Search

Data, Materials, and Software Availability: All study data are included in the article and/or supporting information available online at https://www.pnas.org/lookup/doi/10.1073/pnas.2220124120#supplementary-materials . Copyright © 2023 the Author(s). To address claims of human exceptionalism, we determine where humans fit within the greater mammalian distribution of reproductive inequality. We show that humans exhibit lower reproductive skew (i.e., inequality in the number of surviving offspring) among males and smaller sex differences in reproductive skew than most other mammals, while nevertheless falling within the mammalian range. Additionally, female reproductive skew is higher in polygynous human populations than in polygynous nonhumans mammals on average. This patterning of skew can be attributed in part to the prevalence of monogamy in humans compared to the predominance of polygyny in nonhuman mammals, to the limited degree of polygyny in the human societies that practice it, and to the importance of unequally held rival resources to women’s fitness. The muted reproductive inequality observed in humans appears to be linked to several unusual characteristics of our species—including high levels of cooperation among males, high dependence on unequally held rival resources, complementarities between maternal and paternal investment, as well as social and legal institutions that enforce monogamous norms. This work was conducted as a part of the “Emergence of Hierarchy and Leadership in Mammalian Societies” group at the National Institute for Mathematical and Biological Synthesis, supported by NSF Award DBI-1300426 and the University of Tennessee, Knoxville. It was supported by NSF awards SMA-1329089 and SMA-1743019, and the Santa Fe Institute, as well as the Max Planck Institute for Evolutionary Anthropology, Department of Human Behavior, Ecology and Culture. S.G. was supported by the US Army Research Office grants W911NF-14-1-0637, W911NF-17-1-0150, and the Office of Naval Research grant W911NF-18-1-0138. Additional funding for data collection was provided by the Wenner-Gren Foundation for Anthropological Research awards: 8913 and 7970, by NSF awards: BCS-0924630, BCS-0925910, BCS-0848360, BCS-0514559, BCS-0613226, BCS-0827277, SES-9870429, and DDRIG-1357209, by the National Geographic Society awards: HJ-099R-17, 20113909, 8671-09, and 7968-06, by the Kone Foundation awards: 086809, 088423, and 088423, and by the Jacobs Foundation, the UCSB Broom Center for Demography, and the UCSB Department of Anthropology.

Cooperative resource sharing is widespread across cultures, and it was likely critical during much of human evolutionary history for pooling risk. Need-based sharing specifically pools risk by following two cooperative rules: help others when asked, and only request help when in need. In a two-part study, we first expanded an agent-based model of need-based sharing partnerships, adding two types of defection and varying partnership sizes. We show that refusing to help always has a long-term cost, which increases with larger partnerships. In contrast, "greedy" requests that are not based on survival risk carry little-to-no cost. We then conducted an experimental vignette study of osotua, a need-based sharing tradition, with Tanzanian Maasai pastoralists. We found that participants generally complied with osotua requests, but shared larger amounts for requests that were based on survival risk. We conclude by proposing an expanded framework for evolutionary models involving need and fitness interdependence, where the cost asymmetry among types of defection generally favors a decision heuristic where individuals prefer sharing with those in need, but err on the side of generosity when need is uncertain.

Climate change adaptation involves the management of the climate-related risks, and the IPCC says we must prioritize adaptation immediately. However, researchers and policymakers have little systematic understanding of which adaptations actually reduce risks, including under different climate conditions. Drawing on data from human communities past and present, we review how features of climate variability—temporal autocorrelation, frequency, and severity—may predict which candidate climate change adaptations communities innovate or adopt. Using a case study of climate and remittances in Africa, we outline how researchers can characterize climate data by their autocorrelation, frequency, and severity, and then qualitatively compare these data to candidate adaptations. We include suggestions for how to involve communities in these explorations, from setting climate thresholds to identifying impactful hazards. By better understanding the relationship between climate variability and common solutions used by communities, researchers and policymakers can better support communities as they adapt to contemporary climate change.

