CoChat

Intrinsic Functional Connectivity is Organized as Three Interdependent Gradients

Open access

The DOI matches its Crossref record; 153 of the 154 references it cites were found in a registry or an index.

Abstract

The intrinsic functional architecture of the brain supports moment-to-moment maintenance of an internal model of the world. We hypothesized and found three interdependent architectural gradients underlying the organization of intrinsic functional connectivity within the human cerebral cortex. We used resting state fMRI data from two samples of healthy young adults (N’s = 280 and 270) to generate functional connectivity maps of 109 seeds culled from published research, estimated their pairwise similarities, and multidimensionally scaled the resulting similarity matrix. We discovered an optimal three-dimensional solution, accounting for 98% of the variance within the similarity matrix. The three dimensions corresponded to three gradients, which spatially correlate with two functional features (external vs. internal sources of information; content representation vs. attentional modulation) and one structural feature (anatomically central vs. peripheral) of the brain. Remapping the three dimensions into coordinate space revealed that the connectivity maps were organized in a circumplex structure, indicating that the organization of intrinsic connectivity is jointly guided by graded changes along all three dimensions. Our findings emphasize coordination between multiple, continuous functional and anatomical gradients, and are consistent with the emerging predictive coding perspective.

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/s41598-019-51793-7

Cites

153 of 153 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Both goodness-of-fit estimates were highly replicable across the discovery and replication groups and suggested that a three-dimensional solution was optimal. Stress was plotted as a function of the number of estimated dimensions. Lower stress indicates better fit. The scree plot of normalized stress had an “elbow” when dimensionality was at 3, since further addition of dimensions did not substantially reduce normalized stress. Dashed line indicates normalized stress of 0.05. DAF was plotted as a function of the number of estimated dimensions. Higher DAF indicates better fit. Dashed line indicates DAF of 0.98.
Figure 1 Both goodness-of-fit estimates were highly replicable across the discovery and replication groups and suggested that a three-dimensional solution was optimal. Stress was plotted as a function of the number of estimated dimensions. Lower stress indicates better fit. The scree plot of normalized stress had an “elbow” when dimensionality was at 3, since further addition of dimensions did not substantially reduce normalized stress. Dashed line indicates normalized stress of 0.05. DAF was plotted as a function of the number of estimated dimensions. Higher DAF indicates better fit. Dashed line indicates DAF of 0.98.CC BY · PubMed Central
Figure 2 MDS results revealed that intrinsic connectivity patterns followed a circumplex structure of similarity. We calculated intrinsic connectivity maps based on 109 seeds across five most frequently identified network motifs in the published literature: attention (red), default mode (yellow), executive (green), exteroceptive (light blue), and salience (dark blue). Each point in the scatterplot represents a connectivity map. We plotted (A) Dimension 1 vs. Dimension 2, (B) Dimension 1 vs. Dimension 3, and (C) Dimension 2 vs Dimension 3 to facilitate interpretation.
Figure 2 MDS results revealed that intrinsic connectivity patterns followed a circumplex structure of similarity. We calculated intrinsic connectivity maps based on 109 seeds across five most frequently identified network motifs in the published literature: attention (red), default mode (yellow), executive (green), exteroceptive (light blue), and salience (dark blue). Each point in the scatterplot represents a connectivity map. We plotted (A) Dimension 1 vs. Dimension 2, (B) Dimension 1 vs. Dimension 3, and (C) Dimension 2 vs Dimension 3 to facilitate interpretation.CC BY · PubMed Central
Figure 3 Gradient maps visualized on the brain surfaces. For each dimension, we created a gradient map by weighting every connectivity map by its dimension loading and summing across all weighted maps to create a composite (i.e., a weighted sum akin to factor scores). Gradient 1 (external vs. internal) captured a functional contrast between processing information from the external environment and the internal milieu. Gradient 2 (modulation vs. representation) captured a functional contrast between attentional modulation and content representation. Gradient 3 (anatomical centrality) captured a structural contrast between spatially central nodes and peripheral nodes. We visualized the gradient maps on inflated brain surfaces using Caret154.
Figure 3 Gradient maps visualized on the brain surfaces. For each dimension, we created a gradient map by weighting every connectivity map by its dimension loading and summing across all weighted maps to create a composite (i.e., a weighted sum akin to factor scores). Gradient 1 (external vs. internal) captured a functional contrast between processing information from the external environment and the internal milieu. Gradient 2 (modulation vs. representation) captured a functional contrast between attentional modulation and content representation. Gradient 3 (anatomical centrality) captured a structural contrast between spatially central nodes and peripheral nodes. We visualized the gradient maps on inflated brain surfaces using Caret154.CC BY · PubMed Central
Figure 4 Gradient 3 corresponded to anatomical centrality. (A) Dimension 3 loadings correlated positively with anatomical centrality (N = 109; r(107) = 0.676, p < 0.001). (B) Anatomical centrality for each seed was computed as (maximal distance – node distance)/maximal distance, where maximal distance is the distance for the node with maximal distance from the anterior commissure, with 0 indicating minimum anatomical centrality (at maximum distance from the anterior commissure) and 1 indicating maximum centrality (at the anterior commissure). The anterior commissure was used as a proxy for the center of the brain since it is approximately equidistant from the most distal points of the cerebrum. We visualized anatomical centrality values on inflated brain surfaces using Caret154.
Figure 4 Gradient 3 corresponded to anatomical centrality. (A) Dimension 3 loadings correlated positively with anatomical centrality (N = 109; r(107) = 0.676, p < 0.001). (B) Anatomical centrality for each seed was computed as (maximal distance – node distance)/maximal distance, where maximal distance is the distance for the node with maximal distance from the anterior commissure, with 0 indicating minimum anatomical centrality (at maximum distance from the anterior commissure) and 1 indicating maximum centrality (at the anterior commissure). The anterior commissure was used as a proxy for the center of the brain since it is approximately equidistant from the most distal points of the cerebrum. We visualized anatomical centrality values on inflated brain surfaces using Caret154.CC BY · PubMed Central
Figure 5 Predictive processing framework. Starting with initial conditions in the body and in the world (T0), the brain is thought to continually predict forward in time (T1), preparing changes in the body’s internal systems to support upcoming motor actions. Efferent copies of these motor and visceromotor preparations function as their predicted sensory consequences, cascading to sensory systems to modulate the firing of sensory neurons in advance of incoming sensory inputs. Sensory inputs from the body and the world are continuously compared to prediction signals. If different, prediction errors are sent to update the brain’s internal model for future occasions. This framework is based on a structural model of cortico-cortical connections whereby predictions flow from less to more laminated (i.e., layered) cortices (‘feedback connections’), whereas prediction errors flow in the opposite direction (‘feedforward connections’)155,156. This structural model is consistent with a gradient- but not module-based organization scheme for intrinsic connectivity. This is because the whole brain is thought to participate in predictive processing, not by separating into mental modules, but by operating on a continuous two-way hierarchy. In the feedback direction, abstract predictions are unpacked into particular sensory simulations; at the same time in the feedforward direction, sensory information is compressed to be integrated with the brain’s internal model (see review in7). Figure adapted from104.
Figure 5 Predictive processing framework. Starting with initial conditions in the body and in the world (T0), the brain is thought to continually predict forward in time (T1), preparing changes in the body’s internal systems to support upcoming motor actions. Efferent copies of these motor and visceromotor preparations function as their predicted sensory consequences, cascading to sensory systems to modulate the firing of sensory neurons in advance of incoming sensory inputs. Sensory inputs from the body and the world are continuously compared to prediction signals. If different, prediction errors are sent to update the brain’s internal model for future occasions. This framework is based on a structural model of cortico-cortical connections whereby predictions flow from less to more laminated (i.e., layered) cortices (‘feedback connections’), whereas prediction errors flow in the opposite direction (‘feedforward connections’)155,156. This structural model is consistent with a gradient- but not module-based organization scheme for intrinsic connectivity. This is because the whole brain is thought to participate in predictive processing, not by separating into mental modules, but by operating on a continuous two-way hierarchy. In the feedback direction, abstract predictions are unpacked into particular sensory simulations; at the same time in the feedforward direction, sensory information is compressed to be integrated with the brain’s internal model (see review in7). Figure adapted from104.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 153 references

