A rostral prefrontal mediolateral gradient predicts creativity in frontotemporal dementia

Abstract

Creative thinking is a fundamental aspect of human cognition, enabling the production of novel and useful ideas. It is hypothesized to emerge from the binding and reconfiguration of existing knowledge, through the generation of remote semantic associations and their combination in original and meaningful ways, respectively supported by the default mode (DMN) and executive control (ECN) networks. At the crossroads of these two networks, the rostral prefrontal cortex (PFC) is proposed as a key hub for DMN-ECN interactions, possibly supporting the interplay between generative and combinatory creative processes. However, the specific contributions of its medial and lateral subdivisions to creativity remain unclear.

In this study, we aimed to characterize the involvement of the rostral PFC in creative cognition through the lens of behavioural variant frontotemporal dementia (bvFTD), a relevant pathological model as it primarily affects the rostral PFC and alters intrinsic connectivity within the DMN and ECN. Using whole-brain voxel-based morphometry, we explored brain regions critical for the generation and combination of remote semantic associates, as well as for creative abilities, thought to involve both types of processes. Using resting-state functional connectivity and gradient mapping techniques, we also explored functional connectivity profiles within the rostral PFC and how variations in connectivity within this region predict creative performance.

As a result, we found a critical role for the rostromedial PFC in generating remote semantic associations and for the rostrolateral PFC in combining semantic associates, while both regions were critical for creative abilities. Moreover, we showed that the intrinsic connectivity of the rostral PFC is organised along a mediolateral functional gradient, segregating the rostromedial PFC, connected to the DMN, and the rostrolateral PFC, connected to the ECN. Finally, we showed that the range of this functional gradient, representing the functional differentiation between the ECN and DMN, predicts creative abilities.

Overall, this study advances our understanding of creative cognition, its relationships to the anatomical and functional organization of the PFC, and its impairment in bvFTD.

Altmayer et al. examine how creativity is affected in frontotemporal dementia. They show that distinct regions of the rostral prefrontal cortex support generative versus combinatory creative processes, and that disruption of the functional gradient connecting these regions impairs creative abilities in patients.

Introduction

Creative thinking is a fundamental aspect of human cognition, allowing us to produce original and useful ideas.^1,2 This ability is essential for addressing the challenges of a complex and dynamic environment and is recognized as crucial at both individual and societal levels. However, the mechanisms by which creative ideas arise in the brain remain an active area of research and debate. It is hypothesized that creative ideas do not emerge de novo but rather from the binding and reconfiguration of existing semantic knowledge.³ According to the influential associative theory of creativity,^4,5 creative ideas result from both the generation of remote semantic associations and their combination in original and meaningful ways.^3,4,6,7 The generation of original semantic associations through the activation of mutually remote semantic concepts is thought to rely on associative thinking processes.^3,5,8 Spontaneous or undirected activation of one’s semantic associations stored in memory can explain individual differences in creativity.^5,7,9-11 However, remote thinking also involves goal-directed associative processes for the intentional retrieval of remote associates.^5,8,12-15 Recent research has suggested that low-constrained generation of remote associates involves both spontaneous and goal-directed associative processes.^3,5,12 In more complex creativity tasks requiring a strategic memory search, especially in constrained contexts,^3,16-21 executive control processes including relational integration,^16,22-28 multitasking or branching²⁹ allow the effective combination of multiple semantic associations to meet task or real-life constraints.

It is commonly accepted^5,8,30-32 that the cooperation of low-constrained generative and high-constrained combinatory processes in creativity relies on the respective contributions of the default mode network (DMN), known for its involvement in self-generated thoughts,^33,34 and the executive control network (ECN), supporting executive control processes^35,36 including those involved in creative thinking. As a major component of these two networks, the prefrontal cortex (PFC) is the cortical region most consistently associated with creativity in studies using various experimental tasks and exploring different creativity domains.^37-50 However, the PFC is the largest and most functionally and structurally diverse association cortex in the human brain,^51,52 so reporting PFC involvement in creativity lacks specificity.

