Introduction
Life history (LH) theory posits that adverse childhood environments marked by high and variable threats of death and disability, stemming from external factors like predation, accidents, violence, and infectious diseases, are associated with a fast or faster LH strategy (Ellis et al., Reference Ellis, Shakiba, Adkins and Lester2009). This strategy prioritizes mating and reproduction over efforts to mitigate mortality and improve the living environment. Conversely, safe and stable childhood living environments lead to a slow or slower LH strategy and related behaviors (e.g., disease control effort; Chang et al., Reference Chang, Liu, Lu, Lansford, Bornstein, Steinberg, Deater-Deckard, Rothenberg, Skinner and Dodge2021) that aim at reducing and improving the external mortality conditions at the expense of delayed reproduction. Using different terminologies, Belsky et al. (Reference Belsky, Steinberg and Draper1991) first identified this environment-LH contingency with human development (Also see Belsky, 2012, 2019). However, alternative perspectives, such as internal LH models (Nettle and Bateson, Reference Nettle and Bateson2015; Nettle et al., Reference Nettle, Frankenhuis and Rickard2013), propose that an individual’s internal body state may play a role in shaping LH strategies or regulating the impact of the external environment on LH. Notably absent in existing internal models is the consideration of caregiver–child attachment – a crucial internal system evolved specifically to address extrinsic mortality threats (e.g., predation; Bowlby, Reference Bowlby1969). The current study seeks to explore caregiver–child attachment as an internal mental state in the calibration and modulation of LH strategies. Similar to other internal models where early environments influence individuals’ internal somatic state, determining LH, in our internal mental state model, attachment is hypothesized to mediate and moderate the effects of external environmental factors on LH strategies.
Intrinsic somatic models of LH
LH theory distinguishes between intrinsic and extrinsic components of morbidity and mortality (Williams, Reference Williams1957). The intrinsic component refers to functional degradation stemming from decay of an individual’s internal system (physiological and psychological functions). Aging-related wear and tear of the body and mind, degenerative diseases (e.g., heart and kidney problems), and unhealthy habits (e.g., smoking) exemplify intrinsic mortality risks (Carnes et al., Reference Carnes, Holden, Olshansky, Witten and Siegel2006). Extrinsic morbidity and mortality refer to disability and death that befall an individual due to external and mostly uncontrollable factors such as predation, accidents, and infectious diseases. These two components of morbidity and mortality interact in bringing about an individual’s eventual demise (Carnes et al., Reference Carnes, Holden, Olshansky, Witten and Siegel2006), and the internal state and the external environment together determine LH strategies (e.g., Nettle and Bateson, Reference Nettle and Bateson2015; Nettle et al., Reference Nettle, Frankenhuis and Rickard2013). Research on human and nonhuman animals has documented the interplay between intrinsic and extrinsic conditions. For example, trematoda parasite (external environment) causes deadly diseases (internal state) that make marine snails mature faster (faster LH, Lafferty, Reference Lafferty1993), whereas great tits unaffected by parasite infestation engage in disease control effort and delay reproduction (slower LH, Oppliger et al., Reference Oppliger, Richner and Christe1994). In humans, extrinsic adversities (e.g., socioeconomic deprivation) compromise the intrinsic somatic state (e.g., chronic illness and poor health) that calibrates faster LH such as an early age of pregnancy (Waynforth, Reference Waynforth2012), early onset of menarche, tendency to engage in risky conduct (Hartman et al., Reference Hartman, Li, Nettle and Belsky2017), and reckless sexual and aggressive behavior (Chang et al., Reference Chang, Lu, Lansford, Bornstein, Steinberg, Chen, Skinner, Dodge, Deater-Deckard, Bacchini, Pastorelli, Alampay, Tapanya, Sorbring, Oburu, Al-Hassan, Di Giunta, Malone, Uribe Tirado and Yotanyamaneewong2019a; Ellis et al., Reference Ellis, Shakiba, Adkins and Lester2021).