Research works printed but not linked: 4

  1. Davison, M. L. Multidimensional Scaling. (John Wiley & Sons Inc, 1983).
  2. Guttman, L. In Mathematical thinking in the social sciences (ed Lazarsfeld, P. F.) 258–348 (Columbia University Press, 1957).
  3. Larsen, R. J. & Diener, E. Promises and problems with the circumplex model of emotion., Vol. 13 25–59 (Sage, 1992).
  4. Mesulam, M. M. Patterns in behavioral neuroanatomy. Association area, the limbic system, and hemispheric specialization., 1–70 (Oxford University Press, 1985).

Cited by

See all 48

Metadata

Authors
Zhang, J., Abiose, O., Katsumi, Y., Touroutoglou, A., Dickerson, B. C., & Barrett, L. F.
Affiliations
  • Northeastern University (US)
  • Massachusetts General Hospital (US)
  • Harvard University (US)
Corresponding author
Barrett, Lisa Feldman
Venue
Scientific Reports
Published
2019-11-04
Volume
9
Issue
1
Pages
15976-15976
ISSN
2045-2322 (ISSN-L)
OpenAlex
W2986902590
Cited by
47
References
154
Citations by year
2026: 1 · 2025: 11 · 2024: 7 · 2023: 4 · 2022: 9 · 2021: 8 · 2020: 6 · 2019: 1
FWCI
3.49
Type
journal article
Language
en
Open access
Licence
Locations
  • Scientific Reports (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
Funders
  • U.S. Department of Health & Human Services | NIH | National Institute of Mental Health (R01 MH113234, R01 MH109464)
  • U.S. Department of Health & Human Services | NIH | NCI | Division of Cancer Epidemiology and Genetics, National Cancer Institute (U01 CA193632)
Topics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Functional Brain Connectivity Studies
  • Health Sciences › Medicine › Radiology, Nuclear Medicine and Imaging › Advanced Neuroimaging Techniques and Applications
  • Health Sciences › Medicine › Radiology, Nuclear Medicine and Imaging › Advanced MRI Techniques and Applications
Keywords
intrinsic functional connectivity, resting-state fMRI, cerebral cortex, functional gradients, predictive coding, content representation, attentional modulation
MeSH terms
Adolescent, Adult, Brain, Brain Mapping, Female, Humans, Magnetic Resonance Imaging, Male, Stress, Physiological, Young Adult, Connectome

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: Intrinsic Functional Connectivity is Organized as Three Interdependent Gradients. https://app.cochat.ai/papers/10.1038/s41598-019-51793-7
The paper
Zhang, J., Abiose, O., Katsumi, Y., Touroutoglou, A., Dickerson, B. C., & Barrett, L. F. (2019). Intrinsic Functional Connectivity is Organized as Three Interdependent Gradients. Scientific Reports. https://doi.org/10.1038/s41598-019-51793-7

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.