Neurocognitive reorganization between crystallized intelligence, fluid intelligence and white matter microstructure in two age-heterogeneous developmental cohorts

Ivan L. Simpson-Kent, Delia Fuhrmann, Joe Bathelt, Jascha Achterberg, Gesa Sophia Borgeest, Rogier Kievit

Abstract

Despite the reliability of intelligence measures in predicting important life outcomes such as educational achievement and mortality, the exact configuration and neural correlates of cognitive abilities remain poorly understood, especially in childhood and adolescence. Therefore, we sought to elucidate the factorial structure and neural substrates of child and adolescent intelligence using two cross-sectional, developmental samples (CALM: N=551 (N=165 imaging), age range: 5-18 years, NKI-Rockland: N=337 (N=65 imaging), age range: 6-18 years). In a preregistered analysis, we used structural equation modelling (SEM) to examine the neurocognitive architecture of individual differences in childhood and adolescent cognitive ability. In both samples, we found that cognitive ability in lower and typical-ability cohorts is best understood as two separable constructs, crystallized and fluid intelligence, which became more distinct across development, in line with the age differentiation hypothesis. Further analyses revealed that white matter microstructure, most prominently the superior longitudinal fasciculus, was strongly associated with crystallized ($g_c$) and fluid ($g_f$) abilities. Finally, we used SEM trees to demonstrate evidence for developmental reorganization of $g_c$ and $g_f$ and their white matter substrates such that the relationships among these factors dropped between 7-8 years before increasing around age 10. Together, our results suggest that shortly before puberty marks a pivotal phase of change in the neurocognitive architecture of intelligence.

Introduction

Psychologists have long debated the fundamental structure of human intelligence. A central question is whether cognitive ability is a single "general" factor ($g$) or if it is better described by distinct components: crystallized intelligence ($g_c$), representing acquired knowledge like vocabulary, and fluid intelligence ($g_f$), the capacity to solve novel problems through reasoning. While these structures are well-mapped in adults, their emergence and specialization during childhood remain less clear. Furthermore, the "neural substrates"—the physical brain structures supporting these abilities—are not fully understood. This study investigates how the brain's white matter (the neural "wiring") supports $g_c$ and $g_f$ and how this relationship evolves during development.

Methods

The researchers employed Structural Equation Modelling (SEM) to model "latent" variables of intelligence inferred from a battery of cognitive tests. A key innovation was the use of SEM trees, an algorithmic approach that recursively partitions data to identify specific ages where the relationship between brain structure and behavior significantly changes. This avoids the biases of traditional "median split" methods.

To assess brain structure, Diffusion Tensor Imaging (DTI) was used to measure Fractional Anisotropy (FA), a proxy for white matter microstructure integrity. Higher FA values generally indicate more organized neural pathways.

Datasets and Cohorts

The study utilized two distinct developmental cohorts to ensure robustness:

  1. CALM (Centre for Attention, Learning and Memory): $N=551$ (165 with imaging), ages 5–18. This cohort includes children referred for attention or memory difficulties.
  2. NKI-Rockland Sample: $N=337$ (65 with imaging), ages 6–18. This community-based sample is representative of the general population.
Age distribution of the CALM and NKI-Rockland study cohorts.

Key Insights

The Specialization of Intelligence

The findings support the age differentiation hypothesis. As children mature, $g_c$ and $g_f$ become increasingly distinct. Initially appearing as an undifferentiated "blob" of cognitive ability, these factors specialize into separate toolkits as the brain matures.

Neural Correlates: The Brain’s Superhighway

The Superior Longitudinal Fasciculus (SLF), a major tract connecting frontal and posterior brain regions, emerged as the most consistent neural predictor of intelligence across both cohorts. White matter microstructure explained approximately 32.9% of the variance in crystallized ability and 33.6% in fluid ability within the CALM sample.

The "Puberty Pivot"

The study identified a non-monotonic "reorganization" of neurocognitive structures. The connection strength between white matter and intelligence dipped around age 7–8 before rebounding and increasing around age 10. This suggests that the period shortly before puberty is a critical window for biological "pruning" and hormonal shifts that reshape the architecture of intelligence.

Results

In both samples, a two-factor model ($g_c$ and $g_f$) provided a significantly better fit than a single $g$-factor model. SEM trees identified a significant drop in the $g_c/g_f$ correlation at age 9.12 in the CALM group.

Developmental trajectories of cognitive scores across age for both CALM (red) and NKI (blue) cohorts.

Limitations and Future Directions

As a cross-sectional study, these findings represent snapshots of different children at various ages. Longitudinal research following the same individuals over time is necessary to confirm these reorganization patterns. Additionally, while the CALM sample provided high power, the imaging subset of the NKI-Rockland sample was smaller ($N=65$), highlighting the need for larger datasets like the ABCD study to further explore these developmental transitions.

Summary updated

Abstract

Abstract Despite the reliability of intelligence measures in predicting important life outcomes such as educational achievement and mortality, the exact configuration and neural correlates of cognitive abilities remain poorly understood, especially in childhood and adolescence. Therefore, we sought to elucidate the factorial structure and neural substrates of child and adolescent intelligence using two cross-sectional, developmental samples (CALM: N=551 (N=165 imaging), age range: 5-18 years, NKI-Rockland: N=337 (N=65 imaging), age range: 6-18 years). In a preregistered analysis, we used structural equation modelling (SEM) to examine the neurocognitive architecture of individual differences in childhood and adolescent cognitive ability. In both samples, we found that cognitive ability in lower and typical-ability cohorts is best understood as two separable constructs, crystallized and fluid intelligence, which became more distinct across development, in line with the age differentiation hypothesis. Further analyses revealed that white matter microstructure, most prominently the superior longitudinal fasciculus, was strongly associated with crystallized (gc) and fluid (gf) abilities. Finally, we used SEM trees to demonstrate evidence for developmental reorganization of gc and gf and their white matter substrates such that the relationships among these factors dropped between 7-8 years before increasing around age 10. Together, our results suggest that shortly before puberty marks a pivotal phase of change in the neurocognitive architecture of intelligence.
BibTeX
@inproceedings{simpsonkent2019neurocognitive,
  author = {Ivan L. Simpson-Kent and Delia Fuhrmann and Joe Bathelt and Jascha Achterberg and Gesa Sophia Borgeest and Rogier Kievit and Rogier A. Kievit},
  title = {{Neurocognitive reorganization between crystallized intelligence, fluid intelligence and white matter microstructure in two age-heterogeneous developmental cohorts}},
  booktitle = {Developmental},
  year = {2019},
  doi = {10.1016/j.dcn.2019.100743},
}
Figure