MSense On 27 March 2026, Zhuanghua Shi gave an invited talk at the Max Planck Institute for Biological Cybernetics in Tübingen, hosted by Prof. Li Zhaoping. The talk, titled One Process, Two Biases: A Precision-Weighting Account of the Vierordt Effect and Serial Dependence, presented converging behavioral, computational, and neuroimaging evidence that central tendency and serial dependence in time perception arise from a single precision-weighted inference process rather than separate mechanisms.
Fittingly, the story begins in Tübingen itself. In 1868, Karl von Vierordt discovered here that short durations tend to be overestimated and long ones underestimated. What went unnoticed for over 150 years is that this classic finding resulted from an accidental methodological choice — randomizing stimulus sequences — that amplifies temporal biases. When the sequence is kept stable, the effect largely vanishes. The talk traced how this historical puzzle led to a unified Kalman filter model that captures both central tendency and serial dependence through precision-dependent modulation, supported by fMRI evidence showing the hippocampus gates how strongly prior trials influence current judgments.
The visit also brought stimulating discussions with Prof. Li Zhaoping and Dr. Assaf Breska. Zhaoping presented her latest findings on the central-peripheral dichotomy — a framework proposing that peripheral vision primarily serves orienting (“looking”) while central vision serves recognition (“seeing”), driven by a severe attentional bottleneck for top-down feedback in the brain. Particularly fascinating were the visual illusions her theory predicts: effects that appear in central vision but vanish in the periphery, or vice versa, offering direct experimental tests of the framework.