Serial Dependence: Connecting past and present
A neural signature of serial dependence has been found, which mirrors the attractive bias of visual information seen in behavioral experiments.
- University of California, San Diego, United States
When looking at a scene, we frequently blink, make eye and head movements, and have our vision blocked by moving external objects. Yet, instead of changing abruptly in response to these events, our visual experience remains fluid and continuous.
Research suggests that previously seen stimuli influence what we perceive in the present. A phenomenon, called serial dependence, indicates that the visual system anticipates that an object seen now is the same as the one seen a moment ago (Fischer and Whitney, 2014). This could make new objects look more similar to recently viewed stimuli than they are. For example, a green object may be perceived as teal by someone who has recently seen a similar blue object.
This so-called "attractive bias" applies to many features, from color and orientation to aesthetic value and facial expression – but only when successive stimuli are reasonably similar (Manassi et al., 2023). Rather than taking a rolling average of recent moments, it is thought that past and current visual inputs may be weighted in the brain and selectively integrated to strengthen relevant signals (Cicchini et al., 2024). This could explain why our visual experience seems continuous.
Scientists have been disputing the true ecological purpose of serial dependence and when it emerges in the processing stream. Determining the neural functions that give rise to the behavior would illuminate whether serial dependence actually changes what we see, or influences our memories and decisions that guide actions. Now, in eLife, Cora Fischer, Jochen Kaiser and Christoph Bledowski at Goethe University Frankfurt report having identified a neural signature that tracks the hallmark serial dependence effect (Fischer et al., 2024).
Recent research has detected neural traces of previous stimuli that may linger to induce attractive bias (Barbosa et al., 2020; Ranieri et al., 2022). Others have found that while behavioral reports showed bias towards previous stimuli, response patterns in the visual cortex showed bias away from the previous stimuli instead (Hajonides et al., 2023; Sheehan and Serences, 2022). This hints that a later, post-perceptual process must override this repulsion, but it had not been identified. Fischer et al. found evidence for later neural representations that are attractively shifted toward previous features, matching the attractive behavioral effect.
The team used whole-brain magnetoencephalography (MEG) and quantitative modelling to study the neural responses of humans completing a series of behavioral tasks. Participants underwent multiple trials in which they were presented with two consecutive stimuli (dots moving in one direction and different colored dots moving in another direction) and asked to remember the dot directions. After a short pause, they were cued to recall the direction of one of the dot colors. Memory of the cued stimulus showed typical serial dependence, where the reported directions were shifted a few degrees in the direction of the dots cued for recall in the previous trial (Figure 1).
Figure 1
Neural representations of current stimuli are biased by previously relevant content.
Participants completed a visual working memory task consisting of multiple trials (top). Each trial consisted of two sequential arrays of moving dot stimuli of different colors and motion directions (stimulus 1 and 2; represented as green and purple). Participants were asked to remember the direction of motion of each of the arrays. After a brief pause, a colored recall cue (purple) signaled participants to recall the direction of motion of the corresponding colored dots. Behavioral reports for this target stimulus (right) were biased toward the target direction of the recalled stimulus from the previous trial (blue; left), consistent with serial dependence. The second row shows schematics of the measured neural representations triggered by each dot stimulus below the corresponding stimulus, focusing only on the stimulus representation that was most relevant at each major task phase (also corresponding to the highlighted quantitative model reconstructions shown below). The neural response to the recall cue was influenced by the information for the previous trial target (right; represented as a combination of blue and purple dots). Schematic reconstructions of neural activity across each phase of the trial (bottom) display a peak that reflects the feature value (i.e., motion direction) with the greatest neural evidence. The current trial target reconstruction peaks at a value that is biased in the motion direction of the previous trial target. The reconstruction shows no bias when stimulus 2 (purple) is initially seen. When asked to recall the motion direction, the peak moves to reflect bias in the motion direction of the previous trial target (blue).
Fischer et al. then applied an inverted encoding model to the MEG data to measure representations of both the current remembered motion directions and the motion direction from the previous trial (Sprague et al., 2018). This analysis reconstructed an estimate of the stimulus information in the pattern of neural signals as a proxy for the underlying representation: a reconstruction with a peak centered around the true feature value would reflect an unbiased representation, but one shifted off-center would suggest a representation bias one way or another.
