CoChat

The DOI matches its Crossref record; 52 of the 53 references it cites were found in a registry or an index.

Discrete and continuous mechanisms of temporal selection in rapid visual streams

Open access
Marti, Sébastien, Dehaene, Stanislas · 2017 · Nature Communications

Abstract

Humans can reliably detect a target picture even when tens of images are flashed every second. Here we use magnetoencephalography to dissect the neural mechanisms underlying the dynamics of temporal selection during a rapid serial visual presentation task. Multivariate decoding algorithms allow us to track the overlapping brain responses induced by each image in a rapid visual stream. The results show that temporal selection involves a sequence of gradual followed by all-or-none stages: (i) all images first undergo the same parallel processing pipeline; (ii) starting around 150 ms, responses to multiple images surrounding the target are continuously amplified in ventral visual areas; (iii) only the images that are subsequently reported elicit late all-or-none activations in visual and parietal areas around 350 ms. Thus, multiple images can cohabit in the brain and undergo efficient parallel processing, but temporal selection also isolates a single one for amplification and report.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s41467-017-02079-x

Cites

52 of 52 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Experimental design and behavioral evidence for gradual temporal selection. a Schematic representation of parallel and serial model of temporal selection. From top to bottom are represented the stream of stimuli, the activity in sensory areas, the gradual selection hypothesis, and the discrete selection hypothesis. Each colored curve represents the predicted brain response for one stimulus in the RSVP at a certain stage of processing. In both models, stimuli are integrated in parallel at the sensory level. In the gradual selection model, attention selects multiple stimuli simultaneously. All stimuli benefit from attentional enhancement but at varying degrees. By contrast, in the discrete selection model, attention is all-or-none and only one stimulus is selected for further processing. b Schematic representation of the tasks. In the localizer task (upper part), one stimulus was presented and subjects had to indicate their category. During the dual-task (lower part), stimuli were presented in a rapid visual stream. Subjects were instructed to first determine the color of the very first stimulus (T1) and then identify the image surrounded by a green square (T2). Subjects made systematically three guesses for T2 by order of preference (i.e., from the most probably correct response to the least probably correct). For copyright reasons, the stimuli used in the experiment are replaced by representative images. c Distribution of subjects’ reports. Rows represent Guess 1, 2, and 3 and columns represent inter-target lags (blue, magenta, red and orange for Lag 1, 3, 7, and 9, respectively). For each position in the RSVP, the proportion of trials was compared to the chance level. A significant difference is indicated by an asterisk (*P < .05; **P < .01; ***P < .001). Error bars represent standard error to the mean. d Average ( ± standard error) mode and variance (e) of subjects’ report distributions as a function of inter-target lag. For Guess 2 and 3 panels, trials were split according to Guess 1 position. Closed symbols: subjects reported T-1 as Guess 1; Open symbols: subjects reported T + 1 as Guess 1. Asterisks represent results of a repeated-measures ANOVA with the inter-target lag as a within-subject factor: *P < .05; **P < .01; ***P < .001
Fig. 1 Experimental design and behavioral evidence for gradual temporal selection. a Schematic representation of parallel and serial model of temporal selection. From top to bottom are represented the stream of stimuli, the activity in sensory areas, the gradual selection hypothesis, and the discrete selection hypothesis. Each colored curve represents the predicted brain response for one stimulus in the RSVP at a certain stage of processing. In both models, stimuli are integrated in parallel at the sensory level. In the gradual selection model, attention selects multiple stimuli simultaneously. All stimuli benefit from attentional enhancement but at varying degrees. By contrast, in the discrete selection model, attention is all-or-none and only one stimulus is selected for further processing. b Schematic representation of the tasks. In the localizer task (upper part), one stimulus was presented and subjects had to indicate their category. During the dual-task (lower part), stimuli were presented in a rapid visual stream. Subjects were instructed to first determine the color of the very first stimulus (T1) and then identify the image surrounded by a green square (T2). Subjects made systematically three guesses for T2 by order of preference (i.e., from the most probably correct response to the least probably correct). For copyright reasons, the stimuli used in the experiment are replaced by representative images. c Distribution of subjects’ reports. Rows represent Guess 1, 2, and 3 and columns represent inter-target lags (blue, magenta, red and orange for Lag 1, 3, 7, and 9, respectively). For each position in the RSVP, the proportion of trials was compared to the chance level. A significant difference is indicated by an asterisk (*P < .05; **P < .01; ***P < .001). Error bars represent standard error to the mean. d Average ( ± standard error) mode and variance (e) of subjects’ report distributions as a function of inter-target lag. For Guess 2 and 3 panels, trials were split according to Guess 1 position. Closed symbols: subjects reported T-1 as Guess 1; Open symbols: subjects reported T + 1 as Guess 1. Asterisks represent results of a repeated-measures ANOVA with the inter-target lag as a within-subject factor: *P < .05; **P < .01; ***P < .001CC BY · PubMed Central
