1. Introduction
The idea that climate may affect the nature of society and culture is an old one and goes back to the writings of Aristotle in ancient Greece and Montesquieu in early eighteenth-century France [see discussion in Carleton and Hsiang (2016)]. Notions of environmental determinism were prevalent in scholarship of the late nineteenth and early twentieth centuries, particularly, the idea that temperate climates allowed advanced cultures to flourish while the tropics diminished this likelihood. Reacting strongly against what they perceived as racist ideas, social scientists in the last half of the twentieth century rarely explored the possibility that climate might influence the form of social and cultural traits (Carleton and Hsiang 2016; Harden 2012). There were, however, some scholars who pursued these connections, albeit without racial stereotyping. For example, Whiting (1964) was perhaps the first cross-cultural anthropologist to conduct a systematic study of the possible effects of the physical environment on specific aspects of culture. He found support for the theory that tropical environments might indirectly explain the long postpartum sex taboo because it would allow a mother to breastfeed a child longer and help prevent children succumbing to kwashiorkor, a tropical disease. Whiting (1964, p. 514) also suggested that climate might partially explain customary sleeping arrangements. In warm climate societies, husbands and wives tend to sleep in different beds or bedding. Conversely, in cold climate societies, husbands and wives tend to sleep together and babies typically sleep in separate containers, such as a cradleboard or cradle.
More recently, research on the possible effects of climate on culture has looked at the cultural realms of language, patterns of violence, personality traits, and religion. With regard to language, research has found that in warmer climates, languages have proportionately more consonant–vowel syllables and more sonorous sounds [Munroe et al. 1996, 2000; Munroe and Silander 1999; Fought et al. 2004; see also Ember and Ember (2000) on the interaction between dense plant cover and sonority]. Higher humidity predicts a higher proportion of vowels (Everett 2017) and more use of tones in speech (Everett et al. 2016); also, high altitude predicts more ejective consonants (Everett 2013).
A considerable body of research now suggests relationships between climate/weather events and conflict. Two analyses of previous diachronic studies (Hsiang et al. 2013; Burke et al. 2015) conclude that deviations from moderate temperatures and precipitation increase both the short-term risk of interpersonal violence and the long-term risk of intergroup conflict (riots, organized political conflicts, civil war, and interethnic conflict). More specifically, from their meta-analysis of the post-1950 time period, Hsiang et al. (2013, p. 1212) conclude that for each standard deviation, “change in climate toward warmer temperatures or more extreme rainfall increases the frequency of interpersonal violence by 4% and intergroup conflict by 14% (median estimates).” Some findings from prehistory and history also support the relationship between climate anomalies and/or periods of resource unpredictability and higher degrees of war (Lambert 1997; Lekson 2002; Kang 2000; Zhang et al. 2007) as does some cross-cultural research in nonindustrial samples (Ember and Ember 1992a; Ember et al. 2013). Other research has explored and found relationships between country-level aggregated personality traits (such as individualism and expressiveness) and aspects of climate (e.g., Pennebaker et al.1996; Van de Vliert and Van Lange 2019; but see Van Hemert et al. 2007).
Cross-cultural studies focusing on religious beliefs and climate have found that the presence of moralizing high gods is more likely in drought-prone areas (Snarey 1996; Roes and Raymond 2003) and in areas of increased environmental instability (Botero et al. 2014; Brown and Eff 2010).1 Also, increased religiosity is found in areas that experience extreme weather events or natural disasters (Sinding Bentzen 2019; Hayden 1987; Gibson and Connell 2015). Most of the cross-cultural studies on religion have suggested that moralizing high gods are more likely found in the face of uncertainty, perhaps because they promote more cooperation and solidarity. A cross-cultural study by Skoggard et al. (2020) found that resource stress predicts high god beliefs related to weather, and the authors theorize this may be due to the anxiety associated with such stress. In so far as resource stress is caused by, or exacerbated by, weather events such as drought or flood, we suggest here that climate, too, might be related to high god beliefs. Scholars of religion point out that events beyond human control will often be attributed to anthropomorphized supernatural agents (Guthrie 1995). Such agents are believed to act helpfully and/or harmfully, and humans will try to influence them with propitiation, magic, or collective ritual. Even if gods cannot be expected to provide help, their actions provide humans with explanations for such calamities.
Many of the religious beliefs and practices associated with climate are concerned with water availability and scarcity—whether in the form of rain, groundwater, or both—and its impact on food supply. In Namibia, the Damara word for drought translates as “there is no food” (Schnegg 2019, p. 839). Rain magic and rituals found widely across cultures are attempts to secure such a vital resource for food production (Malinowski 1948). Anxiety over water can also trigger a moral and collective response by communities who see the lack of rain as divine retribution for moral transgressions (Fortes 1987). In their saints’ days celebrations, the Nahua of Mexico acknowledge the importance of water: “Water is our life. It is what gives us life. Without water, we would die” (Taggart 2019, p. 49). The more famous ethnographic examples of ritualized water management are the fiesta complex of Latin America (Reina 1967) and the water temple system of Bali (Lansing 1987).
Overall plan and justification of research
Following up on the Skoggard et al. (2020) study, which found that resource stress predicts beliefs that high gods are associated with weather, the present study explores whether climate measures, particularly those tapping aridity and wetness, would show similar patterns. Climate measures are based on observed weather records; in contrast, the measures of resource stress used by Skoggard et al. were coded from ethnographies (typically based on interviews and participant observation). While ethnographers sometimes have historical reports of famines and droughts, they mostly rely on events that people tell them about. Thus, such measures likely reflect people’s perception of events serious enough to warrant recording. The use of independent weather records also gives us the opportunity to assess the predictability or variability of weather over time, which may not be reported in the ethnography. If both types of variables (resource stress and climate variables) are predictive, how might they be related? More specifically, is resource stress the mediator between climate and high god beliefs, or does climate have independent effects? These questions are the main focus of the cross-cultural research reported here.
In exploring the cross-cultural relationships between climate and religious beliefs involving weather, it is important to note that by examining climate relationships to religious beliefs, we do not necessarily postulate direct effects of climate, and we acknowledge, as do Cane et al. (2014) and Van de Vliert and Conway (2019), the importance of socioeconomic and cultural contexts as well as mediating factors that might account for the correlations found. For example, coping strategies can mitigate effects and perception of rainfall variability (Diem et al. 2017), and resource stress associated with ecosystem collapse may have multiple causations, not just weather (Breshears et al. 2011). But even if the reasons for the relationships are not known, we consider it useful to explore connections between climate and cultural traits. This is a way of making progress toward understanding human–environment interactions. Wherever possible, we try to suggest mechanisms that might explain the relationships we find between climate and beliefs about gods.
To study the possible influence of climate on the belief that high gods are associated with weather, we use a sample of largely nonindustrial societies designed to focus (where possible) on time periods prior to significant culture changes such as placement on reservations, imposition of strong colonial rule, and commercialization (see section 2). To give our cross-cultural study some context, Table 1 displays examples of the varied beliefs about how high gods are believed to be associated with weather. By a high god, we mean a god believed to be the sole creator or governor of the universe who may or may not be actively involved in the world, and if the former, may or may not be concerned with human morality (Swanson 1964, p. 56). Those high gods concerned with morality and meting out punishment for norm violations are referred to here and in the literature on cooperation as “moralizing high gods.” We note that not all societies believe in a high god, and, in some societies, lesser gods and spirits are believed to influence weather. However, in this study, we have chosen to focus only on high god beliefs. One reason is that the high god concept typically has societal reach; that is, people in the society believe that what the high god did or does applies to all of them. Second, high gods are usually the most powerful deity, and therefore following a law of proportionality (Piaget and Inhelder 1997) might be associated with the most powerful climate effects, such as prolonged drought. A third reason to focus on high gods is that in a previous study (Skoggard et al. 2020), which examined the relationships between ethnographically reported resource stress and god beliefs, the results regarding high gods were relatively strong and those involving beliefs about superior gods and minor spirits were very weak. Since our aim is to evaluate the effects of resource stress against independent measures of climate, there is little justification for examining all types of god/spirit beliefs. The examples in Table 1 show various high gods beliefs involving weather.