Attachment as an internal mental regulator of LH
In these models, the intrinsic body state mediates extrinsic mortality conditions and calibrates LH either by initiating mortality-reduction strategies (e.g., disease control efforts) and slowing other aspects of life such as reproduction, or by accelerating reproduction and disregarding mortality threats. The resulting predicted adaptive LH tailors to the vitality status of the internal body state (Nettle et al., Reference Nettle, Frankenhuis and Rickard2013) and represents adaptive strategies in managing mortality threats of the external environment (Clutton-Brock, Reference Clutton-Brock1984). Extending these internal somatic models of LH calibration, we conceptualize the attachment system as a possible internal mental state in organizing and modulating LH. Similar to how physical state is involved in registering the external environment, the mental state of attachment mediates and regulates the maternally socialized environment and formulates LH (Chisholm, Reference Chisholm1996). Mammalian species first experience the external environment through interactions with their mothers or other primary caregivers (Bornstein, Reference Bornstein and Bornstein2019). Through innumerable caregiver–child interactions, the individual develops an internal working model as an internalized appraisal of the environment and of the self (Bowlby, Reference Bowlby1969) that subsequently guides, organizes, and automates behavior (Zimmermann, Reference Zimmermann1999). Evolved to protect from predation (Bowlby, Reference Bowlby1969), attachment and the internal working model are especially involved in processing extrinsic mortality information (Chisholm, Reference Chisholm1996) and the individual’s ability to cope with mortality threats (Lu et al., Reference Lu, Liu and Chang2022). “Henceforward, the two working models each individual must have are referred to respectively as his environmental model and his organismic model” (Bowlby, Reference Bowlby1969, p. 82). Other early writings similarly refer to attachment and the internal working model as “an organism’s capacity to interact effectively with its environment” (White, Reference White1959, p. 297), “the infant’s belief that its actions affect his environment” (Lewis & Goldberg, Reference Lewis and Goldberg1969, p. 82), “broadly conceived competence” (Arend et al., Reference Arend, Gove and Sroufe1979, p. 951), and “the ability to negotiate with the environment” (Cassidy, Reference Cassidy1986, p. 331). Similar to other internal LH models, attachment and the internal working model represent the internal mental state (vis-à-vis the physical state) that provides cognitive and perceptual information (vis-à-vis sensory and interoceptual information) concerning an individual’s cognitive (rather than physical) status. Accordingly, attachment and its internal working model calibrate LH strategies by attempting to reduce and outlive (slow LH) or disregard and outgrow (fast LH) extrinsic mortality risks.
Once formed, attachment operates outside consciousness (Chisholm, Reference Chisholm1996; Main, Reference Main, Parkes, Stevenson-Hinde and Marris1991). We propose two mechanisms to account for this operation. In one, attachment acts as an intermediary that conveys external information and engenders the species-general contingency between adverse environments and fast LH (Belsky et al., Reference Belsky, Steinberg and Draper1991; Chisholm, Reference Chisholm1996; Del Giudice & Belsky, Reference Del Giudice, Belsky, Buss and Hawley2011). In the other, attachment as an internal mental regulator has the additional effect of breaking away from the external contingency and generates new pathways primarily in the slow LH direction of mortality reduction (Lu et al., Reference Lu, Liu, J., Guo, Zhu, Chen, Lansford and Chang2021). In the first mediation process, a child inherits the same extrinsic mortality conditions from his/her caregiver and is rendered the same effects through caregiving behavior and the caregiver’s other LH manifestations. According to the pioneering work by Belsky et al. (Reference Belsky, Steinberg and Draper1991) and other LH researchers (e.g., Chisholm, Reference Chisholm1996; Del Giudice & Belsky, Reference Del Giudice, Belsky, Buss and Hawley2011; Simpson & Belsky, Reference Simpson, Belsky, Cassidy and Shaver2008), the actual environment and caregiver-mediated childhood experience should yield similar effects on LH calibration. Hence, attachment should (statistically) mediate the species-general coupling between the environment and LH. We postulate that a stable living environment is aligned with consistent caregiving, secure attachment, an internal working model that regards the world as predictable and the self as efficacious, and slow LH calibrations (Chisholm, Reference Chisholm1993, Reference Chisholm1996; Simpson & Belsky, Reference Simpson, Belsky, Cassidy and Shaver2008) aimed at mortality reduction. A secure internal working model inscribes the mortality-reduction mindset of slow LH that manifests through insight, planning, and control (Figuredo et al., Reference Figueredo, Jacobs, Gladden, Bianchi, Patch, Kavanagh, Beck, Sotomayor-Peterson, Jiang and Li2018; Thompson, Reference Thompson, Thompson, Simpson and Berlin2021). By contrast, fast LH strategies and a mortality-carefree mindset are associated with environmental harshness and unpredictability, neglectful and inconsistent caregiving, and an insecure internal working model rendering the individuals doubtful of their abilities and fearful and skeptical of the world (Belsky et al., Reference Belsky, Houts and Fearon2010; Chen & Chang, Reference Chen and Chang2012).
