At first glance, not much—you’d think. And that’s partly true. Cognitive tests in infancy are notoriously unreliable, influenced by all sorts of things that have little to do with thinking: motor skills, fatigue, motivation, distractibility. But in an impressive longitudinal study published in PNAS, Gustavson and colleagues show that some early behaviours are surprisingly predictive of later cognitive ability.

In their study, the researchers followed over a thousand twins from infancy into their thirties. At several points in time—7 and 9 months, 1 to 2 years, 3 years, 7 years, 16 years, and around age 29—they assessed general cognitive ability (GCA). They combined classical twin analyses with modern genetic prediction techniques based on polygenic scores. The result? General cognitive ability already shows striking stability from toddlerhood onward.

Genetic influences already explained a substantial share of adult GCA differences by age three. Nearly half of those adult differences were genetically “predicted by age seven.” But genes weren’t the whole story. Shared environmental factors—home, neighbourhood, early schooling—also left a lasting mark. The environment children grew up in around age two still explained about 10% of their cognitive performance at age 29.

Some baby behaviours were modest but meaningful predictors of later intelligence. For instance, how much infants showed interest in new objects (“novelty preference”) or how focused they seemed to a trained observer predicted adult GCA, with small but significant correlations (around 0.16 to 0.20). Other behaviours, like how vocal or active the baby was, only predicted performance in early childhood, not beyond.

Still, there were clear limits. Genetic scores for intelligence and educational attainment started predicting reasonably well from toddlerhood onward, but not in infancy. And how many years of school someone would eventually complete could hardly be predicted by anything measured at nine months. So no, baby behaviour doesn’t offer a straight line to future academic outcomes—but it might be the first faint sketch.

What do we take from this? Two things stand out. First, this study confirms how stable general cognitive ability is over time, and how genetic predispositions and early environments contribute to that stability. Second, it underscores that if we want to support cognitive development, we shouldn’t wait too long. Early childhood doesn’t set everything in stone, but it lays more foundation than expected.

Abstract of the longitudinal study:

Measures of general cognitive ability (GCA) are highly stable from adolescence onward, particularly at the level of genetic influences. In contrast, measurement of GCA in early life (before 3 y old) is less reliable and less is known about the stability of GCA across this period, including its relation to adult GCA. Using data from the Colorado Longitudinal Twin study (N = 1,098), we examined the stability of GCA measures across 5 time-points (years 1 to 2, 3, 7, 16, and 29), including how an array of cognitive measures given at 7 and 9 mo relate to later GCA. We then examined the genetic and environmental stability of GCA across the first 30 y of life using complementary methods: twin analyses and polygenic scores (PGSs). Two infant cognition measures, object novelty and tester-rated task orientation, predicted GCA in adulthood (r = 0.16 and 0.18, respectively). Correlational analyses were consistent with a pattern of increasing stability across development for GCA measures between year 1 to 2 and adulthood (r = 0.39 to 0.85). Subsequent twin analyses revealed that 22% of variance in adulthood GCA was captured by genetic influences on GCA from year 3 or earlier, with an additional 10% explained by shared environmental influences on GCA at year 1 to 2. PGSs for adulthood GCA and educational attainment predicted GCA from 1 to 2 y onward (βs = 0.09 to 0.44) but not infant cognition. Findings suggest that genetic and environmental influences on GCA demonstrate considerable stability as early as age 3 y, but that measures of infant cognition are less predictive of later cognitive ability.