Fritsche et al., 2017, Opposite effects of recent history on perception and decision, Curr Biol
Working memory (WM) responses are influenced by perceptual distractors, be it an attractive or repulsive bias. The findings from serial dependence literature – where it’s proposed that perception is biased towards the stimulus from the previous trial – is highly relevant to this discussion. Additionaly, there is a finding of “tilt after-effect”, which contrasts with serial dependence and proposes that perception of an orientation is repelled away from a recently perceived orientation. This article does some really fascinating work to resolve these discrepant findings. Their ultimate resolve is that the attractive bias found in the serial dependence literature is not a result of altered perception, but instead a result of biased representations from the decision process (i.e., working memory representations).
The first experiment is a straight-forward replication of previous serial-dependence findings (i.e., orientation responses are biased towards that of the previous trial). The crucial element though is the response probe. In this common design, participants perceive a grating and then are prompted with a continuous probe and have a reasonable amount of time to select their response (about 3 seconds to response in the current study). What the current authors highlight nicely is that it is unclear if that attractive bias is happening during the perception of the next grating, or during the response period (i.e., decision process). Enter experiment 2.
Task diagram for the first 2 experiments. (experiment three's task was a slight modification of experiment two's, and four's a modification of one's)
Experiment 2 has a fantastic task design. To investigate if the bias is affecting perception, they suggest that they can’t use the continuous (and lengthy) response probe used in experiment 1. Thus, they switch to focusing on a 2AFC decision. Note that participants are still performing the continuous response, but now they are analyzing how the orientation from the first half of the trial biases perception during the 2AFC task. In this case, perception on nthe 2AFC task is repulsed away from the previous probe (i.e., “inducer” in the text).
Experiment 3 involves minor changes in task from experiment 2. Beyond replicating results from the previous experiments, I think the most critical finding here is that there is an attractive bias in the continuous responses towards the continuous responses from the previous trial but not the stimuli from the 2AFC decision, which occurs between the two continuous responses. That’s crazy!! It’s like there is a running tab of the decision representations that is separate from the briefly perceived probes. I think this is the kicker experiment that dissociates the two separate biases, and they do as well: “Therefore, these findings add strong support for the hypothesis that perception and post-perceptual decisions are concurrently biased into opposite directions.”
Experiment 4 provides strong evidence to support the “decision-based” or WM attractive biasing, by showing that the original biasing effects in experiment 1 are largely dependent on delay period. The longer the delay period, the stronger the attractive bias.
They have nice discussion about the different functional roles of perception and WM:
We speculate that these opposite effects on perception and post-perceptual decision may derive from the distinct goals of perception and decision-making processes: whereas perception may be optimized for detecting changes in the environment, decision processes may integrate over longer time periods to form stable representations.
…we speculate that the adaptation effect reported in the current study might be a consequence of one important goal of the visual system, namely to maximize sensitivity to changes in the physical environment.
Concurrently to adaptation effects in perception, the positive biases on post-perceptual working memory content and decisions might serve an important, yet distinct, functional role. Although it is vital to remain sensitive to changes in the immediate present, the physical environment does usually not change drastically over short timescales. Consequently, biasing working memory representations and thus perceptual decisions toward similar recent decisions would make those more robust to random fluctuations that do not reflect actual changes in the external environment.
- Just a note about how great the writing is. There is no Introduction, and the motivations are baked very nicely into the Results section.