Visual serial dependence

Judge a tilted line, then another. Your second answer may drift toward the first; under other conditions it can drift away. This influence of recent judgments or images is visual serial dependence. Does the past change what you see, what you remember, or the answer you choose?[1] [2] [3]

Imagine a line at 20 degrees followed by one at 30 degrees. An answer of 25 degrees to the second line would be pulled toward the past; 35 degrees would be pushed away. These invented numbers illustrate attraction and repulsion, not measured effects. What determines their balance when a person reports a new orientation?

One possible contributor is a changed sensory state after earlier stimulation. Memory of the earlier image or a tendency to repeat a decision could also affect the report. A behavioral bias alone cannot tell these routes apart. Bias describes a systematic displacement; scatter describes how much repeated answers vary. Either can change without the other improving.

The 2025 mega-analysis challenged a claimed superiority effect in its tasks. In 2026, a study of four-to-nine-image sequences found that input predictability affected attraction. A separate EEG abstract reports less attraction, more repulsion and stronger subsequent sensory responses after omitted responses. Those neural associations do not demonstrate a causal mediator.[1] [2] [3]

Does the trace alter what you see, what you remember, or how you choose an answer? Omitting the earlier response changes later bias, but that result alone cannot locate the cause.[1] [2] [3]

Skipping a response changes more than the visible sequence: it changes the observer’s recent action and may alter engagement with the next image. The EEG result therefore leaves open whether a changed sensory state, a lingering decision, or both produce the new bias. The sequence experiment also found effects of current and previous input reliability under particular conditions; its four-to-nine-image range does not settle what happens in shorter sequences. An explanation must account for these boundaries rather than treat all past input as a single helpful trace.[1] [2] [3]

Seeing each image clearly does not make a series of judgments independent. The past can introduce a systematic error into the next answer. Understanding this matters for how perception balances continuity with detecting a real change; a smoother sequence of answers is not automatically a more accurate one.[1] [2] [3]

Visual judgments depend on a stream of information, not just an isolated snapshot. This problem links the ability to keep perception continuous with the ability to register novelty: treating a new tilt as similar to the old one can stabilize a report while obscuring a real difference.

The distinction between bias and scatter makes that tradeoff concrete. A series of answers can become less variable yet consistently miss the new orientation. Conversely, a directional bias can weaken without all uncertainty disappearing. Explaining both quantities would clarify when history helps a perceptual estimate and when it distorts it, without extending the tested tasks to every real-world judgment.[1] [2] [3]