In the second regulatory (statistical moderation) process, the attachment system has the additional effect of buffering or underregistering rather than overregistering or amplifying environmental risks, and of undercalibrating rather than overcalibrating environmental adversities into fast LH strategies (Sung et al., Reference Sung, Simpson, Griskevicius, Kuo, Schlomer and Belsky2016). The overall net effect of the attachment system is expected to attenuate the coupling between extrinsic mortality risks and fast LH strategies (Lu et al., Reference Lu, Liu and Chang2022). Specifically, secure attachment should downregulate, and insecure attachment should maintain, the coupling between extrinsic mortality and fast LH. Therefore, the attachment system may direct children toward two separate developmental pathways. One is a slowing LH pathway wherein species-general coupling between extrinsic mortality and fast LH is downregulated by secure attachment and the related mortality-reduction effort. The other is a species-general LH pathway perpetuated by insecure attachment of individuals who continue to be shaped by environmental adversities into mortality-carefree fast LH strategists.
Our theorizing (statistical mediation and moderation) about attachment is supported by the empirical literature, which has mainly examined retrospective measures of the childhood environment in relation to concurrent measures of adolescent and adult attachment. For example, a retrospective questionnaire measure of early environmental predictability was positively correlated with secure adult attachment, which itself was positively correlated with long-term (rather than short-term) intimate relationships (Hill et al., Reference Hill, Young and Nord1994). Another retrospective study yielded similar findings that childhood environmental unpredictability was positively correlated with insecure adult attachment, and the latter was positively correlated with intimate partner violence consisting of psychological aggression, physical assault, and sexual coercion (Barbaro & Shackelford, Reference Barbaro and Shackelford2019). The same mediating effect was also reported in additional studies of relations between childhood adversity and fast LH profiles such as psychological distress, harmful drinking, and criminal thinking (Corcoran & McNulty, Reference Corcoran and McNulty2018; Le et al., Reference Le, Levitan, Mann and Maunder2018; Yang & Perkins, Reference Yang and Perkins2020). To a lesser extent, studies also show that attachment measures statistically moderate the relation between environmental adversity (family income-to-needs ratio, parental stress, maternal depression) and fast LH manifestations (early onset of menarche, aggressive behavioral problems, internalizing behavior; Sung et al., Reference Sung, Simpson, Griskevicius, Kuo, Schlomer and Belsky2016; Tharner et al., Reference Tharner, Luijk, van IJzendoorn, Bakermans-Kranenburg, Jaddoe, Hofman, Verhulst and Tiemeier2012; Whittenburg et al., Reference Whittenburg, Stern, Brett, Straske and Cassidy2022).
Present study
Here we tested the statistical mediating and moderating hypotheses about attachment in relation to environmental adversity and fast LH behavioral profiles (Figure 1) in a longitudinal and cross-cultural sample consisting of 1,169 adolescents and their primary caregivers from nine countries. Information about childhood environmental adversity, which was indicated by three proxies, namely unsafe neighborhoods, chaos in the home, and unpredictable life events, was obtained from the adolescents and one parent when the adolescents were 10 years old on average. However, data collection began when the children were 8 years old, and family environments remained stable throughout the subsequent data collection phases. Secure attachment was measured based on the adolescents’ self-reports and reports from both parents when the adolescents were on average 14 years old. Fast LH behavioral profiles consisting of aggression, impulsivity, and risk taking were obtained from adolescents when they were 17 years old. Structural equation modeling was conducted to test relations among these variables with a focus on attachment as both a mediator and a moderator. Attachment was hypothesized to mediate the species-general coupling between environmental adversity and fast LH behavioral profiles. In testing the statistical moderation of attachment, we expected an attenuated association between environmental adversity and fast LH behavioral profiles for higher levels of secure attachment and we expected the strength of the association to be unchanged at lower levels of secure attachment.
Figure 1. Childhood environmental adversity, secure attachment, and their interaction in relation to fast life history behavioral profile. ***p < .001.