They found that motion representations appeared unbiased during encoding (i.e., perception). However, target representations became attractively biased during memory recall, when neural activity shifted toward the target direction from the previous trial. This would suggest participants were remembering a different direction than shown, i.e., they were remembering a direction that was more similar to the previous trial.
The findings of Fischer et al. are unique in capturing a neural bias that emerges at later, post-encoding time points and support the idea that serial dependence lags behind the initial perception. The timing could signify a decision-making phenomenon or indicate that memory representations become noisier and subject to bias over time. Neural representations of the motion direction from the previous trial target also appeared reactivated at the start of both a new trial and the cue for recall in the new trial, and representations of the first stimulus in a trial became attractively biased at the second stimulus onset – hinting that transition points between perceptual episodes may spur (or unveil) the integration process (Sols et al., 2017). However, these MEG signals were not specific to one brain region, so the critical circuits and computations warrant further investigation.
Serial dependence has captivated the vision science community and triggered burgeoning research into how this occasional bug imparts adaptive function. Now, neuroscience is catching up to reveal the physiology that supports serial dependence and our effortless perceptual experience. The work of Fischer et al. paves the way to probe how the visual system determines what information to assimilate, how it manifests at different levels of processing, and how it interacts with other features of the context and natural world.
References
-
Serial dependence in perceptionAnnual Review of Psychology 75:129–154.https://doi.org/10.1146/annurev-psych-021523-104939
-
Serial dependence in visual perceptionNature Neuroscience 17:738–743.https://doi.org/10.1038/nn.3689
-
Multiple and dissociable effects of sensory history on working-memory performanceThe Journal of Neuroscience 43:2730–2740.https://doi.org/10.1523/JNEUROSCI.1200-22.2023
-
Evidence of serial dependence from decoding of visual evoked potentialsThe Journal of Neuroscience 42:8817–8825.https://doi.org/10.1523/JNEUROSCI.1879-21.2022
Article and author information
Author details
Anastasia Kiyonaga
Publication history
Copyright
© 2024, Yang and Kiyonaga
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.
Metrics
-
- 475
- views
-
- 46
- downloads
-
- 0
- citations
Views, downloads and citations are aggregated across all versions of this paper published by eLife.
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Serial Dependence: Connecting past and present
eLife 13:e101212.
https://doi.org/10.7554/eLife.101212
Further reading
-
- Neuroscience
Recent studies suggest that calcitonin gene-related peptide (CGRP) neurons in the parabrachial nucleus (PBN) represent aversive information and signal a general alarm to the forebrain. If CGRP neurons serve as a true general alarm, their activation would modulate both passive nad active defensive behaviors depending on the magnitude and context of the threat. However, most prior research has focused on the role of CGRP neurons in passive freezing responses, with limited exploration of their involvement in active defensive behaviors. To address this, we examined the role of CGRP neurons in active defensive behavior using a predator-like robot programmed to chase mice. Our electrophysiological results revealed that CGRP neurons encode the intensity of aversive stimuli through variations in firing durations and amplitudes. Optogenetic activation of CGRP neuron during robot chasing elevated flight responses in both conditioning and retention tests, presumably by amyplifying the perception of the threat as more imminent and dangerous. In contrast, animals with inactivated CGRP neurons exhibited reduced flight responses, even when the robot was programmed to appear highly threatening during conditioning. These findings expand the understanding of CGRP neurons in the PBN as a critical alarm system, capable of dynamically regulating active defensive behaviors by amplifying threat perception, ensuring adaptive responses to varying levels of danger.
-
- Neuroscience
The central amygdala (CeA) has emerged as an important brain region for regulating both negative (fear and anxiety) and positive (reward) affective behaviors. The CeA has been proposed to encode affective information in the form of valence (whether the stimulus is good or bad) or salience (how significant is the stimulus), but the extent to which these two types of stimulus representation occur in the CeA is not known. Here, we used single cell calcium imaging in mice during appetitive and aversive conditioning and found that majority of CeA neurons (~65%) encode the valence of the unconditioned stimulus (US) with a smaller subset of cells (~15%) encoding the salience of the US. Valence and salience encoding of the conditioned stimulus (CS) was also observed, albeit to a lesser extent. These findings show that the CeA is a site of convergence for encoding oppositely valenced US information.