According to the IPCC, we must immediately prioritize climate change adaptation—change in response to or anticipation of risks from climate change. Some researchers and policymakers urge “transformative change”, a complete break from past practices, yet report having little data on whether new practices reduce the risks communities face, even over the short term. However, much data exists: human communities have long generated solutions to changing climate, and scientists who study culture have examples of effective and persistent solutions. This special issue overviews cultural adaptation to climate change, and in this paper, we review how processes of adaptation, including innovation, modification, selective retention, and transmission, shape the landscapes decision-makers care about—from which solutions emerge in communities, to the spread of effective adaptations, to regional or global collective action. We include a comprehensive portal of data and models on cultural adaptation to climate change, and we outline ways forward, including how to do collaborative research with communities.

Analysis Preregistration Collective action is often studied in the context of common-pool resource systems--systems in which resources can be overharvested but excluding users is costly (Ostrom, 2003). This cost creates a dilemma: preventing overharvesting requires successful cooperation among many common pool resource users (Ostrom, 2008), or one or more individuals willing to foster cooperation among users and/or punish overuse (Gutiérrez et al., 2011; Olson, 1965). The difficulty of successful common pool resource management is compounded by the fact that many common pool resources, like forests, watersheds, grasslands, and fisheries, can be accessed by multiple communities (Ramirez-Sanchez & Pinkerton, 2009; Varughese & Ostrom, 2001), such that monitoring and enforcing rules governing the common pool resource requires multiple communities working together. That is, these systems present a problem of between-community collective action, which may be more difficult to achieve than within-community collective action. People are more likely to participate in collective action when the perceived benefit-cost ratio (that is, net benefits) of participation are greater than the net benefits of not participating. The same principle applies to between-community collective action, and many of the same factors that influence the perceived net benefits of within-community collective action – community-level characteristics like institutional effectiveness and individual-level characteristics, like wealth and gender – are anticipated to also influence the net benefits of between-community collective action. That said, between-community collective action is likely to be affected by other factors not shared with within-community collective action – namely, relationships between individuals in different communities. Long-distance social relationships, which often span community boundaries, can foster interdependence, which can increase individuals’ willingness to generate benefits and avoid costs affecting members of other communities (Pisor et al., 2023; Pisor & Ross, 2022). However, long-distance relationships can also undercut successful collective action around natural resource management by undermining successful monitoring or encouraging outgoing resource flows, away from a forest or fishery (Pisor et al., 2023). In short, long-distance relationships present a unique challenge to the management of large common pool resources, as they can both promote and undercut successful collective action. Here, we investigate the relative roles of (1) long-distance relationships and (2) local institutional effectiveness on participation in collective action in common pool resource management, while controlling for other variables that may modulate the potential net benefits of participation. Specifically, we draw on data from 28 villages in Tanga, Tanzania, each with a nearby Beach Management Unit (BMU) that manages local coastal fisheries, to explore whether an individual’s number of long-distance relationships and community-level perceptions of institutional effectiveness predict individual-level involvement in fishery-promoting activities. Predictions We expect factors that commonly affect within-community collective action around natural resource management will likewise impact between-community collective action. Namely, we predict that: (P1) Individual-level perceptions that BMU management is more effective will be associated with greater individual participation in BMU-related activities. Further, we expect that long-distance relationships will impact between-community collective action, such that: (P2) An individual’s number of long-distance relationships (family, friends, and business associates) will be associated with greater individual participation in BMU-related activities. We estimate the effects of other variables likely to affect the benefits to and constraints on BMU participation, including age, gender, profession, wealth, immigration to the community, and trust of members of the same community and one randomly selected nearby community. We previously preregistered the hypotheses and data collection methods prior to data collection on 2022-01-23 at https://osf.io/bvrty. In that preregistration, we also hypothesized that heterogeneity in communities would be associated with decreased participation in BMU-related activities; however, because of our sampling methods, we do not think we can validly estimate heterogeneity within the communities, and have dropped that hypothesis in these analyses. We have outlined our a priori expectations of the relationships between these predictors, controls, and the outcome – BMU participation – a directed acyclic graph (DAG). The DAG was constructed in a feedback process of proposing a priori expected relationships between variables and then checking the conditional independencies against the data, and then repeating the process until we were satisfied the DAG represented the data and expectations based on ethnographic observations. Throughout this process, we did not examine the outcome variable, BMU participation, or any relationships with it, but only examined the relationships between independent variables. In the attached document, we include figures of the DAG. Relationships are distinguished between (1) family relationships and (2) voluntary friendships