Method
Participants
Data for the present study were drawn from an ongoing longitudinal study that originally recruited children, their mothers, and their fathers in 2008 when the children were 8 years old on average. Families were sampled from 10 cities in nine countries: Shanghai, China (n = 101), Medellín, Colombia (n = 100), Naples, Italy (n = 95), Rome, Italy (n = 99), Zarqa, Jordan (n = 112), Kisumu, Kenya (n = 95), Manila, Philippines (n = 100), Trollhättan/Vänersborg, Sweden (n = 95), Chiang Mai, Thailand (n = 100), and Durham, North Carolina, United States (n = 101 European Americans, n = 94 African Americans, n = 77 Latin Americans). Participants were recruited from schools and communities. Most parents lived together (82%), and were biological parents (97%); nonresidential and non-biological parents also provided data. Sampling included families from each country’s majority ethnic group, except in the United States, where we sampled equal proportions of White, Black, and Latino families. Families from different socioeconomic backgrounds were sampled in proportions representative of each recruitment area. The present sample consisted of 1,169 adolescents (51% girls), their mothers (n = 1,150), and their fathers (n = 1,048). In the last data collection wave of the present study, Time 3, the adolescents were 17 years of age on average (M = 17.27 years, SD = .63). They were 10 (M = 10.29 years, SD = .65) at Time 1 and 14 years old on average (M = 14.28 years, SD = .63) at Time 2 of the present study. At Time 3 of the present study, 78% of the initial sample at Time 1 were with the study eight years later. Participants who provided complete data across the eight years did not differ from the initial sample with respect to adolescent gender, parent marital status, education, and all the substantive variables used in the study. Adolescent age and gender did not vary across sites.
Interview procedures
The primary data collection method for the present study was face-to-face interviews. Interviews lasted 1.5 to 2 hours at each of the three times of data collection and were conducted in participating adolescents’ homes, schools, or at other locations chosen by the participants. Interviews and procedures were approved by the XX University Institutional Review Board (IRB, protocol number 2032) and by the local IRBs at universities in each participating country. Mothers and fathers provided written informed consent, and adolescents provided assent. Family members were interviewed separately to ensure privacy. Measures used in the interviews were administered in the official language of each country, following forward- and back-translation of all instruments. For the present study, adult participants were given the choice of completing the measures in writing or orally, with the interviewer reading the questions aloud and recording the participants’ responses (with a visual aid to ensure that participants understood the response scales). At Time 1, the adolescents, then children, were administered the measures orally, and for the two subsequent assessments they were given the option of completing the measures orally or in writing. Parents completed the questionnaire measures in writing. To thank them for their participation, adolescents were given small gifts or monetary compensation, parents were given modest financial compensation, families were entered into drawings for prizes, and modest financial contributions were made to adolescents’ schools.
Measures
Environmental adversity measured at time 1 when adolescents were 10 years old
In the empirical human LH literature, levels (harshness) and variations (unpredictability) of extrinsic morbidity and mortality are measured by sampling proxies from the current living environment that are believed to cue adverse environmental conditions of the ancestral past (Young et al., Reference Young, Frankenhuis and Ellis2020). Environmental harshness and unpredictability are both predictive of LH in the same direction (Lu et al., Reference Lu, Liu and Chang2022), but cues of unpredictability appear to be stronger predictors than cues of harshness (Hartman et al., Reference Hartman, Sung, Simpson, Schlomer and Belsky2018). Thus, we sampled more unpredictability cues and combined the two kinds of cues to form a single construct of environmental adversity. Three indicators, unsafe neighborhood, chaos in the home, and unpredictable life events, form the environmental adversity construct.
Unsafe neighborhoods
Mothers and children separately reported on the 7-item questionnaire measuring the perceived safety and livability of a neighborhood (Murray & Greenberg, Reference Murray and Greenberg2006; e.g., “My neighborhood is a dangerous place to live,” “My neighborhood is a nice place to live” (reverse coded), and “I feel scared in my neighborhood”). Using a 4-point scale ranging from 0 = “almost never true” to 3 = “almost always true,” the items were measured or recoded in the direction of unsafe neighborhood. Internal consistency reliability estimates were .86 for mother reporting and .77 for child reporting. The correlation between the two ratings was .42. For the structural equation modeling and other analysis reported later, the average of the two ratings was used as an indicator of environmental adversity.
Chaos in the home
We adopted 5 items from the Confusion, Hubbub, and Order Scale (Matheny et al., Reference Matheny, Wachs, Ludwig and Phillips1995) to measure confusion, chaos, and disorder in the home (e.g., “It’s a real zoo in our home,” “The atmosphere in our home is calm” (reverse coded), and “You can’t hear yourself think in our home”). Mothers and children responded to these questions on a 5-point scale ranging from 1 = “definitely untrue” to 5 = “definitely true.” Internal consistency reliability estimates were .67 for mothers and .61 for children. The correlation between the two ratings was .34. In subsequent analyses, the average of the two ratings formed an indicator of environmental adversity.
