Gaussian process linking functions for mind, brain, and behavior

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Psychometrics

Background:

• The relationship between the mind, brain, and behavior is a longstanding scientific challenge.
• Hierarchical latent variable models have linked brain (EEG, fMRI) and behavioral data, but were limited by inflexible linking functions.
• Understanding complex brain dynamics requires more flexible models for brain-behavior associations.

Purpose of the Study:

• To propose a novel data-driven, nonparametric approach for modeling the mind-brain-behavior link.
• To allow complex, emergent linking functions for greater flexibility in analyzing neural dynamics.
• To incorporate spatial and temporal structures for biological plausibility in cognitive models.

Main Methods:

• Developed a nonparametric, hierarchical latent variable model for multivariate, multimodal data.
• Integrated spatial and temporal constraints to ensure biologically plausible system dynamics.
• Validated the model using simulation studies and applied it to simultaneous fMRI and behavioral data.

Main Results:

• The model accurately fits simulated data and recovers latent dynamics effectively.
• In experimental data (fMRI, motion tracking), the model accurately recovered neural and behavioral data.
• The approach revealed complex latent cognitive dynamics, offering insights into experimental task aspects.

Conclusions:

• The proposed nonparametric approach offers a flexible and powerful method for modeling the mind-brain-behavior relationship.
• This method advances cognitive neuroscience by enabling the analysis of complex, dynamic neural processes.
• The findings demonstrate the model's utility in both simulated and real-world experimental neuroscience research.