Fatherhood and Child–Father Attachment in Two Small-Scale Societies

Variation in attachment quality was assessed in 92 one- to three-year-olds’ (49 female, 43 male) attachment to their father, within and across two small-scale, subsistence populations: the Tsimane of lowland Bolivia (n = 29 dyads) and Tannese of Vanuatu (n = 63 dyads). Overall, the Tsimane and Tannese populations show similarity in the patterning of child–father attachment relative to samples from industrialized societies, with 52% of Tsimane dyads and 62% of Tannese dyads showing secure attachment. However, the insecure-ambivalent attachment was more frequent among Tsimane and Tannese dyads relative to samples from industrialized societies, and attachment security was greater for sons than for daughters in Tanna. Of several attributes of fathers, only reports by mothers of how frequently fathers play with their children are associated with child–father secure attachment. Fathers who scored highly on this measure were more likely to have children with secure attachments. Studies of child–father attachment are rare outside of industrialized populations yet are critical to understanding the mechanisms underlying child–father relationships.

In his seminal attachment theory, Bowlby (1969Bowlby ( /1982 posited that infants form an attachment bond with their caregiver, one that is critical to the survival of children. However, important individual differences exist in the quality of child attachment, which are expected to reflect variations in children's early caregiving environments (Ainsworth et al., 1978). Developing a secure or insecure attachment bond may contribute to different trajectories of social and emotional development (Groh et al., 2017). What begets secure attachment? It is when Mary Ainsworth developed a lab-based paradigm to assess the quality of infants' attachment to their caregivers, known as the Strange Situation Procedure (SSP; Ainsworth et al., 1978), that progress on this question was made. The resulting literature suggests that attachment security develops according to the sensitivity with which attachment figures respond to infant's physical and emotional needs (Ainsworth et al., 1971;Lamb & Lewis, 2013;Lucassen et al., 2011). Infants, in turn, develop expectations regarding their caregivers' sensitivity and responsiveness (Bigelow & Rochat, 2006;Bowlby, 1969Bowlby, /1982Bretherton & Munholland, 2016).
Although the universality of attachment and the cultural appropriateness of its most common measure, the SSP, have been criticized (e.g., Rothbaum et al., 2000), the vast majority of available studies find evidence of universality. Every study to date, whether of "Western" or less frequently observed "non-Western" populations, has found that children show attachment behavior when stressed and have a preferential bond with one or more caregivers (see Mesman et al., 2016). Moreover, in their review, Mesman and colleagues (2016) reported that the majority of infants in almost all studies are classified as being securely attached. Of course, understanding how secure attachment emerges in disparate cultures requires much more work. In addition, the SSP should be culturally adapted to ensure that it is a culturally appropriate procedure and that children in all populations only face mild distress. For example, sticking too rigidly to a 2-min rule for separations, or including a stranger in the SSP may create too much distress in children in some societies (van IJzendoorn & Sagi, 2001). Thus, there are reasons to consider the SSP a universal measure of attachment security as well as reasons to suspect that the SSP may require adaptation in particular cultural contexts. Resolving questions of the universality of the SSP requires much more research on attachment in rural, less-studied sample populations, particularly in the Global South. Furthermore, it is ethically imperative that, in general, researchers move beyond studying a narrow slice of the world's population when making claims about human psychology or behavior (Broesch et al., 2020).

Child-Father Attachment
While mothers have been the source of most primary caregiving to children in human populations cross-culturally (Broesch et al., 2021), childcare from other individuals, including fathers, is quite common and has been shown to be integral to family and child well-being (Diniz et al., 2021;Hrdy, 2011). Previous studies of "Western" populations have shown that not only do children often develop secure attachment toward their fathers but also that children who experience more sensitive parenting from their fathers are more likely to develop a secure child-father attachment (Lucassen et al., 2011). Furthermore, emerging research suggests that child-father attachment plays an important role in children's socio-emotional development . For example, studies in various "Western" populations have found that attachment security to fathers (but less so to mothers) is associated with higher parent-and teacher-reported social competence (Bolt et al., 2014) and a greater number of reciprocated friendships (Verissimo et al., 2010) as well as fewer self-reported externalizing symptoms in middle childhood (Bureau et al., 2020). In one study, the quality of child-father attachment assessed at 2 years of age was more predictive of child adjustment at age 16 years than the quality of child-mother attachment (Grossmann et al., 2002). However, the patterning and importance of child-father attachment may show significant cross-cultural variation, given the cross-cultural variation in the extent to which paternal care of children is normative and considered instrumental for child well-being (Gray & Anderson, 2010;Hewlett & MacFarlan, 2010;Shwalb et al., 2013). Observational data indicate considerable variation in paternal care: across small-scale, subsistence societies, fathers account for as little as 1% and as much as 22% of the time that infants are held or attended to by individuals (Kramer & Veile, 2018). Men tend to spend less time in childcare in societies where either gender contributes the majority to subsistence (Quinlan & Quinlan, 2007), where fathers have little impact on childrens' social success (Kaplan & Lancaster, 2003;Schacht et al., 2018), when mothers have older children or other kin who provide significant alloparenting (Griffin & Griffin, 1992;Kramer & Veile, 2018;Winking et al., 2009), where men have more reproductive options outside their marriage (Blurton Jones et al., 2000;Schacht et al., 2014), or where there are cultural norms favorable to divorce or polygyny (Lawson et al., 2015). It remains largely unknown how the range of cultural variation in the above factors influences the patterning of child-father attachment or its importance for child development.