and associations. ● An unobserved lineage variable - familial effects - causes family relationships, regardless of distance. That is, we assume that the size of one’s family is largely determined by family behavior rather than a person's characteristics. ● Friendships and associations are then independently caused by unobserved preferences and constraints to form friends (friendliness) and associations (associativeness), but relationships of close- and long-distance share also share a common set of unobserved preferences and constraints causing more close- and long-distance voluntary relationships (close sociality, distant sociality). ● Finally, more long-distance family is associated with more long-distance friends, possibly because of mutual connections in other villages or the residential mobility enabled by having distant family. The data suggests this is specific to long-distance friends and there is no relationship between long-distance family and associates – thus no connection between the two in our DAG. Next, we add demographic associations relating to relationships: ● The population is weakly patrilocal, with (1) women more likely to move near their husband’s family, and (2) inter-village marriage not being uncommon. This results in women having fewer close-distance family members than men, but this does not affect long-distance family because that is still largely a function of extended family members’ reproductive decisions (see lineage, above). This of course also affects how long women have lived in a village, given men are more likely to have lived in a village their whole lives. ● Gender also structures daily social relationships, and women are likely to have fewer friends and fewer long-distance relationships. ● Age and residence length both work together to predict close-distance family, with people who have lived in the village longer than expected (given their age) having more close-distance family and people who are younger than expected (given how long they have lived in the village) having more close-distance family; this might represent processes of a growing extended family within the village with more individuals dispersing to other villages as a person grows older. ● Similarly, longer residence length is associated with more long-distance family, even when adjusting for age; this is less straightforward but might also be a process of close-distance family dispersing. ● Finally, one’s profession (which is determined in part by gender) is associated with their tendency to make friends and associates. Different professions lead to different opportunities and incentives to form different relationships; for example, fishermen might have more long-distance relationships because they visit different landing sites and meet more people, and traders might have more associates because they provide access to more market opportunities. After exploring the data, we find that profession is associated with close- and long-distance friends and close-associates, but not long-distance associates. Thus, our data suggests profession is a causal factor in friendliness and close-sociality, but not distant-sociality. Next, we add variables expected to be related to collective actions around the fishery: We examine four variables related to collective action: how effective one perceives the BMU to be, within- and between-community trust, and whether one is an elected BMU leader. ● Within- and between-community trust share a common set of unobserved causes as generalized trust, which is likely caused by gender - women are less trusting than men in this sample - and age - older people are more trusting. Within-community trust then also is caused by close-friends and associates, as well as residence length - knowing more people within the village and living longer in the village leads to more trust. Likewise, between-community trust is greater for people with more long-distance friends and associates. However, while we expected this to also be the case for family, we did not find any relationship in this sample between trust and family for either distance. ● BMU leader is a position elected by members of the village, and people with more family members are more likely to be a BMU leader, regardless of whether the family lives within or outside the village, possibly representing inherited prestige markers that they can leverage during elections – in short, a causal relationship between the unobserved variable “lineage” and BMU leader. BMU leader is also associated with more long-distance friends and greater between-community trust. Because leaders are elected by local village members, we expect that these do not cause BMU leaders to be elected but rather BMU leaders have more long-distance friends and between-community trust because they are more likely to visit other villages while performing their duties. ● BMU efficacy is caused by greater perceived trust within- and between-community, as well as by having more close-friends and being a BMU leader. Finally, we add in the outcome variable: We anticipate nearly every measured variable to have a direct effect on our outcome (in blue) BMU participation, except for age, as age did not come up during ethnographic interviews as a factor in whether someone participated in the BMU, but the other variables did. Controlling for the full set of variables allows us to estimate the direct effect of each long-distance relationship in a single model; we can also estimate the direct and total effect of BMU efficacy in the same model. Our key assumptions are that there is no direct connection between the unobserved relationship variables (distant sociality, friendliness, and associativeness) and BMU participation, and that there are no intervening variables between the relationship variables and BMU participation. Data collection We collected data from 2022-01-25 to 2022-04-03 in 28 villages with a BMU in the Tanga region of Tanzania. A total of 1,399 participants were interviewed. Exclusion criteria Participants who do not live in the focal village will be excluded. In our earlier preregistration, we stated we would sample and include non-resident participants; however, this was based on observations during the pilot phase in a different region, where most non-residents where dago fishers who camp at the landing site and are expected to cooperate with the BMU. Most of our non-resident participants were not dago fishers, but transient fishers who