Fig. 2 Time-resolved decoding of stimulus categories. a Right: Example of event-related magnetic fields elicited by the presentation of images during the localizer task. Left: Subtraction between brain sources of each category (face, place, body parts and objects) and the mean of the other categories presented in Z scores according to baseline and projected on a flattened standard brain. Left to right columns represent bottom and left views of the brain at specific time points (50–450 ms from bottom to top). b Performance of classifiers trained to separate the four stimulus categories. Top graph: One classifier was trained at each time sample and tested on the same time sample. The shaded gray area represents the standard error to the mean across subjects. Right graphs: Classifiers were trained at specific time samples (from 120 to 370 ms) and tested on all other time samples. Results from signed rank tests comparing classification performance to chance are represented by the thick line below the x axis with darker colors representing lower P-values. False discovery rate (FDR) correction for multiple comparisons was applied across specific training times (120 to 370 ms), all testing times and conditions (RSVP and localizer). Matrix plot: Classifiers were systematically trained on each time sample and tested on all others. The color code represents the classification performance and the dotted line the diagonal of the matrix. c Classifiers trained on the localizer task were directly applied to the RSVP task. Line plots, matrix plot and statistics are as in b. d Contour plot representing classification performance significantly above chance obtained for each stimulus and averaged across subjects. Colors represent the successive stimuli in the RSVP (except T1) depicted below by a small rectangle. Dotted lines represent the diagonals of each matrix plot. FDR correction for multiple comparisons was applied across all training times, testing times and stimulus positions
Fig. 2 Time-resolved decoding of stimulus categories. a Right: Example of event-related magnetic fields elicited by the presentation of images during the localizer task. Left: Subtraction between brain sources of each category (face, place, body parts and objects) and the mean of the other categories presented in Z scores according to baseline and projected on a flattened standard brain. Left to right columns represent bottom and left views of the brain at specific time points (50–450 ms from bottom to top). b Performance of classifiers trained to separate the four stimulus categories. Top graph: One classifier was trained at each time sample and tested on the same time sample. The shaded gray area represents the standard error to the mean across subjects. Right graphs: Classifiers were trained at specific time samples (from 120 to 370 ms) and tested on all other time samples. Results from signed rank tests comparing classification performance to chance are represented by the thick line below the x axis with darker colors representing lower P-values. False discovery rate (FDR) correction for multiple comparisons was applied across specific training times (120 to 370 ms), all testing times and conditions (RSVP and localizer). Matrix plot: Classifiers were systematically trained on each time sample and tested on all others. The color code represents the classification performance and the dotted line the diagonal of the matrix. c Classifiers trained on the localizer task were directly applied to the RSVP task. Line plots, matrix plot and statistics are as in b. d Contour plot representing classification performance significantly above chance obtained for each stimulus and averaged across subjects. Colors represent the successive stimuli in the RSVP (except T1) depicted below by a small rectangle. Dotted lines represent the diagonals of each matrix plot. FDR correction for multiple comparisons was applied across all training times, testing times and stimulus positionsCC BY · PubMed Central
Fig. 3 MEG evidence for two types of target-specific brain responses: sustained activation of sensory processes, and access to an additional late processing stage. a Temporal generalization matrices for target (upper panel) and non-target (lower panel) stimuli averaged across inter-target lags. Gray areas correspond to below chance classification performance. Dashed arrows represent selected training time depicted in b. b Classification performance as a function of time for classifiers trained at 170 and 370 ms. Results from signed rank tests comparing classification performance to chance are represented by the thick line below the x axis. Signed rank tests comparing target and non-target conditions are represented by areas filled with red, darker colors representing lower P-values. FDR correction for multiple comparisons was applied across specific training times (120–370 ms), testing times and conditions. c Source localization of activation patterns computed from the support vector machine (SVM) weights (see method) indicating the location of the most informative activity for each stimulus category; 200–250 ms and 350–400 ms averaged training times are represented. For display purposes, data points were smoothed using a moving average with a window of three samples
