The impact of intrinsic connectome dynamics on perception is context-dependent

The functional relevance of time-averaged (static) functional connectome patterns is well recognized. However, the real-time relationship between ongoing connectome dynamics and behavioral outcomes is not well understood. It is particularly unclear whether behavior is linked to connectivity dynamics mainly among a common set of connections regardless of the task (suggesting fluctuations in task-common processes like general arousal), or if it varies with different connections depending on the task (indicating fluctuations in context-dependent processes). To investigate this, we compared fMRI data of healthy participants across three cognitive tasks (total N=35): deciding between faces or a vase in the ambiguous Rubin figure, detecting near-random motion in a dot kinematogram, and detecting a near-threshold auditory tone. By using long inter-stimulus intervals (>20s), we examined how pre-stimulus connectome states influenced the perception of upcoming ambiguous stimuli on a trial-by-trial basis across these tasks. Using Support Vector Machine (SVM) models, we demonstrated that pre-stimulus connectome states can predict the perception of upcoming ambiguous or threshold stimuli. At the connection level, we found that distinct sets of task-specific connections enabled these predictions for each task. No single connection was associated with perceptual outcomes across multiple tasks. Predictive connections in all tasks spanned both the task-relevant sensory modality network and high-order cognitive control networks. Only when averaged to the level of intrinsic connectivity networks did perceptually predictive connectivity show some similarity among pairs of tasks. Our findings highlight the functional significance of ongoing connectome states for moment-to-moment behavior, demonstrating that this relationship is largely context-dependent.