Research Group Neural Circuits, Consciousness and Cognition, Max Planck Institute for Empirical Aesthetics, Frankfurt/Main, Germany.
Department of Neurology, NYU Grossman School of Medicine, New York, New York, USA.
J Vis. 2023 Aug 1;23(8):4. doi: 10.1167/jov.23.8.4.
According to a Bayesian framework, visual perception requires active interpretation of noisy sensory signals in light of prior information. One such mechanism, serial dependence, is thought to promote perceptual stability by assimilating current percepts with recent stimulus history. Combining a delayed orientation-adjustment paradigm with predictable (study 1) or unpredictable (study 2) task structure, we test two key predictions of this account in a novel context: first, that serial dependence should persist even in variable environments, and, second, that, within a given observer and context, this behavioral bias should be stable from one occasion to the next. Relying on data of 41 human volunteers and two separate experimental sessions, we confirm both hypotheses. Group-level, attractive serial dependence remained strong even in the face of volatile settings with multiple, unpredictable types of tasks, and, despite considerable interindividual variability, within-subject patterns of attractive and repulsive stimulus-history biases were highly stable from one experimental session to the next. In line with the hypothesized functional role of serial dependence, we propose that, together with previous work, our findings suggest the existence of a more general individual-specific fingerprint with which the past shapes current perception. Congruent with the Bayesian account, interindividual differences may then result from differential weighting of sensory evidence and prior information.
根据贝叶斯框架,视觉感知需要根据先前的信息主动解释嘈杂的感觉信号。一种这样的机制,即序列依赖,被认为通过将当前感知与最近的刺激历史同化来促进感知稳定性。我们结合了延迟定向调整范式和可预测(研究 1)或不可预测(研究 2)的任务结构,在一个新的环境中测试了该解释的两个关键预测:首先,即使在多变的环境中,序列依赖也应该持续存在;其次,在给定的观察者和环境中,这种行为偏差应该从一次到另一次是稳定的。我们依靠 41 名人类志愿者和两个单独的实验会话的数据,证实了这两个假设。即使面对具有多种不可预测类型任务的不稳定环境,群体水平的吸引力序列依赖仍然很强,并且尽管存在相当大的个体间变异性,但从一个实验会话到下一个实验会话,吸引力和排斥性刺激历史偏差的个体内模式非常稳定。与假设的序列依赖的功能作用一致,我们提出,与以前的工作一起,我们的发现表明存在一种更普遍的个体特异性指纹,过去塑造了当前的感知。与贝叶斯解释一致,个体间差异可能是由于对感觉证据和先验信息的不同加权造成的。