Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 90 of the 90 references it cites were found in a registry or an index; 45 of the 61 sources its claims cite were checked for support.

Deviations in effective connectivity explain different hallucination subtypes in Parkinson's disease psychosis

Open access
Vignando, Miriam, Ffytche, Dominic, Mazibuko, Ndabezinhle, Palma, Giulio, Bhat, Anjali, Montagnese, Marcella, Dave, Sonali, Tai, Yen F., Batzu, Lucia, Leta, Valentina, Chaudhuri, K. Ray, Williams Gray, Caroline H., Mehta, Mitul A. · 2026 · Nature Mental Health

Published version of a preprint

Abstract

Abstract Psychosis and visual hallucinations (VH) in Parkinson’s disease (PD) substantially impact patient outcomes, yet the underlying neural mechanisms remain unclear, limiting effective treatments. Here we used dynamic causal modeling to leverage the fast temporal dynamics captured with electroencephalography data during a visual mismatch negativity task in people with PD with ( N = 20) and without ( N = 18) VH to examine effective connectivity. We found reduced top-down and enhanced bottom-up connectivity in ventral visual and prefrontal regions during task performance in PD-VH, suggesting deficits in sensory prediction updating and an overreliance on visual input. Connectivity patterns differed with hallucination complexity, with complex VH being associated with altered top-down and bottom-up right-hemisphere connectivity, and multimodal hallucinations to more widespread bilateral disruption. Increased task activity, as computed with source reconstruction, correlated positively with normative serotonergic 5-HT 2A receptor distribution. These findings highlight specific neural targets for early therapeutic interventions, supporting a transdiagnostic computational architecture of hallucinations.

Overview

CoChat Summary

## TL;DR

Using EEG during a visual mismatch negativity task, this study found that Parkinson’s disease psychosis with visual hallucinations (PD-VH; N = 20) showed reduced top-down and increased bottom-up effective connectivity in ventral visual and prefrontal regions versus Parkinson’s disease without hallucinations (N = 18). Connectivity patterns also varied by hallucination subtype and severity, with complex hallucinations linked to more right-hemisphere disruption and multimodal hallucinations to broader bilateral disruption.

## Key findings

  • In PD-VH, the task condition (rare vs standard) was associated with increased bottom-up connectivity and decreased self-inhibition for V1, plus decreased top-down connection strength; the top-down effect had posterior probability (pp) > 0.99.
  • Hallucination severity tracked connectivity patterns: complex visual hallucinations (CVH) were associated with reduced top-down connectivity from right PFC to right ITG and enhanced bottom-up connectivity from right V1 to right PFC and right ITG.
  • In a regression/cross-validation model, right-hemisphere top-down and bottom-up connectivity predicted CVH severity with R2 = 0.49, adjusted R2 = 0.4, P = 0.014, RMSE = 2.26, and MAE = 177.
  • Source-localized task activity correlated with normative receptor maps: 5-HT2A showed a positive correlation (r = 0.284, P = 0.037), while D1 showed a negative correlation (r = −0.29, P = 0.017); VAChT was negative in the exploratory analyses and remained directionally similar in sensitivity analyses.

## Methods

  • Design and sample: Case-control comparison of 20 PD-VH participants and 18 PD-noVH participants; ethics-approved, clinicaltrials.gov registered (NCT03661125). Hallucination phenomenology and subtype were assessed with SAPS-PD and NEVHI.
  • Task and data: Participants completed a visual mismatch negativity (vMMN) EEG task with 1,000 trials; analyses focused on standard vs rare deviant trials. EEG was preprocessed in SPM12, with source reconstruction over 60–400 ms.
  • Main analysis: Dynamic causal modeling (DCM) with parametric empirical Bayes (PEB) tested task-modulated effective connectivity in ventral and dorsal pathways; follow-up analyses included linear regression, leave-one-out cross-validation, neural mass model simulations, and exploratory receptor atlas correlations.

