CoChat

Word contexts enhance the neural representation of individual letters in early visual cortex

Open access
Heilbron, Micha, Richter, David, Ekman, Matthias, Hagoort, Peter, de Lange, Floris P. · 2020 · Nature Communications

The DOI matches its Crossref record; 68 of the 68 references it cites were found in a registry or an index.

Abstract

Visual context facilitates perception, but how this is neurally implemented remains unclear. One example of contextual facilitation is found in reading, where letters are more easily identified when embedded in a word. Bottom-up models explain this word advantage as a post-perceptual decision bias, while top-down models propose that word contexts enhance perception itself. Here, we arbitrate between these accounts by presenting words and nonwords and probing the representational fidelity of individual letters using functional magnetic resonance imaging. In line with top-down models, we find that word contexts enhance letter representations in early visual cortex. Moreover, we observe increased coupling between letter information in visual cortex and brain activity in key areas of the reading network, suggesting these areas may be the source of the enhancement. Our results provide evidence for top-down representational enhancement in word recognition, demonstrating that word contexts can modulate perceptual processing already at the earliest visual regions.

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/s41467-019-13996-4

Cites

68 of 68 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Experimental paradigm. a Example stimuli for each condition. Participants observed words or nonwords (i.e. orthographically illegal, unpronounceable strings) with a U or N as middle letter, resulting in four conditions. b Functional localiser. During the functional localiser, the key letters (U and N) were presented in isolation and without visual noise, while participants performed an irrelevant task at fixation. c Trial structure. We used a blocked design, in which each 14-s trial consisted of ten words or nonwords with a fixed middle letter. Participants performed an orthographic discrimination task on specific, prelearned targets that occurred once or occasionally twice per trial. Participants were trained in a separate session to perform the task while maintaining fixation at the centre of the screen.
Fig. 1 Experimental paradigm. a Example stimuli for each condition. Participants observed words or nonwords (i.e. orthographically illegal, unpronounceable strings) with a U or N as middle letter, resulting in four conditions. b Functional localiser. During the functional localiser, the key letters (U and N) were presented in isolation and without visual noise, while participants performed an irrelevant task at fixation. c Trial structure. We used a blocked design, in which each 14-s trial consisted of ten words or nonwords with a fixed middle letter. Participants performed an orthographic discrimination task on specific, prelearned targets that occurred once or occasionally twice per trial. Participants were trained in a separate session to perform the task while maintaining fixation at the centre of the screen.CC BY · PubMed Central
Fig. 2 Probing representational enhancement in neural network models and the brain. a Modelling representational enhancement in a hierarchical neural network model6,14. Stimuli used in the experiment were encoded into vectors of visual features and overlaid with Gaussian noise (bottom rows). Inputs were presented to a network with or without word-to-letter feedback connections. For both networks, representational strength was quantified from the distribution of activity levels of letter units for the third position (principle illustrated for the fifth letter, E). Solid circles indicate units (representing features, letters or words); lines indicate feedforward connections, and dotted lines with arrows indicate feedback connections. Note that we used a predictive coding formulation of the network14 but for simplicity only the state estimator (prediction) units are shown in the schematic (see Methods for details). b Quantifying representational letter enhancement using multivariate pattern analysis (MVPA). To probe letter representations in the brain, we used two MVPA techniques: classification (upper panel) and pattern correlation (lower panel).
Fig. 2 Probing representational enhancement in neural network models and the brain. a Modelling representational enhancement in a hierarchical neural network model6,14. Stimuli used in the experiment were encoded into vectors of visual features and overlaid with Gaussian noise (bottom rows). Inputs were presented to a network with or without word-to-letter feedback connections. For both networks, representational strength was quantified from the distribution of activity levels of letter units for the third position (principle illustrated for the fifth letter, E). Solid circles indicate units (representing features, letters or words); lines indicate feedforward connections, and dotted lines with arrows indicate feedback connections. Note that we used a predictive coding formulation of the network14 but for simplicity only the state estimator (prediction) units are shown in the schematic (see Methods for details). b Quantifying representational letter enhancement using multivariate pattern analysis (MVPA). To probe letter representations in the brain, we used two MVPA techniques: classification (upper panel) and pattern correlation (lower panel).CC BY · PubMed Central
Fig. 3 Word contexts enhance letter representations. a Theoretical predictions. We simulated 34 artificial ‘runs’ in which we exposed a network with (top–down model) and without (feedforward model) word-to-letter feedback connections to the experimental stimuli, and computed the average representational strength of the middle letter in word and nonword contexts. Note that the strong dissociation was observed despite the fact that the middle letter was well-above threshold in all conditions for both models (Supplementary Fig. 2). b Letter representation in early visual cortex of 34 human observers. Two multivariate pattern analysis methods (see Methods) revealed that neural representations of letters were enhanced in word compared to nonword contexts, supporting the top–down model. In both panels, grey dots represent individual simulated ‘runs’ (a) or individual participants (b). Lines represent paired differences. White dots, boxes and whiskers represent between-subject medians, quartiles and 1.5 interquartile ranges, respectively. Significance levels correspond to p < 0.01 (**) or p < 0.001 (***) in a paired, two-tailed Student’s t or Wilcoxon sign rank test.
Fig. 3 Word contexts enhance letter representations. a Theoretical predictions. We simulated 34 artificial ‘runs’ in which we exposed a network with (top–down model) and without (feedforward model) word-to-letter feedback connections to the experimental stimuli, and computed the average representational strength of the middle letter in word and nonword contexts. Note that the strong dissociation was observed despite the fact that the middle letter was well-above threshold in all conditions for both models (Supplementary Fig. 2). b Letter representation in early visual cortex of 34 human observers. Two multivariate pattern analysis methods (see Methods) revealed that neural representations of letters were enhanced in word compared to nonword contexts, supporting the top–down model. In both panels, grey dots represent individual simulated ‘runs’ (a) or individual participants (b). Lines represent paired differences. White dots, boxes and whiskers represent between-subject medians, quartiles and 1.5 interquartile ranges, respectively. Significance levels correspond to p < 0.01 (**) or p < 0.001 (***) in a paired, two-tailed Student’s t or Wilcoxon sign rank test.CC BY · PubMed Central
Fig. 4 Information-activation coupling analysis. a ROI-based coupling analysis. For two ROIs, GLMs were fitted to estimate coupling between early visual cortex classification evidence and regional BOLD amplitude for words and nonwords separately. We then tested for increased coupling (higher coefficients) in words compared to nonwords. b Illustration for example participant. A single (averaged) timecourse was extracted from each ROI and regressed against visual cortex classification evidence to test for increased slopes in words compared to nonwords. For illustration purposes, only the predicted slopes based on the regressor of interest are shown. Note that classification evidence was defined as probability, but here expressed in arbitrary units due to the whitening operation. c Whole-brain results. Same analysis as in a and visualised in b, but performed for each voxel independently. Resulting contrast images (word-nonword) were tested at the group level for increases in coupling in words compared to nonwords. This revealed statistically significant clusters (p < 0.05 FWE corrected), in the left pMTG and in the left IFG. Glass brain plot rendered with nilearn63. For a non-thresholded slice-by-slice rendering of the whole-brain results in c, see Supplementary Fig. 11. Grey dots indicate coefficients of individual participants, and lines the within-subject differences; white dots, boxes and whiskers are between-subject medians, quartiles and interquartile ranges, respectively. Significance stars correspond to p < 0.01 (**) or p < 0.001 (***) in a paired two-tailed t-test or Wilcoxon sign rank test. pMTG posterior medial temporal gyrus, IFG inferior frontal gyrus.
Fig. 4 Information-activation coupling analysis. a ROI-based coupling analysis. For two ROIs, GLMs were fitted to estimate coupling between early visual cortex classification evidence and regional BOLD amplitude for words and nonwords separately. We then tested for increased coupling (higher coefficients) in words compared to nonwords. b Illustration for example participant. A single (averaged) timecourse was extracted from each ROI and regressed against visual cortex classification evidence to test for increased slopes in words compared to nonwords. For illustration purposes, only the predicted slopes based on the regressor of interest are shown. Note that classification evidence was defined as probability, but here expressed in arbitrary units due to the whitening operation. c Whole-brain results. Same analysis as in a and visualised in b, but performed for each voxel independently. Resulting contrast images (word-nonword) were tested at the group level for increases in coupling in words compared to nonwords. This revealed statistically significant clusters (p < 0.05 FWE corrected), in the left pMTG and in the left IFG. Glass brain plot rendered with nilearn63. For a non-thresholded slice-by-slice rendering of the whole-brain results in c, see Supplementary Fig. 11. Grey dots indicate coefficients of individual participants, and lines the within-subject differences; white dots, boxes and whiskers are between-subject medians, quartiles and interquartile ranges, respectively. Significance stars correspond to p < 0.01 (**) or p < 0.001 (***) in a paired two-tailed t-test or Wilcoxon sign rank test. pMTG posterior medial temporal gyrus, IFG inferior frontal gyrus.CC 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 68 references

