Correlation between structural connectivity (SC) from diffusion MRI and functional connectivity (FC) from fMRI is widely used in neuroimaging psychiatry as a measure of structure-function coupling (SFC) and is often taken as evidence of tight coupling between anatomy and neural communication. This Perspective examines conceptual and biophysical limitations of interpreting SFC as evidence of neural coupling and mechanistic dysfunction in psychiatric disorders. SFC quantifies statistical dependence between network summaries but does not capture biophysical causation. Local circuit properties such as excitation-inhibition balance, synaptic gain, and neuronal responsiveness can alter FC without detectable changes in macroscale white matter structure. Global, static correlations further ignore temporal constraints such as conduction delays and frequency-dependent dynamics. FC is additionally filtered through neurovascular coupling, often altered in psychiatric conditions, weakening the interpretability of SFC as a proxy for neural coupling. Biophysically constrained generative models offer an alternative framework in which SC provides a fixed anatomical scaffold for simulating neural dynamics and explaining FC differences via explicit circuit parameters. SFC remains useful as a descriptive network measure, but its interpretation as a marker of biophysical coupling is unwarranted. Generative modeling approaches are needed to support mechanistic inference and translational relevance.
Richard O. Nkrumah (Wed,) studied this question.