One of the most consistent and well-established empirical regularities in the study of contemporary hunter-gatherers (foragers) is the existence of a flexible sexual/gendered division of labor, within which men and women make different but complementary subsistence contributions. Men tend to spend much of their time hunting medium-to-large-sized game, but often return empty-handed, whereas women spend much of their time caring for young children, gathering plant foods, and sometimes hunting small game, typically more reliable food sources (R. B. Bliege Bird and Codding, 2015; Kelly, 2013; Marlowe, 2005, Marlowe, 2007; Murdock and Provost, 1973). Despite these complementary divisions of labor, which vary as a function of ecological factor (Marlowe, 2007), gender roles are not rigidly deterministic and vary with cultural context (Brightman, 1996; Noss and Hewlett, 2001). For instance, women make capable hunters (Goodman et al., 1985; Noss and Hewlett, 2001), with reports of female participation in hunting going back centuries (Hoffman, Farquharson, & Venkataraman, 2024), and men regularly gather and are involved in childcare (K. Endicott, 1974; Kramer, 2018; Marlowe, 2007; Turnbull, 1965; Wood and Marlowe, 2013). Among most foraging populations, however, women hunt rarely, and sometimes not at all. One estimate placed the frequency of women's hunting across foraging societies at roughly 7% (Gurven and Hill, 2009).

A recent study has questioned any such gendered division of labor in ethnographically known foragers (Anderson et al., 2023). We follow Anderson et al. (2023) in referring to sex as biologically denoted traits whereas gender refers to culturally denoted traits reflecting the intersection of social norms and personal expression. Based on a survey of ethnographies that contained explicit descriptions of hunting, Anderson et al. (2023) reported that women's hunting occurs in 79% of foraging societies and that this involves large game in 33% of foraging societies. With these findings, they highlight the “…significant role females have in hunting, thus dramatically shifting stereotypes of labor…”, and go on to question the existence of any gendered divisions of labor. Other forms of labor common in foraging societies (e.g. gathering, food processing, childcare) were not analyzed by gender to bolster this claim. These claims have been widely reported in the press, e.g.(Aizenman, 2023; Alex, 2023; Criado, 2023; Miller, 2023), and are cited in several scientific articles (Atanasiu and Fornaro, 2023; Bebber et al., 2023; Friant, 2023; Hora et al., 2023; Lacy and Ocobock, 2023; Miszaniec et al., 2023; Muñoz-Reyes et al., 2023; Smallwood et al., 2023).

Scientific paradigms must always be challenged. We agree that, historically, hunting and men's labor have been over-emphasized in research among forager populations. For instance, see (Bliege Bird and Codding, 2021) for a discussion of gender bias in historical datasets such as the Ethnographic Atlas. We applaud Anderson et al. (2023) for their willingness to challenge an orthodoxy that is often misused to justify misogyny and limit the opportunities of women (Bachaud and Johns, 2023; Vallerga and Zurbriggen, 2022). Nevertheless, their claims about contemporary foraging societies are contradicted by the large existing literature on female hunting and the gendered division of labor. Misrepresentation of the ethnographic record devalues the many essential contributions of forager women (and men) beyond hunting and dispenses with a century of hard-won empirical research. In this piece, we critique the claims of Anderson et al. (2023), showing that multiple methodological failures all bias their results in the same direction.

Anderson et al. (2023) coded their data at the society level. They provide no evidence that they conducted a paragraph-level analysis, which has been standard in cross-cultural research for some years (Ember, 2007). It was not possible to ascertain where the relevant text was located that was used to determine women's hunting. As a result, their analysis is not replicable. Further, their coding scheme does not account for the frequency of women's hunting in a society in terms of effort allocated or the amount of prey acquired. Women's hunting was coded as a binary variable and recorded as ‘present’ whether the case was a single report or habitual involvement. There was no minimum threshold for inclusion and, e.g., societies where hunting constituted only 1.2% of women's returns (Marlowe., 2010) were coded as having women's hunting. Beyond brief commentary on ‘opportunism’ in hunting, the results of Anderson et al. (2023) give little indication about the importance or broader societal context of women's hunting activities.