Behavioral Antecedents of Child-Father Attachment
Although attachment theory has primarily examined the association between parental sensitivity (i.e., the ability to identify, interpret, and respond effectively to the child's needs) and the child's sense of security, empirical evidence suggests only a modest association between these two concepts in child-mother and child-father dyads (Deneault et al., 2022). Without denying the importance of parental sensitivity, researchers have started exploring other parenting behaviors that may contribute to child-parent attachment relationships, particularly for child-father dyads (e.g., Brown & Aytuglu, 2020;Cabrera et al., 2018). For example, fathers' motivation to engage in "rough-and-tumble" play with their children may be integral to the child-father bond and any unique effects of fathers on children's social development (Lamb, 2010). Fathers generally demonstrate physical play as an interactive mode with their children, and paternal sensitivity is often displayed in the context of monitored Deneault and controlled excitement (Grossmann et al., 2008). Studies have suggested that fathers more than mothers pursue quick and sudden peaks of positive arousal in play with their children (Feldman, 2003). However, more recent research in "Western" samples finds less evidence of a difference between fathers and mothers in their use of challenging play with young children (e.g., "rough and tumble" play), which challenges children to be assertive or take risks (Majdandzic et al., 2016;Olofson & Schoppe-Sullivan, 2022).
Previous studies suggested that less absolute time spent with children is not necessarily a risk for the development of attachment insecurity (Aviezer & Sagi-Schwartz, 2008). Thus, attachment researchers paid almost no attention to the degree of exposure to a child in attachment development, perhaps assuming that only a minimum threshold of exposure is necessary to produce an association between parental care and child attachment (Ainsworth as cited in Duschinsky, 2020). Because childcare by fathers tends to be more variable (compared with mothers) in many ways (Gray & Anderson, 2010;Hewlett & MacFarlan, 2010;Kramer & Veile, 2018), it remains unclear which aspects of childcare by fathers may have significance for childfather attachment.

Contextual Effects on Child-Father Attachment
The child-father attachment has been found to be more vulnerable than child-mother attachment to various contextual factors (Bureau et al., 2017;Cabrera et al., 2000;Schoppe-Sullivan et al., 2006), including child personality (McBride et al., 2002) and mother-father communication style and relationship satisfaction (Brown et al., 2010;Coley & Schindler, 2008;Goldberg & Carlson, 2014;Lamb, 2010). In addition, fathers' attitudes toward their parental role have consequences for child-father attachment (Cabrera et al., 2000;Cox et al., 1992). Within and across societies, fathers' attitudes toward parenting are typically more variable relative to mothers, due in part to greater range and less rigidity in cultural expectations that fathers engage in the direct care of children (Gray & Anderson, 2010;Hewlett & MacFarlan, 2010). Such gender-specific cultural expectations may explain why child-father attachment is often more sensitive to contextual effects in general, relative to child-mother attachment.
In addition, research among families in small-scale societies has found that paternal status is negatively associated with fathers' contributions toward childcare and family life (Hewlett, 1988;Stieglitz et al., 2012). A study among Aka hunter-gatherers of central Africa found that men who have greater influence in their community engaged in less childcare (Hewlett, 1988). In the Tsimane, a recent study found that men with higher incomes tend to shunt a greater percentage of their income toward market purchases that do not benefit the household . Although no research has examined the effect of paternal status on child attachment in smallscale societies, evidence in "Western" societies indicates that child-father attachment is influenced by mother-father communication and relationship satisfaction-both of which have been demonstrated to be affected by paternal status in small-scale societies (Brown et al., 2010;Coley & Schindler, 2008;Goldberg & Carlson, 2014;Lamb, 2010;Stieglitz et al., 2012).

