and significance for shamanistic practice before it will attract rock art” (J. Deacon 2001, p. 249). Almost all researchers working on San rock art and landscape have followed her second answer. As will become apparent, we favour a variation on her first suggestion for the answer. We are sympathetic to arguments that the places where the art was made must have been significant to the San in some way; how could it be otherwise? Nevertheless, we hold that arguments about the “power of place” and rock art as a form of “marking” or “signing” the landscape, when employed as a reason for the presence of iconography at specific places, tend to have logical issues of circularity that need to be overcome.

One of the reasons, we think, that such arguments tend to circularity is because of the emphasis they place on trying to construct a San cosmology that incorporates landscape from the ethnography when that record is largely silent about specific locations of iconography. This silence leads to general models that, while useful to a degree, are not as compelling as one might hope. David Lewis-Williams (2010b), for example, sets out to build a model of how San peoples might have perceived the terrain of the Drakensberg mountains of South Africa, where so many rock paintings are found. He concludes that the San believed in a three-tiered universe and that this belief was “projected onto the Maloti-Drakensberg landscape” (Lewis-Williams 2010b, p. 15). This tiered cosmos comprises this world, another above and a third one below and, importantly, San ritual specialists (who were the artists) mediated these tiers (Lewis-Williams 1996). Importantly, as Mazel (2011, p. 285) points out, Lewis-Williams’s effort to extract San concepts of landscape from their mythology does not link specific images to topography but, rather, builds a general model for rock art locales in the Drakensberg mountains. Here, the relationship between imagery and topography is subsumed into a broad cognitive model that is imposed on terrain. Art is then placed within this space that is already pre-populated with locational significance. In this way, then, arguments situated within a deep understanding of San cosmology are like those that consider the view, viewshed, acoustics, soundshed, and boundary-marking in that they begin from the position that the locale must have prior significance. The art then ‘marks’ the location as significant.

The problem with treating cosmology as primary over both iconography and topography is that it implies a cognitive template which guides (or in the worst cases determines) the actions of agents. This problem is most pronounced in arguments that are based on the so-called “ontological turn”. Although there are a growing number of New Animist arguments in San rock art research, there are only a few that deal with the issue of rock art and landscape. These handful of landscape studies are typically oblivious of any of the earlier phenomenological literature on rock art and landscape and tend to deal with the relationship between image to landscape only tangentially, if at all. Andrew Skinner (2023), for example, extrapolates a ‘navigational theory’ from /Xam folklore but does not apply this extrapolation to San rock art. Sam Challis (2019, p. 1), on the other hand, working with a single image from a site at a high altitude in Lesotho, suggests that the place, “having potency, is therefore both powerful and dangerous”. There is ethnographic evidence, as we shall see later on, for rock art images being potent, because they are made with bodily elements from supernaturally potent animals, such as the blood of eland; however, there is no ethnographic support that, through some process of mystical osmosis, the imagery transferred that potency to the locale itself. There is in these and other New Animist research on San rock art, an emphasis on the construction of a broad “San idiom” or “Bushman Mind”—a cognitive template—and then, if discussed, rock art is situated within these constructions. Such templates are, of course, reminiscent of structuralism and are objectivist in that the model or template is constructed independently of individual agency and, in some cases, is treated as a determining mechanism. A recent example of this type of determinism is the work of Stewart and Challis (2023). Starting from a New Animism perspective, they argue for an “environmental ontology”, which they define as:

the way in which the landscape would have been seen by indigenous people—the San, their ancestors and descendants—an ethnographically-informed indigenous perspective on the environment and its resources, especially relating to peoples’ relationships with the entities that embody and influence those resources.

(Stewart and Challis 2023, p. 124)

They then set out to construct what the San landscape “idiom” might be, concluding, as does Lewis-Williams, that the Maloti-Drakensberg mountains “was arguably the San tiered cosmos manifest, giving physical form to the axis mundi by reaching up into the sky” (Stewart and Challis 2023, p. 131). Having established the San landscape idiom, they proceed to map out environmental changes over the last 4500 years or so for Maloti-Drakensberg. They then situate a single site from Lesotho within this environmental history. The site of Likoaeng shows a shift in faunal remains over time; fish bones replace mammal bones over time and, as time passes, the remains of a single species of fish dominate (Stewart and Challis 2023, p. 136). Stewart and Challis then provide interpretations about San hunter–gatherer responses to these changes. For example, they state that the “arrival of the rains, the subsequent running of the fish, and finally the arrival of the herds would have created conditions in which a number of actions—according to San idiom—were necessary” (Stewart and Challis 2023, p. 138). In another example, they say that a “response to the more temporally-focused eland aggregation would, in the San idiom, have required a mediation with the forces on the landscape…” (Stewart and Challis 2023, p. 139). In these examples, the word “idiom” can be read as “cognitive template” and, when Stewart and Challis use words such as “necessary” and “required” in these contexts, it reads as if the cognitive template demanded and/or determined action and responses in a particular way. There is very little room in their objectivist (in the sense of the word as used by Giddens 1984) argument for individual agency and variability and, in such arguments, people are treated as automatons who are governed by the rules of a cognitive template.

The various arguments for rock art and landscape that we have discussed so far—boundary marking, viewsheds, soundsheds, the power of place, and cognitive templates—all have fundamental problems with logic, and many suffer from a paucity of relevant data. The issues with these various arguments, stem, in part, from the question that many researchers—both in southern Africa and beyond—begin from: “Why did hunter-gatherers paint or engrave at some locations but not at others?” In the southern African case, this question eventually leads to an impasse. Asking why some rock art ridges were engraved or shelters painted and others not is seemingly reasonable and, even more specifically, asking why a certain area of a ridge is engraved or a certain part of a shelter painted and other areas and parts are not is, also, seemingly sensible. Ask why one boulder in that exact area on that specific ridge is engraved and an adjacent one not or why a small patch of rock in a shelter is painted while a contiguous section is not, and we get into the realm of idiosyncratic artistic choice and individual agency. Asking any questions at a scale smaller than this clearly becomes absurd because, ultimately, the relationship of image to location is arbitrary in the semiotic sense that there is no necessary bond between them. Their relationship with one another probably depended on multiple social, individual, and serendipitous factors that varied across space and through time. This is not, however, to say that the locales of rock art had no commonly shared significance to the communities who made images there. Moreover, we would not like to suggest that asking why some places have rock art and others do not is a meaningless question to ask in every rock art region of the world. There are examples, most notably from North America, where asking that question may yield positive insights (see, for example, Norder 2018).