## Limitations

  • The paper states that the study did not include healthy controls, so it cannot fully determine whether findings are specific to PD or disentangle aging-related effects.
  • It also states that EEG cannot robustly assess subcortical sources, and the receptor-binding analyses are exploratory because both EEG source reconstruction and receptor-density maps have low spatial resolution.
  • Evident from the text: the receptor-atlas/PET analyses rely on normative maps rather than individual PET data, so they are indirect and hypothesis-generating rather than participant-specific.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s44220-026-00669-7

Cites

90 of 90 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1 Summary of the vMMN task and findings. The overt task required participants to press a button when the fixation cross became bigger and a different button when the fixation cross became smaller. a, Visual summary of a trial (600 ms): the fixation cross remained on the screen for the whole duration of the experiment; changes in cross size lasted 200 ms; bars flashed peripherally every 50 ms, changing in orientation (90°, 130° and 160°). b, vMMN ERP for PD with and without VH (POZ shown (F(1,36) = 4.32, MS = 289.821, P = 0.045); data published20). c, Density plots showing the distribution of the amplitude of MMN in PD and PD-VH at electrode POZ: PD without VH showed overall greater negativity at POZ compared with PD-VH; PD-VH (left column) showed almost no negativity in this time frame, as confirmed by the analyses reported in our ERP study and summarized in b. The density plots highlight how PD-noVH show a consistency in negativity that is lost or variable in PD-VH. Panel a adapted with permission from ref. 20 courtesy of the authors. Source data
Fig. 1 Summary of the vMMN task and findings. The overt task required participants to press a button when the fixation cross became bigger and a different button when the fixation cross became smaller. a, Visual summary of a trial (600 ms): the fixation cross remained on the screen for the whole duration of the experiment; changes in cross size lasted 200 ms; bars flashed peripherally every 50 ms, changing in orientation (90°, 130° and 160°). b, vMMN ERP for PD with and without VH (POZ shown (F(1,36) = 4.32, MS = 289.821, P = 0.045); data published20). c, Density plots showing the distribution of the amplitude of MMN in PD and PD-VH at electrode POZ: PD without VH showed overall greater negativity at POZ compared with PD-VH; PD-VH (left column) showed almost no negativity in this time frame, as confirmed by the analyses reported in our ERP study and summarized in b. The density plots highlight how PD-noVH show a consistency in negativity that is lost or variable in PD-VH. Panel a adapted with permission from ref. 20 courtesy of the authors. Source dataCC BY · PubMed Central
Fig. 2 Source reconstruction methods, results and ventral model specification. a, The main steps of source reconstruction: cortical mesh creation, projection on structural template, co-registration of EEG signal to the template and maximum intensity projection (MIP) for each participant, allowing us to conduct analyses in sensor space (Supplementary Information 3 for details) and to generate a 3D image for the time window of interest (here, 60–400 ms). b, Source reconstruction results for the rare deviant trials are shown (P < 0.001 uncorrected in the top row and PFWE < 0.05 in the bottom row). The one sample t-test for vMMN in the whole sample for signal source reconstruction for the rare deviant, on images generated for the 60–400-ms interval (P < 0.001, height threshold F = 11.74, d.f. (1.74), k > 200) returned clusters in the inferior and anterior temporal gyrus, bilaterally and in lateral occipital regions (the occipital–calcarine cluster remains significant after FWE correction: P = 0.026 cluster and P = 0.031 peak). We also found a large medial thalamic cluster (k = 725, P = 0.07). We also conducted paired-samples t-tests in each group separately, comparing deviant with standard, which returned a significant peak (k = 145, P < 0.001 uncorrected and PFWE = 0.014) in the bilateral calcarine cortex, mirroring the results of the standard and deviant source reconstruction analysis (Supplementary Fig. 2b; see Supplementary Information 4 for coordinates). For the PD-VH group, a significant peak (k = 83, P < 0.001 uncorrected and PFWE = 0.021) was found in the left ITG (fusiform region) (Supplementary Information 4). All participants were included in the analysis (PD-VH N = 20, and PD-noVH N = 18). The brain maps presented have been produced with xjview (https://www.alivelearn.net/xjview). c, Dipole locations selected on the basis of the results presented in a, here presented on a template of the brain. Dipole coordinates: V1 left –10, –82, –0.5, right 10, –82, –0.5; left ITG –44,–4, –34, right 44, –4, –34; and left PFC –38, 33, 35, right 38, 33, 35. d, A schematic of model connections: AF, forward connections in blue; AB, backward connections in fuchsia; AL, cross-hemispheric connections in orange; input (C matrix). Connections included in the task modulatory matrix (B matrix) for the winning model v3 specified forward connections from V1 to ITG, V1 to PFC and ITG to PFC, and backward connections from ITG to V1, PFC to ITG and PFC to V1. Lateral (inter-hemispheric) connections between regions were also specified between left and right V1, ITG and PFC. Green circles represent connectivity on the same region, testing for self-connectivity during task conditions. e, A visual representation of the ERP neural mass used as described in ref. 69. Source data
