Caregiving and Adolescent Cognitive Outcomes

The Trude et al. (2021) analysis published in The Lancet Child & Adolescent Health tested whether two operationally distinct components of nurturing care—responsive caregiving and learning opportunities—buffer the cognitive consequences of cumulative early adversity. Using birth cohorts in Brazil and South Africa, the study delivered an answer with an unexpected asymmetry: nurturing care does buffer adolescent IQ outcomes, but the operative component differs by setting. In South Africa, responsive caregiving carried the protective effect; in Brazil, structured learning opportunities did. The asymmetry has implications for how nurturing-care interventions should be targeted, and it raises measurement questions about whether the two components are as separable as the framework asserts.

The cumulative-adversity dose-response problem

The question of whether early adversity damages later cognition is settled (Black et al., 2017, in the Lancet Early Childhood Development series, summarized this evidence base). What is unsettled is which child-environment factors modify the dose-response relationship and at which developmental window. Cumulative adversity scores—indices that sum exposures across domains such as poverty, maternal mental health, low birthweight, and low maternal education—make it tractable to ask “for each additional adversity, how much IQ?” but they hide variability in the duration, intensity, and timing of individual exposures.

Trude et al. (2021) used a 0-to-9 cumulative adversity score that aggregated household wealth, household crowding, maternal schooling, maternal height, maternal age at child’s birth, maternal mental health, child birthweight, gestational age, and child length at 12 months. Their primary IQ outcomes were assessed at age 18 in the 1993 Pelotas Birth Cohort (Brazil, N = 632, four WAIS-III subtests: Similarities, Picture Completion, Arithmetic, Symbol Coding) and at age 16 in the Birth to Twenty Plus cohort (South Africa, N = 1,130, Raven’s Standard Progressive Matrices). The use of different IQ instruments across cohorts is a methodological constraint to which the discussion will return.

The dose-response, before moderation

Without conditioning on caregiving environment, the cohort-specific adversity slopes were both negative and substantially different in magnitude. Each one-Z-score increase in cumulative adversity predicted a 5.89-point IQ decrement (95% CI: −7.29 to −4.50) in the Pelotas cohort and a 2.69-point decrement (95% CI: −4.52 to −0.86) in the Bt20+ cohort. The Brazilian effect size is more than twice the South African one in absolute terms. Two non-exclusive explanations are plausible: the Wechsler short form used in Pelotas may be more sensitive to crystallized-knowledge effects of educational deprivation than is Raven’s, which is more purely fluid; alternatively, the joint distribution of adversities may differ across the two settings such that comparable Z-score increments index different absolute adversity loads.

The bottom line of the unmoderated analysis is that adversity’s cognitive cost is real in both settings but not equivalent in size, and that comparing absolute β coefficients across cohorts requires interpretive caution.

The asymmetric moderation finding

The headline result of the paper is that nurturing care, measured at preschool ages with the Home Observation for Measurement of the Environment (HOME) Inventory, did moderate the adversity-IQ slope—but the active component differed by cohort.

In the Bt20+ South African cohort, responsive caregiving moderated the adversity-IQ association (interaction β = 2.24, 95% CI 0.94 to 3.54, p = 0.0075). Decomposing the interaction: under low-caregiving conditions, each Z-score of cumulative adversity predicted a 5.4-point IQ decrement (p = 0.020); under medium and high caregiving, the adversity coefficient was no longer statistically distinguishable from zero. Functionally, responsive caregiving absorbed the cognitive cost of cumulative adversity in the South African data.

In the 1993 Pelotas Brazilian cohort, learning opportunities (rather than responsive caregiving) carried the protective effect (interaction β = 1.74, 95% CI 0.43 to 3.04, p = 0.0092). Under low learning-opportunity conditions, each Z-score of adversity predicted a 6.66-point IQ decrement (p < 0.0001); under high learning opportunities, the adversity coefficient became non-significant (p = 0.89). The same protective pattern—moderation that essentially zeroes out the adversity slope at high nurturing levels—appeared, but the moderator was the structural-environment component rather than the relational-caregiving component.

Crucially, neither responsive caregiving nor learning opportunities moderated outcomes for psychosocial adjustment or height-for-age in either cohort. The protective effect was specific to cognitive outcomes, despite cumulative adversity having direct effects on all three outcome domains.

Why the asymmetry?

The cohort-specific moderation pattern is the most theoretically interesting feature of the paper, and the authors are appropriately cautious about it. Three explanations are worth distinguishing.

Measurement timing. The HOME Inventory was administered at age 4 in Pelotas and at age 2 in Bt20+. Responsive caregiving may be more salient at age 2, when verbal and self-regulatory scaffolding from caregivers is dominant, while structured learning opportunities may matter more at age 4, when children begin to engage with materials and pre-academic content. If so, the cohort difference reflects developmental timing rather than population difference.

Floor and ceiling effects. Pelotas was a relatively higher-SES sample than Bt20+ on the relevant adversity dimensions. Responsive caregiving may have hit a ceiling in the Brazilian sample (most parents already exhibited adequate responsiveness, leaving little variance to moderate adversity effects), while learning opportunities, which depend on tangible resources (books, toys, structured play), retained variance to detect a moderation effect. In Bt20+, the more strained environment may have left greater between-family variance in caregiver responsiveness, and less variance in learning opportunities (because materials and structured time were uniformly scarce).

Construct overlap. Responsive caregiving and learning opportunities are conceptually distinct but empirically correlated; HOME Inventory subscale intercorrelations are typically in the .40 to .60 range. The cohort-specific moderation result may reflect, in part, which component carries more unique variance after the other is held constant in the specific data structure of each cohort. This is a measurement-model concern that the analysis cannot fully resolve.