Current Research
In this study, we examine child-father attachment quality using a modified version of the SSP, in two small-scale, rural, societies: the Tsimane of Amazonian Bolivia and ni-Vanuatu on Tanna Island, Vanuatu. Both societies were selected as a result of long-term research collaborations between the authors and Tsimane and Tannese communities. Our principal aim is to determine whether distributions of child-father attachment security are similar across societies where many other factors influencing child-father relationships may vary. To that end, we also compare child-father attachment among the Tsimane and Tannese with previously published results from "Western," industrialized populations. In addition, we examined variation in attachment quality within and between the Tsimane and Tanna families, as a function of several measures of paternal investment in their children as well as other contextual factors. This secondary (and somewhat exploratory) aim was to help us better understand the antecedents of child-father attachment in these societies. Child-father attachment has rarely been studied outside the context of industrialized populations. Given cross-cultural variability in expectations of fathers and fathering behavior, it is critical that our knowledge of father effects on children's attachment, and on children's development more generally, encompass diverse cultural settings.

The Tsimane
The Tsimane people live in villages of 20 to 500 individuals in the neotropics of central Bolivia. Their economy is based on swidden horticulture, hunting, fishing, and gathering. Since the 1990s, primary schools have spread to most villages and secondary schools now exist in several villages, but average educational attainment remains low, particularly among older individuals. The research for this study was conducted in a village with a longer history of formal education and greater proximity to a market town, with approximately 25% to 40% of adults completing primary school. Food sharing and collaboration in productive activities are mostly confined to extended families residing in the same or nearby households (Hooper et al., 2015). The extended family, not the village, remains the central unit of social organization. On the other hand, unrelated community members will regularly visit each other to socialize and drink shocdye' (chicha), an alcoholic beverage fermented from manioc and maize. Women specialize in chichamaking and so play a pivotal role in social integration across families.
Less than 5% of men are polygynously married, almost always to two sisters, and few marriages end in divorce . None of the fathers involved in this study were polygynously married. Postmarital residence is often but not always matrilocal or bilocal at first, with a greater likelihood of patrilocal or neolocal residence after the birth of children. Tsimane women average approximately nine offspring over their lifetime and do the large majority of direct childcare and food processing (McAllister et al., 2012). When parents are in their 20s, direct childcare is at its peak and consumes roughly 25% and 3% of women's and men's time, respectively . During the first 6 years of a child's life, mothers provide 69% of direct care, siblings 10%, fathers 7%, and the rest from other kin including grandparents . Parenting is fairly permissive and even toddlers spend much of their time in unstructured and unsupervised play. Fathers tend to focus their direct care on older children, particularly instruction in normative behavior, tool-making, and subsistence practice. Fathers will increase their time in direct care when mothers are absent or occupied or they have few other children to assist in childcare ).

Tanna
Tanna Island is an island in the archipelago of Vanuatu in the South Pacific. People on Tanna live in villages of fewer than 100 people, comprising several hamlets, with one chief leading each village. Each village is centered around a nakamal (meeting place) where kava is prepared and drunk by males in the village. This research was conducted in Western Tanna (approximately 20km inland) in multiple host communities that have welcomed our research projects with families and children for over a decade. Day-to-day life is similar to other rural and small-scale south Pacific islands in many ways-with a subsistence lifestyle, few publicly provided amenities, and a strong emphasis on community-oriented living. The homes are made of locally available materials such as bamboo and coconut leaves. At the time of this study, the rural villages did not have publicly funded electricity and the road to the villages made travel difficult during wet weather. Tannese children live a collective and group-oriented lifestyle with an emphasis on the success and well-being of the extended family and local village more than individual self-interest. One of the authors of this study has had a strong collaborative relationship with the communities since 2012 and has been discussing topics concerning child development and parenting-and our strange methods for examining these issues-throughout the course of the research program. It should be noted that the SSP was discussed extensively with our collaborators on Tanna prior to the start of this study.
Tannese mothers are typically the primary caregivers in the households in this society, providing full-time care to the children with the help of other relatives, community members, and children (Broesch et al., 2021). It is common for child siblings to care for each other, with smaller infants and children usually within earshot of the mother. Fathers play a secondary role in caregiving with mothers as the primary caregivers. When fathers are engaged with their children, they attend to and interact with infants in ways that resemble mothers around the world (see Broesch & Bryant, 2018). Fathers play a special role in the informal education of their sons, providing instruction in house-building, clearing bushes for gardening, traditional dances, and preparing kava.

