The visual system extracts statistical information about the environment to manage noise, ensure perceptual stability and predict future events. These summary representations are able to inform sensory information received in subsequent times or in other regions of the visual field. This has been conceptualized in terms of Bayesian inference within the predictive coding framework. Nevertheless, contextual influence can also drive anti-Bayesian biases, as in sensory adaptation. Variance is a crucial statistical descriptor, yet relatively overlooked in ensemble vision research. We assessed the mechanisms whereby visual variability exerts and is subject to contextual modulation over time and across the visual field. Perceptual biases over time: serial dependence (SD) In a series of visual experiments, we examined SD on visual variance: the influence of the variance of previously presented ensembles in current variance judgments. We encountered two history-dependent biases: a positive bias exerted by recent presentations and a negative bias driven by less recent context. Contrary to claims that positive SD has low-level sensory origin, our experiments demonstrated a decisional bias requiring perceptual awareness and subject to time and capacity limitations. The negative bias was likely of sensory origin (adaptation). A two-layer model combining population codes and Bayesian Kalman filters replicated positive and negative effects in their approximate timescales. Perceptual biases across the visual field: Uniformity Illusion (UI) In UI, presentation of a pattern with uniform foveal components and more variable peripheral elements results in the latter taking the appearance of the foveal input. We studied the mechanistic basis of UI on orientation and determined that it arose without changes in sensory encoding at the primary visual cortex. Conclusions We studied perceptual biases on visual variability across space and time and found a combination of sensory negative and positive decisional biases, likely to handle the balance between change sensitivity and perceptual stability.