Fig. 2 Source reconstruction methods, results and ventral model specification. a, The main steps of source reconstruction: cortical mesh creation, projection on structural template, co-registration of EEG signal to the template and maximum intensity projection (MIP) for each participant, allowing us to conduct analyses in sensor space (Supplementary Information 3 for details) and to generate a 3D image for the time window of interest (here, 60–400 ms). b, Source reconstruction results for the rare deviant trials are shown (P < 0.001 uncorrected in the top row and PFWE < 0.05 in the bottom row). The one sample t-test for vMMN in the whole sample for signal source reconstruction for the rare deviant, on images generated for the 60–400-ms interval (P < 0.001, height threshold F = 11.74, d.f. (1.74), k > 200) returned clusters in the inferior and anterior temporal gyrus, bilaterally and in lateral occipital regions (the occipital–calcarine cluster remains significant after FWE correction: P = 0.026 cluster and P = 0.031 peak). We also found a large medial thalamic cluster (k = 725, P = 0.07). We also conducted paired-samples t-tests in each group separately, comparing deviant with standard, which returned a significant peak (k = 145, P < 0.001 uncorrected and PFWE = 0.014) in the bilateral calcarine cortex, mirroring the results of the standard and deviant source reconstruction analysis (Supplementary Fig. 2b; see Supplementary Information 4 for coordinates). For the PD-VH group, a significant peak (k = 83, P < 0.001 uncorrected and PFWE = 0.021) was found in the left ITG (fusiform region) (Supplementary Information 4). All participants were included in the analysis (PD-VH N = 20, and PD-noVH N = 18). The brain maps presented have been produced with xjview (https://www.alivelearn.net/xjview). c, Dipole locations selected on the basis of the results presented in a, here presented on a template of the brain. Dipole coordinates: V1 left –10, –82, –0.5, right 10, –82, –0.5; left ITG –44,–4, –34, right 44, –4, –34; and left PFC –38, 33, 35, right 38, 33, 35. d, A schematic of model connections: AF, forward connections in blue; AB, backward connections in fuchsia; AL, cross-hemispheric connections in orange; input (C matrix). Connections included in the task modulatory matrix (B matrix) for the winning model v3 specified forward connections from V1 to ITG, V1 to PFC and ITG to PFC, and backward connections from ITG to V1, PFC to ITG and PFC to V1. Lateral (inter-hemispheric) connections between regions were also specified between left and right V1, ITG and PFC. Green circles represent connectivity on the same region, testing for self-connectivity during task conditions. e, A visual representation of the ERP neural mass used as described in ref. 69. Source dataCC BY · PubMed Central
Fig. 3 PEB results. a, A bar plot of estimated effective connectivity for each connection surviving the pp > 0.99 (free energy-based model evidence) threshold. Each panel shows the posterior estimates (model-derived quantities) from PEB of the DCM parameters: the x axis indexes the PEB parameter and the y axis shows the posterior estimate for the group effect. The background has been shaded to indicate direction only, with purple indicating a decreased strength in PDVH, and yellow indicating an increased strength in PD-VH (negative values correspond to increased connectivity in PD-VH, as the PD-VH group was classified as −1 in the design matrix, as opposed to 1 for PD). The red bars indicate the standard error. b, The connectivity matrix of results presented in a. The matrix shows how top-down effective connectivity is reduced versus increased bottom-up connectivity in VH in this task. From, regions sending the influence; To, regions receiving the influence (the first column in the matrix shows connectivity from left V1 to left V1, right V1, the second from right V1 to right ITG and right PFC, the 4th and 5th columns show decreased connectivity from left and right PFC at task). c, A schematic of BMA results with PEB values. Yellow represents increased strength in PD-VH and purple indicates decreased strength in PD-VH (negative values represent increased connectivity in PD-VH as the PD-VH group was classified as −1 in the design matrix, as opposed to 1 for PD). The negative value for self-connection in the left V1 (gray) area indicates less negative self-inhibition. This is consistent with the recursive PEB analysis reported in the main text, as well as in Supplementary Information 6 and Supplementary Fig. 7. d, A graphical