Hypotheses
We first compared the patterning of child-father attachment on Tanna with the Tsimane and with two meta-analytic reviews of child-father attachment in large-scale, industrialized societies globally: a distribution of infant-father dyads (k = 12; N = 1,114; Madigan et al., 2021), and distribution of father-preschooler dyads (k = 6; N = 514; Deneault et al., 2022). We expected our child-father attachment patterns to be similar across Tanna, the Tsimane, and the two distributions from industrialized populations (Hypothesis 1). However, we anticipated variation with respect to child gender. The Tsimane recognize bilateral inheritance and have little wealth to transmit to children, while the Tannese recognize patrilineal inheritance in which primarily sons inherit the land and other valuable property. To the extent patrilineal inheritance associates with fathers investing more time in their sons from an early age, it is more likely we find a son-bias in child-father attachment in Tanna, relative to the Tsimane (Hypothesis 2). A recent study found Tsimane fathers engaged in the same amount of verbal interaction with daughters and sons (Cristia et al., 2017).
Through maternal interviews, we next compared child-father attachment to attributes of fathers that may impact the amount or quality of childcare they provide: how frequently the father plays with their children, father's income, father's years of formal education, and father's informal political influence in the community. First, we expected that how frequently fathers play with their children positively associates with secure child-father attachment because the measure of play is most proximate to behavioral antecedents of secure attachment previously identified in the literature (Hypothesis 3). Second, we anticipated that the children of fathers who have greater status due to their income, formal education, or political influence within their community would be less likely to have a secure child-father attachment (Hypothesis 4). Time engaged in the pursuit or maintenance of status may directly trade-off with time spent in childcare, the status may beget greater bargaining power within marriages over domestic tasks, or individuals less invested in their families may self-select into status pursuit.
Finally, consistent with the literature in "Western" societies (Brown et al., 2010;Coley & Schindler, 2008;Goldberg & Carlson, 2014;Lamb, 2010), we anticipate that lower quality of fathers' relationship with the mother-as reported by mothers-will be linked to reduced likelihood of secure child-father attachments (Hypothesis 5).

Participants
We tested 92 fathers with their 1-to 3-year-old children (49 female and 43 male): 63 Tannese (24 males) and 29 Tsimane (19 males). Tanna families were recruited from 15 different villages while Tsimane families were recruited from one more market-integrated village. Child participants on Tanna were significantly younger compared with the Tsimane (16.1 months vs. 22.2 months); t(90) = 4.137, p < .000, and child participants on Tanna included a greater percentage of girls (62%) relative to the Tsimane participants (34%). Combining the samples, fathers were on average 31.2 years of age, had achieved 5.0 years of formal education, earned US$253 (USD) in annual income, and had 1.9 additional children in the household, with substantial variation (see Table 1). Fathers did not differ in age between Tanna and the Tsimane, t(80) = -.880, p = .382. Tannese participating fathers had attained fewer years of formalized education than Tsimane participating fathers, t(88) = 4.339, p = .016, earned less income, t(83) = 3.768, p < .001, and had fewer children in their household, t(88) = 2.857, p = .005.

General Procedure
In June and July 2015, we traveled to Bolivia and Vanuatu and obtained permission to conduct the study in each region with local political representatives and governing organizations. Village meetings were also held to discuss the study and solicit interest. We discussed our interest in better understanding fathers and their young children and described the study methods and design with local women leaders (Tanna). We recruited fathers and their children to participate in the study by word of mouth throughout the villages. The experimenter was a local adult who spoke in the local language to the fathers and children. The local experimenters in each region worked closely with and were trained and supervised by the authors of the study. The study was conducted in empty houses in the participating communities. Once informed consent was obtained by the local experimenter, the entire interaction was video recorded using two separate cameras. The cameras were placed in inconspicuous locations to not draw attention-in two corners of the room. The child-father dyad participated in an adapted separation-reunion procedure, which is a hybrid between the SSP (Ainsworth et al., 1978) and the separation-reunion procedure (Preschool Attachment Coding System [PACS]; Cassidy et al., 1992). In this adapted version, the separations were shorter (2 min) because pilot testing with 12 Tannese children in 2014 indicated that longer separations induced considerable distress in most cases. No strangers were included in the study design, because available strangers were the study authors, and we anticipated the ethnic differences would provoke increased anxiety and distress in the children. Furthermore, the use of stranger anxiety to test attachment quality to caregivers may be prejudicial in traditional cultures where interaction with strangers is rare, particularly for infants (Keller, 2018). During the childfather attachment episodes, we recruited the child participant's mother to interview her about paternal investment and other attributes of the father (details found in supplemental online material [SOM]).