Research works printed but not linked: 1

  1. Dehaene, S. Reading in the Brain: The New Science of How We Read (Penguin, 2009).

Cited by

See all 48

Metadata

Authors
Heilbron, M., Richter, D., Ekman, M., Hagoort, P., & de Lange, F. P.
Affiliations
Radboud University Nijmegen (NL)
Corresponding author
Heilbron, Micha
Venue
Nature Communications
Published
2020-01-16
Volume
11
Issue
1
Pages
321-321
ISSN
2041-1723 (ISSN-L)
OpenAlex
W2991650583
Cited by
48
References
68
Citations by year
2026: 2 · 2025: 7 · 2024: 5 · 2023: 9 · 2022: 11 · 2021: 6 · 2020: 8
FWCI
12.81
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Communications (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Radboud Repository (Radboud University) (submittedVersion)
  • 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)
  • MPG.PuRe (Max Planck Society) (submittedVersion, other-oa)
  • Radboud Repository (Radboud University) (submittedVersion, other-oa)
  • Data Archiving and Networked Services (DANS) (submittedVersion)
Funders
  • Nederlandse Organisatie voor Wetenschappelijk Onderzoek (Uitslag beoordeling MaGW Onderzoekstalent 2017 406-17-516)
  • EC | Horizon 2020 Framework Programme (ERC Starting Grant 678286)
Topics
  • Social Sciences › Psychology › Developmental and Educational Psychology › Reading and Literacy Development
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Visual perception and processing mechanisms
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Neurobiology of Language and Bilingualism
Keywords
word context, top-down modulation, early visual cortex, word recognition, reading, pseudowords, bottom-up model, top-down model, reading network, perceptual enhancement
MeSH terms
Cognitive Neuroscience, Adult, Brain Mapping, Female, Form Perception, Humans, Language, Magnetic Resonance Imaging, Male, Pattern Recognition, Visual, Reading, Visual Cortex, Visual Perception, Neural Networks, Computer, 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: Word contexts enhance the neural representation of individual letters in early visual cortex. https://app.cochat.ai/papers/10.1038/s41467-019-13996-4
The paper
Heilbron, M., Richter, D., Ekman, M., Hagoort, P., & de Lange, F. P. (2020). Word contexts enhance the neural representation of individual letters in early visual cortex. Nature Communications. https://doi.org/10.1038/s41467-019-13996-4

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.