Henrike M. Jungeblut

Young scientist talk \ Manfred Eigen lecture theatre

White matter microstructure is a candidate neurobiological substrate underlying individual differences in cognitive abilities. However, measurement inaccuracies in MRI-derived metrics often attenuate observed effect sizes, obscuring true brain-behavior relationships. Here, we demonstrate how latent variable modeling of neuroimaging data can mitigate these reliability concerns by estimating latent effect sizes free from measurement error.
We analyzed data of N = 365 individuals (age range: 18−74) drawn from two independent samples (Dortmund Vital Study: N = 150, Clinicaltrials.gov: NCT05155397; Mainz Network Study: N = 215). Using Structural Equation Modeling, we derived latent measurement models for three distinct microstructural markers across 52 white matter tracts: white matter integrity (fractional anisotropy, FA), neurite density (intra-neurite volume fraction, INVF), and myelin content (magnetization transfer ratio, MTR). All three microstructural markers were adequately captured by general-factor models, indicating that they all reflect brain-general rather than purely localized characteristics. Crucially, when relating the latent factors of white matter microstructure to a latent variable of fluid intelligence, we found robust latent positive associations for white matter integrity (β = 0.26, p < .001) and myelin content (β = 0.25, p = .017).
These findings not only establish anatomically informed measurement models for the microstructural properties of the white matter, but they also demonstrate that integrating psychometric methods with neuroimaging data provides a powerful, scalable framework to uncover the biological underpinnings of human cognition with greater precision.