Beliefs about high gods and weather. Here ID indicates the identifier number.
To summarize, our main goal in this paper is 1) to explore whether climate as well as climate variability (i.e., predictability) may predict beliefs that high gods are associated with weather; and 2) if climate variables are predictive, to evaluate whether climate adds to our understanding beyond resource stress. Our major expectation was that such beliefs would most likely be found in drier and drought-prone climates where greater anxiety and uncertainty about food failure or water shortage may lead people to look to religion for help or explanations.
2. Methods
In this section we describe the sample for the study (section 2a), the measures for the four dependent variables on beliefs about the high gods’ association with weather (section 2b), the measures for two sets of independent variables (section 2c)—the climate measures [section 2c(1)] and the measures for resource stress [section 2c(2)]. Section 2d outlines the analyses.
a. Sample
We use the same starting sample as Skoggard et al. (2020) who examined the relationships between resource stress and beliefs that supernatural beings are associated with weather, except that here we focus exclusively on beliefs about high gods. As explained in Skoggard et al. (2020), the original sample of 96 societies fulfilled the following conditions: 1) it consisted of all 186 Standard Cross-Cultural Sample (SCCS) cases for which food-destroying natural hazards were more reliably coded by Ember and Ember (1992b) for a 25-yr period (from −15 years to +10 years around the ethnographic present) and 2) could be coded for god/spirit involvement with weather. The 46 societies in the present study (listed in Table S1 of the online supplemental material) were all those cases from the Skoggard et al. sample that had a high god present and were ratable on whether or not the high god was believed to be associated with weather. Note that the original SCCS sample of 186 mostly nonindustrial societies was constructed to be a representative worldwide sample that maximized historical independence of cases (Murdock and White 1969). Each society has a specified time and place focus (an “ethnographic present”); most of the societies (86%) have focuses that range from the 1860s to the 1960s. The complete range of focuses is from 1820 to 1980, and we assume a range of 15 years prior and 10 years after. Figure S2 in the online supplemental material shows the ethnographic present time frame compared with the GHCN weather data time frames. The creators of the original sample tried to pick earlier time periods and place focuses that had as little disruption to the group’s way of life as possible (Murdock and White 1969, p. 340).
Figure 1 shows the geographical distribution of the 46 societies with an indication of whether they believe their high god is associated with weather. We have overlaid the high god beliefs over average annual rainfall because it is a proxy measure for the wet and dry factor scores. In addition, we have noted with an “x” the 50 societies in our 96 starting sample that either lacked high gods or could not be coded for their association with weather. Note that the belief in a high god does not have a uniform worldwide distribution, because the belief in a high god is strongly related to higher levels of political hierarchy (Swanson 1964; Peregrine 1996), and regions such as North America generally had small-scale populations.
Distribution of societies highlighting whether high gods are associated with weather (black) or not (blue) as well as societies where high gods are absent (x). Underlying shading is total annual precipitation (mm).
Citation: Weather, Climate, and Society 13, 2; 10.1175/WCAS-D-20-0080.1
Distribution of societies highlighting whether high gods are associated with weather (black) or not (blue) as well as societies where high gods are absent (x). Underlying shading is total annual precipitation (mm).
Citation: Weather, Climate, and Society 13, 2; 10.1175/WCAS-D-20-0080.1
Distribution of societies highlighting whether high gods are associated with weather (black) or not (blue) as well as societies where high gods are absent (x). Underlying shading is total annual precipitation (mm).
Citation: Weather, Climate, and Society 13, 2; 10.1175/WCAS-D-20-0080.1
b. Measures of dependent variables—High gods association with weather
Our measures of beliefs that high gods are associated with weather come from the previous study by Skoggard et al. (2020). The underlying information comes from ethnography, descriptive accounts of cultural and social life largely written by anthropologists. The procedures and variables may be summarized as follows: 1) Coders read relevant passages from the ethnographic record, largely from the online “eHRAF World Cultures” database (HRAF 2020) for the specified time and place focuses to make ratings of high god beliefs; 2) to find relevant information, the coders mostly used advanced search in eHRAF World Cultures to both look at subjects on religion, disasters, or ethnometeorology combined with keywords about weather events (e.g., drought, flood, rain) to get a picture of how people thought about the causes of weather and its association with gods and spirits; 3) four variables were coded for high gods: (i) “associated with weather” is a general category that identifies whether or not a high god is involved with weather, (ii) used weather to harm food supply, (iii) used weather to help food supply, and (iv) used weather to punish humans for some moral transgression. Table 2 gives the distribution of each high god category for the present study (see details in Skoggard et al. 2020). Note that the coders were not allowed to infer the lack of association of the gods with weather merely from the absence of information on the subject. The coder must have had information on beliefs about the causes of weather events to infer that high gods were not involved. Also, note that the category “associated with weather” would be considered present if any of the other attributes [(ii)–(iv)] were present. However, association with weather could be coded as present for other reasons, such as statements that the high god was the god of wind, or that the high god brought rain, without any specification that it helped or harmed food supply; or that the god was “malevolent,” “mischievous,” or “capricious” in bringing bad weather, but not “punitive.” For a “punitive” designation to be applied, ethnographic evidence had to clearly indicate a god’s moralistic intent to punish.
Summary statistics about high gods and their association with weather.
c. Measures of independent variables
1) Climate variables
As explained in Felzer et al. (2020), our original interest in collecting climate data was to provide alternative ways of assessing the frequency of food-destroying natural hazards for the anthropological record where ethnographic and historical reports were lacking. This effort was part of a larger project on the possible effects of natural hazards on culture. Since hazards can include droughts, floods, and killing frosts, we developed a variety of specific indices to try to capture weather extremes based on daily precipitation, minimum and maximum temperature data from weather stations, and monthly gridded data for drought indices. As the Felzer et al. (2020) paper made clear, the goal of using weather data to better predict droughts and floods although laudable, did not achieve the precision that we wanted. The Felzer et al. study also calculated predictability of these extreme indices (Jiang et al. 2016) based on the hypothesis that it may be the predictability or variability, not just the occurrence, of extreme weather events that may influence culture. However, given the large number of indices included, Felzer et al. used factor analysis to reduce the dimensionality of wet, dry, and cold indices [see Felzer et al. (2020) for overall details]. For the present study, we did not have any strong a priori expectations for high god beliefs regarding weather except for the idea that aridity and drought might be more anxiety producing than wetness or flooding (Gibson and Connell 2015; Rosinger and Brewis 2020; Taggart 2019). We used three wet factor scores (overall wet factor, relative wet factor, and absolute wet factor) and cold factor scores from Felzer et al. (2020) for the present study but developed slightly new factor scores for dryness and dryness predictability. The new dry factor scores were intended to better represent drought, rather than general dryness.
To give context to these climate variables, this section summarizes 1) the sources and indices for the climate data, 2) spatial and temporal mapping of climate data to ethnographic sites, 3) predictability or variability calculations, 4) the dimension reduction procedures and the resultant factor scores, and 5) a sensitivity analysis that narrows the geographic scope of weather stations and other analyses to explore the temporal assumptions.
