Stibbard-Hawkes has shown that despite regular acknowledgement of gaps and bias in the archaeological record, theorists of the human past, me included, have too readily reasoned from absent evidence. In particular, we have accepted that the lack of evidence for symbolic technology in the Late Pleistocene (excepting the ambiguous case of ochre) is evidence that pre-Late Pleistocene hominins did not use such technologies; in some cases even inferring that they lacked the cognitive capacities to use material symbols. Trace erosion can always cause false negatives, but Stibbard-Hawkes’ ethnographic data (admittedly, only African data) shows that symbolic technology is particularly likely to erode; particularly subject to the risk of false negatives.
The paper thus poses an important methodological challenge: In archaeological contexts, when can we infer from lack of evidence for X to the lack of X? The answer surely cannot be: Never. We have very good reason to believe, for example, that Late Pleistocene hominins did not live in towns and villages (or work metal); a secure conclusion based on lack of evidence for settled lives (or metal working). But what makes this inference sound? I will explore this through a less extreme example. Consider Acheulean tool using communities between about 1 mya and 800 kya. This region of space and time saw the emergence both of more complex and symmetrical Acheulean tools (Kuhn, Reference Kuhn2020) and of very large-brained hominins, with Heidelbergensis-grade hominins encephalisation overlapping the range of modern hominins (Klein, Reference Klein2009). But despite the emergence of enhanced knapping skills and the first direct evidence of domesticated fire (Alperson-Afil, Richter, & Goren-Inbar, Reference Alperson-Afil, Richter and Goren-Inbar2007), there is no evidence of hafted tools. I suggest that this is good evidence that, minimally, hafted tools were not regularly part of these hominins’ technical repertoire and, more strongly, that they lacked the capacity to haft tools.
The following considerations support this analysis: (i) From 1 mya on, sites preserving a lithic signature are found over a wide range of regions and habitats: We are not sampling just a small fraction of hominin subsistence economies from this period; (ii) while hafted tools (in general) require more time, materials, and skill to make, theoretical considerations suggest that this would be profitable over a broad range of environmental conditions. Hafted tools bring mechanical advantage and safety (by placing distance between the working component of the tool and the operator). I know of no formal behavioural ecology models exploring these economics, but I predict that such models would show a clear hafting profit. (iii) Theoretical considerations are supported by ethnographic data: Hafted tools are used ubiquitously, even by foragers under selection for simple, lightweight generalist tools, like those of the Australian Western Desert (Gould, Reference Gould1980). (iv) While many hafted tools are made entirely of perishable materials (some bow/arrow designs), a wide range of hafted tools have resistant, identifiable components (spears, javelins, arrows, knives, denticulate tools).
The moral: We can trust absence of evidence as (defeasible) evidence of absence when (a) an archaeological landscape has been reasonably well sampled; (b) we have independent evidence that if a capacity was available to agents in that landscape, they would have had incentives to exercise that capacity: The inference is more robust if the incentive is strong, and not tied to a specific constellation of factors within that landscape; (c) if the capacity were exercised with some regularity, at least some traces would be archaeologically visible.
Can we export considerations of this kind to material symbols? Not so well, as function does not constrain material substrate to the same degree as it does with subsistence technologies. There are more degrees of freedom, but perhaps to some degree. The most persuasive accounts of the function of material symbols see them as signals of social identity and role, becoming important in denser social contexts (see, e.g., Kuhn, Reference Kuhn2014; Kuhn & Stiner, Reference Kuhn, Stiner, Mellars, Boyle, Bar-Yosef and Stringer2007; McDonald & Harper, Reference McDonald and Harper2016). In such contexts, individual interactions become mediated by signals of identity and role, rather than solely by direct personal knowledge, and social aggregates signal rights of place and cohesion to one another.
Given this, the most plausible absence of evidence as evidence of absence is perhaps coastal South Africa between about 150 kya and 100 kya. This archaeological landscape has been reasonably well sampled and has been argued to be a refuge area (in part because of coastal resources) in Late Pleistocene aridity pulses (Marean, Reference Marean, Bicho, Haws and Daviis2011; Marean, Reference Marean2016). If so, it may well have had threshold densities, perhaps not in residential groups, but of densely networked groups, with their potentials for conflict and cooperation. Geographically widespread ethnographic report suggests that shells are often used as a signalling medium; and in identifiable ways, even at some cost. They were indeed so used in Southern Africa somewhat later. We could not reasonably conclude that the peoples represented by Pinnacle Point material culture did not use material symbols. But perhaps we can conclude that they did not use material symbols for the purposes to which shell-based signals are well adapted (Kuhn & Stiner, Reference Kuhn, Stiner, Mellars, Boyle, Bar-Yosef and Stringer2007). Shells are durable (much more so than, say, seed pods) and come in varieties, such that the instances of a specific variety are very similar to each other, while contrasting with others. This makes it possible to use them in complexly patterned, precise, low amplitude signals within extended social networks in which agents shared common cultural norms, but where it is not true that everyone knows everyone well. We might infer that the social world of South Africa before about 100 kya was simpler and more fragmented than it became, below a threshold of signal-mediated social interaction. This inference is less secure than that to an Acheulean without handles, given the looser relationship between function and form; that evidence for social density is difficult to read, and that we do not know the density thresholds above which within-group signals become important. But it has some traction.