Fig. 3 MEG evidence for two types of target-specific brain responses: sustained activation of sensory processes, and access to an additional late processing stage. a Temporal generalization matrices for target (upper panel) and non-target (lower panel) stimuli averaged across inter-target lags. Gray areas correspond to below chance classification performance. Dashed arrows represent selected training time depicted in b. b Classification performance as a function of time for classifiers trained at 170 and 370 ms. Results from signed rank tests comparing classification performance to chance are represented by the thick line below the x axis. Signed rank tests comparing target and non-target conditions are represented by areas filled with red, darker colors representing lower P-values. FDR correction for multiple comparisons was applied across specific training times (120–370 ms), testing times and conditions. c Source localization of activation patterns computed from the support vector machine (SVM) weights (see method) indicating the location of the most informative activity for each stimulus category; 200–250 ms and 350–400 ms averaged training times are represented. For display purposes, data points were smoothed using a moving average with a window of three samplesCC BY · PubMed Central
Fig. 4 Evidence for early gradual selection: early brain responses are selectively sustained over time as a function of target proximity. a Panels represent temporal generalization matrices for the target stimulus (T) and distractors at positions T−4 to T + 2 at long inter-target lag. The color scale is as in Fig. 3a except that below-chance classification performance is not represented for visibility purposes. b Classification performance as a function of time for a classifier trained at 170 ms after stimulus onset. Filled areas represent time samples at which the classification performance was significantly different from chance (Signed-rank tests, corrected for multiple comparisons using FDR across stimulus positions, training times (170 and 370 ms) and testing times). Colored dotted lines represent stimuli onsets. c Mean classification performance of the same classifier but averaged over time (400–550 ms). The dotted line represents chance. d Effect of inter-target lag. Classification performance averaged over Guess 1 and 2 and over time (400–550 ms) as a function of inter-target lag. e Effect of report. Mean classification performance between 400 and 550 ms for stimuli reported as Guess 1, 2, 3 and unreported stimuli. The red dotted line represents chance. Asterisks represent results from signed rank tests (*P FDR < 0.05; **P FDR < 0.01; ***P FDR <0.001, FDR corrected across stimulus positions). For display purposes, data points were smoothed using a moving average with a window of three samples
Fig. 4 Evidence for early gradual selection: early brain responses are selectively sustained over time as a function of target proximity. a Panels represent temporal generalization matrices for the target stimulus (T) and distractors at positions T−4 to T + 2 at long inter-target lag. The color scale is as in Fig. 3a except that below-chance classification performance is not represented for visibility purposes. b Classification performance as a function of time for a classifier trained at 170 ms after stimulus onset. Filled areas represent time samples at which the classification performance was significantly different from chance (Signed-rank tests, corrected for multiple comparisons using FDR across stimulus positions, training times (170 and 370 ms) and testing times). Colored dotted lines represent stimuli onsets. c Mean classification performance of the same classifier but averaged over time (400–550 ms). The dotted line represents chance. d Effect of inter-target lag. Classification performance averaged over Guess 1 and 2 and over time (400–550 ms) as a function of inter-target lag. e Effect of report. Mean classification performance between 400 and 550 ms for stimuli reported as Guess 1, 2, 3 and unreported stimuli. The red dotted line represents chance. Asterisks represent results from signed rank tests (*P FDR < 0.05; **P FDR < 0.01; ***P FDR <0.001, FDR corrected across stimulus positions). For display purposes, data points were smoothed using a moving average with a window of three samplesCC BY · PubMed Central
Fig. 5 Evidence for late discrete selection: Late brain responses are sustained only for reported stimuli. a Classification performance as a function of time for target stimulus and nearby distractors for a classifier trained at 370 ms (colors and statistics are as in Fig. 4b). Colored dotted lines represent stimuli onsets. b Mean classification accuracy over a time period of 400–550 ms (as in Fig. 4c). c Effect of inter-target lag. Classification performance averaged over Guess 1 and 2 and over time (400–550 ms) as a function of inter-target lag. d Effect of report. Mean classification performance between 400 and 550 ms for stimuli reported as Guess 1, 2, 3, and unreported stimuli. The red dotted line represents chance. Asterisks represent results from signed rank tests (*P < 0.05; **P < 0.01; ***P < 0.001, FDR corrected across conditions). For display purposes, data points were smoothed using a moving average with a window of three samples
Fig. 5 Evidence for late discrete selection: Late brain responses are sustained only for reported stimuli. a Classification performance as a function of time for target stimulus and nearby distractors for a classifier trained at 370 ms (colors and statistics are as in Fig. 4b). Colored dotted lines represent stimuli onsets. b Mean classification accuracy over a time period of 400–550 ms (as in Fig. 4c). c Effect of inter-target lag. Classification performance averaged over Guess 1 and 2 and over time (400–550 ms) as a function of inter-target lag. d Effect of report. Mean classification performance between 400 and 550 ms for stimuli reported as Guess 1, 2, 3, and unreported stimuli. The red dotted line represents chance. Asterisks represent results from signed rank tests (*P < 0.05; **P < 0.01; ***P < 0.001, FDR corrected across conditions). For display purposes, data points were smoothed using a moving average with a window of three samplesCC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 52 references