(i) Sources and indices
Our indices are based both on daily weather station data and monthly gridded data. We first describe the weather station indices (see Table 3). We used extreme indices from the Expert Team on Climate Change Detection and Indices (ETCDI) (Karl et al. 1999; Peterson et al. 2001) to represent the potential for floods plus one index of drought. These annual indices are based on daily precipitation and minimum and maximum temperature averaged over all the years and are derived from weather station data from the Global Historical Climatology Network (GHCN) (Menne et al. 2012). The temporal range of these data is starting years from 1861 to 1955 and ending years from 2007 to 2017 (Fig. S2 in the online supplemental material). There is overlap in 24% of the cases between the SCCS sites and GHCN weather station data, as the ethnographic data are generally prior to the weather station data. These data were gap filled by procedures described in Felzer et al. (2020). The wet indices (R10, R20, R95P, R99P, RX1, RX5, SDII, CWD, and PRCPTOT), were subdivided into those based on relative extremes (R95P, R99P, and CWD) and those based on absolute extremes (R20, R10, RX1, and RX5). The drought index was the number of consecutive dry days (CDD). We also developed several indices of extreme cold [number of frost days (FD), cold-spell duration index (CSDI), and Tmin] and one measure of extreme heat [number of summer days (SU)] from the GHCN data.
List of indices used for climate variables (frequency and percentages shown). Note that all data are from daily GHCN weather station data, except for P-Eth, P-Epm, and PDSI. An asterisk indicates that the index is not an extreme index.
The remaining drought indices are from the Climatic Research Unit (CRU) at the University of East Anglia, United Kingdom, which has a spatial resolution of 0.5° × 0.5° (CRU3.23) (Harris et al. 2014). The widely used Palmer drought severity index (PDSI) contains information about antecedent and current soil moisture (Alley 1984) and is computed from 1901 to 2009 (Van der Schrier et al. 2013). The two precipitation minus evaporation indices, which describe the deficit of soil moisture were developed by Felzer et al. (2020) from monthly temperature, precipitation, and Penman–Monteith potential evapotranspiration (PET_pm) data from 1901 to 2014. One version of the precipitation minus evaporation (P-E) index is based on the PET_pm (P-Epm), and a second version (P-Eth) is based on the Thornthwaite evapotranspiration (Willmott et al. 1985), which depends on temperature. The monthly values of each index for each site were averaged to provide a mean annual value and then averaged over all years to provide one value per site.
(ii) Spatial and temporal mapping of climate data to ethnographic sites
Except for a few SCCS societies that had a community focus, for most SCCS sites, we obtained a centroid coordinate and used this coordinate as a proxy of their focus’ location. We then searched the GHCN database to find one primary and eight secondary weather stations to provide climate to the SCCS site. For any GHCN station to be considered in our analysis, it must have more than 10 years of data with less than 10% missing data. The primary station is the GHCN station that is closest to the SCCS coordinate, and we used data from the eight secondary stations to fill data gaps in the primary station. Most of the primary stations are within the 300-km radius to the SCCS coordinate, but we acknowledge that for some SCCS sites, obtaining a representative weather station for its climate may be challenging due to spatial or temporal coverages. See the distribution of the sites and the weather stations in Fig. S3 of the online supplemental material. We subsequently performed a sensitivity test by limiting the weather station to a radius of 300 km to the focuses of the SCCS society and eliminating all SCCS sites that do not meet this criterion. Furthermore, we developed an alternative approach of using four secondary stations to fill data gaps in the primary station. As we discuss later in section 3, this sensitivity test gives us confidence in our original approach. Details can be found in section 3 of the online supplemental material, and a map of the weather stations for the sensitivity analysis is in Fig. S4 of the online supplemental material.
Temporal overlap was a challenge inasmuch as some societies had EPs that predated the 1900s; 24% of our SCCS sites have a weather station with some temporal overlap. The degree of overlap is shown in Fig. S2 of the online supplemental material. However, as we show in online supplemental Figs. S1a–c for the case of our drought indices, there has been no significant shift toward more extreme drought from 1901 to 2014, so even though the temporal periods are not always overlapping, the climate should be representative for the geographic location.
In section 4, we note some limitations posed by the lack of temporal and geographic fit for some of our climate measures. Much of the error resulting from lack of fit is likely to be random rather than systematic error. Random error usually depresses the size of correlations. Therefore, there is more concern that we might have depressed a “true” result because of such error. On the other hand, “true” correlations are likely to be higher than the observed correlations if there is random error (Blalock 1972, 413–414).
(iii) Predictability measures
Predictability measures based on Colwell (1974) were calculated for all the precipitation indices based on GHCN weather station data as well as the three other drought measures (PDSI and two P-E indices). Predictability is the variation among successive periods in the pattern of a periodic phenomenon, composed of contingency (seasonality) and constancy (low interannual variability). Predictability is high if the precipitation has a high recurrent pattern in seasonality (i.e., large contingency) or low interannual variability (large constancy).
(iv) Dimension reduction
In both Felzer et al. (2020) and here, we use principal components analysis for dimension reduction (conducted in IBM’s SPSS 24 statistics software). The three wet factor scores are described in more detail in Felzer et al. (2020) where dimension reduction was performed on all nine wet indices, relative extremes, and absolute extremes—three factors in total [the factors labeled “overall wet factor,” “relative wet factor,” and “absolute wet factor” are described in section 2c(1)(i)] An overall wet predictability factor was also derived from the predictability scores for the nine indices. For ease of interpretation of the results we have reverse coded (multiplied by “−1”) the wet predictability factor so that a high score conveys more predictably wet.
For this study, dry factors were computed slightly differently from Felzer et al. (2020). A dry or drought factor was computed from only four indices here: PDSI, two precipitation minus evaporation measures (P-Eth and P-Epm), and the log10 of CDD. Using principal component analysis in SPSS 24, the overall variance explained by this dry factor was 71%. We have also recalculated the dry predictability factor based on the two P-E predictability measures and the log10 of the PDSI predictability, with resulting variance explained by this factor of 70%. The factor loadings for these variables are shown in Table S2 of the online supplemental material. For ease of interpretation of the results, we have reverse coded (multiplied by “−1”) the dry factor score so that a high dry factor score conveys more dryness.
(v) Sensitivity analysis
Because some of our gap-filling procedures may have introduced some measurement error by including data from weather stations that were not in that close proximity, we redid our analyses with more stringent criteria for how close the stations needed to be to the focus and for the weather stations to be used for gap filling. The procedures used for this “sensitivity analysis” are explained more fully in section 3, Fig. S4, and Table S6 of the online supplemental material.
2) Resource stress
This section describes the four basic measures of resource stress and a combined score of resource stress used to test our overall model.
We employ the same four basic measures of resource stress used in the Skoggard et al. (2020) study; however, the fourth measure of resource abundance was modified for this study. The first three measures (famine, natural hazards, and chronic scarcity) were the same variables used by Skoggard et al. These variables were originally coded by two independent coders from ethnographer reports by Ember and Ember (1992a,b) for a 25-yr time period around the ethnographic present (EP)—ranging from 15 years prior to 10 years after the EP. For these three measures, a higher score indicates more resource stress. To minimize measurement error, the coders were given explicit instructions about when inferring absence was or was not appropriate and were asked to discuss any difference in coding before creating a resolved score. The fourth measure (titled “resource abundance” in Skoggard et al. 2020) was originally based on dimension reduction procedure for 11 environmental/climate variables following Botero et al. (2014). The rationale for including this fourth measure in the past and present study is the desire to have an etic measure of environmental conditions predictive of more difficulty in collecting or growing plants and animals. However, for the present study, we have redone the dimension reduction procedure, narrowing it to include just the variables relating to the number of plant and animal species in the environment and net primary productivity. This is because in this study we need to separate climate variables from resource stress variables. More specifically, for the whole SCCS sample we performed a principal component analysis (PCA) on the variables of net primary productivity, vascular plant richness, amphibian richness, mammal richness, and bird richness; we dropped annual mean precipitation, annual precipitation variance, annual temperature variance, temperature predictability, annual mean temperature, and precipitation predictability. The variables were downloaded from the Database of Places, Language, Culture and Environment (D-PLACE) website (https://d-place.org; Kirby et al. 2016). Before doing the analyses, we excluded cases that did not match the SCCS focuses used for the other resource stress measures. Moreover, following Skoggard et al. (2020) we used imputation procedures for the four variables with a small amount of missing data—vascular plant richness, amphibian richness, mammal richness, and bird richness. (Note that cases with extensive missing information were eliminated so that no variable would contain more than 5% missing data, which was then imputed using a predictive mean matching procedure.) See section 4 of the online supplemental material for details of the procedures used to create the factor variable “plant and animal richness.” We have reverse coded the variable and relabeled it as “low plant and animal richness” so that a high score reflects more potential resource stress. See Table S4 of the online supplemental material for the PCA loadings.