Stibbard-Hawkes has shown that despite regular acknowledgement of gaps and bias in the archaeological record, theorists of the human past, me included, have too readily reasoned from absent evidence. In particular, we have accepted that the lack of evidence for symbolic technology in the Late Pleistocene (excepting the ambiguous case of ochre) is evidence that pre-Late Pleistocene hominins did not use such technologies; in some cases even inferring that they lacked the cognitive capacities to use material symbols. Trace erosion can always cause false negatives, but Stibbard-Hawkes’ ethnographic data (admittedly, only African data) shows that symbolic technology is particularly likely to erode; particularly subject to the risk of false negatives.
The paper thus poses an important methodological challenge: In archaeological contexts, when can we infer from lack of evidence for X to the lack of X? The answer surely cannot be: Never. We have very good reason to believe, for example, that Late Pleistocene hominins did not live in towns and villages (or work metal); a secure conclusion based on lack of evidence for settled lives (or metal working). But what makes this inference sound? I will explore this through a less extreme example. Consider Acheulean tool using communities between about 1 mya and 800 kya. This region of space and time saw the emergence both of more complex and symmetrical Acheulean tools (Kuhn, Reference Kuhn2020) and of very large-brained hominins, with Heidelbergensis-grade hominins encephalisation overlapping the range of modern hominins (Klein, Reference Klein2009). But despite the emergence of enhanced knapping skills and the first direct evidence of domesticated fire (Alperson-Afil, Richter, & Goren-Inbar, Reference Alperson-Afil, Richter and Goren-Inbar2007), there is no evidence of hafted tools. I suggest that this is good evidence that, minimally, hafted tools were not regularly part of these hominins’ technical repertoire and, more strongly, that they lacked the capacity to haft tools.
The following considerations support this analysis: (i) From 1 mya on, sites preserving a lithic signature are found over a wide range of regions and habitats: We are not sampling just a small fraction of hominin subsistence economies from this period; (ii) while hafted tools (in general) require more time, materials, and skill to make, theoretical considerations suggest that this would be profitable over a broad range of environmental conditions. Hafted tools bring mechanical advantage and safety (by placing distance between the working component of the tool and the operator). I know of no formal behavioural ecology models exploring these economics, but I predict that such models would show a clear hafting profit. (iii) Theoretical considerations are supported by ethnographic data: Hafted tools are used ubiquitously, even by foragers under selection for simple, lightweight generalist tools, like those of the Australian Western Desert (Gould, Reference Gould1980). (iv) While many hafted tools are made entirely of perishable materials (some bow/arrow designs), a wide range of hafted tools have resistant, identifiable components (spears, javelins, arrows, knives, denticulate tools).
The moral: We can trust absence of evidence as (defeasible) evidence of absence when (a) an archaeological landscape has been reasonably well sampled; (b) we have independent evidence that if a capacity was available to agents in that landscape, they would have had incentives to exercise that capacity: The inference is more robust if the incentive is strong, and not tied to a specific constellation of factors within that landscape; (c) if the capacity were exercised with some regularity, at least some traces would be archaeologically visible.
Can we export considerations of this kind to material symbols? Not so well, as function does not constrain material substrate to the same degree as it does with subsistence technologies. There are more degrees of freedom, but perhaps to some degree. The most persuasive accounts of the function of material symbols see them as signals of social identity and role, becoming important in denser social contexts (see, e.g., Kuhn, Reference Kuhn2014; Kuhn & Stiner, Reference Kuhn, Stiner, Mellars, Boyle, Bar-Yosef and Stringer2007; McDonald & Harper, Reference McDonald and Harper2016). In such contexts, individual interactions become mediated by signals of identity and role, rather than solely by direct personal knowledge, and social aggregates signal rights of place and cohesion to one another.
Given this, the most plausible absence of evidence as evidence of absence is perhaps coastal South Africa between about 150 kya and 100 kya. This archaeological landscape has been reasonably well sampled and has been argued to be a refuge area (in part because of coastal resources) in Late Pleistocene aridity pulses (Marean, Reference Marean, Bicho, Haws and Daviis2011; Marean, Reference Marean2016). If so, it may well have had threshold densities, perhaps not in residential groups, but of densely networked groups, with their potentials for conflict and cooperation. Geographically widespread ethnographic report suggests that shells are often used as a signalling medium; and in identifiable ways, even at some cost. They were indeed so used in Southern Africa somewhat later. We could not reasonably conclude that the peoples represented by Pinnacle Point material culture did not use material symbols. But perhaps we can conclude that they did not use material symbols for the purposes to which shell-based signals are well adapted (Kuhn & Stiner, Reference Kuhn, Stiner, Mellars, Boyle, Bar-Yosef and Stringer2007). Shells are durable (much more so than, say, seed pods) and come in varieties, such that the instances of a specific variety are very similar to each other, while contrasting with others. This makes it possible to use them in complexly patterned, precise, low amplitude signals within extended social networks in which agents shared common cultural norms, but where it is not true that everyone knows everyone well. We might infer that the social world of South Africa before about 100 kya was simpler and more fragmented than it became, below a threshold of signal-mediated social interaction. This inference is less secure than that to an Acheulean without handles, given the looser relationship between function and form; that evidence for social density is difficult to read, and that we do not know the density thresholds above which within-group signals become important. But it has some traction.
Financial support
This research was supported by grant DP210102513, “Beyond Scenarios: Testable Models of the Evolution of Norms.”
Competing interest
None.