representation of the connections; color coded as in a,b. The brain and nodes were generated with the BrainNetviewer (https://www.nitrc.org/projects/bnv/). e, Plots showing the distribution of Ep in PD with and without VH for each connection found differing in the model: the violin and box and whisker plots are color-coded with the color of the node from which the connection starts. The dots are color-coded by group: noVH participants (18) with black dots and VH participants (20) with white dots. In each box plot, the centerline indicates the median, the box bounds indicate the interquartile range (25th–75th percentiles) and the whiskers extend to the minimum and maximum values. f, Connectivity differences in PD-VH and PD-noVH: a bar plot of estimated effective connectivity for each connection surviving the pp > 0.99 (free energy) threshold. Yellow represents increased connectivity in PD-VH and purple represents decreased connectivity in PD-VH. Connectivity is overall increased in PD-VH in all the connections modelled (all with pp >0.99). The brain and nodes were generated with BrainNet viewer. Latent connectivity A matrix: a graphical representation of the connections, color-coded as in a,b. g, Latent connectivity A matrix: a graphical representation of the connections, color-coded as in a,b. The brain and nodes were generated with BrainNet viewer. Source data
Fig. 3 PEB results. a, A bar plot of estimated effective connectivity for each connection surviving the pp > 0.99 (free energy-based model evidence) threshold. Each panel shows the posterior estimates (model-derived quantities) from PEB of the DCM parameters: the x axis indexes the PEB parameter and the y axis shows the posterior estimate for the group effect. The background has been shaded to indicate direction only, with purple indicating a decreased strength in PDVH, and yellow indicating an increased strength in PD-VH (negative values correspond to increased connectivity in PD-VH, as the PD-VH group was classified as −1 in the design matrix, as opposed to 1 for PD). The red bars indicate the standard error. b, The connectivity matrix of results presented in a. The matrix shows how top-down effective connectivity is reduced versus increased bottom-up connectivity in VH in this task. From, regions sending the influence; To, regions receiving the influence (the first column in the matrix shows connectivity from left V1 to left V1, right V1, the second from right V1 to right ITG and right PFC, the 4th and 5th columns show decreased connectivity from left and right PFC at task). c, A schematic of BMA results with PEB values. Yellow represents increased strength in PD-VH and purple indicates decreased strength in PD-VH (negative values represent increased connectivity in PD-VH as the PD-VH group was classified as −1 in the design matrix, as opposed to 1 for PD). The negative value for self-connection in the left V1 (gray) area indicates less negative self-inhibition. This is consistent with the recursive PEB analysis reported in the main text, as well as in Supplementary Information 6 and Supplementary Fig. 7. d, A graphical representation of the connections; color coded as in a,b. The brain and nodes were generated with the BrainNetviewer (https://www.nitrc.org/projects/bnv/). e, Plots showing the distribution of Ep in PD with and without VH for each connection found differing in the model: the violin and box and whisker plots are color-coded with the color of the node from which the connection starts. The dots are color-coded by group: noVH participants (18) with black dots and VH participants (20) with white dots. In each box plot, the centerline indicates the median, the box bounds indicate the interquartile range (25th–75th percentiles) and the whiskers extend to the minimum and maximum values. f, Connectivity differences in PD-VH and PD-noVH: a bar plot of estimated effective connectivity for each connection surviving the pp > 0.99 (free energy) threshold. Yellow represents increased connectivity in PD-VH and purple represents decreased connectivity in PD-VH. Connectivity is overall increased in PD-VH in all the connections modelled (all with pp >0.99). The brain and nodes were generated with BrainNet viewer. Latent connectivity A matrix: a graphical representation of the connections, color-coded as in a,b. g, Latent connectivity A matrix: a graphical representation of the connections, color-coded as in a,b. The brain and nodes were generated with BrainNet viewer. Source dataCC BY · PubMed Central