As noted in Skoggard et al. (2020), the resource stress variables are highly intercorrelated (α = 0.77), and analyzing each form of resource stress separately could yield biased estimates as a result of multiple comparisons. If each comparison is testing the same hypothesis, these multiple comparisons can lead to type-I error. To address this issue, we averaged the z score of each resource stress variable to create one combined variable measuring each society’s degree of resource stress.
d. Analyses
There are two main types of analyses: bivariate correlations and path models. We conducted bivariate correlations to first determine if there are relationships between climate and beliefs about high gods’ involvement with weather. Our second set of correlations assess the relationship between climate and resource stress to gauge whether they might be measuring the same or different concepts. Our last set of correlations (shown in Table S3 of the online supplemental material) show the associations among the climate factor scores.
In our path models we use the dependent variable “high gods associated with weather” in order to retain as many cases as possible. Since beliefs about gods and resource stressors are not likely to influence climate in any appreciable way, we consider climate to be an exogenous variable. However, certain climate variables (such as very low, very high, or unpredictable rainfall) likely play a role in increasing the likelihood of resource stress, which in turn has been shown to be associated with the belief that high gods are associated with weather (Skoggard et al. 2020). Consequently, we consider resource stress as a possible mediating factor that may help us explain the relationship between climate and high god beliefs. To test this general model, we estimate mediational path models that include the direct effect of climate on high gods associated with weather (controlling for resource stress), and the indirect effect of climate on high gods associated with weather through resource stress.
3. Results
We first turn to the relationships between the climate variables and the four high god variables shown in Table 4. Looking at the precipitation variables first (columns 1–6), 20 of the 24 or 83% of the correlations are significant (p < 0.05) or marginally significant (p < 0.1); 19 of the 24 correlations or 79% are significant. None of the relationships with the cold factor (column 7) or the thermal stress indicator are significant or marginally significant (column 8).
Pearson correlations between climate variables and high god variables (limited to societies with high gods). Note that p < 0.1 is indicated by a plus sign, and p ≤ 0.05, p ≤ 0.01, or p ≤ 0.001 is indicated by one, two, or three asterisks, respectively.
Moreover, the directions of the relationships in Table 4 are consistent with the idea that people living in drier climates are more likely to believe that high gods are associated with weather. The correlations in columns 1 and 2, indicating more dryness and more dryness predictability, are all positive. Conversely, columns 3–6, with three wet indicators and one predictability of wetness indicator, are all negative, suggesting that wetter and more predictably wet climates are generally associated with the absence of beliefs that high gods are associated with weather. Note that the predictability of dryness factor score (column 2) has the weakest results; the dryness factor (column 1) has the strongest.
Now we turn to the question of whether resource stress might mediate the results. We first look at the relationships between the measures of resource stress and the climate variables and then we examine path models to evaluate various causal models. As noted above, Skoggard et al. (2020) found that resource stress generally predicted beliefs that high gods were associated with weather. This raises the question of whether the climate variables are tapping the same constructs as the resource stress variables. Looking at the correlations between resource stress and precipitation measures (columns 1–6 of Table 5), only about one-half (53%) are significant or marginally significant in the expected direction (drier with more resource stress). Only 43% are significant. Moreover, these correlations are moderate at best (with the strongest being 0.52) suggesting little redundancy between constructs.2 Only the overall wet factor score is significantly related to all resource stress variables (more wetness, less resource stress). Of the resource stress variables, low plant and animal richness and the combined resource stress measure had the strongest correlations with the precipitation factor scores. Looking at the temperature-related climate variables (columns 7–8 of Table 5), the cold factor score was positively and significantly correlated with both low plant and animal richness and the overall measure of resource stress. Thermal stress, on the other hand, was negatively and significantly correlated with low plant and animal richness.
Pearson correlations between climate and resource stress measures. Note that p < 0.1 is indicated by a plus sign, and p ≤ 0.05, p ≤ 0.01, or p ≤ 0.001 is indicated by one, two, or three asterisks, respectively.
In the second part of our analyses, we sought to determine if the relationship between climate and high god beliefs may be mediated/explained by resource stress. Do drier climates predict beliefs that high gods are associated with weather because these climates have increased resource stress? In other words, is resource stress the mediator (indirect effect) through which climate has an effect on high god beliefs? Note that we do not include path models with the temperature-related climate variables because the correlations between these variables and high gods associated with weather are essentially zero (−0.05 and 0.01; see Table 4).
Figure 2 depicts the results of our mediational path analyses. In answer to the question of whether resource stress is the mediator through which climate affects beliefs in high gods, the results suggest not. The only model coming close to showing a considerable mediational effect of resource stress is shown in Fig. 2a. In that model, the indirect effect of resource stress shown in square brackets is almost significant; the 95% confidence interval varies from −0.05 to 0.35 (it would be significant if it did not cross the zero boundary). In general, the results are generally consistent with the idea that the precipitation climate variables may have direct effects on beliefs about gods as well as some indirect effects via resource stress. In three of the six models (Figs. 2b,d,e), the direct effect of climate on gods controlling for resource stress is significant. (The direct effect is inside the parentheses in the bottom arrow.) And, in four models (Figs. 2a,b,d,f), the combined direct and indirect effects of the climate variables (shown outside the parentheses in the bottom arrow) are not only significant, but the indirect path coefficient is also considerably higher than the direct path coefficient, suggestive of a partial mediation effect of resource stress. Also note that resource stress has a significant independent effect on high god beliefs in three models (Figs. 2a,c,d; marginally significant in Fig. 2f). Resource stress, even if not the mediator for the relationships between climate and high god beliefs, is predicted by the precipitation variables; in five of the models (Figs. 2a,b,c,e,f) the relationship is significant or marginally so (Fig. 2d).
Mediational path models, showing a representation of how climate factors and resource stress predict beliefs that high gods are associated with the weather. Indirect effects were calculated using the “lavaan” R package, which uses the delta method. The direct effect c′ of climate factors on high gods associated with weather, controlling for resource stress, is in parentheses. The effect to the left of the parentheses is the total effect c of various climate factors on high gods associated with weather. Inside the square brackets are the estimates of the lower and upper bounds of the effect of climate on high god beliefs through resource stress (indirect 95% confidence interval). All effects have been standardized so that they can be interpreted as effect sizes. Dark-blue arrows indicate significant effects (p ≤ 0.05), and light-blue arrows indicate nonsignificant effects (p > 0.05); p < 0.1 is indicated by a plus sign, and p ≤ 0.05, p ≤ 0.01, or p ≤ 0.001 is indicated by one, two, or three asterisks, respectively.