Attachment Procedure
There were a series of five separation and reunion episodes, modeled on the procedure by Cassidy and Marvin for preschool-age children (Cassidy & Marvin, 1992). First, child-father dyads were invited into the testing location where we provided a series of age-appropriate toys (such as a dinosaur, a ball, some wooden building sticks, a plush finger puppet, a car, and a wooden farm animal set). Although the toys were not culturally relevant toys as they were manufactured items, they were not flashy and represented a simplistic set of stimuli that was transportable across cultures and comparable to those used in other studies. Fathers did not receive any specific instruction as to how the dyad should interact (see SOM for instructions). Second, following these 5 min of free play interaction, the local experimenter signaled the father to leave (by knocking on the door of the testing room) for a 2-min separation. Third, after the separation, the father returned to the testing room with the child for a 5-min reunion. Fourth, the father and child were then separated a second time for 2 min. Fifth, the father and the child were reunited for another 5 min. In 46% and 58% of the first and second separations, respectively, the child expressed considerable distress before the end of the separation period, so those particular separations were shortened. Following each separation, the local experimenter instructed the parent to reunite with the child for free-play interaction but did not provide further instruction on how to interact with the child. For a systematic review of this procedure, see O'Neill et al. (2021).

Attachment Coding
The quality of the child-father attachment relationship was assessed using the adapted procedure described above. Attachment classifications for infants (<18 months old) were based on the coding system developed by Ainsworth and her colleagues (1978). The reliability and validity of the Strange Situation have been widely outlined (Ainsworth et al., 1978;Lyons-Ruth & Jacobvitz, 2008;Solomon & George, 2008). Although the validity of the attachment procedure and coding system has been largely derived from Western samples, studies conducted in non-Western contexts suggest that similar attachment patterns can be identified across various cultures and contexts (Mesman et al., 2016;van IJzendoorn & Sagi, 2001). The Strange Situation has been validated by fathers (Umemura et al., 2013;van IJzendoorn & De Wolff, 1997). Recent meta-analytic results of child-father attachment in infancy  including 12 samples (1,114 children) showed a similar distribution for child-father dyads as compared with child-mother dyads. Attachment classifications for early preschoolers were coded according to the PACS (Cassidy et al., 1992). Validity of the PACS has also been demonstrated in recent systematic reviews exploring associations with parental sensitivity , parental well-being (Badovinac et al., 2018), and child social adaptation . The PACS also demonstrated validity through associations with family interactions observed at home (National Institute of Child Health & Human Development Early Child Care Research Network [NICHD], 2001) and triadic play with parents . A recent meta-analysis (Deneault et al., 2022) of 97 samples (8,186 children) revealed a distribution of attachment classifications strikingly similar to the infant system. Although most research comes from Europe and North America, this review also includes data from Africa, Asia, the Middle East, and Australia/New Zealand. The analysis did not reveal strong evidence of variability in distributions based on geographical locations. Finally, the PACS has been specifically validated with fathers as well (Bureau et al., 2017). The meta-analysis by Deneault and colleagues (2022) covering 6 samples (514 children) suggested that child-father attachment classification shows a similar distribution as child-mother attachment. In the current study, coding was performed several months after videos were recorded by a trained coder blind to our research hypotheses. While coding involved both main and sub-classifications of attachment (see SOM), estimates of interrater reliability are based on the main classifications.

Coding of Infants
Securely attached (B) infants respond to the father's return by demonstrating a desire for proximity and contact, whereas insecure-avoidant (A) infants avoid proximity or interaction upon reunion with their fathers, and insecure-ambivalent (C) infants demonstrate angry resistant behavior toward their fathers upon reunion (Ainsworth et al., 1978). Infants classified as insecure-disorganized (D) exhibited incoherent, conflicted, or contradictory behaviors in the presence of their father (Main & Solomon, 1986). Roughly 19% of cases (n = 10) in this subsample were randomly selected to be double-coded, resulting in 80% agreement (kappa = 0.62) for the four-way classification (B, A, C, and D).