Anderson et al. (2023) found evidence for women's hunting in 50/63 (79%) of forager societies surveyed. For this estimate to be representative of actual contemporary forager diversity, the sampling procedure would need to be unbiased. Anderson et al. (2023) describe their sampling procedure as follows: they chose 391 forager societies from the D-PLACE (https://d-place.org/) ethnographic database (Kirby et al., 2016) and selected that subset of 63 in which ‘explicit’ information about hunting was available. To account for autocorrelation (Galton's problem), societies were chosen in geographically diverse locations. Little further information about sampling was provided in the text of the article. However, the senior author told the newspaper El Pais — for an article entitled ‘Women have always hunted as much as men’ — that societies were only included where sources were explicitly “detailing hunting behavior and strategies”. Studies lacking tables, statistics or details, were excluded from the sample (Criado, 2023). The methods did not incorporate other statistical procedures that are common in quantitative ethnographic analysis, and the inclusion criteria were neither well described nor clearly operationalized. Given that the resulting estimate of 79% provided by Anderson et al. (2023) is substantially higher than previous quantitative assessments of 7% (Gurven and Hill, 2009), it is worthwhile to consider selection bias.

“Reproduction” is defined as obtaining the same results as a given study using the same data and same methods, and “replication” as obtaining similar results using new data and similar methods as one or more previous studies (National Academies of Sciences, Engineering, and Medicine, 2019). We did both. We first tried to reproduce Anderson et al. (2023) by following their methods to obtain the same texts from the same database. When this failed, we tracked down most of their sources and recoded them. In the interest of reproducing the study of Anderson et al. (2023) as closely as possible, we do not limit our attention to foragers. While Anderson et al. (2023) refer to all societies in their study as foragers, they did not distinguish between subsistence types in their study, labeling several horticulturalist and agriculturalist groups as foragers.

First, we explored whether Anderson et al. (2023) followed their own reported exclusion criteria. In this section of our critique, we worked with population names in the datasets as they were presented, not adjusting for potential cases of pseudoreplication. We considered, first, whether societies were included in the analysis that were not found in D-PLACE (Table S1). We identified a significant number of societies in the Anderson et al. (2023) sample that were not in the D-PLACE database (which comprises the Ethnographic Atlas, the Binford dataset, the Standard Cross-Cultural Sample (SCCS), and the Western North American Indians dataset). Specifically, of those 63 societies with “explicit data on hunting” (p.3), 22/63 (35%) were not present in D-PLACE. Of the 50/63 societies with putative evidence of women's hunting, 18 (36%) were not found in D-PLACE. This raises the question of how societies outside of D-PLACE were chosen, as no details were given. We found that of those 22 societies gathered from outside of D-PLACE, 18/22 (81%) had evidence of women's hunting, according to the authors. In the absence of methodological details about how bias was avoided, this result implies biased selection.

We next considered whether there were societies among the initial 391 D-PLACE samples for which explicit descriptions of hunting were available, but were excluded nonetheless. Here, we found evidence that Anderson et al. (2023) did not follow their own reported selection criteria and excluded some groups that should have been included, again inflating the frequency of women's hunting. Our attempts to acquire from D-PLACE (Kirby et al., 2016) the same list of 391 societies that Anderson et al. (2023) began with did not meet with success due to a lack of sufficient information in their paper, and so we could not reproduce their analysis. Instead, to assess the possibility that some relevant ethnographies were wrongly excluded, we constrained our search to two recent large-scale and comprehensive quantitative syntheses of human foraging behavior (Koster et al., 2020; Kraft et al., 2021). We explored the overlap between these two datasets, D-PLACE, and the Anderson et al. (2023) dataset (Table S1). We found 18 societies that were described in D-PLACE, in which comprehensive explicit data on hunting existed, but which were excluded from the Anderson et al. (2023) analysis. This provides evidence that the study did not follow its own exclusion criteria. Moreover, of these 18 societies (Anbarra, Arnhem, Bari, Bororo, G/wi, Gadio Enga, Kanela, Mamainde, Maya, Mekranoti, Nukak, Onge, Piro, Shipibo, Xavante, Yanomamö, Ye'kwana, Yukpa), 9 (G/wi, Nukak, Onge, Yanomamö, Arnhem, Shipibo, Mamainde, Bari, and Maya) reported explicitly that women did not hunt, and in the other 9 only men's hunting was mentioned. This again raises the possibility of biased selection and again serves to artificially inflate estimates of women's hunting. A comprehensive re-analysis of the entire D-PLACE database would likely reveal more cases of improper exclusion.