would fish for the day before returning home, and are not expected to cooperate with the BMU. There are no other exclusion criteria. Variables BMU Participation Our outcome variable will be self-reported participation in BMU activities in the past twelve months. Participants were asked 9 questions about their participation and answered either yes or no, with a “yes” indicating they had participated in the activity. The questions were about participation in meetings, beach cleanings, educating others, projects, and monitoring and enforcement of regulations toward neighbors and people from outside the village. Responses will be analyzed in a multilevel Bernoulli model, estimating the probability of responding yes to each of the participation questions, with varying intercepts for participants and varying intercepts and slopes for questions. Close-distance Relationships Participants were asked the number of adult family, friends, and business associates they had in their village. The total number of each relationship will be log-transformed for analysis because we expect the effects of relationships to be non-linear - going from zero to one relationship is likely to have a larger effect than going from two to three relationships. Log-transforming will also help to handle outliers in the data. Long-distance relationships Participants were asked the number of adult family, friends, and business associates they had in each other village in the CFMA. For each relationship type, we will sum the number of that relationship in each village. The total number of each relationship will be log-transformed for analysis because we expect the effects of relationships to be non-linear - going from zero to one relationship is likely to have a larger effect than going from two to three relationships. Log-transforming will also help to handle outliers in the data. BMU efficacy Participants were asked 6 questions about how effective the local BMU at working toward different goals, such as cleaning the beach or developing the landing site. Participants were asked whether the BMU does anything to advance that goal, and participants could respond yes or no. A principal components analysis (PCA) of the questions finds that reducing the responses to a single dimension accounts for 43% of the variance in responses, and that the second dimension had an eigenvalue of less than 1, suggesting the second dimension captures less variability than a single question. We will thus use computed scores from the first dimension of the PCA as our measure of BMU efficacy. Within- and Between-Community Trust To measure within-community trust, we asked participants how many people in the village they trust to keep their promises and how many they trust to cooperate, with responses of none, a few, half, most, and all. We will code these responses as 1 (none) to 5 (all), sum them, and subtract 1 to produce a scale ranging from 1 to 9. Participants were also asked the same questions for a randomly selected village within the CFMA; the same coding procedure will be used to calculate a participant’s between-community trust in this randomly-selected village. These variables will be modeled as monotonic ordinal effects in the analysis. Profession Participants were asked how they were involved in the fishery as a source of income. Participants could give multiple responses. We will collapse profession categories into boat owner, fisher, trader, processor, seaweed farmer, and other. People who gave multiple responses will be assigned a primary profession based on observations during ethnographic interviews that some professions often involved ancillary professions (e.g., traders also often processed their own goods). This procedure was used in a previously published paper with this data set (see: https://doi.org/10.1016/j.evolhumbehav.2022.09.004). Profession will be included in the model as five indicator variables comparing each other profession to fisher. Residency length Participants were asked the year they moved to their current primary residence. We will compute the years they have lived in the current village by computing the difference between 2022 and the year they moved into the village. Before analysis, residency length will be standardized across participants. We include age in the model, so residency length will be adjusted by age. Age We asked participants what year they were born. Age is the difference between 2022 and their birth year. Before analysis, age will be standardized across participants. Gender Participants could respond either “man” or “woman.” We will use indicator coding, with man coded as 0 and woman coded as 1. Leader Some participants were elected leaders of the BMU. We will include in the model an indicator variable of whether people were a leader, with non-leader coded as the baseline. Note that the last BMU elections in every village were two years prior to the survey, so participation in the BMU in the past twelve months could not have caused a person to be elected as a BMU leader. Analysis Missing Data Missing data in responses likely results from errors during the interview as researchers accidentally missed or passed a question. As such, we expect the data to be missing completely at random. Rather than throwing out responses, we will use the mice package in R using predictive mean matching to impute missing data prior to any transformations above. We will use the predictive mean matching algorithm in mice to impute 5 data sets, which will be pooled together and analyzed using brms. Model Specification We will analyze the data in a Bayesian multilevel regression model, regressing BMU participation on the variables described above, using a Bernoulli likelihood. The Bernoulli distribution is a special case of the binomial distribution with only one trial, which describes the probability of observing a “success,” or in our model, the probability of saying yes to a given participation question; we use this distribution because the outcome variable is a single yes/no question and we are estimating the probability of answering yes to that question. The model will include varying intercepts for participants, and varying intercepts and slopes for relationship for BMU participation question, BMU, CFMA, and researcher; the random slopes for predictors of interest are included to examine generalizability across actions, sites, and researcher characteristics. See attached script for model specification.

