Rhythmic Sampling and Competition of Target and Distractor in a Motion Detection Task

Reviewer #1 (Public review):

Summary:

Using a combination of EEG and behavioural measurements, the authors investigate the degree to which processing of spatially-overlapping targets (coherent motion) and distractors (affective images) are sampled rhythmically and how this affects behaviour. They found that both target processing (via measurement of amplitude modulations of SSVEP amplitude to target frequency) and distractor processing (via MVPA decoding accuracy of bandpassed EEG relative to distractor SSVEP frequency) displayed a pronounced rhythm at ~1Hz, time-locked to stimulus onset. Furthermore, the relative phase of this target/distractor sampling predicted the accuracy of coherent motion detection across participants.

Strengths:

(1) The authors are addressing a very interesting question with respect to sampling of targets and distractors, using neurophysiological measurements to their advantage in order to parse out target and distractor processing.

(2) The general EEG analysis pipeline is sensible and well-described.

(3) The main result of rhythmic sampling of targets and distractors is striking and very clear even on a participant level.

(4) The authors have gone to quite a lot of effort to ensure the validity of their analyses, especially in the Supplementary Material.

(5) It is incredibly striking how the phases of both target and distractor processing are so aligned across trials for a given participant. I would have thought that any endogenous fluctuation in attention or stimulus processing like that would not be so phase aligned. I know there is literature on phase resetting in this context, the results seem very strong here and it is worth noting. The authors have performed many analyses to rule out signal processing artifacts, e.g., the sideband and beating frequency analyses.

Weaknesses:

(1) In general, the representation of target and distractor processing is a bit of a reach. Target processing is represented by SSVEP amplitude, which is most likely going to be related to the contrast of the dots, as opposed to representing coherent motion energy, which is the actual target. These may well be linked (e.g., greater attention to the coherent motion task might increase SSVEP amplitude), but I would call it a limitation of the interpretation. Decoding accuracy of emotional content makes sense as a measure of distractor processing, and the supplementary analysis comparing target SSVEP amplitude to distractor decoding accuracy is duly noted.

(2) Comparing SSVEP amplitude to emotional category decoding accuracy feels a bit like comparing apples with oranges. They have different units and scales and probably reflect different neural processes. Is the result the authors find not a little surprising in this context? This relationship does predict performance and is thus intriguing, but I think this methodological aspect needs to be discussed further. For example, is the phase relationship with behaviour a result of a complex interaction between different levels of processing (fundamental contrast vs higher order emotional processing)?

Reviewer #2 (Public review):

Summary:

In this study, Xiong et al. investigate whether rhythmic sampling - a process typically observed in the attended processing of visual stimuli - extends to task-irrelevant distractors. By using EEG with frequency tagging and multivariate pattern analysis (MVPA), they aimed to characterize the temporal dynamics of both target and distractor processing and examine whether these processes oscillate in time. The central hypothesis is that target and distractor processing occur rhythmically, and the phase relationship between these rhythms correlates with behavioral performance.

Major Strengths:

(1) The extension of rhythmic attentional sampling to include distractors is a novel and interesting question.

(2) The decoding of emotional distractor content using MVPA from SSVEP signals is an elegant solution to the problem of assessing distractor engagement in the absence of direct behavioral measures.

(3) The finding that relative phase (between 1 Hz target and distractor processes) predicts behavioral performance is compelling.

Major Weaknesses and Limitations:

(1) Incomplete Evidence for Rhythmicity at 1 Hz: The central claim of 1 Hz rhythmic sampling is insufficiently validated. The windowing procedure (0.5s windows with 0.25s step) inherently restricts frequency resolution, potentially biasing toward low-frequency components like 1 Hz. Testing different window durations or providing controls would significantly strengthen this claim.

(2) No-Distractor Control Condition: The study lacks a baseline or control condition without distractors. This makes it difficult to determine whether the distractor-related decoding signals or the 1 Hz effect reflect genuine distractor processing or more general task dynamics.

(3) Decoding Near Chance Levels: The pairwise decoding accuracies for distractor categories hover close to chance (~55%), raising concerns about robustness. While statistically above chance, the small effect sizes need careful interpretation, particularly when linked to behavior.