We next considered the inclusion of non-D-PLACE sources by Anderson et al. (2023) (Table S1). Koster et al. (2020) selected datasets that characterized hunting across the lifespan, and Kraft et al. (2021) selected studies that characterized subsistence contributions in-depth at the societal level, partly to characterize gender-based differences in foraging. As these studies summarize some of the most authoritative datasets on forager subsistence behavior, they constitute perhaps the best explicit descriptions currently available. Between these two publications, there were 60 unique societies. Of these, Anderson et al. (2023) included 18 in their study (!Kung, Ache, Agta, Aka, Baka, Batek, Bofi, Cree, Efe, Gunwinngu, Hadza, Hiwi, Inuit, Martu, Mbuti, Pume, Punan, Tsimane), and, of these, 15 were coded as positive for the occurrence of women's hunting. 14 of these societies were contained in D-PLACE. This leaves 42 societies with comprehensive descriptions of subsistence behavior, including hunting, that they did not investigate.

Of the 60 unique societies between Koster et al. (2020) and Kraft et al. (2021), 34 were contained in D-PLACE, and 26 of these societies were not (Achuara, Batek, Bofi, Dolgan, Etolo, Hiwi, Inujjuamiut, Kaul, Lufa, Machiguenga, Maku, Maring, Mvae, Mayangna, Nen, Nimboran, Nuaulu, Nunoa, Quichua, Tatuyo, Tsembega, Tsimane, Waorani, Wola, Wayana, Yassa). Given that Anderson et al. (2023) themselves appear to have sought sources outside of D-PLACE, it is not clear why many of the most comprehensively described societies were not chosen. In these populations, based on published work (Kaplan et al., 2000; Koster et al., 2020; Kraft et al., 2021), women's hunting occurs rarely or is not reported at all. For example, Koster et al., 2020 presented data on 40 small-scale societies. The data are largely unbiased toward the description of female hunting (J. Koster, pers comm). Only 3/40 societies (Baka, Tsimane, Martu) showed evidence for women's hunting. Similarly, the SI of Kraft et al. (2021) has detailed information on the source data, including gendered aspects of foraging.

Of course, the absence of evidence does not constitute evidence of absence, and it is always possible that women's hunting was underreported. However, given that the Koster et al. (2020) and Kraft et al. (2021) datasets drew on studies where long periods of fieldwork were conducted, in the first case with particular emphasis on hunting, it is specifically these societies where evidence of women's hunting is least likely to have been missed. This shows that even if Anderson et al. (2023) had strictly followed their own exclusion criteria, by limiting their search to D-PLACE, they would have missed a great deal of highly pertinent evidence that runs contrary to their conclusions.

We acknowledge that this is not a precise reproduction, as Anderson et al. (2023) did not provide enough detail in their methods to make this possible. Given the challenges of working with large ethnographic databases, we sympathize with the efforts of Anderson et al. (2023) and do not believe they were intentionally cherry-picking. Our point is that selecting different societies would have resulted in a far different (lower) estimate. Moreover, adding more societies to the denominator of the frequency calculation of Anderson et al. (2023) would change their estimate drastically. This is without examining the >300 other societies that were in their original (unsourced) sample, in some of which hunting is likely described in detail. Even putting aside their numerous coding errors (see next section), Anderson et al. (2023) estimate of 79% is unreliable and probably heavily inflated. A full analysis of several hundred ethnographic reports would be necessary to arrive at a defensible estimate.