Citation: Weather, Climate, and Society 13, 2; 10.1175/WCAS-D-20-0080.1
Mediational path models, showing a representation of how climate factors and resource stress predict beliefs that high gods are associated with the weather. Indirect effects were calculated using the “lavaan” R package, which uses the delta method. The direct effect c′ of climate factors on high gods associated with weather, controlling for resource stress, is in parentheses. The effect to the left of the parentheses is the total effect c of various climate factors on high gods associated with weather. Inside the square brackets are the estimates of the lower and upper bounds of the effect of climate on high god beliefs through resource stress (indirect 95% confidence interval). All effects have been standardized so that they can be interpreted as effect sizes. Dark-blue arrows indicate significant effects (p ≤ 0.05), and light-blue arrows indicate nonsignificant effects (p > 0.05); p < 0.1 is indicated by a plus sign, and p ≤ 0.05, p ≤ 0.01, or p ≤ 0.001 is indicated by one, two, or three asterisks, respectively.
Citation: Weather, Climate, and Society 13, 2; 10.1175/WCAS-D-20-0080.1
Mediational path models, showing a representation of how climate factors and resource stress predict beliefs that high gods are associated with the weather. Indirect effects were calculated using the “lavaan” R package, which uses the delta method. The direct effect c′ of climate factors on high gods associated with weather, controlling for resource stress, is in parentheses. The effect to the left of the parentheses is the total effect c of various climate factors on high gods associated with weather. Inside the square brackets are the estimates of the lower and upper bounds of the effect of climate on high god beliefs through resource stress (indirect 95% confidence interval). All effects have been standardized so that they can be interpreted as effect sizes. Dark-blue arrows indicate significant effects (p ≤ 0.05), and light-blue arrows indicate nonsignificant effects (p > 0.05); p < 0.1 is indicated by a plus sign, and p ≤ 0.05, p ≤ 0.01, or p ≤ 0.001 is indicated by one, two, or three asterisks, respectively.
Citation: Weather, Climate, and Society 13, 2; 10.1175/WCAS-D-20-0080.1
Results for the more restricted sample (the “sensitivity” analysis) using closer geographic coordinates for weather stations [see section 2c(1)(v), along with Table S5 in the online supplemental material] generally show the same pattern of bivariate results despite the smaller sample size. For the precipitation variables (columns 1–6 of Table S5), 18 of the 24 or 75% of the correlations are significant (p < 0.05) or marginally significant (p < 0.1); 16 of the 24 or 67% are significant. The directions of the correlations are the same as in the main analysis; drought or dryness predicts more belief that high gods are associated with weather, and wetness predicts less belief. The major difference in the sensitivity analysis is that the wet predictability factor scores are now much weaker predictors of high god beliefs. Only two of the relationships are significant or marginal in column 6 of Table S5 as compared with four significant results previously. Since the predictability results are now very weak, with neither dry predictability nor wet predictability showing any significant relationships to the belief that the high god is associated with weather, we concentrate our discussion of the path mediation results for the sensitivity analyses on four models (Figs. S5a,b,e,f). Of these models, only the model in Fig. S5a comes close to showing a considerable mediational effect of resource stress. In that model, the indirect effect of resource stress (shown in square brackets) is almost significant judging by the 95% confidence interval that varies from −0.04 to 0.50 (it would be significant if it did not cross the zero boundary). However, three of the four other path models suggest direct effects of the climate variables. The absolute wet factor score (Fig. S5e) has a very high negative (−0.82; p ≤ 0.05) direct path correlation with the belief that the high god is involved with weather and two other factor scores—the dry factor score and the overall wet factor score have marginally significant direct effects (path coefficients of 0.48 and −0.70, respectively). And, in two models (Figs. S5a,f), the direct and indirect effects of the climate variables combined are not only significant, but the combined coefficients are considerably higher than the direct effects (cf. the path coefficient outside the parentheses with the one in parentheses—see the bottom arrow). Also note that resource stress has a significant independent effect on high god beliefs in the model in Fig. Fig. S5.a.
4. Discussion and conclusions
Our research has found relatively strong and significant associations between some aspects of climate, particularly indicators of dryness or drought, and a society’s belief that their high god is associated with weather. More telling than simply the significance of the associations is the consistent direction of those associations—the more the indicators suggest drier climates or greater predictability of dryness (including the negative relationships with the wetness scores), the more likely the society has such beliefs. In contrast, wetter climates are associated with the reduced likelihood of beliefs that high gods are associated with weather. Moreover, these relationships do not appear to be the result of a mediation effect from resource stress. Most of our discussion below addresses possible mechanisms that might explain why the climate indicators have some effects independent of resource stress, but we also address the issue of drought and wetness predictability versus unpredictability and some limitations of our methods. We close with questions for future research.
a. Explanations for climate effects
Previous research found support for the theory that greater anxiety about food supply will increase the likelihood that people will look to their god for help with weather or, at least look to explain their difficulties from godly actions (Skoggard et al. 2020). Does “anxiety” theory more broadly help us understand the relationships we found with climate? We think so. Recall that the Nahua of Mexico say, “Water is our life. It is what gives us life. Without water, we would die” (Taggart 2019, p. 49), suggesting that too little water is more frightening than too much water. We suggest that a drought may generally be more frightening to a population than flood. First, flooding is usually a fast-onset hazard and typically affects people living near rivers; it is rarely a societywide or even regionwide event. Second, many societies with frequent flooding have adapted to floods by situating buildings away from flood plains, building houses on stilts, catching water in reservoirs, or building raised fields, or canals to channel water, and so on. Although societies have also adapted to drought through practices such as planting drought-resistant crops, storing grain in case of crop failure, and moving when necessary, drought is a slow-onset event that can be more devastating, affecting wide regions and persisting for years. Thus, even if societies have strategies to mitigate drought, they may not succeed if droughts are very prolonged. If drought is more anxiety-arousing than flood, this may explain why more aridity (both a high dry factor score and low wetness factor score) predicts beliefs that gods are associated with weather.
An additional explanation as to why dryness and drought may influence behavior independent of resource stress is found in recent research on resource scarcity that distinguishes between water and food insecurity as separate but related phenomena (Wutich and Brewis 2014; Brewis et al. 2020). Water scarcity can have an immediate health and life threatening impact independent of food scarcity (Rosinger and Brewis 2020; Schuster et al. 2020). In this case dry weather that may not affect food supply could still affect thirst, thus triggering its own psychological stresses and coping mechanisms, including those associated with religious beliefs and practices.
A third reason why the climate scores may affect the results independently from resource stress is that the climate factor scores presumably tap much broader and more long-term conditions than the resource stress measures. Recall that three of the four resource stress measures rate the conditions (i.e., food-destroying natural hazards, famine, or chronic scarcity) within a specific 25-yr time frame (from −15 to +10 years around the EP). Events like prolonged drought and famine do not usually occur that frequently and therefore different 25-yr time frames will often have different values. While climate obviously does change and climate change has accelerated in recent years, a society living in a relatively dry and drought-prone environment will likely have lived in such an environment for a long time and will continue to do so in the future. Figures S1a–c in the online supplemental material compare some of our drought indices and their standard deviations in two 50-yr time periods (1901–50 and 1960–2009). The means and standard deviations of the precipitation minus evaporation indices (P-Eth and P-Epm) are virtually unchanged over time. While the mean for the Palmer drought severity index (PDSI) does change some (from −0.42 to −0.78) from the first to the second 50-yr periods indicative of some more drought over time, the PDSI is on a scale from −10 to 10, so this change is not that substantial. In fact, NOAA’s own classification has from −1.9 to 1.9 representing midrange conditions (https://www.ncdc.noaa.gov/temp-and-precip/drought/historical-palmers/overview), so these values do not represent a change from normality. However, none of these measures suggest a profound shift from dry to wet. If climate measures tap longer spans of time, this could partially account for why the two sets of variables have partially independent effects. A society coded as having no apparent resource stress in the 25 years might still experience considerable dryness within a longer time period and therefore people may have anxiety about drought stemming from the past. Ember and Ember (1992a), analyzing the relationship between war frequency and natural hazards, found that fighting frequency was just as high for societies reporting a “fear of a food-destroying natural hazard” as those with one or more actual hazards in 25 years. If ethnographers report that people are afraid of a hazard, it almost certainly means that the hazard has occurred in the past. This suggested to the Embers that it was fear of loss, rather than actual loss, that might have driven populations to fight for additional resources. Similarly, if the climate measures are better indicators of a long time span, this might explain why most of the precipitation climate variables remain significant even when controlling for the resource stress measures. Note that harshness of climate per se does not seem to predict the belief that gods are associated with weather inasmuch as the cold factor score and the thermal stress measure were not predictive.