(4) No Clear Correlation Between SSVEP and Behavior: Neither target nor distractor signal strength (SSVEP amplitude) correlates with behavioral accuracy. The study instead relies heavily on relative phase, which - while interesting - may benefit from additional converging evidence.

(5) Phase-analysis: phase analysis is performed between different types of signals hindering their interpretability (time-resolved SSVEP amplitude and time-resolved decoding accuracy).

Appraisal of Aims and Conclusions:

The authors largely achieved their stated goal of assessing rhythmic sampling of distractors. However, the conclusions drawn - particularly regarding the presence of 1 Hz rhythmicity - rest on analytical choices that should be scrutinized further. While the observed phase-performance relationship is interesting and potentially impactful, the lack of stronger and convergent evidence on the frequency component itself reduces confidence in the broader conclusions.

Impact and Utility to the Field:

If validated, the findings will advance our understanding of attentional dynamics and competition in complex visual environments. Demonstrating that ignored distractors can be rhythmically sampled at similar frequencies to targets has implications for models of attention and cognitive control. However, the methodological limitations currently constrain the paper's impact.

Additional Context and Considerations:

(1) The use of EEG-fMRI is mentioned but not leveraged. If BOLD data were collected, even exploratory fMRI analyses (e.g., distractor modulation in visual cortex) could provide valuable converging evidence.

(2) In turn, removal of fMRI artifacts might introduce biases or alter the data. For instance, the authors might consider investigating potential fMRI artifact harmonics around 1 Hz to address concerns regarding induced spectral components.

Author response:

Reviewer 1:

(1) In general, the representation of target and distractor processing is a bit of a reach. Target processing is represented by SSVEP amplitude, which is most likely going to be related to the contrast of the dots, as opposed to representing coherent motion energy, which is the actual target. These may well be linked (e.g., greater attention to the coherent motion task might increase SSVEP amplitude), but I would call it a limitation of the interpretation. Decoding accuracy of emotional content makes sense as a measure of distractor processing, and the supplementary analysis comparing target SSVEP amplitude to distractor decoding accuracy is duly noted.

We agree with the reviewer. This is certainly a limitation and will be acknowledged as such in the revised manuscript.

(2) Comparing SSVEP amplitude to emotional category decoding accuracy feels a bit like comparing apples with oranges. They have different units and scales and probably reflect different neural processes. Is the result the authors find not a little surprising in this context? This relationship does predict performance and is thus intriguing, but I think this methodological aspect needs to be discussed further. For example, is the phase relationship with behaviour a result of a complex interaction between different levels of processing (fundamental contrast vs higher order emotional processing)?

Traditionally, the SSVEP amplitude at the distractor frequency is used to quantify distractor processing. Given that the target SSVEP amplitude is stronger than that for the distractor, it is possible that the distractor SSVEP amplitude is contaminated by the target SSVEP amplitude due to spectral power leakage; see Figure S4 for a demonstration of this. Because of this issue we therefore introduce the use of decoding accuracy as an index of distractor processing. This has not been done in the SSVEP literature. The lack of correlation between the distractor SSVEP amplitude and the distractor decoding accuracy, although it is kind of like comparing apples with oranges as pointed out by the reviewer, serves the purpose of showing that these two measures are not co-varying, and the use of decoding accuracy is free from the influence of the distractor SSVEP amplitude and thereby free from the influence by the target SSVEP amplitude. This is an important point. We will provide a more thorough discussion of this point in the revised manuscript.

Reviewer 2:

(1) Incomplete Evidence for Rhythmicity at 1 Hz: The central claim of 1 Hz rhythmic sampling is insufficiently validated. The windowing procedure (0.5s windows with 0.25s step) inherently restricts frequency resolution, potentially biasing toward low-frequency components like 1 Hz. Testing different window durations or providing controls would significantly strengthen this claim.

This is an important point. We plan to follow the reviewer’s suggestion and repeat our analysis using different window sizes to test the robustness of the observed 1Hz rhythmicity. In addition, we plan to also apply the Hilbert transform to extract time-point-by-time-point amplitude envelopes, which will provide a window-free estimation of the distractor strength and further validate the presence of the low-frequency 1Hz dynamics.