It is our contention, then, that in a southern African context and in some other rock art places also, a more profitable question to ask is: “How does rock art construct landscape?” While a small shift in emphasis, changing the question here actually has profound consequences for studies of both painted and engraved places. Asking this question allows us to see that space is given significance through the act of placing images, not the other way around. Making engraved or painted images then is not so much about marking the landscape but rather about creating the landscape. We are not the first to suggest this type of approach to rock art and landscape. David Lewis-Williams, for example, has argued that the Upper Palaeolithic peoples of Europe saw a parallel between entering the deep caves in which rock art imagery occurs and the vortex experienced in deep altered states and that, as such, the caverns were seen as the entrails of the spirit realm (Lewis-Williams 2002, p. 209ff.). Various types of images and ways of producing those images contributed to the construction of the caves as the netherworld (Lewis-Williams 2010a, chap. 8). David Whitley (e.g., Whitley 1998, 2024; see also Whitley et al. 2004) has similarly argued for the construction of landscapes using rock art imagery in a North American context. Despite these efforts, and the influence of Lewis-Williams on southern African rock art research, such arguments, however, have been absent from the literature on that region. Shifting perspective from ‘marking’ to ‘creating’ landscape for this region, however, means that topography becomes socially and symbolically significant by placing pictures at a certain locale or, to put it another way, ‘space’ becomes ‘place’ by investing it with iconography. If we accept that locales are made important through the act of placing images there, and we leave the seemingly unanswerable question—in southern Africa at least—of why some places are engraved or painted and others are not, it allows for the exploration of how rock art images construct landscape. To conduct this exploration, we use the triadic framework of iconography, cosmology and topography mentioned earlier. As research on rock paintings throughout southern Africa—including those of the Matobo Hills and the Cape Fold Mountains—has a solid foundation in iconographic analysis, we begin our exploration with the images. We discuss the two motifs mentioned earlier that offer a way into a deeper analysis of how painted shelters construct landscape. One motif is extensively painted in the Matobos (but variations might also occur in other parts of Zimbabwe, Namibia and northern South Africa) while the other is—apparently—exclusive to a small part of the Cape Fold Mountains.

3. The Iconography of Formlings and Bird-Tailed Therianthropes

The Matobo Hills are a 2.6-billion-year-old series of granitic dome-shaped mountains of approximately 4500 km² in south-eastern Zimbabwe (Hubbard 2018, p. 76). The geomorphology has allowed for vegetation that supports high densities of animals. The vegetation, animal life, and the shelters in the Hills have attracted human occupation since the Early Stone Age. By the Later Stone Age, hunter–gatherers living here made rock paintings, and some 800 painted shelters are known (Walker 1996, p. 60). Archaeologists believe the paintings cover a period from 13,000 years ago to about 1500 years ago (see Walker 1996, pp. 11–14 for a discussion on the chronology of the Matobo Hills rock art). Painted spalls in the deposits, suggest that rock painting was at its peak between 10,000 and 8000 years ago (Garlake 1995; Walker 1996, pp. 11–14). The subject matter includes anthropomorphic images, plants and trees, insects, fish, and a diversity of mammals that includes elephant (Loxodonta africana), giraffe (Giraffa camelopardalis), kudu (Tragelaphus strepsiceros), duiker (Sylvicapra grimmia), impala (Aepyceros melampus), and tsessebe (Damaliscus lunatus) (Walker 1996, p. 81). The images are of a fine-line technique consistent with San rock art in other parts of southern Africa.

By contrast to the Matobo Hills, the Cape Fold Mountains are a belt of complex folded and faulted quartzitic sandstone layers stretching for 850 km on a north–south axis along South Africa’s Atlantic Coast and for 650 km along the country’s southern shoreline (Newton et al. 2006, p. 521). Although the sediments that form these mountains date to the Cambrian–Ordovician periods (510 to 330 million years ago), the mountains only began appearing 120 million years ago with continental uplift and the erosion of the overlying deposits. Today, the mountains are covered with fynbos, an extremely diverse and endemic group of plants that are part of the Cape Floral Kingdom. Archaeological evidence shows a human presence from at least the Early Stone Age (H. Deacon 1993) with significant occupation from the Middle Stone Age onwards (see, for example, J. Deacon 1984; Henshilwood et al. 2004; Wurz 2008) but it is likely that many rock art sites found in these mountains date from the second half of the Holocene (S. Hall 1994). Although there are no dates yet for the parietal art, archaeologists have found over 40 painted stones, many associated with burials (Rudner 1971; Pearce 2005). These stones, a unique feature of the southern Cape, date to between 6400 ± 75 BP (H. Deacon et al. 1976, p. 142) and 1925 ± 33 BP (Rudner 1971, pp. 54–55). As with the Matobo Hills, the images in the Cape Fold Mountains have been painted in a fine-line technique consistent with a San rock art tradition and include anthropomorphic images, fish, and a diversity of large mammals such as hartebeest (Alcelaphus buselaphus), elephant, and eland (Taurotragus oryx).

The rock paintings of the Matobo Hills and the Cape Fold Mountains are so-called classic San fine-line images, which through well-established research, are implicated in San religious belief and ritual practice (see, for example, Lewis-Williams 2019). While the two rock arts share commonalities, they also show differences (on regionality in southern African rock art see Laue 2014, 2021a, 2021b). One particularly distinctive image from the Matobo Hills, for example, is a depiction that Leo Frobenius called a ‘formling’ (Frobenius 1931; Mguni 2005, p. 35). Formlings describe paintings that are seemingly unidentifiable, but which share common features that suggest they depict the same thing. Typically, a formling is a series of “vertically or horizontally compartmentalized stacks of ovals, oblong or tubular cores” (Mguni 2005, p. 34). These series of cores are often (but not always) accompanied by crenelations, castellations, microdots, flecks, and orifices. Importantly, each formling is embellished in unique ways (Mguni 2015, p. 17). Their uniqueness, however, does not detract from recognising that they belong to the same category of image (Figure 2).

Scholars have variously identified formlings as depictions of waterfalls (R. Hall 1912), boulders (Frobenius 1929) and hilltops, skin clothing (Goodwin 1946, p. 17), villages/huts/rain clouds (Rudner and Rudner 1970), cornfields/quivers/mats/xylophones (Cooke 1969, p. 42), grain bins (Holm 1957, p. 9), pools of water (Breuil 1966, pp. 115–16), beehives, background decor (Mason 1958, p. 363), royal symbols (Frobenius 1930, p. 193), metaphorical maps and symbols of supernatural potency (Huffman 1983, pp. 50–51; Garlake 1987). The diversity of interpretation is a testament to the elusive nature of these images to modern viewers. Although their identity has long been subject to dispute, researchers have long agreed that they are a central and important motif in the rock art of the Matobo Hills (and other areas in Zimbabwe) and the accurate identification of them is key to further research on the imagery in this area.