The management of large common-pool resources, like fisheries and forests, is more difficult when more people and more communities can access them—a particular problem given increased population sizes and higher mobility in the Anthropocene. Social relationships spanning communities, such as kin relationships, business or trade relationships, and friendships, can make management even more challenging by facilitating overharvesting, transmitting norms for overharvesting, and encouraging corruption. However, these long-distance relationships can also bolster management by transmitting norms for sustainability, promoting interdependence, and laying the groundwork for nested management systems. Here, we review the positive and negative impacts of long-distance relationships on natural resource management, providing illustrative examples from our field research on forest and fisheries management in Tanzania. Drawing on the evolutionary literature, the development literature, and our field data, we offer suggestions for how development partners may leverage, or even deliberately foster long-distance relationships to promote successful natural resource management.

Note: This preregistration was created during data collection but before examining the data. Hypothesis Across cultures, pairs of friends usually help each other without strict bookkeeping of who did what and without making their help explicitly contingent on being repaid. However, research on friendships has mostly focused on close-distance friends, and people often form friendships with people across vast distances. Long-distance friendships can be more costly to maintain; for example, infrequent interactions and not belonging to the same institutions, which are important for enforcing rules, can make defections between long-distance friends more likely. Because of these challenges, long-distance friends might only be willing to help each other when help is assured to be repaid. We hypothesize that help between long-distance friends is more explicitly contingent than help between close-distance friends. However, distance isn’t binary: friendships exist on a spectrum of close-distance to long-distance. For example, while on average long-distance friends will be less likely to interact frequently and less likely to belong to the same religious institution, close-distance friends who interact infrequently or belong to different religious institutions will be more likely to make their help more explicitly contingent. Likewise, if some long-distance friends interact more frequently or belong to the same institution, their help may be less explicitly contingent. We hypothesize that when friends, regardless of distance, interact more frequently or belong to the same religious institution, help between friends will be less explicitly contingent. Method Sample KMS, BA, KB, PF, and RK collected data from January 2022 to April 2022 in 28 fishing villages across five Collaborative Fisheries Management Areas (CFMA) in the Tanga Region in northeastern Tanzania. We anticipate recruiting approximately 45 participants per village, with an expected final sample of 1,260 participants. Participants were recruited by local leaders, who brought fishery users to us to be interviewed. Procedures Interviews were conducted in Kiswahili using Open Data Kit (ODK) on a cellphone to record responses. Participants were asked the number of friends living in each village in the CFMA. ODK randomly selected one friend from the focal village and one friend from another village in the CFMA for follow-up questions about characteristics of the friendship and the help they have received from the friend. Participants were first asked about the friend in the focal village followed by the friend in the other village. Due to software constraints rather than deliberate design, questions were always asked in the same order. This study was declared exempt by the Washington State University IRB and was approved by the Tanzanian Commission for Science and Technology (COSTECH). Measures Help received from friend We asked participants whether their friend has ever given or loaned them more than 10,000 TSH. We specified an amount greater than 10,000 TSH because people are expected to regularly help their neighbors in need of help by providing a few thousand shillings; per pilot work in the Lindi and Pwani Regions of Tanzania, 10,000 TSH is usually gifted and loaned only between family and friends. When participants say a friend helped by giving or loaning money, we asked follow-up questions about the largest amount of money given or loaned and what that gift or loan was for. For participants who answered no to receiving a gift or a loan, we impute 0 as the max amount received. We also asked how frequently participants receive gifts or loans from the friend, with response options of monthly, seasonally, yearly, or less than once a year. For participants who answered no to receiving a gift or a loan, we impute a response category of never. Interaction frequency We asked participants how often they see each other face-to-face, with response options of daily, every other day, weekly, monthly, seasonally, yearly, and less than once a year. Belong to same religious institution We asked participants their religion, their friend’s religion, and whether they and their friend attend the same mosque or church, to which they could respond yes or no. Participants will be categorized into “not belonging to the same religious institution,” “belonging to the same religion but not in the same congregation,” and “belonging to the same congregation.” Profession We asked participants their profession in the fishery. Different professions differ in how much they might travel and the importance of business connections in other villages. Options for profession are fisher, captain, boat owner, processor, trader, agent (someone who buys and transports fish, usually sardines, on behalf of an investor or company), seaweed farmer, and other. Participants could choose all options that apply to them. Wealth To measure wealth, we asked participants whether they own a house with amenities and own personal items associated with wealth in the local context. For house amenities, we asked whether their house has: a cement or tile floor, block or stone walls, a metal roof, electricity, piped water, an indoor toilet, and a satellite dish. For personal possession, we asked if they or someone in their household own: a cellphone, a smartphone, a radio, a television, an electric fan, a refrigerator, a generator, a bicycle, and a motorcycle. Responses could be yes or no. We will use multiple correspondence analysis (MCA)of these items to create a standardized-wealth index, extracting the first 1-2 dimensions such that at least 40% of the variance is summarized. Pre-analysis Exclusion criteria We have no exclusion criteria for participants. Participants who completed the friendship section of the survey and had at least one close- or long-distance friend will be included in analyses. Missing data Gift size and frequency. Due to a coding error in the survey, in the first seven villages we did not ask follow-up questions about the gifts of help people received from close- and long-distance friends. We will exclude these participants from analyses. Wealth. Our measure of wealth is constructed from a number of questions using MCA, which generally requires complete data. Because of the number of questions, it is possible that some questions were occasionally skipped due to researcher error, and thus would require excluding these participants from analysis. If more than 1% of participants would be excluded because of missing wealth data, we will use multiple imputation for multivariate analysis via the missMDA R package. We will impute 10 data sets and run the models on each dataset, pooling the results together into one posterior distribution to estimate effects. Causal models Base model Our primary outcome variables are whether the participant has been given (give) or loaned (loan) money from their friend. Our primary exposure variable is whether the friend is close- or long-distance (D). However, many participants will not have long-distance friends, and participants who do not have long-distance friends may interact with close-distance friends differently than people with long-distance friends. As such, we include in our analyses variables that might affect both whether one has a long-distance friend and whether one receives help in a gift or loan. These variables are wealth (w), profession (p), gender (g), and age (a). Economic and social activities are highly gendered in this context, and gender determines what professions are available to a person, their opportunity to generate and accumulate wealth, and the kind of help they receive. Profession likely affects wealth, the opportunity and importance of making long-distance friends, and the kind of help one asks for and receives. Wealth likely affects the ability to travel to meet and maintain long-distance friends and the kind of help one receives. Age influences what profession one can do, what wealth a person has accumulated, their opportunity of making long-distance friends, and their likelihood of receiving help. These relationships result in the directed acyclic graph (DAG) in Figure 1. To estimate the direct effect of distance on help, we will adjust for gender, profession, wealth, and age (see Figure 2). Full model For this model, our primary outcome variables are whether the participant has been given (give) or loaned (loan) money from their friend. Our exposure variables are whether the friends belong to the same religious institution (R) and how frequently they interact (F). Other relevant variables are whether the friend is close- or long distance (d), wealth (w), profession (p), gender (g), and age (a). The relationships between gender, profession, wealth, age, distance, loan, and give are the same as the base model. Only distance determines whether friends belong to the same religious institution; because almost all participants will be Muslim, gender, profession, wealth, and age are not expected to affect whether they attend the same mosque. Frequency of interaction is determined by gender, profession, wealth, age, distance, and whether friends belong to the same religious institution. This results in the DAG in Figure 3. To test the direct effects of the exposure variables of interest, we will adjust for gender, profession, wealth, age, and distance (see Figure 4). Predictions 1. Comparing gifts and loans, people are more likely to receive help, receive more money, and receive help more frequently in the form of a gift than a loan from close-distance friends than long-distance friends. 2. People who interact more frequently with their friends are more likely to receive help, receive more money, and receive help more frequently in the form of a gift than a loan. 3. People who belong to the same religious institution as their friends are more likely to receive help, receive more money, and receive help more frequently in the form of a gift than a loan. For Prediction 1, we use a difference-in-difference approach, testing the interaction between distance and the kind of help, comparing the difference between receiving help in the form of a gift and a loan between close- and long-distance friends. Our prediction is that the difference between receiving help as a gift or a loan is greater for close-distance friends than long-distance friends. For Predictions 2 - 3, we predict frequency of interaction and belonging to the same religious institution will interact only with the form of help, not with distance, so we will directly compare the effect of each variable on receiving help