(2) No-Distractor Control Condition: The study lacks a baseline or control condition without distractors. This makes it difficult to determine whether the distractor-related decoding signals or the 1 Hz effect reflect genuine distractor processing or more general task dynamics.

We agree with the reviewer. This is certainly a limitation and will be acknowledged as such in the revised manuscript.

(3) Decoding Near Chance Levels: The pairwise decoding accuracies for distractor categories hover close to chance (~55%), raising concerns about robustness. While statistically above chance, the small effect sizes need careful interpretation, particularly when linked to behavior.

This is a good point. In addition to acknowledging this in the revised manuscript, we will carry out two additional analyses to test this issue further. First, we will implement a random permutation procedure, in which the trial labels are randomly shuffled and the null-hypothesis distribution for decoding accuracy is built, and compare the decoding accuracy from the actual data to this distribution. Second, we will perform a temporal generalization analysis to examine whether the neural representations of the distractor drift over the course of an entire trial, which is 11 seconds long. Recent studies suggest that even when the stimulus stays the same, their neural representations may drift over time.

(4) No Clear Correlation Between SSVEP and Behavior: Neither target nor distractor signal strength (SSVEP amplitude) correlates with behavioral accuracy. The study instead relies heavily on relative phase, which - while interesting - may benefit from additional converging evidence.

We felt that what the reviewer pointed out is actually the main point of our study, namely, it is not the overall target or distractor strength that matters for behavior, it is their temporal relationship that matters for behavior. This reveals a novel neuroscience principle that has not been reported in the past. We will stress this point further in the revised manuscript.

(5) Phase-analysis: phase analysis is performed between different types of signals hindering their interpretability (time-resolved SSVEP amplitude and time-resolved decoding accuracy).

The time-resolved SSVEP amplitude is used to index the temporal dynamics of target processing whereas the time-resolved decoding accuracy is used to index the temporal dynamics of distractor processing. As such, they can be compared, using relative phase for example, to examine how temporal relations between the two types of processes impact behavior. This said, we do recognize the reviewer’s concern that these two processes are indexed by two different types of signals. We plan to normalize each time course, make them dimensionless, and then compute the temporal relations between them.

Appraisal of Aims and Conclusions:

The authors largely achieved their stated goal of assessing rhythmic sampling of distractors. However, the conclusions drawn - particularly regarding the presence of 1 Hz rhythmicity - rest on analytical choices that should be scrutinized further. While the observed phase-performance relationship is interesting and potentially impactful, the lack of stronger and convergent evidence on the frequency component itself reduces confidence in the broader conclusions.

Impact and Utility to the Field:

If validated, the findings will advance our understanding of attentional dynamics and competition in complex visual environments. Demonstrating that ignored distractors can be rhythmically sampled at similar frequencies to targets has implications for models of attention and cognitive control. However, the methodological limitations currently constrain the paper's impact.

Thanks for these comments and positive assessment of our work’s potential implications and impact. We will try our best in the revision process to address the concerns.

Additional Context and Considerations:

(1) The use of EEG-fMRI is mentioned but not leveraged. If BOLD data were collected, even exploratory fMRI analyses (e.g., distractor modulation in visual cortex) could provide valuable converging evidence.

Indeed, leveraging fMRI data in EEG studies would be very beneficial, as having been demonstrated in our previous work. However, given that this study concerns the temporal relationship between target and distractor processing, it is felt that fMRI, given its well-known limitation in temporal resolution, has limited potential to contribute. We will be exploring this rich dataset in other ways where the two modalities are integrated to gain more insights not possible with either modality used alone.

(2) In turn, removal of fMRI artifacts might introduce biases or alter the data. For instance, the authors might consider investigating potential fMRI artifact harmonics around 1 Hz to address concerns regarding induced spectral components.

We have done extensive work in the area of simultaneous EEG-fMRI and have not encountered artifacts with a 1Hz rhythmicity. Also, the fact that the temporal relations between target processing and distractor processing at 1Hz predict behavior is another indication that the 1Hz rhythmicity is a neuroscientific effect not an artifact. However, we will be looking into this carefully and address this in the revision process.