One region in particular, the rostral PFC [also known as the frontopolar cortex, Brodmann area (BA) 10], seems essential to creative thinking.^16,42,53-55 Indeed, the rostral PFC presents the appropriate characteristics to play a central role in DMN-ECN interactions, possibly supporting the interplay between generative and combinatory creative processes. Phylogenetically one of the most recently and widely developed brain areas, with a broad range of complex intra- and inter-areal interactions,^29,51,56 the rostral PFC is thought to sit at the apex of the prefrontal functional and architectural hierarchy, allowing the most complex and integrated degree of information processing.^52,57-60 For instance, this region has been related to a number of high-order cognitive functions, including abstract thinking,^57,61-63 analogical reasoning,^{23,24,26,64,65} working memory and multitasking.^29,66-70 Overall, its role could be summarized as the capacity to process and integrate information that is temporally or semantically distant.^51,71 Yet the rostral PFC itself is not a homogeneous region either. Substantial evidence indicates medial and lateral subdivisions of the rostral PFC, characterized by differential cytoarchitectonic,⁷² structural^73,74 and functional connectivity properties,^73,75,76 as well as distinct cognitive functions.^72,75,77 Specifically, the rostromedial PFC is described as a core node of the DMN, involved in internally focused, self-generated thoughts,^11,33 while the rostrolateral PFC is associated with the ECN and controlled cognition.^11,73,75 Yet, the respective boundaries between the rostromedial and rostrolateral PFC remain a matter of debate.⁷³

Whatever the structural and functional differences between these rostral regions, substantial evidence supports the involvement of both the rostromedial and rostrolateral PFC in creative thinking.^{16,26,42,47,53} Several patient studies demonstrated a causal relationship between brain structure and function in a manner that complements functional MRI (fMRI) findings, showing the critical role of the rostral PFC in distinct creative processes.^16,23,54,78 In particular, Bendetowicz and colleagues¹⁶ demonstrated a cognitive dissociation between the rostromedial and rostrolateral PFC in patients with focal frontal lesions. The rostromedial PFC was shown to be critical for generating remote semantic associations in a word-to-word generation task, the Free Generation of Associates Task (FGAT).¹⁶ Conversely, the rostrolateral PFC was shown to be critical to efficiently combining remote concepts to find solutions in a French adaptation of Mednick’s Remote Associates Task,⁴ the Combination of Associates Task (CAT),^16,27 which was replicated in a recent study.⁷⁹ Furthermore, brain lesions overlapping with the DMN impacted the generation task (FGAT) and lesions overlapping with the ECN the combination task (CAT). Overall, while most studies have described the rostromedial and rostrolateral PFC as nodes of separate networks displaying opposite cognitive roles and connectivity profiles, the hypothesis of their joint critical involvement in human creative thinking remains to be tested. In addition, the question of the reciprocal functional interactions between these two regions and their implications for the cooperation of the DMN and ECN during creative cognition^31,80 remains unaddressed. As argued by some authors,^81-84 low-constrained generative and high-constrained combinatory processes might not be completely separable but rather function as an integrated continuum to support individual creative abilities. Hence, we hypothesized that the anatomical proximity of the DMN and ECN within the rostral PFC could support their interaction, possibly through gradual functional connectivity changes within this region rather than discrete functional boundaries, with implications for the coordination of generative and combinatory processes involved in creativity.