At the turn of the millennium, however, a new exegesis emerged on formlings that, as we shall see, allows for an investigation of landscape in the Matobo Hills. Drawing on earlier suggestions that formlings represent insect nests (Pager 1976; Walker 1996), Siyakha Mguni (2006), has shown, through a careful consideration of their individual components and their contextual associations, that formlings depict termite nests. Mguni points to ten corresponding features in formling images and termite nests that are convincing. With so many matching features, one would think that formlings would have long since been identified as termitaria. As Mguni points out, however, it is difficult to determine the ‘correct’ orientation of formlings. Most rock art, globally, is depicted from a lateral view (Deregowski 1995). This is true of San rock art as well, although there are examples of views of animals from the front, back, and even above (Figure 3). In the Matobo Hills, most rock paintings conform to this principle, and it, therefore, seems a safe assumption that formlings also depict something from a side view. However, when viewing termite mounds from side-on in nature, they don’t look anything like a formling. It is only when one cuts away a termite nest and looks inside that one sees the formling-like structure. Painted formlings, then, are depictions of a “cross-section” of a termite mound.

As with the Matobo ‘formlings’, the Cape Fold Mountains also have distinctive, localized images. Images of human subjects with elongated arm-like appendages but with fork-like lower limbs in place of legs or no legs at all, for example, have long intrigued scholars (Alexander 1837; Hollmann 2005a; Lewis-Williams et al. 1993; Lewis-Williams 1977). These images are found in an area of about 350 km in length between the towns of Montagu and Joubertina (Hollmann 2005a). Recent work expands the geographical spread of these motifs 150 km farther east (Laue 2021a) so that they are now understood to cover an area of some 500 km (Figure 1). As with formlings, these forked-tailed images are unique to specific sites but share commonalities that make them readily recognizable. Variations in convention do occur—sometimes the figure’s bodies are decorated and at other times not; head shape varies, and the arm-like appendages hold equipment in some examples but in others, they appear more wing-like. The images are sometimes connected by lines, which have been interpreted as threads of light (Lewis-Williams et al. 2000; Hollmann 2005a, p. 26) and they are occasionally depicted exiting cracks or steps in the rock face (Hollmann 2005a, p. 30). As with formlings, the variations do not detract from them being readily identifiable as a distinctive category of image (Figure 4).

Similarly to ‘formlings’, the therianthropic motifs from the Cape Fold Belt have elicited many interpretations. Previous efforts, however, were limited to the images from the site of Ezeljagdspoort (see Figure 1). Scholars suggested that images from this single site represented fish-tailed people, which alluded to “the amphibious nature attributed to the whites by natives in the olden days” (Alexander 1837, p. 317); “watermaidens” (W. Bleek 1875, p. 20); “creatures, fish-like up to the waist and human above” possibly referring to the dugong (Willcox 1963, plate 11); illustrations of folk-tales (Leeuwenburg 1970); a group of rain-shamans trying to pacify the angry rain animal (Lewis-Williams 1977, p. 167); “an intrinsically powerful reification of a Bushman shaman-artist’s hallucination of swallow-shamans engaged in a rain-production ritual” (Lewis-Williams et al. 1993, p. 287).

Recent work by Jeremy Hollmann (2005a, 2005b), based on more examples, shows that these images reference swifts (Apidodae) rather than swallows (Hirundinidae); they have swift bodies, wings and tails but have human heads. Hollmann labels them as “swift people” and argues that we should understand them “as depicting a shifting spectrum of spirit beings, comprised of temporarily transformed living ritual practitioners, dead ritual practitioners and mythological beings” (Hollmann 2005a, p. 27). Unlike formlings, swift people are painted from the normal lateral perspective and are not a cross-section of anything. Their placement on the rock surface, however, is significant. As with formlings, swift-people are constructs that have a basis in something “real”—termite nests for Matobo formlings and swifts for the therianthropic images of the Cape Fold Mountains—but are re-imagined through the process of image-making into something “really made-up”1. Having a rudimentary but sufficient grasp of the iconography of the two areas, we now need to consider how these images relate to San cosmology before considering how they link to the rock shelters in the Matobos and the Cape Fold Mountains.

4. The Cosmology of Spirit and Human Worlds

The San ethnographic record is diverse, and while by no means complete, covers different peoples and spans more than 100 years (Barnard 1992). Sources taken from extinct San languages in South Africa (e.g., Hollmann 2004; Lewis-Williams 2015), as well as those from extant groups in Namibia and Botswana (e.g., Guenther 1999; Katz et al. 1997; Marshall 1999), show significant correspondence in religious belief and ritual practice across space and through time (Lewis-Williams and Biesele 1978; Lewis-Williams 1984). While there are differences and variations, most San groups believe in dual gods (a lesser and greater deity), spirits-of-the-dead, and another ‘world’ in which these beings exist and to which people go when they die (Blundell 2004, p. 90ff). San accessed this ‘spirit’ world primarily through a ritual that most extant groups still practice today, and which appears to extend back into deep time (Lewis-Williams 1984). At this ritual, variously called the medicine dance, trance dance, or great dance, the San harness a supernatural potency, known as n/om in the north and !gi in the south, that permeates the universe but is found in great quantities in certain animals, such as the eland (Lewis-Williams 1981a). Importantly, unlike the case with Native Americans, this supernatural power does not appear to reside in topographical features but rather in organisms and may be one of the reasons for differences in hunter–gatherer landscapes between southern Africa and North America. From an etic perspective, this harnessing of potency is an emic explanation of an altered state of consciousness induced through audio driving and hyper-ventilation during a dance. The San interpret the experiences of altered states as revelations of another religious reality (Lewis-Williams and Dowson 1988; Dowson 1988). This religious reality forms the context of the art.

Research on altered states of consciousness is traditionally a phenomenological one in which people, either in an ethnographic context or within a Western laboratory situation, describe their experiences. A cross-cultural survey of these accounts shows a remarkable similarity in the description of these experiences (Lewis-Williams and Dowson 1988; Winkelman 1989). At first, people see a limited range of geometric forms, which can appear projected onto a surface. These forms are a product of the optic nervous system and all Homo sapiens sapiens experience them. People begin to interpret these geometric forms in terms of their emotional state and their cultural predilections. After some time, people talk about passing through a tunnel into another world or place where they may see and experience the complex effects of altered states (Lewis-Williams 2002). The various stages of the process are not discrete but build one on the other (Lewis-Williams 2003). The literature on altered states provides a type of universal ethnography with which to extend the specific accounts of San ritual belief and practice. It is the great dance, the associated symbols of supernatural potency, the interactions between the living and the spirit world, and the embodied experiences of altered states of consciousness that are widely depicted in southern African San rock art. This includes the rock art of the Matobos and the Cape Fold Mountains. Any interpretation of either termitaria or swift people needs to begin with a consideration of how such images fit into San religious belief and practice. We begin with the formlings of the Matobos.