Unpredictable life events
Using the Social Readjustment Rating Scale (Holmes & Rahe, Reference Holmes and Rahe1967), mothers reported on whether 10 unpredictable negative life events happened in the last 2 years in the family to which the child was likely to be exposed (e.g., “severe and/or frequent illness,” “accidents and/or injuries,” and “death of other important person”). The 10 items were summed to create a scale. Internal consistency reliability estimate was .65.
Secure attachment measured at time 2 when adolescents were 14 years old
Ratings from both parents and from the children provided three indicators of secure attachment.
Parent ratings
We employed a single-item measure of secure attachment, similar to the widely used measure of adolescent and adult romantic attachment by Hazan and Shaver (Reference Hazan and Shaver1987). Hazan and Shaver (Reference Hazan and Shaver1987) adapted Ainsworth et al.’s (Reference Ainsworth1978) verbal description of secure attachment to better suit the adult population. In our study, we utilized Ainsworth et al.’s (Reference Ainsworth1978) original description, separately asking both parents to rate, on a 7-point scale ranging from “1 = not at all fit (0%)” to “7 = complete fit (100%),” the extent to which their child fits the description of secure attachment.
Child rating
The Security Scale (Kerns et al., Reference Kerns, Klepac and Cole1996, Reference Kerns, Tomich, Aspelmeier and Contreras2000) originally has 15 but also uses 8 self-report items (Kerns et al., Reference Kerns, Schlegelmich, Morgan, Abraham, Kerns and Richardson2005; Pauletti et al., Reference Pauletti, Cooper, Aults, Hodges and Perry2016). It adopts Harter’s format to measure children’s perceptions of a secure attachment relationship with a parent. We used 8 items to measure children’s relationship with their mother or primary caregiver. Each item includes two statements, and participants first decide which statement better describes them and then rate the statement on two scales of “sort of true for me” or “really true for me.” Sample items include “Some kids worry that their mom might not be there when they need her – Other kids are sure their mom will be there when they need her,” and “Some kids find it easy to trust their mom – Other kids are not sure if they can trust their mom.” The rating scale for each item was converted to 1 to 4, with higher scores indicating greater secure attachment. Internal consistency reliability estimate was .77.
Fast LH behavioral profiles measured at time 3 when adolescents were 17 years old
Aggressive and exclusive sociality that is adaptive in a precarious environment to address immediate survival concerns aligns with fast pace of life (Chang et al., Reference Chang, Lu, Lansford, Skinner, Bornstein, Steinberg, Dodge, Chen, Tian, Bacchini, Deater-Deckard, Pastorelli, Alampay, Sorbring, Al-Hassan, Oburu, Malone, Di Giunta, Uribe Tirado and Tapanya2019b; Figueredo et al., Reference Figueredo, Jacobs, Gladden, Bianchi, Patch, Kavanagh, Beck, Sotomayor-Peterson, Jiang and Li2018), and impulsivity and risk preference covary in predictable ways with fast LH traits (Sear, Reference Sear2020). Thus, we used these three indicators to form the fast LH behavioral profiles construct.
Aggression
Fathers and mothers completed 20 items of the Child Behavior Checklist (CBCL; Achenbach, Reference Achenbach1991) to measure aggression (e.g., “argues a lot,” “gets in many fights,” and “threatens people”). A 3-point scale ranging from 0 = “never” to 2 = “often” registers the frequency an adolescent engaged in each of these behaviors. Internal consistency reliability estimates were .88 and .87 for fathers and mothers, respectively. The correlation between the two ratings was .56, well justifying averaging the two parental ratings to form the aggression indicator.
Impulsivity
Adolescents completed an 8-item scale of impulsivity selected from the 30-item Barratt Impulsiveness Scale (Patton et al., Reference Patton, Stanford and Barratt1995; Steinberg et al., Reference Steinberg, Sharp, Stanford and Tharp2013). Sample items include “I do not pay attention” and “I plan what I have to do (reverse coded).” The items were rated on a 4-point scale ranging from 1 (never true) to 4 (always true). The internal consistency reliability estimate was .67.
Risk preference
Following the literature (e.g., Duell et al., Reference Duell, Steinberg, Chein, Al-Hassan, Bacchini, Lei, Chaudhary, Di Giunta, Dodge, Fanti, Lansford, Malone, Oburu, Pastorelli, Skinner, Sorbring, Tapanya, Uribe Tirado and Alampay2016), we adapted a self-report measure of risk preference (Benthin et al., Reference Benthin, Slovic and Severson1993). Adolescents were asked about the following nine scenarios involving risky behavior: smoking cigarettes, drinking alcohol, vandalizing property, going to dangerous places, riding in cars with drunk drivers, having unprotected sex, stealing from stores, engaging in gang fights, and using weapons to threaten someone. For each scenario, adolescents rated two questions on 4-point scales: “How would you compare the benefits of this activity with the risks?” (1 = the risks are far greater than the benefits; 4 = the benefits are far greater than the risks), “If something bad happened because of this activity, how serious would it be?” (1 = not at all serious; 4 = very serious). The average of the two ratings for the nine scenarios formed the construct, with a higher score indicating a greater degree of risk preference. The internal consistency reliability estimate was .88.