To address this question, we explored creative cognition and its neural underpinnings in behavioural variant of frontotemporal dementia (bvFTD),⁸⁵ the third most common neurodegenerative dementia after Alzheimer’s and Lewy body⁸⁶ diseases, and a highly relevant pathological model for several reasons. First, research on brain-lesioned patients offers unique insights into brain regions whose integrity is critical for creativity, which are not provided by purely correlational studies in healthy individuals.^38,87-90 Second, progressive neurodegenerative diseases, which cause widespread brain changes, may be more appropriate than focal brain lesions for studying gradual functional connectivity alterations.^91,92 Finally, bvFTD is a unique model for studying creative cognition since the disease predominantly involves PFC atrophy^93-96 and intrinsic connectivity modifications in both the DMN and ECN,^94,97,98 key players in creativity.^54,88,90

The present study aimed to refine the characterization of rostral PFC involvement in creative cognition through the lens of patients with bvFTD. First, using whole-brain voxel-based morphometry (VBM), we aimed to extend previous results showing, in other pathologies, the critical roles of the rostromedial and rostrolateral PFC in the generation (FGAT) and combination (CAT) of remote semantic associates, respectively,¹⁶ and to test the hypothesis of the joint critical contribution of these regions in overall individual creative abilities, as assessed by the referential Alternative Uses Task (AUT).^99,100 Second, the current study aimed to understand the intrinsic functional connectivity of the rostral PFC and its implications in creative abilities. To test our hypothesis of graded functional connectivity changes between rostromedial and rostrolateral PFC related to creativity, we applied novel functional gradient mapping techniques^101,102 to resting-state functional connectivity (RSFC) data within the rostral PFC and investigated the main axis of functional variation in this brain region in healthy controls. We anticipated identifying a mediolateral principal connectivity gradient linking medial and lateral poles of the rostral PFC. Third, we aimed to explore how gradient modifications, expected in bvFTD,¹⁰³ related to brain atrophy and to creative abilities.

Materials and methods

Participants

Participants were recruited between September 2017 and December 2023 within the ECOCAPTURE protocol (clinicaltrials.gov: NCT03272230; INSERM, C16-87).^104-110 Eligible participants were French-native speakers aged 40 to 85, with Mini-Mental State Examination (MMSE¹¹¹) scores above 17 and free from neurological, psychiatric or substance abuse disorder (except bvFTD, for patients).

Patients meeting probable bvFTD diagnostic criteria⁸⁵ were recruited prospectively from two tertiary referral centres at Salpêtrière and Fernand-Widal hospitals in Paris. Controls were recruited through public announcement and matched patients for age, sex and education (Barbizet scale, from 1, illiterate, to 7, university degree¹¹²).

Fifty-six participants were screened: 27 patients and 29 controls. Nine patients and one control were excluded because of significant semantic deficits impacting instructional understanding, as judged by the trained neuropsychologists who assessed them. One control was excluded due to a psychiatric condition. Thus, the final sample for the behavioural and VBM analyses included 45 participants: 18 bvFTD patients and 27 controls. From this sample, one patient did not complete the FGAT due to technical issues, and two other patients refused to perform the AUT, but all were included in the analyses for the other creativity tasks. Additionally, three controls and three patients were excluded from fMRI analyses due to excessive movements, resulting in a sample of 39 participants (15 patients, 24 controls) for fMRI analyses. Analyses relating brain structure or function to creative behaviour were performed on the whole dataset, combining patients and controls. Analyses characterizing the functional organization of the rostral PFC were restricted to controls, to capture the intrinsic organization of this region as it appears in healthy brains, without confounds from disease-related alterations. The strategy for each set of analyses is summarized in Fig. 1.

Figure 1.