Fig. 4 Hallucination phenomenology regression models. a, An UpSit plot summarizing the co-occurrence of hallucination subtypes. Bars at the top indicate the number of participants experiencing each specific combination of subtypes (intersection size). Filled dots in the matrix show which subtypes are part of each combination. Bars on the left show the total number of participants with each subtype (set size on the right and actual N on the left). Complex, presence, passage, pareidolia and simple VH scores are derived from the NEVHI (described in detail in Supplementary Information 1); multimodal hallucination scores are derived from the SAPS-PD questionnaire described in Supplementary Information 1, and the modality of multimodal hallucination is further subtyped in c. All participants with VH were included in the analysis (PD-VH N = 20). b, Linear regression analyses on CVH for matrix B. Yellow represents increased connectivity in the PEB in PD-VH, and purple represents decreased connectivity. c, Multimodal score (derived from SAPS-PD information) regression analyses. The mosaic plot shows the proportion of participants with hallucinations in each specific modality (light blue represents that the symptom is present, dark blue represents that it is not present). Reported in all the plots are the t values and the individual regression uncorrected and corrected (FDR) P values. Complete details are provided in S7. Yellow is used for regions showing increased connectivity in PD-VH and purple for those showing decreased connectivity in PD-VH. d, The matrix A connectivity from left V1 (LV) to left PFC (LPFC) and multimodal hallucinations individual regression. All PD-VH participants were included in the regression analyses (N = 20) prior to outlier removal. Source data
Fig. 4 Hallucination phenomenology regression models. a, An UpSit plot summarizing the co-occurrence of hallucination subtypes. Bars at the top indicate the number of participants experiencing each specific combination of subtypes (intersection size). Filled dots in the matrix show which subtypes are part of each combination. Bars on the left show the total number of participants with each subtype (set size on the right and actual N on the left). Complex, presence, passage, pareidolia and simple VH scores are derived from the NEVHI (described in detail in Supplementary Information 1); multimodal hallucination scores are derived from the SAPS-PD questionnaire described in Supplementary Information 1, and the modality of multimodal hallucination is further subtyped in c. All participants with VH were included in the analysis (PD-VH N = 20). b, Linear regression analyses on CVH for matrix B. Yellow represents increased connectivity in the PEB in PD-VH, and purple represents decreased connectivity. c, Multimodal score (derived from SAPS-PD information) regression analyses. The mosaic plot shows the proportion of participants with hallucinations in each specific modality (light blue represents that the symptom is present, dark blue represents that it is not present). Reported in all the plots are the t values and the individual regression uncorrected and corrected (FDR) P values. Complete details are provided in S7. Yellow is used for regions showing increased connectivity in PD-VH and purple for those showing decreased connectivity in PD-VH. d, The matrix A connectivity from left V1 (LV) to left PFC (LPFC) and multimodal hallucinations individual regression. All PD-VH participants were included in the regression analyses (N = 20) prior to outlier removal. Source dataCC BY · PubMed Central
Fig. 5 Source-reconstructed signal–receptor binding atlases exploratory analysis. a, As described in the Methods and in Supplementary Information 13, we parcellated the normative PET receptor density atlases (maps derived from independent PET datasets, not collected in this cohort) and the source-reconstructed EEG signals with the Freesurfer aparc parcellation inclusive of the Desikan–Killiany atlas (converted to MNI template space); PET atlases and individual participant data were aligned to the brain atlas, and registration was checked examining PET atlas and brain atlas sizes and the transformation matrix, as well as visually with ‘checkreg’, before data extraction. Parcellated data were then used to run linear regression models (one per receptor) on a subset of regions of the Freesurfer aparc parcellation inclusive of the Desikan–Killiany atlas (cerebellar, white matter and ventricular regions were removed) to identify and remove outliers using Cook’s distance. P values from the regressions were corrected for multiple comparisons. Then, the BrainSMASH toolbox38 was used to generate MNI coordinates for the Desikan–Killiany atlas; we removed the cerebellar and ventricular regions and the regions that were marked as outliers from the regression models before proceeding to compute the Pearson’s actual correlations of the two maps. Only the maps surviving this threshold were finally entered in linear mixed-effects models, which had the purpose of examining individual-level data taking age, LEDD, MoCA score and disease duration into account. b, Linear regressions results; the regression lines are color coded to match the color selected for each PET map, for ease of display. Insets: the PET map images were rendered by overlapping the atlas and the relevant PET map; the colors were arbitrarily chosen to highlight the binding distribution in the different maps. The data points in the scatterplot represent the N of regions entered in the analysis after nonrelevant regions (cerebellar, white matter and ventricular regions) and outliers were removal. The MMN data presented on the x axis represents the mean difference between the MMN signal in PD-VH and PD-noVH. c, LMM results. The x axis represents mean centered receptor density, and the y axis represents the mean centered MMN signal; we mean centered the data as the scale of the different predictors varied greatly. The blue line represents PD-VH, and the red line represents PD-noVH. The shading around the lines represent the s.e.m. Source data