Another reason why the precipitation variables may be predictive of high god beliefs independent from resource stress is that weather can have detrimental impacts not directly related to food supply, such as impeding travel, destroying homes, or causing personal accidents. These events can invoke in humans the desire to seek supernatural aid and comfort. Also, weather is a powerful natural force outside human control, a force that humans attribute to supernatural agents as part of a cognitive worldview that dissolves boundaries between the natural and supernatural realms and ascribes agency to natural events (Guthrie 1995). Last, insofar as weather might have a direct effect on psychological states such as anxiety (Howarth and Hofman 1984), it could very well have an indirect influence on religious beliefs for the reasons described above.
We noted in section 3 that the dry predictability factor score was a weaker predictor of high god beliefs than the dry factor score. And in the path models, particularly the sensitivity analysis, the overall wet predictability factor also became much less predictive than the overall wet factor. This suggests that predictability of precipitation may be less anxiety-arousing than unpredictability and therefore the former would be expected to be less predictive of high god beliefs involving weather. If the anxiety theory has merit, this is precisely the outcome that we would expect.
b. Methodological limitations
To minimize error, particularly random error in correlational designs, cross-cultural researchers are advised to pick an ethnographic focus and code all the variables for each studied society for approximately the same time period and the same location (Ember and Ember 2009). While we would have liked to have time and place controls for all of our measures, and we did so for three of the resource stress measures and the high god measures, the climate/environmental data presented more a challenge because many of the societies studied in the ethnographic record have both early time focuses and are often in very rural regions without many weather stations. In addition, ethnographers are not always clear about their exact locations in the field. Sometimes this is because they are generalizing to the society as a whole (and therefore we pick a centroid), but sometimes it is because they are deliberately obscuring the community location for privacy reasons. If all the variables are coded from the same ethnography, the exact location is not needed, but obviously for the purpose of picking weather stations it is better to have exact locations. We presume that the lack of precise fits for weather stations will create random error. As noted above, random error almost always reduces correlations (Blalock 1972), which should give us more, not less, confidence that the correlations we find are likely stronger than the ones obtained in our research.
Our analysis suggests critical data gaps and uncertainties needed to capture the full spectrum of climate influence on societal perception of high gods. First, spatial and temporal coverage of weather data is key for an accurate assessment of climate influences, but, given the nonindustrial nature of these mostly remotely located societies, not all SCCS societies have good records of weather data in close proximity. Some longer-term regional weather analyses or modeling products such as reanalyses would help to improve the confidence of the relationships derived in this study. Second, climate extremes could change rapidly under the influence of the large-scale climate oscillation (e.g., El Niño–Southern Oscillation) and/or climate change (Breshears et al. 2011; Diem et al. 2017; Wainwright et al. 2019). However, as evaluated in Jiang (2016), the scores of temperature and precipitation predictability do not differ significantly in El Niño and La Niña years globally. Also, societal perception of climate and the associated religious beliefs may change over a relatively short period of time (i.e., within several human generations). We were not able to provide a time series assessment of climate influence on society perception of high gods due to limited data availability, but this is a crucial research direction for future work. Third, we did not consider the possible uncertainties in the historical climate data in our analysis. For example, it has been shown that precipitation datasets could be unreliable in areas of low density of weather station coverage (e.g., Hanel et al. 2020; Dinku 2019; Adam and Lettenmaier 2003; Yin and Gruber 2010), and as such, uncertainties in the meteorological data may lead to different interpretations of the climate influence on societal beliefs. Furthermore, the use of different climate products may also lead to different results. Therefore, we suggest future research should assess the discrepancy among different data products and uncertainties in weather data and their possible influences on the relationship derived in this study.
c. Future research questions
Although we have gained insight into the relationship between climate and culture, we are left with some puzzles that require more research. Is there support for the idea that drought is more anxiety arousing than flooding? Are people who believe that high gods influence weather less anxious about weather-related hazards than those without such beliefs? How much time does it take to influence or change beliefs regarding high gods’ involvement with weather—one, two, or three generations? Are there other differences in beliefs about gods, including other types of gods, that are associated with different types of climate?
Acknowledgments
This material is based upon work supported by the National Science Foundation (NSF) Interdisciplinary Behavioral and Social Science (IBSS) under Grant 1416651 (program solicitation NSF 12-614). We acknowledge Rui Cheng for her contributions to the development of the climate extreme and predictability indices and Christina Carolus for her help in processing the data on growing seasons and biomes. We also thank Christina Carolus for help in coding the high god variables. We appreciate Joshua Conrad Jackson’s advice on the mediational path analysis model.
Data availability statement
Most analyses were performed in R (http://www.r-project.org/index.html). Code to gap fill and process the climate data is available via GitHub (https://github.com/mingkaijiang/IBSS_climate). All other data and code are available at hrafarc.org so that others may reproduce our analyses. The R package “ltm,” version 1.1-1 (Rizopoulos 2006), was used to calculate Cronbach’s alpha; “stats,” version 2.6.0 (https://www.R-project.org/), was used to calculate correlations; and the R package “lavaan,” version 0.6-5 (Rosseel 2012), was used to run path models.
REFERENCES
Adam, J. C., and D. P. Lettenmaier, 2003: Adjustment of global gridded precipitation for systematic bias. J. Geophys. Res., 108, 4257, https://doi.org/10.1029/2002JD002499.
Alley, W. M., 1984: The Palmer drought severity index: Limitations and assumptions. J. Climate Appl. Meteor., 23, 1100–1109, https://doi.org/10.1175/1520-0450(1984)023<1100:TPDSIL>2.0.CO;2.
Ammār, Ḥ., 1954: Growing Up in an Egyptian Village: Silwa, Province of Aswan. Routledge and Kegan Paul, Ltd., 316 pp.
Blalock, H. M., Jr., 1972: Social Statistics. 2nd ed. McGraw Hill, 583 pp.
Botero, C. A., B. Gardner, K. R. Kirby, J. Bulbulia, M. C. Gavin, and R. D. Gray, 2014: The ecology of religious beliefs. Proc. Natl. Acad. Sci. USA, 111, 16 784–16 789, https://doi.org/10.1073/pnas.1408701111.
Breshears, D. D., L. López-Hoffman, and L. J. Graumlich, 2011: When ecosystem services crash: Preparing for big, fast, patchy climate change. Ambio, 40, 256–263, https://doi.org/10.1007/s13280-010-0106-4.
Brewis, A., and Coauthors, 2020: Household water insecurity is strongly associated with food insecurity: Evidence from 27 sites in low- and middle-income countries. Amer. J. Hum. Biol., 32, e23309, https://doi.org/10.1002/ajhb.23309.
Brown, C., and E. A. Eff, 2010: The state and the supernatural: Support for prosocial behavior. Struct. Dyn., 4, 1–21.
Burke, M., S. M. Hsiang, and E. Miguel, 2015: Climate and conflict. Annu. Rev. Econ., 7, 577–617, https://doi.org/10.1146/annurev-economics-080614-115430.