Overview of analysis plan. Analyses relating brain structure (top row) or function (bottom row) to creative behaviour were performed on the whole dataset combining patients and controls (left column). Analyses characterizing the functional (bottom row) organization of the rostral prefrontal cortex (PFC) in healthy brains were restricted to controls (right column). Top left: Structural brain correlates of creative cognition: Whole-brain voxel-based morphometry (VBM) analyses were performed on the whole cohort [18 behavioural variant frontotemporal dementia (bvFTD) patients and 27 healthy controls] to identify the brain regions critical for the generation (FGAT) and combination (CAT) of remote associations, and for individual creative abilities (AUT) thought to involve both generation and combination processes. We hypothesized that distinct rostral prefrontal regions would be critical for creative cognition. Bottom right: Intrinsic functional organization of the rostral PFC in healthy controls: Gradient-mapping analyses of resting-state functional MRI (rs-fMRI) data were performed in healthy controls (n = 24) to explore the rostral PFC’s principal gradient, i.e. the principal axis of functional variations inside this region. This principal gradient analysis was replicated in two independent datasets of young healthy adults. Bottom left: Gradient disruption in bvFTD and relation with creative cognition: Individual-level rostral PFC principal gradients were extracted for all participants from rs-fMRI data (15 patients + 24 controls), and we explored how gradients’ range, i.e. the functional segregation between both gradient ends, related to both brain atrophy and creative cognition. AUT = Alternative Uses Task; CAT = Combination of Associates Task; FGAT = Free Generation of Associates Task.

All participants completed the entire ECOCAPTURE protocol with the same schedule, without deviations from the preregistered protocol. The current study focuses on creativity tasks, neuropsychological tests, and structural and fMRI data. A subsample (14 patients and 20 controls) of the current study sample has been the focus of a detailed analysis of FGAT cognitive components in a preceding study.¹² An approved ethics committee authorized the study, all participants provided written informed consent, and anonymity was preserved, in accordance with French regulations and the Declaration of Helsinki.

Creativity tasks

To explore creative thinking, we used two tasks aiming to disentangle different aspects of associative thinking. To explore the generation of remote semantic associates,^5,13,16,53 an essential process of creative cognition,^{4,13,16,53,113,114} we used the FGAT, a simple word-association task previously validated in patients¹⁶ and controls.^50,115,116 To assess the additional controlled processes that allow the goal-directed combination of multiple semantic associations in agreement with given constraints,^4,16,27 we used the CAT, a French adaptation^16,21,27,117 of Mednick’s Remote Associates Task.⁴ Additionally, the AUT, a classical divergent thinking task, was used as a reference creativity task capturing individual creative abilities.^83,118 For all three tasks, we used comparable ‘Efficiency’ scores computed as the ratio of task creative performance to the cumulative response time across all task trials, reflecting the proportion of unique (FGAT), correct (CAT) or creative (AUT) answers by time unit (minutes).

Detailed information about the rationale, verbal material, experimental setup, data cleaning and variable computation for each task is provided in Supplementary material, ‘Method 1’.

Neuropsychological testing

Participants’ general and executive cognitive functioning were assessed using the MMSE¹¹¹ and Frontal Assessment Battery (FAB¹¹⁹), respectively.

By summing scores from the short French versions of a naming and a semantic matching test commonly used in routine clinical practice,¹²⁰ we also calculated an individual semantic score, which was covaried out in brain imaging analyses to control for the effect of potential semantic deficits on creativity tasks.

Behavioural measures of apathy and mood disturbances were also assessed and correlated with creativity task performance to explore potential associations. Apathy was measured using the Starkstein Apathy Scale (SAS¹²¹) and the Dimensional Apathy Scale (DAS¹²²), which explores distinct Executive, Emotional and Initiation dimensions of apathy. Mood was assessed using the Hospital Anxiety and Depression Scale (HADS^123,124), providing separate anxiety and depression scores.

Neuroimaging analyses

Voxel-based morphometry analysis

We used VBM to explore, without a priori, the brain regions critical for efficiency in each creativity task. Voxel-wise whole-brain multiple linear regressions revealed regions in which grey matter volume (GMV) showed a significant relationship with CAT, FGAT and AUT Efficiency, using separate models for each task. Regressions were performed on the whole dataset, combining patients and controls (n = 45, 18 patients and 27 controls). Voxel-wise whole-brain group differences in GMV were assessed using a one-tailed two-sample t-test. Results are reported at a family-wise error (FWE) corrected P_FWE < 0.05 statistical threshold applied at the voxel level. For all analyses, age, sex, education and semantic score were covaried out, and individual differences in total intracranial volume were accounted for. To rule out the possibility that significant findings are mostly driven by group differences and do not represent a generalizable brain-behaviour relationship, each regression model was also explored after adding participants’ clinical status (bvFTD or controls) as a covariate, with a more lenient uncorrected P_uncorrected < 0.001 statistical threshold applied at the voxel level.