Early European travellers to southern Africa observed San eating termites and, in the Kalahari, termites are still eaten (Mguni 2006, p. 58ff). They were (and are still) valued by the San for their fat (D. Bleek 1928, pp. 16–17; Mguni 2015) as San believe fat is especially supernaturally potent. Depictions of termites in rock paintings then would almost certainly connote their link to supernatural energy. Termite nests, however, refer to a wider spectrum of associations than simply supernatural energy. As Mguni (2015) has demonstrated, termitaria make remarkably suitable conceptual vehicles for considering the interconnections between the world of the living and the spirit realm among the San. The natural properties of the nests, which were keenly observed by the San, made them particularly resonant symbols for San religious belief and practice. For one, termite nests, once abandoned, are re-inhabited by other creatures, most notably by snakes, including the Puff Adder (Bitis arietans) and Cape Cobra (Naja nivea) (Branch 2016). Both these serpents are intimately associated with rain animals (D. Bleek 1933; see also Mallen 2005). Rain is a supernatural animal amongst San groups and the animal is believed to live in the spirit world—something that Deacon pointed out in her work on the landscapes of /Xam engravings (J. Deacon 1988; see also J. Deacon and Foster 2005). This spirit world is variously described as being both in the sky but also underneath the ground and it can be accessed by San ritual specialists at various places, including waterholes, small features of rock surfaces such as cracks and steps and small holes in the ground (Lewis-Williams and Dowson 1990; Lewis-Williams 1997b).

Termites construct their nests both below and above ground, with the inner core of the nest typically below and the outer mound being above ground. Importantly, trees often grow from termite nests and in forested areas, the exterior of the nest can reach great heights when these trees are used for support. There are trees painted in the rock art of the Matobo Hills and within Zimbabwean rock art in general and they often occur with formlings. Elsewhere in southern Africa, where San rock paintings occur, images of trees are rare. The association of termite mounds with trees links the world below with this world and the world above; it is a useful columna cerului as demonstrated by the symbolic use of a tree to link worlds in religious cosmologies throughout the world (Eliade 1964). It is not only their association with trees that makes termitaria such poignant symbols in San religious cosmology but also the fact that termites emerge from the nests and fly into the air, particularly during the rainy season (Neoh and Lee 2009). Here again, we see a connection between below, above and this world. The symbolic associations of termitaria, then, led Mguni (2015, p. 147ff.) to argue that, in San thought, the inside of a termite nest was regarded as being God’s House (Mguni 2006, p. 64ff.).

Where Mguni interpreted formlings as illustrations of termitaria representing God’s House, Hollmann looks at the morphology of real birds and compares them to those of the rock paintings. This comparison shows that forked-tailed therianthropes are modelled on swifts. Many of the fork-tailed therianthropes appear to exhibit swift behaviour, such as wing-clapping (a behaviour where the bird’s wings meet above or below the body to create a clapping sound) or circusing, where the birds congregate on flight and ‘scream’ (Lack 1973). Swift behaviour made them suitable symbols for San religious ideas, where clapping is such an integral part of the medicine dance because it controls the availability of supernatural potency (Marshall 1969). There are five species of swifts known in the southern Cape region; all of them have underdeveloped legs and spend most of their lives on the wing. They cannot perch and never land on the ground willingly. They nest in rock crevices, rock overhangs, and abandoned swallow nests, and they approach these nests at high speeds and quickly vanish into the narrow entrance with their wings folded (see Henningsson et al. 2010).

Whereas painted termitaria symbolize God’s House, swift people appear to represent the sensation of flight, so commonly experienced in altered states of consciousness (Lewis-Williams 1985; Lewis-Williams and Dowson 1999). Formlings depict the spirit world as it lies underground, while swift-people connote flight and the idea that the spirit realm lies in the sky. Both termitaria and swift-people, then, connect this world to the one of spirits; both refer to ideas of the tiered cosmos and the space where gods, the dead, and other supernatural entities exist; the termitaria and fork-tailed therianthropes thus express the cosmological iconographically. However, it is not through cosmology alone that we can explain San landscapes. Instead, it is the articulation of cosmology through image that provides the key to unlocking how the topography is transformed into landscape.

5. Topography, Mimesis, Agency

In the third part of this paper, we consider how iconography and cosmology interdigitate with topography. Here, as mentioned earlier, we draw on the ideas of Taussig as well as Gell. Taussig reconsidered the role of sympathetic magic as part of what he calls the ‘mimetic faculty’. Considering James Frazer’s distinction between contagious and homeopathic/imitative forms of sympathetic magic, first espoused in the first edition of The Golden Bough (Frazer 1890) and then developed in subsequent editions, Taussig discusses the nature of magic. He points out that there is a “notion of the copy, in magical practice, affecting the original to such a degree that the representation shares in or acquires the properties of the represented” (Taussig 1993, pp. 47–48). Referring to the earlier work of Yrjö Hirn (1900), Taussig notices that in many cases the distinction between the two types of magic defined by Frazer does not hold and that

…in so many instances of sympathetic magic the copy, far from being a faithful copy, is an imperfect ideogram. What makes up for this lack of similitude, what makes it a “faithful” copy, indeed a magically powerful copy, he declares, are precisely the material connections—those established by attaching hair, nail cuttings, pieces of clothing, and so forth, to the likeness. Thus does the magic of Similarity become but an instance of the magic of Contact—and what I take to be fundamentally important is not just that a little bit of Contact makes up for lack of Similarity, or that some smattering of real substance makes up for a deficiency in the likeness of the visual image, but rather that all these examples of (magical) realism in which image and contact interpenetrate must have the effect of making us reconsider our very notion of what it is to be an image of some thing, most especially if we wish not only to express but to manipulate reality by means of its image.