Cane, M. A., E. Miguel, M. Burke, S. M. Hsiang, D. B. Lobell, K. C. Meng, and S. Satyanath, 2014: Temperature and violence. Nat. Climate Change, 4, 234–235, https://doi.org/10.1038/nclimate2171.
Carleton, T. A., and S. M. Hsiang, 2016: Social and economic impacts of climate. Science, 353, aad9837, https://doi.org/10.1126/science.aad9837.
Carucci, L. M., 1998: Working wrongly and seeking the straight: Remedial remedies on Enewetak Atoll. Pac. Stud., 21, 1–27.
Colwell, R. K., 1974: Predictability, constancy, and contingency of periodic phenomena. Ecology, 55, 1148–1153, https://doi.org/10.2307/1940366.
Diem, J. E., J. Hartter, J. Salerno, E. McIntyre, and A. S. Grandy, 2017: Comparison of measured multi-decadal rainfall variability with farmers’ perceptions of and responses to seasonal changes in western Uganda. Reg. Environ. Change, 17, 1127–1140, https://doi.org/10.1007/s10113-016-0943-1.
Dinku, T., 2019: Challenges with availability and quality of climate data in Africa. Extreme Hydrology and Climate Variability: Monitoring, Modelling, Adaptation and Mitigation, Elsevier, 71–80, https://doi.org/10.1016/B978-0-12-815998-9.00007-5.
Ember, C. R., and M. Ember, 1992a: Resource unpredictability, mistrust, and war: A cross-cultural study. J. Conflict Resolut., 36, 242–262, https://doi.org/10.1177/0022002792036002002.
Ember, C. R., and M. Ember, 1992b: Warfare, aggression, and resource problems: Cross-cultural codes. Behav. Sci. Res., 26, 169–226, https://doi.org/10.1177/106939719202600108.
Ember, C. R., and M. Ember, 2000: High CV score: Regular rhythm or sonority? Amer. Anthropol., 102, 848–851, https://doi.org/10.1525/aa.2000.102.4.848.
Ember, C. R., and M. Ember, 2009: Cross-Cultural Research Methods. AltaMira, 228 pp.
Ember, C. R., T. A. Adem, and I. Skoggard, 2013: Risk, uncertainty, and violence in eastern Africa. Hum. Nat., 24, 33–58, https://doi.org/10.1007/s12110-012-9157-5.
Everett, C., 2013: Evidence for direct geographic influences on linguistic sounds: The case of ejectives. PLOS ONE, 8, e65275, https://doi.org/10.1371/journal.pone.0065275.
Everett, C., 2017: Languages in drier climates use fewer vowels. Front. Psychol., 8, 1285, https://doi.org/10.3389/fpsyg.2017.01285.
Everett, C., D. E. Blasi, and S. G. Roberts, 2016: Language evolution and climate: The case of desiccation and tone. J. Lang. Evol., 1, 33–46, https://doi.org/10.1093/jole/lzv004.
Felzer, B. S., C. R. Ember, R. Chiang, and M. Jiang, 2020: The relationships of extreme precipitation and temperature events with ethnographic reports of droughts and floods in nonindustrial societies. Wea. Climate Soc., 12, 135–148, https://doi.org/10.1175/WCAS-D-19-0045.1.
Fortes, M., 1987: Religion, Morality and the Person: Essays on Tallensi Religion. Cambridge University Press, 347 pp.
Fought, J. G., R. L. Munroe, C. R. Fought, and E. M. Good, 2004: Sonority and climate in a world sample of languages: Findings and prospects. Cross-Cultural Res., 38, 27–51, https://doi.org/10.1177/1069397103259439.
Gibson, C., and J. Connell, 2015: The role of festivals in drought-affected Australian communities. Event Manage., 19, 445–459, https://doi.org/10.3727/152599515X14465748512560.
Grigson, W. V., and V. Elwin, 1949: Maria Gonds of Bastar. Oxford University Press, 427 pp.
Guthrie, S. E., 1995: Faces in the Clouds: A New Theory of Religion. Oxford University Press, 290 pp.
Hanel, M., S. Nasreen, M. Vargas, U. Singh, P. Máca, O. Rakovec, R. Kumar, and Y. Markonis, 2020: Validation of reconstructed hydroclimate variables for past drought assessment. General Assembly 2020, online, European Geosciences Union, EGU2020-17284, https://doi.org/10.5194/egusphere-egu2020-17284.
Harden, C. P., 2012: Framing and reframing questions of human–environment interactions. Ann. Assoc. Amer. Geogr., 102, 737–747, https://doi.org/10.1080/00045608.2012.678035.
Harris, I., P. D. Jones, T. J. Osborn, and D. H. Lister, 2014: Updated high-resolution grids of monthly climatic observations—The CRU TS3.10 dataset. Int. J. Climatol., 34, 623–642, https://doi.org/10.1002/joc.3711.
Hayden, B., 1987: Alliances and ritual ecstasy: Human responses to resource stress. J. Sci. Study Relig., 26, 81–91, https://doi.org/10.2307/1385842.
Howarth, E., and M. Hoffman, 1984: A multidimensional approach to the relationship between mood and weather. Br. J. Psychol., 75, 15–23, https://doi.org/10.1111/j.2044-8295.1984.tb02785.x.
HRAF, 2020: eHRAF World Cultures. Human Relations Area Files, Yale University, https://ehrafworldcultures.yale.edu/.
Hsiang, S. M., M. Burke, and E. Miguel, 2013: Quantifying the influence of climate on human conflict. Science, 341, 1235367, https://doi.org/10.1126/science.1235367.
Jiang, M., 2016: On the history and evidence of the Colwell Index in quantifying environmental predictability, and its applications in characterizing precipitation predictability in the conterminous United States. Ph.D. dissertation, Lehigh University, 217 pp.
Jiang, M., B. S. Felzer, and D. Sahagian, 2016: Characterizing predictability of precipitation means and extremes over the coterminous United States, 1949–2010. J. Climate, 29, 2621–2633, https://doi.org/10.1175/JCLI-D-15-0560.1.
Kang, B. W., 2000: A reconsideration of population pressure and warfare: A protohistoric Korean case. Curr. Anthropol., 41, 873–881, https://doi.org/10.1086/317416.
Karl, T. R., N. Nicholls, and A. Ghazi, 1999: CLIVAR/GCOS/WMO Workshop on Indices and Indicators for Climate Extremes Workshop summary. Weather and Climate Extremes, K. R. Karl, N. Nicholls, and A. Ghazi, Eds., Springer, 3–7.
Kirby, K. R., and Coauthors, 2016: D-PLACE: A global database of cultural, linguistic and environmental diversity. PLOS ONE, 11, e0158391, https://doi.org/10.1371/journal.pone.0158391.
La Barre, W., 1948: Aymara Indians of the Lake Titicaca Plateau. American Anthropological Association, 250 pp.
Lambert, P. M., 1997: Patterns of violence in prehistoric hunter–gatherer societies of coastal Southern California. Troubled Times: Violence and Warfare in the Past, D. L. Martin and D. W. Frayer, Eds., Gordon and Breach, 77–109.
Lansing, J. S., 1987: Balinese “water temples” and the management of irrigation. Amer. Anthropol., 89, 326–341, https://doi.org/10.1525/aa.1987.89.2.02a00030.
Lekson, S., 2002: War in the Southwest, war in the world. Amer. Antiq., 67, 607–624, https://doi.org/10.2307/1593794.
Malinowski, B., 1948: Magic, Science and Religion. Doubleday, 274 pp.
McIlwraith, T. F., 1948: Bella Coola Indians. Vol. I, University of Toronto Press, 763 pp.
Menne, M. J., I. Durre, R. S. Vose, B. E. Gleason, and T. G. Houston, 2012: An overview of the Global Historical Climatology Network-Daily database. J. Atmos. Oceanic Technol., 29, 897–910, https://doi.org/10.1175/JTECH-D-11-00103.1.