Analytic strategy
Because we used latent constructs rather than directly observed variables, we conducted structural equation modeling using Mplus 7.0 (Muthén & Muthén, Reference Muthén and Muthén1998–2012), and we used full information maximum likelihood estimation to account for missing data (Schafer & Graham, Reference Schafer and Graham2002). We adopted the following recommended cutoff values to assess model fit: chi-square to degrees of freedom ratio (χ2/df <5.0; Kline, Reference Kline1998), comparative fit index (CFI ≥ .90; Marsh et al., Reference Marsh, Balla and McDonald1988), Tucker-Lewis index (TLI ≥ .90; Marsh et al., Reference Marsh, Balla and McDonald1988), root mean squared error of approximation (RMSEA ≤ 0.08; Browne & Cudeck, Reference Browne, Cudeck, Bollen and Long1993), standardized root mean square residual (SRMR ≤ 0.08; Hu & Bentler, Reference Hu and Bentler1999), and minimum factor loading (loading > .32; Tabachnick & Fidell, Reference Tabachnick and Fidell2013).
When testing an interaction or moderation model, the two variables forming an interaction or moderation (environmental adversity and secure attachment) are typically treated as correlated with unspecified causal directions. Because we also formulated a mediation hypothesis involving attachment as the mediator, in the model presented in Figure 1, environmental adversity and secure attachment were specified as one leading to the other rather than as two correlates. Statistical estimation of the interaction or moderation effect remains the same whether the two main effect variables, environmental adversity and secure attachment, represent directional or nondirectional relations. Our moderation hypothesis concerns the directional association of environmental adversity leading to fast LH behavioral profiles just as when the interaction or moderation is formed by environmental adversity and secure attachment as two correlates. We computed the interaction construct by using the default approach of Mplus rather than manually pairing the indicators of the two constructs and multiplying them (Marsh et al., Reference Marsh, Wen and Hau2004). The Mplus default approach does not provide goodness-of-fit statistics (Maslowsky et al., Reference Maslowsky, Jager and Hemken2015; Muthén & Muthén, Reference Muthén and Muthén1998–2012). Instead, Mplus provides a measure, D, of relative fitness of the interaction model compared to the main-effect-only model without the interaction term. D is the difference of the log-likelihood values of the two models (D = −2 × [(log-likelihood for the main effect model) − (log-likelihood for the interaction model)]; Muthén & Muthén, Reference Muthén and Muthén1998–2012). D follows a chi-square distribution with DF being the difference in the number of estimated parameters between the two models, which, in the present case, was 1.
Results
Table 1 presents the Ms, SDs, and correlations of all the variables used in the study. The correlations were based on different informants (i.e., adolescents, mothers, and fathers), over time lags of up to 8 years, and across diverse cultural groups. They showed good convergent and discriminant validity with mono-trait measures more highly correlated with each other than with hetero-trait measures. Inter-trait correlations were also in expected directions, with indicators of environmental adversity (e.g., unpredictable life events obtained from mothers) longitudinally and significantly correlated with indicators of fast LH behavioral profiles (i.e., aggression, impulsivity, risk preference reported by adolescents). These indicators were also correlated with secure attachment in the expected directions. We also present the Ms and SDs of the variables for the two genders in Table 2. Boys scored significantly higher than girls on two of the fast LH behavioral indicators. There were no directional or substantial differences in the zero-order correlations or structural relations between the two genders and across 10 cultural groups.
Table 1. Correlations, means, and standard deviations of variables used in the study
†p < .10, *p < .05, **p < .01, ***p < .001.
Table 2. Gender differences of variables used in the study
*p < .05, ***p < .001.