(Taussig 1993, p. 57)

An example of the blurring between similarity and contact magic that provides a useful analogy for San rock art is that of Navajo sand painting. The Navajo strive for the correct relationship (hózhó) between humans, nature, and the spirit world. On the other hand, the incorrect bond between things (hóchxó) threatens this ideal. An imbalance between things can be caused by a range of factors, including improper relationships with other people, the environment and supernatural beings who inhabit the world (Samuels 1995). Through five clusters of ritual processes that have complex sub-rituals and variations within them, hózhó can be restored. These five ritual clusters are organized around “sings”, “chants”, or “chantways” (Lamphere 1969, p. 279). Chantways have male and female branches and, to make matters more complex, there is significant fluidity in all these rituals and chantways. Each chantway is built around a myth, and the entire entangled process of ritual, chantway, and myth is fundamentally connected to the Holy People—spirit beings who visit this world (Samuels 1995).

Although there are variations between the rituals and amongst different Navajo groups, there are broadly shared characteristics. They typically have a set duration of two, five, or nine days. They involve purification through the ingestion of purgatives and other means as well as fasting and, crucially for our purposes, they include altered states of consciousness and image-making. Indeed, Navajo ritual “constitutes a system of interrelated symbols which are repeatedly patterned in a particular manner” (Lamphere 1969, p. 283). Linking this pattern of symbols is the shaman (Foster 1963, p. 43) who typically meets with a patient and other members of the community in a hogan (a structure made of logs and earth) to perform various rituals over the course of the prescribed number of days. A key aspect of these rituals is sand painting. Sand painting is somewhat misleading as other materials—such as pollen, flower petals, and cornmeal—are used as well and, at times, exclusively so (Foster 1963, p. 43). Whatever the material, sand paintings are made on the floor of the hogan during rituals and there are rules that govern their construction. For example, the paintings are made after sunrise and must be destroyed before sunset (Cummings and Cummings 1936, p. 1). The image is typically thought of as an altar by scholars, rather than a painting.

Two independent accounts collected contemporaneously—one of a Yebitchai ritual by James Stevenson (1891) and the other by Washington Matthews (1888) of a dsilyídje qaçàl (Navajo Mountain Chant)—illustrate the use of sand painting as an altar. In both Stevenson’s and Mathews’ accounts, the sand paintings represent known deities or spirit-beings who are summoned by the ideograms and who then inhabit the icon. The patient sits on the images—identifying with the spirit-being—and the constituent elements of the image are applied to the corresponding body part of the patient as medicine. After the ceremony, the participating group destroys the image. Attendees, however, take the elements used to make the image, because they now believe that they have healing properties. For example, the material used to illustrate the arm of an ideogram is now applied to ailments that affect the arm of a person post-ritual. In the Navajo case, it is the occupation of the image by a cosmological entity that imparts efficacy to the constituent material as a healing salve for patients’ ailments. To use Alfred Gell’s terminology, the ideogram and its constituent materials have acquired secondary agency.

Gell (1998) argues that objects—not just human beings—have their own form of agency. He points out that, “Social agency can be exercised relative to ‘things’ and social agency can be exercised by ‘things’ (and also animals)” (Gell 1998, pp. 17–18). Gell thus makes a distinction:

…between ‘primary’ agents, that is, intentional beings who are categorically distinguished from ‘mere’ things or artefacts, and ‘secondary’ agents, which are artefacts, dolls, cars, works of art, etc. through which primary agents distribute their agency in the causal milieu, and thus render their agency effective.

(Gell 1998, p. 20)

In this view of the ‘distributed agent’, agency is disembodied from the primary agent through the use and creation of material items. For Gell (1998, p. 21), then, agents are “not just where their bodies were, but in many different places (and times) simultaneously.” This is possible because “objectification in artefact-form is how social agency manifests and realizes itself, via the proliferation of fragments of ‘primary’ intentional agents in their ‘secondary’ artefactual forms” (Gell 1998, p. 21). Material items, including rock art images, then, ‘are’ and are not merely ‘used’ by agents. Through interacting and using objects, ones not just made by us, we are constantly engaging with the agency of others through space and across time (on agency in San rock art see Dowson 1994, 1998, 2000; Lewis-Williams 1997a; Blundell 2004).

In the example of the Navajo sand painting that we discussed, it is not just the constituent elements of the altar that have a curative agency after being transformed during the ceremonies, but it is the actual icon that is used as an altar that also has a form of agency. It is simultaneously the vessel that will contain the supernatural ancestor and the vehicle that will merge the human patient with the constrained ancestor. In the case of Navajo sand painting, the agency of the icon-altar is destroyed when the image is erased at the end of the ceremony; the removed ingredients, however, retain their agency as they can still be used in curing. Navajo sand painting provides a useful analogy with which to understand San rock painting. In the example of Navajo sand painting, it is specifically the blurring of separate identities into one another—the bodies of the participants, the deities and the icon—that is a suitable comparison for looking at the relationship between topography, cosmology, and iconography amongst the San. While for Taussig, it is the general ‘mimetic faculty’ that humans possess that allows for the blurring of these elements into each other, it is really the role of altered states of consciousness in breaking down boundaries between things that are of interest when considering San rock painting.

A common feature of altered states is the experience of imagery. These images are not static but are animated and are generated within the nervous system (Siegel and Jarvik 1975; Siegel and West 1975; Siegel 1977); they are thus embodied. They are also commonly experienced as projections onto surfaces (Klüver 1966; Siegel 1977, p. 136). While independent of external stimuli, these images often incorporate aspects of the external world and extend beyond the purely visual. Nevertheless, in many cultures that have institutionalized altered states, the imagery experienced is spoken of as “pictures”. For the Putumayo of South America, for example, healing

…usually requires the healer and the patients taking together, at night, a “hallucinogenic” medicine known as yagé. When strong, this medicine brings forth mental pictures referred to as paintings or pintas, mainly but not exclusively visual, and these images can have curative functions.

(Taussig 1993, p. 60)

More important than seeing images or projecting them onto surfaces, is the breakdown of the boundary between the image and the person experiencing it. A physician, for example, who ingested the powerful hallucinogenic drug mescal described his experience of the breakdown between self and image to researchers:

The subject stated that he saw fretwork before his eyes, that his arms, hands and fingers turned into fretwork. There was no difference between the fretwork and himself, between inside and outside. All objects in the room and the walls changed into fretwork and thus became identical with him. While writing, the words turned into fretwork and there was, therefore, an identity of fretwork and handwriting. ‘The fretwork is I.’

The subject’s identification with the fretwork image eventually extended to engulf all aspects of his experience in the altered state: I am fretwork; I hear what I am seeing; I think what I am smelling; everything is fretwork … I am music, I am climbing in music; I am a touching fretwork; everything is the same.