Coding of Children
In the PACS, children are classified as secure, avoidant, ambivalent, or behaviourally disorganized (Moss et al., 2004). Secure (B) children interact in a calm and comfortable manner upon reunion with their father. Avoidant (A) children preserve neutrality in their interaction with the father and exhibit little emotional expression toward him. Ambivalent (C) children display immature and/or resistant behavior toward their father and contact with the father does not seem to satisfy their need for comfort. Finally, behaviorally disorganized (D) children do not demonstrate a coherent attachment strategy and/or exhibit signs of disorganization according to infancy criteria. There are three subtypes to the (D) classification: Controlling-caregiving children appear excessively caring, concerned, and helpful in their interaction with the father whereas controlling-punitive children show hostility toward the father by giving harsh orders or blatantly ignoring the father. Controlling-mixed children either oscillate between caregiving and punitive behavior or appear overly controlling without demonstrating a clear punitive or caregiving pattern.
Despite differences in the coding criteria from infancy to the preschool system to accommodate for development, the function of attachment behaviors remains similar. Even the new controlling categories emerge from the disorganization pattern as they are thought to be child attempts to prevent frightening parental behavior, contributing to disorganization. Recent metaanalytic work showed significant stability from infancy to the preschool years (Opie et al., 2021) as well as similar attachment classification distribution (Deneault et al., 2022).
Roughly 41% of cases (n = 16) in this subsample were randomly selected for interrater reliability, resulting in 87.5% agreement (kappa = 0.80) for the four-way classification (B, A, C, and D/controlling).

Interview
General Procedure. While the father and the child participated in the separation-reunion procedure, either another local experimenter (Tanna) or one of the study authors versed in the local language recruited the mother of the child for an interview. The interview was done in private, and mothers were informed that their answers would remain confidential.
Measures. We asked mothers to report the ages and years of formal education for the father and all children in the household. For Tsimane, these responses were validated with existing databases based on years of demographic interviews in the community. Mothers also reported father's income over the past year, broken down by income source. In the Tsimane sample only, we validated mothers' reports of fathers' income by comparing them to fathers' reports of their own income; they show a significant correlation, r(25) = .656, p < .001. Mothers were also asked to nominate five influential men in the community. A father who received any nomination was considered influential.
As part of the interview, mothers used a 5-point Likert-type scale to assess how frequently fathers play with children. Mothers used additional 5-point Likert-type scales to express how happy they are in their current relationship with the father, and whether they are currently angry with the father. Mothers also reported the number of children in the household and the number of adults outside the household who provide help in childcare. For the list of interview questions, see SOM.
Missing data from the mother interview resulted from mothers indicating that they did not know the answer or they did not respond. Missing data was minimal for a given measure, with the exception of father's age, which we lack for 10 of the Tannese fathers.
Translations. All scripts, consent, and interview materials were first translated into the local language. They were then back-translated from the local language into English (on Tanna) or Spanish (for the Tsimane) by a local individual to ensure that the meanings were the same. Any discrepancies were discussed and resolved by the experimenter and field research assistants. During interviews, mothers were verbally tested on their comprehension of the Likert-type scale response options before their answers were solicited.

Analysis
Statistical tests and graphical representations were produced using IBM SPSS Statistics 26. To test hypothesis 1, we compared distributions of attachment categories across samples using a proportion test. We opted for Fisher's exact test in lieu of chi-square since expected values were low for three of the cells. To facilitate tests of our other hypotheses, we transformed our attachment measure into a binary variable (securely attached=1, all else=0). Dichotomizing the attachment measure in this way accords with previous research using the attachment classifications for infants (Brown et al., 2012;Demers et al., 2010;Stupica et al., 2011) and research using the PACS (Bureau et al., 2017;Greenberg et al., 2001). Analyses of attachment quality were then run using binary logistic regression.
Prior to the analysis, we transformed the income and formal education measures into quintiles within each country, to facilitate comparison across countries and to handle skewness in the continuous measures. We then performed an exploratory factor analysis using principal components extraction of the income, formal education, and political influence measures, which yielded a single component explaining 51% of the variation in the data (See SOM for loadings). We labeled this component "status." We also performed an exploratory factor analysis using principal components extraction of the measures of mother's happiness with father and mother's anger toward father, which yielded a single component explaining 56% of the variation in the data (See SOM for loadings). We labeled this component "mother-father relationship."

