Fronto-parietal theta-gamma Phase-Amplitude Dynamics Predict Cognitive Processing Efficiency in Multiple Object Tracking: A Drift Diffusion Model and Bayesian Mixed Model Approach

Objectives: Individual differences in dynamic attention are closely linked to fluid intelligence and real-world performance, yet the neural sources of these differences remain poorly understood. While traditional frameworks emphasize the amount of attentional resources invested, emerging evidence suggests that the efficiency with which resources are integrated across brain regions and frequency bands may be equally critical. This study investigates whether single-trial theta-gamma phase-amplitude coupling (PAC) in the fronto-parietal network predicts cognitive processing efficiency in a multiple object tracking (MOT) task. Specifically, we aim to: (1) examine whether theta-gamma PAC differs between correct and incorrect trials (subsequent memory effect); (2) test whether single-trial PAC strength predicts trial-level drift rate derived from a hierarchical Bayesian drift diffusion model; (3) investigate whether this predictive relationship is modulated by tracking load (2, 4, and 6 targets); and (4) decompose the variance into inter-individual (trait-like) and intra-individual (state-like) components to characterize the sources of neural efficiency. We hypothesize that correct trials will exhibit stronger PAC, with gamma activity preferentially locked to the peak of the theta wave, and that PAC will positively predict drift rate, particularly under high load conditions where cross-frequency coordination is critical for maintaining performance. Methods: A total of 40 healthy adults will be recruited, with an expected final sample of approximately 35 after EEG artifact rejection. Participants will perform a classic MOT task with 336 trials evenly distributed across three load conditions (2, 4, and 6 targets). Each trial consists of a 1250-ms cue phase, a 4000-ms tracking phase, and a response phase where participants judge whether two probed items were targets or distractors. EEG will be recorded from 64 channels and preprocessed using standard pipelines including filtering, artifact subspace reconstruction, ICA, and current source density transformation. Theta-gamma PAC will be quantified using the phase locking value (PLV) between the phase of a multivariate theta component extracted from prefrontal electrodes (Fz, F3, F4, FCz, FC3, FC4) and the amplitude envelope of multivariate gamma components from three functionally defined regions of interest: frontal (executive control), centro-parietal (attentional allocation), and occipital (early visual processing). PAC will be computed using 500-ms sliding windows (50-ms steps) across the tracking period, and trial-level PAC strength will be averaged within each cluster. Trial-level quality control will be performed using permutation testing (1000 iterations) to retain only trials with significant coupling (p < 0.05). Drift rate will be estimated using a hierarchical Bayesian drift diffusion model (HDDM) with drift rate varying by load condition and boundary separation, non-decision time, and starting point held constant. Single-trial drift rates will be extracted via posterior predictive methods. Bayesian linear mixed models will be employed to predict trial-level drift rate from PAC strength, load, and their interaction, with random intercepts and random slopes for PAC at the subject level. Time-resolved analyses will be conducted to identify temporal windows where PAC effects are most pronounced, with cluster-based permutation correction for multiple comparisons. Hypothesis: We hypothesize that: (1) correct trials will show significantly higher PAC than incorrect trials, with gamma activity preferentially coupled to the theta peak; (2) single-trial PAC strength will positively predict drift rate, indicating that stronger cross-frequency coordination facilitates evidence accumulation; (3) this predictive effect will be amplified under high load (6 targets) compared to low load (2 targets), reflecting the increased demand for efficient resource integration under cognitive challenge; and (4) variance decomposition will reveal substantial inter-individual variability in PAC effects (trait-like efficiency baselines) alongside significant intra-individual fluctuations (state-like load adaptation), with trial-level PAC explaining unique variance in drift rate beyond subject-level differences. Innovation and Value: This study introduces several methodological and theoretical innovations. First, it extends the continuous shared resource theory from a focus on "resource amount" to "resource integration efficiency," operationalized through theta-gamma PAC. Second, it is the first to link single-trial, time-resolved PAC to drift rate in MOT, enabling a fine-grained analysis of how cross-frequency coordination dynamically supports cognitive processing efficiency. Third, the functional parcellation of gamma activity into three distinct clusters allows for the dissociation of executive control, attentional allocation, and visual processing contributions to theta-gamma coupling. Fourth, the integration of multivariate PAC methods with hierarchical Bayesian modeling and rigorous trial-level quality control enhances both the sensitivity and reliability of the findings. By demonstrating that neural coordination efficiency predicts trial-by-trial fluctuations in evidence accumulation, this research will provide a novel neurophysiological framework for understanding individual differences in dynamic attention, with implications for both basic cognitive neuroscience and applied domains such as talent identification and cognitive training.