Position in the nurturing-care literature

The Trude et al. (2021) findings extend the foundational case made by Britto et al. (2017) and Black et al. (2017) in the 2017 Lancet Early Childhood Development series, which framed nurturing care—comprising health, nutrition, security, responsive caregiving, and early learning—as the operative protective factor across the first thousand-and-more days of life. Those framework papers established the conceptual case; Trude et al. provided one of the cleanest empirical demonstrations using harmonized longitudinal data across two middle-income countries that the protective effect propagates into adolescent cognition specifically, rather than just into early childhood developmental milestones.

The longitudinal payoff is substantial. Most nurturing-care evidence to date has measured outcomes within early childhood—at age 3, age 5, perhaps age 8. Trude et al. measured outcomes at ages 16-18, when participants are within years of entering adulthood and the labor market. Demonstrating that a HOME Inventory administered before kindergarten predicts adolescent IQ and modifies adversity effects across that developmental span is the kind of evidence that supports policy intervention timing decisions: the window during which nurturing-care effects on cognition are still establishable is wider than narrow-window critical-period accounts suggest.

Limitations the authors flag

Three limitations qualify the interpretation. First, cumulative adversity scores collapse duration and intensity. A child who experienced maternal depression for twelve months is treated identically to one who experienced six months of maternal depression and six months of household crowding, even though the developmental implications differ. Second, parental IQ was not measured. Both the cumulative adversity score and the IQ outcome have heritable components, and the protective effect of nurturing care on the adversity-IQ slope could be partially confounded with parental cognitive ability that drives both nurturing behaviors and child IQ. Third, the moderation analyses may have been underpowered: detecting interaction effects with longitudinal cohort data requires effective sample sizes that the cohort attrition over 16-18 years has substantially reduced.

A fourth limitation the authors do not emphasize but that warrants mention is the use of different IQ instruments in the two cohorts. WAIS-III subtest composites and Raven’s Standard Progressive Matrices are not interchangeable measures of g; they weight crystallized and fluid components differently, and the differential adversity slope sizes across cohorts may be partly attributable to instrument-specific sensitivity rather than to true population differences. A within-cohort replication using a single instrument would strengthen the comparative claim.

Policy and intervention implications

For policymakers in middle-income countries, the practical message is that nurturing-care interventions targeting the preschool window can compensate, partly or fully, for the cognitive consequences of cumulative early adversity. The Trude et al. (2021) results do not, however, settle which component to fund first. In settings closer to the South African profile—high adversity, variable caregiving—programs that train and support responsive caregiving (parenting interventions, home-visiting programs) may carry the larger protective effect. In settings closer to the Brazilian profile—moderate adversity, variable access to materials and structured early childhood education—investment in learning-opportunity infrastructure (books in the home, preschool quality, structured play environments) may yield the larger protective effect. The sensible policy inference is that both components matter and that the relative emphasis should be guided by the local distribution of nurturing-care variance, which a well-designed HOME Inventory needs assessment can identify.

The broader scientific contribution of the paper is to anchor the nurturing-care framework in adolescent cognitive outcomes with cohort-comparable evidence. That anchor strengthens the case for treating early childhood as a high-yield investment window in both cognitive and human-capital terms—an argument the 2017 Lancet series made conceptually and that Trude et al. extend with measurement-grade longitudinal evidence.

Frequently asked questions

What is responsive caregiving?

Responsive caregiving is one of five components of the nurturing-care framework articulated by Britto et al. (2017): the others are health, nutrition, security and safety, and early learning. It refers to caregiver attentiveness, contingent responding, and emotional availability—the relational scaffolding through which young children develop self-regulation and verbal communication. It is typically measured with subscales of the Home Observation for Measurement of the Environment (HOME) Inventory.

What did Trude et al. (2021) find?

Cumulative early adversity predicted lower adolescent IQ in both Brazil (β = −5.89 per Z-score, Pelotas N = 632) and South Africa (β = −2.69 per Z-score, Bt20+ N = 1,130). Nurturing care moderated the adversity-IQ slope, but the active component differed: responsive caregiving carried the protective effect in South Africa (interaction β = 2.24, p = 0.0075), while structured learning opportunities carried it in Brazil (interaction β = 1.74, p = 0.0092). At high nurturing-care levels in either cohort, the adversity coefficient was no longer statistically distinguishable from zero.

Why might responsive caregiving and learning opportunities have different effects in different settings?

The authors highlight three candidate explanations. The HOME Inventory was administered at age 4 in Pelotas and age 2 in Bt20+, so developmental timing may favor different components at different ages. Floor or ceiling effects in each population may leave different components with more variance to detect moderation. And the two components are conceptually distinct but empirically correlated, so unique-variance allocation depends on each cohort’s data structure.

Was the protective effect specific to cognition?

Yes. Neither responsive caregiving nor learning opportunities moderated outcomes for psychosocial adjustment or height-for-age in either cohort, despite cumulative adversity having direct effects on all three outcome domains. The cognitive specificity argues that nurturing care interacts with adversity through pathways relevant to cognitive development rather than through generic resilience.

How were IQ outcomes measured at adolescence?

Pelotas used a four-subtest WAIS-III short form (Similarities, Picture Completion, Arithmetic, Symbol Coding) at age 18; Bt20+ used Raven’s Standard Progressive Matrices at age 16. The two instruments are not interchangeable measures of general intelligence—they weight crystallized and fluid components differently—which is one reason cross-cohort effect-size comparisons require interpretive caution.

What does this imply for early-childhood policy?

The window during which nurturing-care interventions can compensate for early adversity extends into the preschool years and remains predictive into adolescence. The relative emphasis—training responsive caregiving versus expanding access to learning materials and structured early childhood education—should be guided by the local distribution of nurturing-care variance, which a well-designed HOME Inventory needs assessment can identify.