(Klüver 1966, p. 22; originally in German to K. Beringer 1923)

In some cases, the breakdown between self and image in altered states is so extreme that people see themselves projected into the pictures of their visions. For Putumayo healers and patients, sometimes

…the senses cross over and translate into each other. You feel redness. You see music. Thus nonvisual imagery may evoke visual means…. You may also see your body as you feel yourself leaving it, and one can even see oneself seeing oneself—but above all this seeing is felt in a nonvisual way. You move into the interior of images, just as images move into you.

(Taussig 1993, pp. 57–58)

What makes this breakdown of self and image important for understanding the relationship between iconography, cosmology, and topography, is that these sorts of experiences are universally embodied—as is evident from the Navajo sand-painting rituals discussed—and that they occur commonly in altered states of consciousness and thus also occur amongst the San. This is apparent from ethnohistorical accounts.

In one well-known example, Richard Katz (1982, pp. 236–37) asked Kalahari San to draw themselves. Kinachau, an experienced healer, drew himself as a series of spirals and/or zigzags (Katz 1982, p. 237, Figure e). Lewis-Williams and Dowson (1988, p. 211) have suggested that these self-portraits depicted the San ritual specialists as they saw themselves in altered states like the fretwork example discussed. Perhaps the most insightful example of the breakdown between image and self, however, came from Maqoqa Dyantyi (referred to as ‘M’ in some of the literature), a San descendant who was interviewed on multiple occasions in the 1980s (Jolly 1986; Lewis-Williams 1986; Prins 1990; Jolly and Prins 1994). She was the daughter of Lindiso, the last known San rock painter in the Drakensberg (Blundell 2004, p. 40ff). Maqoqa (Figure 5) described how San danced ecstatically before the rock paintings and touched them, thereby harnessing the supernatural potency in the images (Jolly 1986). Mapote, a Baphuti man who lived and painted with San people in the nineteenth century, told Marion How, the wife of a colonial official in 1930, that the art was made with eland blood, corroborating Maqoqa’s description of the paint as having supernatural potency (How 1962). Eland blood, as with the animal’s fat and sweat, was regarded as particularly redolent with supernatural potency (Lewis-Williams 1981a). The most depicted identifiable species of animal in the Drakensberg is the eland (Vinnicombe 1972, 1975). There are large numbers of images of antelope where the species are indeterminate to modern viewers and, following Taussig, if these were to be found to have eland blood in them, they might very well be meant to represent eland through the magic of contact. As he points out, the “wonder of mimesis lies in the copy drawing on the character and power of the original, to the point whereby the representation may even assume that character and that power” (Taussig 1993, p. xiii). This would be true not only for images of eland but for representations of termite nests and swift people. The dancing and touching of images such as the eland that was full of supernatural potency suggest that, under certain circumstances, the San rock paintings were like Navajo sand paintings. San rock paintings “were not simply depictions of other things—animals, people, visions and so forth; rather, they were things in themselves; they had a life and existence of their own” (Lewis-Williams and Dowson 1990, pp. 14–15).

Importantly, what these examples show, be they from the Putumayo, the Navajo, westerners, or the San, is that altered states of consciousness are extreme conduits of the imagination. Real and non-real, the actual and the imagined are combined in novel, and sometimes unexpected, ways. It is this extreme imagination that is often captured in the imagery painted and engraved at various topographical places. Moreover, the testimonies of Maqoqa and Mapote point to the San entering altered states in the shelters before the paintings. Some San who danced in front of the paintings would have incorporated them into their experiences. The ability of extra-somatic static imagery to appear animated in altered states of consciousness would have facilitated this perception. It is thus likely that, under certain circumstances, the painted imagery became ‘inhabited’ by spirit world entities such as the dead, rain-animals, and other beings and creatures from the supernatural realm in the sense that the images were the actual things and not merely depictions.

Moreover, in altered states, some San would have likely experienced their visions projected onto the already-painted surfaces and, crucially, they would have experienced a breakdown of categories, a collapse of the distinction between self and image and they would have believed themselves to be one with the images and, so, the spirit world (see also Blundell 1998). The degree of this breakdown, no doubt, varied from person to person. While some people might have become completely immersed, others may have seen a projection of themselves ‘floating’, ‘swimming’, or ‘flying’ amongst the images (Lewis-Williams 1981b). Others may have not experienced these sensations at all. It is also potentially the case that for some San people experiencing hallucinations, the natural features, like any pre-existing imagery there, have been incorporated into their ecstatic experiences. Such inclusions have been experienced by people taking mescal, who report the sudden jerky movement of a cigarette packet or the back-and-forth movement of the ceilings and walls (Klüver 1966, p. 41). Under experimental conditions, a cigarette moved into the visual field of someone experiencing a hallucination may be interpreted as a source of light in their vision (Klüver 1966, p. 28) and, generally, “the behaviour of visionary objects and of real objects during the mescal state offers many striking parallels. Thus we have not only diplopia and polyopia of visionary, but also of real objects” (Klüver 1966, p. 44). The universality of these experiences in hallucinations suggests that some San at least incorporated images, objects, people, walls, ceilings, and other real features at rock art sites into their altered states. In this way, San rock painting was like Navajo sand painting because, like the Navajo altar icon, the rock art was the vessel that contained supernatural entities as well as the vehicle for merging humans with that supernatural world.

Unlike Navajo sand painting, however, San rock painting was not destroyed after the ritual process of creation. Instead, it appears that the artists intended the art to be long-lived. Many rock art sites, for example, have layers of superimposed images, where artists placed motifs partially over older pictures, in a manner that comments, elaborates or enhances the earlier imagery but does not erase them (Lewis-Williams 1972, 1974). In other cases, in the far northern reaches of the Cape Fold Belt, artists appear to have ‘renewed’ images by re-painting them (Yates and Manhire 1991). While there is no evidence for the ‘renewal’ of swift-people motifs or formlings specifically, what matters here is that San artists revisited rock art sites in different parts of southern Africa and added to, enhanced, or modified images already there. Moreover, recent dating work in the south-eastern mountains shows that many sites were painted over hundreds and sometimes thousands of years (Bonneau et al. 2017a, 2017b). The accompanying research on the constituent elements of pigments shows, in some cases at least, continuity in paint recipes, suggesting longevity of embedded cultural knowledge of art production at sites (Pearce, pers. comm.; see also Bonneau et al. 2014, 2022). The testimonies of Mapote and Maqoqa, the superpositioning, the longevity of usage and paint recipes at some sites all suggest that in many if not all cases, San artists painted with the intention of preparing the locales of their art for future ritual events. Artists thus produced images with the knowledge that future rituals including dancing—or some other method for inducing an altered state—would take place before them and that the iconography would become part of forthcoming experiences; previous generations of San artists, then, distributed their agency through time and across space when they created paintings. We do not here suggest that the images at sites were planned out or that sites were specifically chosen to be prepared (although we cannot preclude this in all cases). This does not fit well with current San practices in the Kalahari, where rituals such as the dance seem to happen organically and fluidly (Marshall 1969). It was likely the same for rock art images and rock art sites. A San rock artist made an image at a site and then others added images and over time the site became a densely painted or engraved locale or, for whatever reason, it did not. In some cases too, it is likely that images were made for future personal consumption rather than for the public. Nevertheless, given the evidence of Maqoqa, Mapote, and the treatment of paintings at the sites that we have mentioned, it seems that images were meant to be viewed in post-painting ritual activities and not merely as passive observers.