Fig. 5 Source-reconstructed signal–receptor binding atlases exploratory analysis. a, As described in the Methods and in Supplementary Information 13, we parcellated the normative PET receptor density atlases (maps derived from independent PET datasets, not collected in this cohort) and the source-reconstructed EEG signals with the Freesurfer aparc parcellation inclusive of the Desikan–Killiany atlas (converted to MNI template space); PET atlases and individual participant data were aligned to the brain atlas, and registration was checked examining PET atlas and brain atlas sizes and the transformation matrix, as well as visually with ‘checkreg’, before data extraction. Parcellated data were then used to run linear regression models (one per receptor) on a subset of regions of the Freesurfer aparc parcellation inclusive of the Desikan–Killiany atlas (cerebellar, white matter and ventricular regions were removed) to identify and remove outliers using Cook’s distance. P values from the regressions were corrected for multiple comparisons. Then, the BrainSMASH toolbox38 was used to generate MNI coordinates for the Desikan–Killiany atlas; we removed the cerebellar and ventricular regions and the regions that were marked as outliers from the regression models before proceeding to compute the Pearson’s actual correlations of the two maps. Only the maps surviving this threshold were finally entered in linear mixed-effects models, which had the purpose of examining individual-level data taking age, LEDD, MoCA score and disease duration into account. b, Linear regressions results; the regression lines are color coded to match the color selected for each PET map, for ease of display. Insets: the PET map images were rendered by overlapping the atlas and the relevant PET map; the colors were arbitrarily chosen to highlight the binding distribution in the different maps. The data points in the scatterplot represent the N of regions entered in the analysis after nonrelevant regions (cerebellar, white matter and ventricular regions) and outliers were removal. The MMN data presented on the x axis represents the mean difference between the MMN signal in PD-VH and PD-noVH. c, LMM results. The x axis represents mean centered receptor density, and the y axis represents the mean centered MMN signal; we mean centered the data as the scale of the different predictors varied greatly. The blue line represents PD-VH, and the red line represents PD-noVH. The shading around the lines represent the s.e.m. Source dataCC BY · PubMed Central
Fig. 6 Summary of the neural mass model used in the study and of main results. Top: each node in our network was fitted with the relevant (ERP) neural mass model69 in which feedforward connections (blue) are thought to encode prediction errors, whereas feedback connections (pink) encode predictions. Middle: summary of the results of the PEB group analysis for the oddball task eliciting our MMN. Purple arrows are connections where estimated connection strength was decreased in PD-VH, and yellow arrows represent connections where such strength was increased in PD-VH. The gray circle describes the weaker self-inhibition in left V1 in PD-VH. Bottom: summary of the results of multiple regression LOOCV machine learning models relating VH severity to effective connectivity in PD-VH. CVH were best predicted by right-hemisphere connections found related to CVH severity (rV1-rPFC(increased), rV1-rITG (increased) and rPFC-rITG (decreased)), supporting the view that these regions operate as a network. Figure created with Adobe Illustrator for visualization purposes.
Fig. 6 Summary of the neural mass model used in the study and of main results. Top: each node in our network was fitted with the relevant (ERP) neural mass model69 in which feedforward connections (blue) are thought to encode prediction errors, whereas feedback connections (pink) encode predictions. Middle: summary of the results of the PEB group analysis for the oddball task eliciting our MMN. Purple arrows are connections where estimated connection strength was decreased in PD-VH, and yellow arrows represent connections where such strength was increased in PD-VH. The gray circle describes the weaker self-inhibition in left V1 in PD-VH. Bottom: summary of the results of multiple regression LOOCV machine learning models relating VH severity to effective connectivity in PD-VH. CVH were best predicted by right-hemisphere connections found related to CVH severity (rV1-rPFC(increased), rV1-rITG (increased) and rPFC-rITG (decreased)), supporting the view that these regions operate as a network. Figure created with Adobe Illustrator for visualization purposes.CC BY · PubMed Central