Merker, M., 1910: Masai: Ethnographic Monograph of an East African Semite People. Dietrich Reimer (Ernst Vohsen), 456 pp.
Munroe, R. L., and M. Silander, 1999: Climate and the consonant-vowel (CV) syllable: A replication within language families. Cross-Cultural Res., 33, 43–62, https://doi.org/10.1177/106939719903300104.
Munroe, R. L., R. H. Munroe, and S. Winters, 1996: Cross-cultural correlates of the consonant-vowel (CV) syllable. Cross-Cultural Res., 30, 60–83, https://doi.org/10.1177/106939719603000103.
Munroe, R. L., C. Fought, and J. Fought, 2000: Rhythmicity or sonority: Response to Ember and Ember’s “cross-language predictors of consonant-vowel syllables.” Amer. Anthropol., 102, 844–848, https://doi.org/10.1525/aa.2000.102.4.844.
Murdock, G. P., and D. R. White, 1969: Standard cross-cultural sample. Ethnology, 8, 329–369, https://doi.org/10.2307/3772907.
Norenzayan, A., A. F. Shariff, W. M. Gervais, A. K. Willard, R. A. McNamara, E. Slingerland, and J. Henrich, 2016: The cultural evolution of prosocial religions. Behav. Brain Sci., 39, e1, https://doi.org/10.1017/S0140525X14001356.
Pennebaker, J. W., B. Rimé, and V. E. Blankenship, 1996: Stereotypes of emotional expressiveness of northerners and southerners: A cross-cultural test of Montesquieu’s hypotheses. J. Pers. Soc. Psychol., 70, 372–380, https://doi.org/10.1037/0022-3514.70.2.372.
Peregrine, P., 1996: The birth of the gods revisited: A partial replication of Guy Swanson’s (1960) cross-cultural study of religion. Cross-Cultural Res., 30, 84–112, https://doi.org/10.1177/106939719603000104.
Peterson, T., C. Folland, G. Gruza, W. Hogg, A. Mokssit, and N. Plummer, 2001: Report on the Activities of the Working Group on Climate Change Detection and Related Rapporteurs. World Meteorological Organization, 143 pp.
Piaget, J. E., and B. Inhelder, 1997: The Child’s Conception of Space. Routledge, 504 pp.
Reina, R. E., 1967: Annual cycle and fiesta cycle. Social Anthropology, Vol. 6, Handbook of Middle American Indians, Manning Nash, Ed., 317–332.
Rizopoulos, D., 2006: ltm: An R package for latent variable modelling and item response theory analyses. J. Stat. Software, 17 (5), 1–25, https://doi.org/10.18637/jss.v017.i05.
Roes, F. L., and M. Raymond, 2003: Belief in moralizing gods. Evol. Hum. Behav., 24, 126–135, https://doi.org/10.1016/S1090-5138(02)00134-4.
Roscoe, J., 1911: Baganda: An Account of Their Native Customs and Beliefs. Macmillan and Co., 547 pp.
Rosinger, A. Y., and A. Brewis, 2020: Life and death: Toward a human biology of water. Amer. J. Hum. Biol., 32, e23361, https://doi.org/10.1002/ajhb.23361.
Rosseel, Y., 2012: lavaan: An R package for structural equation modeling. J. Stat. Software, 48, 1–36, http://www.jstatsoft.org/v48/i02/.
Schapera, I., 1930: Khoisan Peoples of South Africa; Bushmen and Hottentots. G. Routledge, 445 pp.
Schnegg, M., 2019: The life of winds: Knowing the Namibian weather from someplace and from noplace. Amer. Anthropol., 121, 830–844, https://doi.org/10.1111/aman.13274.
Schuster, R. C., M. S. Butler, A. Wutich, J. D. Miller, and S. L. Young, 2020: If there is no water, we cannot feed our children: The far-reaching consequences of water insecurity on infant feeding practices and infant health across 16 low- and middle-income countries. Amer. J. Hum. Biol., 32, e23357, https://doi.org/10.1002/ajhb.23357.
Sinding Bentzen, J., 2019: Acts of God? Religiosity and natural disasters across subnational world districts. Econ. J., 129, 2295–2321, https://doi.org/10.1093/ej/uez008.
Skoggard, I. A., C. R. Ember, E. Pitek, J. C. Jackson, and C. Carolus, 2020: Resource stress predicts changes in religious belief and increases in sharing behavior. Hum. Nat., 31, 249–271, https://doi.org/10.1007/s12110-020-09371-8.
Snarey, J., 1996: The natural environment’s impact upon religious ethics: A cross-cultural study. J. Sci. Study Relig., 35, 85–96, https://doi.org/10.2307/1387077.
Swanson, G. E., 1964: The Birth of the Gods: The Origin of Primitive Beliefs. University of Michigan Press, 260 pp.
Taggart, J., 2019: The Rain Gods’ Rebellion: The Cultural Basis of a Nahua Insurgency. University Press of Colorado, 168 pp.
van der Schrier, G., J. Barichivich, K. Briffa, and P. Jones, 2013: A scPDSI-based global data set of dry and wet spells for 1901–2009. J. Geophys. Res. Atmos., 118, 4025–4048, https://doi.org/10.1002/jgrd.50355.
Van de Vliert, E., and L. G. Conway III, 2019: Northerners and southerners differ in conflict culture. Negotiation Confl. Manage. Res., 12, 256–277, https://doi.org/10.1111/ncmr.12138.
Van de Vliert, E., and P. A. M. Van Lange, 2019: Latitudinal psychology: An ecological perspective on creativity, aggression, happiness, and beyond. Perspect. Psychol. Sci., 14, 860–884, https://doi.org/10.1177/1745691619858067.
Van Hemert, D. A., Y. H. Poortinga, and F. J. R. van de Vijver, 2007: Emotion and culture: A meta-analysis. Cognit. Emotion, 21, 913–943, https://doi.org/10.1080/02699930701339293.
Vreeland, H. H., 1973: Mongol Community and Kinship Structure. Greenwood Press, 327 pp.
Wainwright, C. M., J. H. Marsham, R. J. Keane, D. P. Rowell, D. L. Finney, E. Black, and R. P. Allan, 2019: “Eastern African paradox” rainfall decline due to shorter not less intense long rains. npj Climate Atmos. Sci., 2, 34, https://doi.org/10.1038/s41612-019-0091-7.
Whiting, J. W., 1964: Effects of climate on certain cultural practices. Explorations in Cultural Anthropology: Essays in Honor of George Peter Murdock, W. H. Goodenough, Ed., McGraw Hill, 511–544.
Willmott, C. J., C. M. Rowe, and Y. Mintz, 1985: Climatology of the terrestrial seasonal water cycle. J. Climate, 5, 589–606, https://doi.org/10.1002/joc.3370050602.
Wutich, A., and A. Brewis, 2014: Food, water, and scarcity: Toward a broader anthropology of resource insecurity. Curr. Anthropol., 55, 444–468, https://doi.org/10.1086/677311.
Yin, X., and A. Gruber, 2010: Validation of the abrupt change in GPCP precipitation in the Congo River Basin. Int. J. Climatol., 30, 110–119, https://doi.org/10.1002/joc.1875.
Zhang, D. D., J. Zhang, H. F. Lee, and Y. Q. He, 2007: Climate change and war frequency in eastern China over the last millennium. Hum. Ecol., 35, 403–414, https://doi.org/10.1007/s10745-007-9115-8.
Moralizing high gods are gods who judge and act on the adherence of humans to social norms and promoting sociality: for an overview of the literature see Norenzayan et al. (2016).
Of course, it is also possible that the correlations are modest because the time frames for the resource stress measures and the climate measures are often from different time periods.