We initially examined and found support for measurement invariance across sites. The measurement model was identified by a confirmatory factor analysis in all 10 sites as the best fitting model with adequate measurement properties. Subsequent measurement invariance tests based on the alignment method (Asparouhov & Muthén, Reference Asparouhov and Muthén2014) of Mplus revealed fewer than 4% noninvariant measurements which was far below the 20%–25% minimum noninvariance threshold (Muthén & Asparouhov, Reference Muthén and Asparouhov2014). We then conducted the full structural equation modeling analysis and tested the mediation model without the interaction term. The goodness-of-fit statistics (χ2/df = 4.99, CFI = 0.96, TLI = 0.94, RMSEA = 0.070, SRMS = 0.056) of the model met the recommended cutoff values for adequate model fit. As shown in Figure 1, all parameter estimates were in the expected directions and were statistically significant. The factor loadings were adequate and were robust even though the indicators were obtained from different informants and some (e.g., proxies of environmental adversity) are not expected to be highly correlated in approximating diverse environmental conditions. Similarly, parameter estimates of the structural model were consistent with our hypotheses. Environmental adversity was negatively associated with secure attachment (β = −.18, p < .001) and positively predicted fast LH behavioral profile (β = .35, p < .001); secure attachment was negatively related to fast LH behavioral profiles (β = −.49, p < .001). The mediating effect of secure attachment between environmental adversity and fast LH behavioral profiles was significant (β = .10, 95% CI = [.01, .19]) based on a bootstrapping procedure with 2000 resamples and the maximum likelihood estimation. It represents 26% of the total effect (β = −.44, p < .001) of environmental adversity on fast LH behavioral profiles. These findings support our hypothesis that secure attachment mediates the relation between environmental adversity and fast LH behavioral profiles. Separately, we included family income on a 10-point scale equated across sites and years of education of each parent in the SEM analysis. The results concerning our variables of interest remained the same in terms of statistical significance and relation directionality.
We finally used the default approach of Mplus to test the moderation hypothesis by comparing the interaction or moderation model with the main-effect only or baseline model, which is our mediation model. The log-likelihood for the main-effect-only or baseline model was −9164.07; and that for the interaction model was = −9156.23; D = 15.68, p < .001. The statistically significant reduction of the log-likelihood value indicates substantial improvement in data fit by the hypothesized moderation model over the baseline model. For parameter estimation, the interaction between environmental adversity and attachment was significant (β = −.46, p < .001), supporting internal regulation by attachment of the external influence by environmental adversity on fast LH behavioral profiles. Figure 2 displays the simple slopes of environmental adversity on fast LH behavioral profiles at 1 SD (β = .07, ns) and −1 SD of secure attachment (β = .33, p < .001). Compared to the main effect (β = .35, p < .001), the first simple slope at higher levels of secure attachment was much reduced and nonsignificant, whereas the second simple slope at lower levels of secure attachment remained the same as the main effect. As predicted, secure attachment mainly attenuated or nullified the association of environmental adversity to fast LH behavioral profiles at higher levels of secure attachment, and the negative association was maintained but was not amplified at lower levels of secure attachment (i.e., insecure attachment).
Figure 2. Simple slopes and 95% confidence bands of the regression of fast life history profile on childhood environmental adversity at 1 SD above (light) and 1 SD below (dark) the mean of secure attachment.
Discussion
Belsky et al. (Reference Belsky, Houts and Fearon2010) first identified the effect of caregiver–child attachment on LH outcomes, while Sung et al. (Reference Sung, Simpson, Griskevicius, Kuo, Schlomer and Belsky2016) were the first to report the moderating effect of attachment on the relationship between environmental adversity and fast LH. From an internal (mental) state perspective (Nettle et al., Reference Nettle, Frankenhuis and Rickard2013), our findings align with those of these pioneering studies. Caregiver–child attachment statistically mediated the longitudinal association between earlier environmental adversity and adolescent fast LH behavioral profiles in the same direction as the external environment. This finding supports the notion that some of the effect attributable to the child’s living environment is transmitted to the child through caregiving behaviors and the caregiver’s other LH manifestations, which are normally shaped by the same environment that the child inherits from the caregiver (Belsky et al., Reference Belsky, Steinberg and Draper1991; Chisholm, Reference Chisholm1996). This finding also affirms the logic that internal state is in part caused by and exerts similar influence on LH as the external environment (Chang et al., Reference Chang, Lu, Lansford, Bornstein, Steinberg, Chen, Skinner, Dodge, Deater-Deckard, Bacchini, Pastorelli, Alampay, Tapanya, Sorbring, Oburu, Al-Hassan, Di Giunta, Malone, Uribe Tirado and Yotanyamaneewong2019a). Secure attachment resulting from supportive family environment transmits the same fitness enhancing effect on LH development as the supportive and stable environment. Dysfunctional parenting and insecure attachment that are caused by and in turn transmute environmental adversities into fast LH strategies perpetuate but do not intensify the species-general contingency between the living environment and LH strategies.
