The intrinsic brain activity observed through resting-state functional magnetic resonance imaging (rs-fMRI) offers significant information to investigate underlying brain processes. Since traditional latency analysis models are limited to assessing macroscopic functional dynamics, the physical significance of fMRI-derived latency structures remains unexplored. To fill the gap, the spatiotemporal characteristics of fMRI are investigated using latency structure analysis in 469 neurologically healthy adults. After calculating the lagged cross-covariance of the time series, principal component analysis is applied to generate latency eigenvectors. These eigenvectors are associated with neural parameters derived from the biophysical model, revealing significant correlations with excitatory and inhibitory synaptic gating, recurrent connection, and excitation/inhibition balance. Association analyses with temporal and spatial features revealed that the latency eigenvectors are significantly associated with intrinsic neural timescale, and each latency eigenvector is paired with major brain axes from functional gradients, including the sensory-transmodal, visual-motor, and multiple demand-task-negative systems. These findings indicate that the latency model aligns with a seminal model of cortical hierarchy and intrinsic neural signaling. The clinical implications of latency eigenvectors are validated in autism spectrum disorder. This study enhances the understanding of the spatiotemporal characteristics of fMRI signals, providing insights into the physiology underlying the latency structures of brain signals.