Cited by

See all 127

Metadata

Authors
Marti, S., & Dehaene, S.
Affiliations
Université Paris-Sud (FR)
Corresponding author
Marti, Sébastien
Venue
Nature Communications
Published
2017-12-05
Volume
8
Issue
1
Pages
1955-1955
ISSN
2041-1723 (ISSN-L)
OpenAlex
W2768186777
Cited by
127
References
53
Citations by year
2026: 5 · 2025: 5 · 2024: 20 · 2023: 10 · 2022: 9 · 2021: 18 · 2020: 26 · 2019: 26 · 2018: 8
FWCI
9.41
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Communications (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • Institut National de la Santé et de la Recherche Médicale
  • Fondation pour la Recherche Médicale
Topics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Visual perception and processing mechanisms
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Visual Attention and Saliency Detection
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Neural and Behavioral Psychology Studies
Keywords
temporal selection, rapid serial visual presentation, target detection, visual attention, parallel computing, ventral visual cortex, parietal cortex, multivariate decoding
MeSH terms
Adult, Algorithms, Brain, Female, Humans, Magnetic Resonance Imaging, Male, Parietal Lobe, Photic Stimulation, Visual Cortex, Visual Pathways, Magnetoencephalography, Multivariate Analysis, Young Adult

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Discrete and continuous mechanisms of temporal selection in rapid visual streams. https://app.cochat.ai/papers/10.1038/s41467-017-02079-x
The paper
Marti, S., & Dehaene, S. (2017). Discrete and continuous mechanisms of temporal selection in rapid visual streams. Nature Communications. https://doi.org/10.1038/s41467-017-02079-x

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.