Verification

Claims the paper makes

27 Valid · 4 Not valid · 8 Not checked · 86 Own results and interpretations

40 more claims

Create a free account to see all 125 claims and every source verdict.

See all claims — free account
  • Probably not a substantial claim. 43
  • It may state two separate claims as one. 46

Create a free account to see these claims.

Citations

Cites

Show all 83 references

Metadata

Authors
Vignando, M., Ffytche, D., Mazibuko, N., Palma, G., Bhat, A., Montagnese, M., Dave, S., Tai, Y. F., Batzu, L., Leta, V., Chaudhuri, K. R., Williams Gray, C. H., & Mehta, M. A.
Affiliations
  • King's College London (GB)
  • University of Southampton (GB)
  • University of Exeter (GB)
  • University of Cambridge (GB)
  • University of London (GB)
  • Hammersmith Hospital (GB)
  • King's College Hospital NHS Foundation Trust (GB)
Corresponding author
Vignando, Miriam
Venue
Nature Mental Health
Published
2026-06-10
Volume
4
Issue
6
Pages
994-1009
ISSN
2731-6076 (ISSN-L)
OpenAlex
W7164158954
Cited by
0
References
83
FWCI
0.00
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Mental Health (publishedVersion, cc-by)
  • Research Square (submittedVersion, cc-by)
  • PubMed (publishedVersion)
  • PubMed Central (publishedVersion, cc-by)
  • Apollo (University of Cambridge) (submittedVersion, cc-by)
Funders
  • Alzheimer's Research UK (ARUK-RF2022B-002)
  • RCUK | Medical Research Council (MR/R005931/1)
Topics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Hallucinations in medical conditions
  • Health Sciences › Medicine › Clinical Neurology › Parkinson's Disease Mechanisms and Treatments
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Functional Brain Connectivity Studies
Keywords
Parkinson's disease psychosis, visual hallucinations, effective connectivity, mismatch negativity, hemispheric lateralization, multimodal hallucination, prefrontal cortex, ventral visual cortex

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.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-11, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-11 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

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: Deviations in effective connectivity explain different hallucination subtypes in Parkinson's disease psychosis. https://app.cochat.ai/papers/10.1038/s44220-026-00669-7
The paper
Vignando, M., Ffytche, D., Mazibuko, N., Palma, G., Bhat, A., Montagnese, M., Dave, S., Tai, Y. F., Batzu, L., Leta, V., Chaudhuri, K. R., Williams Gray, C. H., & Mehta, M. A. (2026). Deviations in effective connectivity explain different hallucination subtypes in Parkinson's disease psychosis. Nature Mental Health. https://doi.org/10.1038/s44220-026-00669-7

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.