This intentionality in the creation of images for future rituals encompassed more than just the actual iconography: it included the integration of natural features of the rock surface into the composition of images. The use of cracks and steps in the rock surface to indicate creatures, such as antelope-headed serpents, emerging from or entering the rock surface has been described since the 1990s in San rock art (Lewis-Williams and Dowson 1990). Today, researchers accept that the rock surface was a veil between this world and the spirit one. Subsequently, such aspects were observed in rock art traditions in other parts of the world, such as in the Upper Palaeolithic caves in Europe (Clottes and Lewis-Williams 1998), or the more recent engravings in Alta, Norway (Helskog 2010). In the Upper Palaeolithic cave of Niaux, for example, a natural hollow in the rock surface has been turned into an animal head by adding antlers or horns at the appropriate places (Clottes and Lewis-Williams 1996; Lewis-Williams 2002, pp. 212–13). Knut Helskog (2012), working on the Alta rock engravings, notes that the rock surface seems to have been part of the compositional symbolism of iconography. Natural depressions, for example, some of which collect water, are incorporated into the compositions of engraved two-dimensional imagery to illustrate animals drinking water or to represent caves where bears might enter and exit. Helskog points out that for many years researchers missed these inclusions of natural features because of an over-emphasis on iconography or, as he puts it, “the tyranny of the image”. His point has pertinence to the study of southern African San rock paintings, where the detailed images and the explanatory power of ethnographic-based interpretation led researchers to focus on the iconography; researchers only note natural features when they interact with the imagery explicitly—such as when something appears to enter or leave a crack in the rock surface. As we have mentioned, there are few, if any, studies of San rock art that consider the relationship between image and terrain even when landscape is the topic. Yet, it is not a far leap from using the rock surface in the construction of the image to the use of image to shape the natural topography of the rock surface into a representational form. Examples of such behaviour have been observed in the Upper Palaeolithic. In Pergousset, for example, a protruding rock with a tapering end is transformed into a fish emerging from the wall with the simple addition of gills and an eye (one of us, Blundell, visited this cave in 1995).

This type of modification, where the markings or images give form to natural features in the rock, as far as we are aware, is not yet described for San rock paintings. In the case of the Matobo Hills and the Cape Fold Belt, however, we argue that artists shaped the rock itself through the act of painting. We begin with the Matobo Hills before turning our attention to the Cape Folded Belt. Over time, the granite of the Matobo Hills has eroded to form distinctive “whaleback” hills and other striking shapes (Figure 6). Within these forms, erosion has created shelters over the millennia. The eroded shelters are often—but not always—dome-shaped. Mostly, such domes are on the side of a hill and are half a dome, so to speak; however, in some rare cases, you can walk underneath a hill and emerge in a fully domed shelter within. In such cases, the shelter wall provides a 360-degree area for images rather than the normal 180-degree wall of most semi-domed shelters on the sides of mountains (Figure 7).

While rock paintings occur on numerous surfaces in the Matobos, including small shelters and overhangs, the largest and most densely painted sites are in dome- and semi-dome-shaped shelters. The fine-lined paintings in these shelters, typically, are not so large as to be visible from afar. Moreover, the details on the images are small, even if the overall form is large, thus requiring close-up viewing of motifs. The densely superimposed panels compound the need for close-up viewing to discern one shape from another; the farther back one stands, the more the images tend to blur into an undifferentiated mass. Seeing shapes and details in the images of the Matobo Hills, then, required up-close viewing just as Maqoqa Dyantyi described for the Drakensberg. This is the opposite of what is required in a search for vista, view, and the visual acquisition of nearby sites. The detail suggests that viewers of the art in the Matobos would have to be inside the semi-dome and domed shelters to see the images. Viewers would ‘enter’ the mountain because many of the dome shelters give the impression that one is partly in the mountain when viewing the images. Importantly, these subterranean shelters often include depictions of formlings (Figure 8).

Formlings would have possessed secondary agency; once made they exerted influence over viewers and particularly so over those versed in the cosmological ideas expressed by the iconography. Importantly, the dome-shaped shelters of the Matobo Hills mimic the shape of actual termite mounds. Within these domed and semi-domed shelters are painted “cross-sections” of termite mounds. The reason for painting a cross-section of a termite mound rather than a view of a termite mound from a lateral perspective, then, is because when making or viewing a formling image, one was already inside a metaphoric termite mound and not outside viewing it from a lateral perspective. Through the application of image, the artists of the Matobo Hills invested terrain with meaning by iconographically altering it to mimic the form of a termite mound. They constructed landscape by applying termitaria imagery to topography and, in so doing, they turned the rock shelters of the Matobos into termite mounds or God’s House, much as Medieval Christian cathedrals were turned into God’s House through the manipulation of architecture and art (Spurrell 2020). San rock artists, then, moulded the topography of certain granite domes in the Matobos—already visually like real termite mounds—into termitaria on a grand scale. Through the secondary agency of the formling images, the shelters at which they occur became metaphorical termite mounds for those viewers cognisant with the conventions of the art and allowed for future ritual practices in the shelters that facilitated the experience of being ‘within a termite mound’ and thus within God’s House.