in the form of a gift and a loan. Our prediction is that the effects of frequency of interaction and belonging to the same religious institution are larger for gifts than loans. Models We will analyze the data using multilevel Bayesian regression models. We will use a hurdle lognormal likelihood to estimate the probability one receives a gift or a loan, and if they do receive a gift or a loan, the mean max amount received. For frequency of receiving a gift or loan, we will use an ordered categorical likelihood to estimate the probability of receiving gifts or loans monthly, seasonally, yearly, less than yearly, and never. We will run two versions of each model, one a baseline model estimating the total effect of distance on help received, and one a full model estimating the direct effects of frequency of interactions and belonging to the same religious institution on help received. In all models, we include gender, age, profession, and wealth. All parameters interact with the form of help: gifts vs loans. We include in this preregistration R code for implementing the model (see file model.r); here we describe key decisions. Parameters. We estimate categorical predictors (gender, profession, distance) and their interaction with the form of help (gifts vs loans) as varying intercepts. This is to avoid estimating differences relative to an intercept that would include the effects of other categorical variables. By estimating these effects as varying intercepts, we can compute the difference between the estimated intercepts of receiving help from a close- and long-distance friend as a gift and as a loan while holding the other effects constant. We estimate the effect of continuous predictors - standardized age, standardized wealth, frequency of interaction, belonging to the same religious institution - as fixed-effects. For continuous (age, wealth) and ordered categorical (frequency of interaction, belonging to the same religious institutions) variables, there are two predictors, one for gifts and one for loans. Because religious institutions and frequency of interactions are ordered categories, we model these as monotonic predictors. Finally, we include varying intercepts for participant, CFMA, participant village of residence, and alter village of residence. Priors. In the included script, we specify regularizing priors that make reasonable predictions but still allow for a range of possibilities. These priors predict thick-tailed Gaussian distributions over the range of possible values, but extreme values are somewhat less likely than more moderate values (see Figures 5 and 6 for prior predictive checks for the model of max amount given and frequency of contribution, respectively). With the anticipated data, these priors should have minimal effects on the posterior. The priors also do not specify any of the effects in the model to be in a particular direction, and what effect the priors do have would be to make somewhat more conservative estimates of the effects. However, these priors may change - particularly their sigma may be decreased - to help with model estimation by reducing the state space of potential values.

Intergroup and long-distance relationships are both central features of human social life, but because intergroup relationships are emphasized in the literature, long-distance relationships are often overlooked. Here, we make the case that intergroup and long-distance relationships should be studied as distinct, albeit related, features of human sociality. First, we review the functions of both kinds of relationship: while both can be conduits for difficult-to-access resources, intergroup relationships can reduce intergroup conflict whereas long-distance relationships are especially effective at buffering widespread resource shortfalls. Second, to illustrate the importance of distinguishing the two relationship types, we present a case study from rural Bolivia. Combining ethnography and two different experimental techniques, we find that the importance of intergroup relationships-and the salience of group membership itself-varies across populations and across methods. Although ethnography revealed that participants often rely on long-distance relationships for resource access, we were unable to capture participant preferences for these relationships with a forced-choice technique. Taken together, our review and empirical data highlight that (1) intergroup and long-distance relationships can have different functions and can be more or less important in different contexts and (2) validating experimental field data with ethnography is crucial for work on human sociality. We close by outlining future directions for research on long-distance relationships in humans.

Communities want to determine their own climate change adaptation strategies, and scientists and decision-makers should listen to them — both the equity and efficacy of climate change adaptation depend on it. We outline key lessons researchers and development actors can take to support communities and learn from them. National Socio-Environmental Synthesis Center (NSF award number 1639145).

In the present study, we investigate whether exposure to low levels of precipitation (that is, droughts) and high levels of precipitation (what we call “excess precipitation,” related to flooding, high water levels, and landslides) predicts investment in long-distance social relationships. Specifically, however, we hypothesize that it is not the presence or absence of exposure to hazards like droughts and excess precipitation that will predict long-distance relationships, but specifically whether exposure is (1) chronic, (2) temporally autocorrelated, or (3) otherwise patterned (as opposed to anomalous).

An examination of fictive kin (godparents) as a category of alloparents todetermine their impact, if any, on education levels in godchildren.