Detailed information about MRI acquisitions and VBM preprocessing and analyses is provided in Supplementary material, ‘Method 2’.

Resting state functional connectivity gradient analysis

We used voxel-to-voxel RSFC to explore the rostral PFC’s (BA 10¹²⁵) principal connectivity gradient, i.e. the principal axis of functional variation in this brain region, and to describe its functional characteristics, in healthy individuals (n = 24). Detailed information about MRI acquisitions and RSFC preprocessing and analyses is provided in Supplementary material, ‘Method 2’.

Principal gradient replication

To investigate the external validity of the characterized rostral PFC principal gradient, we ran the same gradient analysis in two independent datasets of 93 [43 females, mean age = 25.6 years, standard deviation (SD) = 3.8]¹²⁶ and 36 (18 females, mean age = 26.6 years, SD = 4.1)⁵⁰ young healthy subjects, respectively. Preprocessing and first-level analyses of the RSFC data of these datasets are described in the corresponding publications.

Principal gradient functional characterization

We quantified the algebraic connectivity of the group-level rostral PFC principal gradient.¹²⁷ This value represents the degree of gradation in functional connectivity changes across the rostral PFC, ranging from 0, indicating hard, discrete boundaries between disconnected subregions within the gradient, to a value of 1, indicating graded transitions between connected subregions.¹²⁸ As gradient analysis only provides topographical information on the rostral PFC principal gradient, we also characterized the different functional connectivity profiles of the two extremes of the principal gradient with the rest of the brain, in healthy participants (n = 24), using seed-based RSFC analyses with the extreme thirds of the identified principal gradient as seeds. We then used paired two-sample t-tests in healthy controls to explore the brain regions with greater functional connectivity to one extreme third of the gradient compared to the opposite third, and vice versa. Results are reported at a FWE corrected P_FWE < 0.05 statistical threshold at the voxel level. As recommended,¹²⁹ Yeo and colleagues’ seven-networks resting-state parcellations of the cortex, cerebellum and striatum^130-132 were used to determine which global-scale intrinsic functional network our findings were affiliated with.

Principal gradient relationship with creative behaviour

Finally, we explored the relationship between principal gradient characteristics and creative behaviour across patients and controls (n = 39, 15 patients and 24 controls). At the participant level, we quantified the range of the rostral PFC principal gradient, i.e. the difference in gradient values between the positive and negative extreme ends of the gradient, representing the functional distance between the two ends of the gradients, with higher values indicating greater functional segregation.^133,134 Based on previous work,¹⁰³ we expected the gradient range to be reduced in patients with bvFTD, as a result of decreased functional differentiation between cortical areas and increased focal connectivity. Thus, we tested whether the gradient range was impaired in patients with bvFTD using a one-sided Mann–Whitney U-test and whether grey matter atrophy within the whole brain or the rostral PFC predicted the rostral PFC principal gradient range. To explore the functional significance of the gradient range, we also explored how it related to creativity task efficiency (CAT, FGAT and AUT).

Control analyses in additional Brodmann areas

To assess the specificity of the rostral PFC (BA 10) gradient analysis, we ran the same gradient analysis in other BAs involving both the DMN and ECN, i.e. BAs in which both the DMN and ECN overlapped with at least 10% of the BA volume.