By comparison with the granite Matobo Hills, the Cape Fold Belt (Figure 9) is characterised by a series of folds (Newton et al. 2006, p. 521) that appear as “a succession of creamy white to ash grey quartzite” ridges (Rust 1988, p. 10). Within these ridges, there are shallow rock shelters and occasional deeper structures that extend into the mountainside but very few can be called a “cave”. The rock art sites in the Cape Fold Belt mountains are found in these shallow shelters and overhangs that were formed during the orogenic folding that gave rise to the mountains (Truswell 1970, p. 107; Newton et al. 2006, p. 521). None of the known shelters have the semi-spherical quality or the relatively smooth surfaces of the Matobos; instead, the quartzite of the Cape Fold Mountains has eroded in a manner that produces gnarled, craggy, undulating, and endlessly variable surfaces (Figure 10). Frequently, San artists positioned images to take advantage of the complex surfaces to create effects such as motifs emerging from or entering cracks or hollows in the rock surface (Hollmann 2005a; Laue 2017). The paintings on these surfaces, as with those of the Matobos, are fine-lined and painted in significant detail. As with the Matobos, viewers need to be close-up to the rock surface to see the images. The terrain further constrains the viewing distance; many shelters are on steep slopes where the hill falls sharply away, preventing observation of the images from afar or from outside the shelter; this is often compounded by dense vegetation. In other cases, artists appear to have placed images in areas where the geology forces viewers into a close encounter with imagery. At one shelter, for example, images are placed on the ceiling in an area where it is a little over a metre high and the viewer sits on the floor to see them (Laue 2017). Whereas the 180-degree topology of the Matobos rock shelters envelops the viewer, here, the limited viewing space and the small, detailed images scattered over the variable surfaces lure viewers closer to the shelters. Together, the iconography and topography act as an enveloping mechanism that draws the viewer up close to the rock surface.

As with formlings in the Matobos, swift-tailed therianthropes had secondary agency. Artists constructed and constrained the viewer’s experience of the images by carefully placing them in relation to the topography of the rock surface. For example, by painting the motifs emerging from cracks and steps in the rock surface, San artists ensured that future viewers understood that the images were from the spirit world behind the rock (Figure 11). In some cases, the painters in the Cape Fold Belt went beyond just using the micro features of the rock surface, and like the artists in the Matobos, used images to shape the rock into landscape. One image of a swift-therianthrope, for example, is painted with its tail abutting a step in the rock but instead of heading out towards the viewer, it faces inwards towards a tunnel at the back of the overhang in which it is painted (Figure 12), creating the impression that the painted bird is flying into the shelter. At another site, two swift-tailed therianthropes are painted on a low ceiling in a shelter that narrows into a deep cleft. The orientation of the images makes them appear to be flying out of the shelter. Indeed, in several places where swift-therianthropes are painted, the shelters have a tunnel-like form that resembles a swallow’s nest (Hollmann 2005a, p. 31) (Figure 13).

Some species of swift use old retort-shaped nests of other species such as swallows and martins (Lack 1956, p. 16) and Hollmann (2005a, p. 32) has recorded paintings of swallow nests associated with swift-tailed therianthropes. Moreover, as with actual termite nests located near or even in the rock shelters of the Matobos, swallow nests are commonly found in the rock shelters of the Cape Fold Belt. It would thus seem that, in some cases, the shelters of the Cape Fold Belt were mimetically altered into bird nests through the application of certain images. A shelter in the Groot Winterhoek Mountains at the eastern edge of the Cape Fold Belt aptly demonstrates this suggestion. Here, in a deep and rounded shelter with a long tunnel that extends into the hillside, artists painted 21 images of flying birds that are identifiable as swifts. By adding swift motifs, San artists created the impression that the shelter is a swallow’s nest; the horizontal cross-section of the cave mirrors a cross-section of a swallow’s nest. The swift images are painted as if they are in the bowl of a swallow’s nest (Figure 13). When swifts do not use the abandoned nests of swallows, they nest in the various nooks and crannies of the rock itself, often flying into shelters at high speed and then landing on the rock surface where they nest and then flying out again when they need to feed. There was thus a clear similarity between “real” swifts flying in and out of shelters and the “really made-up” swift people, painted as if flying in and out of shelters. In the Cape Fold Belt then, through the secondary agency of the swift imagery placed in them, shelters were sometimes fashioned into metaphorical swallow’s nests and participants in subsequent rituals which took place there would have an enhanced sense of being ‘within’ such a nest (Figure 14).

6. Conclusions

We have argued in this paper that many attempts to understand the relationship of San rock art to landscape are hampered by logical flaws. These flaws include circular argumentation and the use of cognitive templates that treat topography as having prior significance. In these views, painted or engraved imagery is seen as simply rendering a terrain pre-populated with meaning as visible; rock art is just a form of ‘marking’ the landscape. We have suggested, instead, that a better way to look at how rock art is related to landscape is to consider the relationship between topography, cosmology, and iconography. Applying this tri-nodal approach, we have argued that San artists manipulated natural terrain features through the act of placing specific motifs that resonated with their beliefs. Specifically, motifs of cross-sections of termite nests in the Matobo Hills shaped some dome-shaped shelters into metaphorical termite mounds and paintings of swift-tailed therianthropes in the Cape Fold Belt fashioned some shelters to appear more like bird nests. The natural topographic features of the shelters in these two areas mimic the form of real termite mounds and real bird nests, respectively. The alteration of topography through iconography in both these areas can be understood through ethnographic exegesis of widespread San religious belief and ritual praxis. For the Matobo artists, the termitaria imagery facilitated ideas of being underground, in the supernatural world while for the Cape Fold Belt artists, it was bird nests that played this role. Images of flying termites and those of flying swifts or swift people would have referred to sensations of flight commonly occurring in altered states of consciousness and referenced in San cosmology as a way of accessing the spirit world. Paradoxically, the San spirit world was simultaneously below the ground and in the sky above in a tiered universe (Lewis-Williams and Pearce 2004).

We believe that there are other examples of imagery used to construct landscape to be found in southern African San rock art as well as in the rock art of hunter–gatherers from other parts of the world. We should not expect that these examples will be the same for every hunter–gatherer group across space or for the same cultural group through time. There are key differences in hunter–gatherer religious beliefs and practices throughout the world. As we pointed out earlier in the paper, for example, Native Americans across both North and South America see supernatural power as residing in topographical features while San see it as primarily residing in biological entities. Moreover, whereas many (but not all) San rock art sites in the Matobo Hills and Cape Fold Belt seem to have been made with the intention of preparing the space for future rituals, many Native American rock art sites seem to have been the result of vision questing, where an image was left behind seemingly without any intention to shape the space for future rituals; that said, rock art sites in North America were used for non-shamanic curing activities afterwards (Whitley 1998, p. 18). Perhaps, what is most important about the argument offered here and also by other researchers not working in rock art contexts (e.g., Whallon & Lovis 2016; Maher 2021) is that—if correct—it suggests that hunter–gatherers were not passively ‘in’ the land, reacting to it, and nor were they simply ‘marking’ significant places with rock art but, instead, they were actively creating landscape. Typically, it is believed that it was only later that agricultural people transformed space meaningfully into landscape through built structures. The sites in the Matobo Hills, in the Cape Fold Belt and, indeed, in other parts of the world suggest that people were shaping their environment into landscape through image-making on a larger scale and from much earlier on than is generally thought to be the case.