Attachment Patterns
We first examined whether the patterning of child-father attachment was similar across cultural contexts (Hypothesis 1): Tannese, the Tsimane, a meta-analytic distribution of infant-father dyads (k = 12; N = 1,114; Madigan et al., 2021), and a meta-analytic distribution of father-preschooler dyads (k = 6; N = 514; Deneault et al., 2022). A comparison of the four samples (see Figure 1) reveals a significant difference in the distribution of the attachment quality categories, p < .001 (Fisher's exact test). However, Bonferroni-corrected post hoc tests show this difference is not found within all attachment categories but rather is driven by differences in insecureambivalent attachments: relatively more in the Tanna sample (p < .001) and in the Tsimane There is no significant difference between the Tanna and Tsimane samples in the patterning of child-father attachment, p = .353 (Fisher's exact test). In the Tanna and Tsimane samples, 62% and 52% of children were identified as securely attached, respectively. In the Tanna and Tsimane samples combined, 54 children (59%) were identified as securely attached, 28 (29%) as insecure-ambivalent, 9 (11%) as disorganized, and 1 (1%) as insecure-avoidant (see Table 2). When transformed into a binary measure, attachment quality still does not differ between the Tanna and Tsimane samples: in a binary logistic regression model of attachment quality (N = 92), attachment quality does not differ by cultural sample (Tsimane=0, Tanna=1), odds ratio (OR) = 1.517, 95% confidence interval (CI) = [0.624, 3.687], p = .358.
In the combined Tanna and Tsimane samples, the binary attachment measure does not vary with the age of the child (N = 92), OR = 0.970, 95% CI = [0.915, 1.028], p = .303 nor gender (female = 0, male = 1) of the child (N = 92), OR = 0.727, 95% CI = [0.315, 1.678], p = .455. To test whether Tannese fathers show more son bias in attachment relative to Tsimane fathers (Hypothesis 2), we then examined whether the effect of child gender interacts with society (see Figure 2). In a binary logistic regression model of attachment security (N = 92),

Father Attributes
We next examined associations among the binary attachment measure and attributes of fathers. We anticipated father's frequency of play with his children positively associates with childfather attachment security (Hypothesis 3). A binary logistic regression model controlling for cultural sample (Tsimane= 0, Tanna= 1), father age, child age, and child gender (N = 80) finds father's play positively predicts child-father attachment security, OR = 1.730, 95% CI = [1.043, 2.867], p = .034. We also tested whether fathers who have higher social status had a less secure attachment to their children (Hypothesis 4). A binary logistic regression model controlling for cultural sample, father age, child age, and child gender (N = 72) suggests father's status does not associate with child-father attachment security, OR = 1.065, 95% CI = [0.639, 1.774], p = .810.
Finally, we tested whether mother-father relationship quality (greater happiness and less anger reported by the mother with regard to the father), associates with child-father attachment security (Hypothesis 5). A binary logistic regression model controlling for cultural sample, father age, child age, and child gender (N = 79) suggests that mother-father relationship quality does not associate with child-father attachment security, OR = 1.084, 95% CI = [0.663, 1.772], p = .747. This non-significant relationship does not appear to be due to lack of external validity to the mother-father relationship measure. Bivariate correlations suggest motherfather relationship quality positively associates with mother reports of how frequently the father plays with his children and negatively associates with childcare burden, as indexed by the number of other children in the household and the number of childcare helpers from other households (see Table 3).