Statistical analyses

All statistical analyses on clinical and neuropsychological data were conducted using RStudio 1.4.1106.¹³⁵ Significance was determined by an alpha level of 0.05. Between-group differences were tested using non-parametric Mann-Whitney U-tests, or χ² tests for nominal variables. U-statistics are reported with SDs. Spearman’s rank partial correlations were performed to assess the relationships between creativity tasks performance and other variables, partialling out the effects of age, sex, education and semantic score. We used false discovery rate (FDR) correction for multiple comparisons. Multiple linear regression analyses were conducted using the Enter method and age, sex, education, semantic score and total intracranial volume as covariates.

Results

Behavioural results

Forty-five participants, 18 bvFTD patients and 27 healthy controls, were included in the behavioural analyses. Patients and controls did not differ in age, sex or education (Table 1). As expected in bvFTD, patients showed lower scores than controls in general cognitive (MMSE), executive efficiency (FAB) and semantic scores. Patients with bvFTD also showed more apathetic (SAS, DAS), anxiety (HADS-Anxiety) and depression (HADS-Depression) symptoms and were the only group using psychotropic medications.

Table 1.

Demographic and cognitive characterization of the participants

	Patients (n = 18)	Controls (n = 27)	Group comparisons
Sex, female/male	8/10	15/12	χ 2 = 0.53, P = 0.465
Age, years	66.28 (9.20)	64.78 (7.25)	U = 289.00, P = 0.291
Disease duration, years	5.11 (2.70)	0.00 (0.00)	NA
Education (Barbizet scale, /7)	6.19 (1.07)	6.32 (0.80)	U = 243.00, P = 1.000
Semantic score (/80)	71.00 (12.83)	79.19 (0.96)	U = 56.50, P < 0.001a
MMSE (/30)	24.28 (2.59)	29.22 (0.97)	U = 15.50, P < 0.001a
FAB (/18)	12.67 (3.38)	17.15 (0.86)	U = 13.50, P < 0.001a
SAS (/42)	16.94 (5.02)	6.63 (3.99)	U = 458.00, P < 0.001a
DAS Total (/72)	31.94 (9.67)	18.78 (7.74)	U = 402.00, P < 0.001a
DAS Executive (/24)	9.67 (4.84)	4.15 (3.27)	U = 400.00, P < 0.001a
DAS Emotional (/24)	11.34 (3.29)	8.63 (3.87)	U = 341.50, P = 0.023b
DAS Initiation (/24)	10.89 (5.48)	7.00 (3.72)	U = 343.50, P < 0.020b
HADS Total (/42)	14.89 (7.75)	5.96 (3.82)	U = 418.50, P < 0.001a
HADS Anxiety (/21)	8.00 (5.04)	4.48 (2.58)	U = 350.00, P < 0.013b
HADS Depression (/21)	6.89 (3.53)	1.48 (1.85)	U = 441.00, P < 0.001a
Psychotropic medication use
Antidepressant, n	12 (66.7%)	0 (0.0%)	NA
Equivalent fluoxetine dose	0.69 (0.60)	0.0 (0.0)	NA
Antipsychotic, n	2 (11.1%)	0 (0.0%)	NA
Benzodiazepine, n	4 (22.2%)	0 (0.0%)	NA
FGAT efficiency	1.47 (1.02)	2.81 (1.06)	U = 81.00, P < 0.001a
CAT efficiency	1.38 (0.81)	3.72 (1.68)	U = 46.00, P < 0.001a
AUT efficiency	0.86 (0.58)	1.80 (0.75)	U = 62.00, P < 0.001a

FGAT efficiency corresponds to the ratio between the number of unique answers and total response time over all trials in the ‘FGAT-Distant’ condition. CAT efficiency corresponds to the ratio between the number of correct answers and total response time over all CAT trials. AUT efficiency corresponds to the ratio between the sum of highest ratings and total response time across all AUT trials. Mean scores (standard deviations) are provided under ‘Patients’ and ‘Controls’, and two-sided Mann–Whitney U-test statistics are given under ‘Group comparisons’. AUT = Alternate Uses Task; CAT = Combination of Associates Task; DAS = Dimensional Apathy Scale; FAB = Frontal Assessment Battery; FGAT = Free Generation of Associates Task; HADS = Hospital Anxiety and Depression Scale; MMSE = Mini-Mental State Examination; NA = not applicable; SAS = Starkstein Apathy Scale.