The more significant finding is the statistical moderation of attachment. Consistent with our hypothesis, secure attachment primarily functions to downregulate or nullify the effects of the external environment on LH behavioral profiles, rather than insecure attachment upregulating or strengthening these effects. The evolutionary function of caregiving and parenting, which forms caregiver–child attachment, is to protect offspring from mortality threats (e.g., protection from predation; Bowlby, Reference Bowlby1969), aligning with the mortality-reduction function of slow LH. Parenting and the resulting attachment foster not only the knowledge and skills necessary to master the environment but also the associated mindset about the self in relation to the world. Established early in childhood as secure attachment and an efficacious internal working model, the internalization of the mental representation of the world as controllable and predictable, and of the self as capable and efficacious, potentially shifts species expectations about extrinsic mortality threats. Instead of perceiving extrinsic mortality threats as uncontrollable and inescapable, the parentally socialized mindset may view them as controllable and reducible. A pervasive belief in the controllability, dependability, and predictability of the external environment, combined with substantive knowledge, skills, and cognitive abilities to conquer nature, enables well-socialized human offspring to manage mortality threats in their living environment. Such an internal state effectively redirects the developmental trajectory from a species-general, adversity-contingent fast track, which disregards mortality and accelerates reproduction, to a slow pathway aimed at reducing extrinsic risks, delaying reproduction, and subsequently slowing LH.
These findings also support the internal models of LH development (e.g., Nettle et al., Reference Nettle, Frankenhuis and Rickard2013). Organisms actively regulate external effects through internal somatic and cognitive adjustments rather than responding passively to the environment. The somatic and mental statuses of individuals, responsible for the internal adjustment of external environmental effects, create phenotypic variations within species, which are the basis of human LH research. In the present study, individual differences in attachment and related internal working models enable individuals to respond to environmental adversities differently, accounting for observed variations in LH behavioral profiles. According to internal models in general and the attachment-specific internal model investigated here, extrinsic mortality risks may not uniformly affect age-specific populations but may respond to the mortality-reduction efforts and abilities of individuals. Extrinsic mortality risks can be rendered controllable depending on the intrinsic physical and cognitive attributes of individuals. This observation challenges the prediction of the species-general LH principle, which emphasizes extrinsic mortality risks causing indiscriminate casualties independent of individuals’ survival abilities and efforts. While existing internal models are based on internal body states (Chang et al., Reference Chang, Lu, Lansford, Bornstein, Steinberg, Chen, Skinner, Dodge, Deater-Deckard, Bacchini, Pastorelli, Alampay, Tapanya, Sorbring, Oburu, Al-Hassan, Di Giunta, Malone, Uribe Tirado and Yotanyamaneewong2019a), the present study includes mental states in intrinsic mortality–vitality determination. Just as a sound internal body state can influence responses to external threats, the internalized secure mental representation of the external environment in relation to the self may shift one’s perception of extrinsic mortality threats from being uncontrollable and inescapable to being controllable and reducible.
The present study has certain limitations. Most notably, the childhood environment variables were collected when the children were 10 years old. However, family environments remained stable throughout the data collection phases of the study that started when children were 8 years old on average. The items (e.g., “My neighborhood is a nice place to live.” “The atmosphere in our home is calm.”) reflect a consistent state of affairs over an extended period up to the time of the interviews reported in this study. Respondents were also asked to reflect on their past experiences. Thus, these variables effectively captured early or earlier childhood environments, potentially including more critical periods of LH development. We measured attachment during adolescence, but attachment, which is primarily formed during early childhood, remains relatively stable throughout life (Main, Reference Main, Parkes, Stevenson-Hinde and Marris1991). Finally, as suggested by one of the reviewers, race is highly relevant to LH processes (Rushton, Reference Rushton1996). Our multi-country data allow for comparative investigations across both racial and cultural dimensions, which future research may explore. Despite these and other limitations, our study represents one of the first theoretical and empirical attempts to conceptualize attachment as an internal state in slowing human LH, in addition to mediating the species-general environmental contingency on LH development.
Acknowledgements
We acknowledge the editor and reviewers for their valuable comments on the previous version of the manuscript.
Funding statement
This research has been funded by the Eunice Kennedy Shriver National Institute of Child Health and Human Development grant RO1-HD054805 and Fogarty International Center grant RO3-TW008141, as well as the Intramural Research Program of the NIH/NICHD. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH or NICHD.
Competing interests
The authors declare that there is no conflict of interest.