Discussion
We examined patterns of child-father attachment in two rural, indigenous societies located in Bolivia (the Tsimane) and in Vanuatu (Tannese). Using a modified strange situation procedure, we tested fathers with their young children (1-to 3-year-olds) using a series of separation and reunion episodes. We coded for attachment quality using infant and PACS coding systems. Our analyses indicate that these two societies did not differ from one another in the distribution of attachment categories. Furthermore, the overall patterning of attachment among Tsimane and Tannese showed similarity to child-father attachment distributions reported in two meta-analyses conducted on infancy  as well as preschool samples (Deneault et al., 2022). On the one hand, there were relatively more child-father attachments reported as insecure-ambivalent and almost zero reports of insecure-avoidant attachment in the Tanna and Tsimane samples. Our sample had relatively more insecure-ambivalent compared with insecureavoidant (compared to the meta-analysis, both infants and preschooler data). This difference is interesting (see Figure 1; insecure-avoidant and insecure-ambivalent) and there are several possible explanations for this difference. One possibility is that we are capturing cultural differences in the behavioral expression of distress to the SSP rather than a meaningful difference in the classification of the child's insecure attachment. A related possibility is a cultural variation in infants' and preschoolers' reactions when in distress, in general. On the contrary, the relative frequency of secure attachment was strikingly similar across the two meta-analyses and the Tanna and Tsimane samples. In general, our results give less credence to suggestions that secure attachment may not be adaptive in all cultures (Simpson & Belsky, 2016), and accord with claims that caregiving crossculturally shows "universality without uniformity" (Mesman et al., 2018). Attachment theory has been critiqued as rooted in a Western middle-class family model, in which family sizes are small, social interactions tend to be dyadic, and communication with children includes a considerable focus on their inner mental states (Keller, 2013). In contrast, Tanna and Tsimane social organization is focused on extended households of kin, in which young children experience social relationships with multiple social partners and come to anticipate less than continuous responsiveness to their demands. Nevertheless, secure attachment to primary caregivers, including to fathers, can emerge across such diverse parenting norms. One must not equate universal behavior with universal processes (Kline et al., 2018). It is possible that the cultural pathway to the same attachment classification may be quite variable both between and within cultures. Therefore, more work is needed to help us better understand the process of attachment in diverse cultural settings.
We did predict and find one significant source of variation in child-father attachment, comparing Tanna and Tsimane. Tannese sons were more likely to be securely attached to their father compared with daughters, and compared with Tsimane sons. It is possible that cultural variation in the security of child-father attachment by child gender reflects cultural variation in inheritance norms. The Tannese who participated in the study tend to follow patrilineal inheritance norms, in which sons inherit land and other valuable property. The Tsimane do not restrict inheritance by gender and furthermore transmit little material wealth at all. In societies characterized by patrilineal inheritance, fathers may have a greater incentive to directly invest in their sons more than in their daughters, with consequences for attachment to even their young children.
We also evaluated several attributes of fathers, which may associate with child-father attachment security. These measures were derived from interviews with mothers. We anticipated and found that frequency of father's play with their children, in both societies, positively associates with secure child-father attachment, while measures of father's social status and father-mother relationship quality bore no relationship to child-father attachment.
Play may be particularly important to child-father attachment if it constitutes a larger fraction of fathers' time in direct childcare, relative to mothers and also if father play tends to have unique effects on children's social development. In Western settings, fathers' engagement in physical, often "rough-and-tumble," play with children has been argued to have unique effects on the child-father bond and on aspects of children's social development such as confidence in novel social settings (Abraham & Feldman, 2018;Grossmann et al., 2008;Lamb, 2010). The effect of father's play is also noteworthy in that fathers cross-culturally tend to dedicate much more time to resource provisioning. In the Tsimane, for example, fathers provide only 7% of the direct care received by their children during the first 6 years of a child's life ) yet provide a majority of the calories children consume (Hooper et al., 2015). Thus, it is perhaps surprising that, in our study, mother-father relationship quality associates at all with fathers' play with children. One possibility is that mothers anticipate benefits to their children's social development from child-father play. Because father-mother relationship quality did not associate with the security of child-father attachment, a complementary if not more likely possibility is that father's play with their children signals to mothers the likelihood of fathers' continuing provisioning of resources to the household.
Our study makes a unique contribution to understanding the roles of caregivers in relation to child-father attachment. Few studies have examined fathers' attachment with their children and its determinants in small-scale, non-industrialized societies. However, we note several limitations. First, although we modified the strange situation protocol to remove the stranger from the procedure, the children still seemed quick to distress. The strange situation paradigm of leaving a child is quite unusual in the Tsimane and in Tanna. To compensate, we shortened the amount of time the child was left alone. Furthermore, the strange situation provides only one measure of the quality of child-father attachment and the child-father relationship more generally. Although there is a body of literature supporting the idea that this measurement tool transcends ages and family styles, there is very little evidence using this paradigm in small-scale, non-Western, traditional societies where individuals live more inter-dependently in dense kin networks. This could also be seen as a strength of our study, as the patterning of child-father attachment did not differ in any large way from comparable research conducted in urban, Western samples. Finally, our interview data are based on maternal perceptions of paternal behavior rather than asking the fathers themselves. Although this is a clear limitation of the study, given our work in these regions, we felt that this procedure would more accurately reflect what fathers are doing with their children and how they are contributing to the family.
In the future, we recommend the evaluation of several measures of child-father relationship quality. As important, we need more studies on child-father attachment, its behavioral and physiological determinants, and its relationship with child outcomes across diverse cultural settings (Bakerman-Kranenburg et al., 2019;Mattison et al., 2019). Our research shows an association between mothers' report of the frequency of father-play and child-father secure attachment. We suggest further exploration of this association to determine precisely which features of childfather relationships are supporting secure attachment across societies. We suspect that our measure of "father-play" reflects a commitment to childcare by the father yet this needs to be explored. A deeper understanding of the features of father-play will aid in developing more comprehensive recommendations to fathers, irrespective of cultural background and social circumstances. A more systematic theory of child-father attachment will explain the full range of cross-cultural variation in paternal care of offspring and predict the effects of such variation on child outcomes in particular cultural contexts.