Choice History Biases Depend on Environmental Stability and State Uncertainty

Perceptual decisions under uncertainty are often biased by the history of preceding events. For example, observers tend to repeat (or alternate) their judgments of the sensory environment more often than expected by chance (Braun, Urai, & Donner, 2018; Frund, Wichmann, & Macke, 2014). We test the idea that such choice history biases arise from the context-dependent accumulation of internal decision signals across trials (Glaze, Kable, & Gold, 2015). Observers performed a standard visual random dot motion discrimination task near psychophysical threshold in several different environments. Those were made up of different levels of auto-correlation between the stimulus categories in successive trials (Repetitive, Random, or Alternating), and the absence or presence of single-trial outcome feedback. Participants adjusted both the strength and the sign of their history biases to the environment. When no feedback was available this adjustment was driven by previous choices modulated by confidence. When feedback was provided the adjustment was predominantly based on previous stimuli.