^a P < 0.001.

^b P < 0.05.

We also showed that patients with bvFTD were impaired in all evaluated aspects of creativity (Table 1 and Fig. 2), performing worse than controls in both the intentional generation of remote semantic associates (FGAT) and the efficient combination of several associations (CAT), as well as in overall creative abilities (AUT).

Figure 2.

Group comparisons of efficiency scores across creativity tasks. Distribution and group comparisons for participants’ efficiency in the Free Generation of Associates Task (FGAT), Combination of Associates Task (CAT) and Alternative Uses Task (AUT) between patients with behavioural variant frontotemporal dementia (bvFTD, in red) and healthy controls (in green). ***P < 0.001.

Behavioural and affective symptoms were not significantly associated with creativity task performance (Supplementary Table 1). Specifically, neither apathy scores (SAS, DAS total or DAS subscales) nor mood disturbance (HADS) correlated with FGAT, CAT or AUT performance. A negative correlation observed between HADS total and anxiety scores and AUT efficiency did not survive FDR correction for multiple testing.

In patients with bvFTD, the equivalent fluoxetine dose, representing antidepressant dose regardless of the type of antidepressant,¹³⁶ showed no significant association with FGAT (ρ = −0.03; P = 0.936), CAT (ρ = 0.38; P = 0.182) or AUT (ρ = −0.09; P = 0.782) performance. Patients under other medications (benzodiazepines, n = 4; antipsychotics, n = 2) were too scarce to allow for reliable correlation statistics.

Voxel-based morphometry analyses

Pattern of atrophy in patients

Group comparisons of GMV showed an atrophy pattern in patients characteristic of bvFTD (Fig. 3 and Supplementary Table 2), mainly affecting the bilateral dorsolateral PFC (dlPFC), orbitofrontal (OFC), cingulate and insular cortices.

Figure 3.

Voxel-based morphometry results.  Top row: Atrophy pattern in patients with behavioural variant frontotemporal dementia (bvFTD) compared to controls. Significant clusters from the whole-brain voxel-based morphometry (VBM) group comparison are colour-coded to show voxels where grey matter volume (GMV) is lower in patients with bvFTD than in controls. The significance threshold was set at P_FWE < 0.05 at the voxel level, with a spatial cluster extent threshold of 100 contiguous voxels. Rows 2–4: Structural brain correlates of each creativity task’s efficiency. Significant clusters for whole-brain non-parametric VBM regressions of GMV on creativity scores in the entire population (n = 45, 18 patients and 27 controls) are colour-coded for each creativity task (FGAT: blue, CAT: red, and AUT: green). Age, sex, education and semantic score were used as covariates. The significance threshold was set at P_FWE < 0.05 at the voxel level, with a spatial cluster extent threshold of 100 contiguous voxels. Bottom row: Overlap of each task’s brain correlates (using the same colour coding) with the rostral prefrontal cortex (Brodmann area 10, in yellow). AUT = Alternative Uses Task; BA = Brodmann area; CAT = Combination of Associates Task; FGAT = Free Generation of Associates Task; FWE = family-wise error; L = left; PFC = prefrontal cortex; R = right.

Neural correlates of creativity task efficiency

We used whole-brain voxel-wise analyses across patients and controls (n = 45, 18 patients and 27 controls) to explore brain regions critical for each creativity task (Fig. 3 and Supplementary Table 3). FGAT efficiency correlated positively with GMV in the left rostromedial PFC and in the right insula and ventrolateral PFC (vlPFC). CAT efficiency correlated with GMV in the bilateral rostrolateral PFC, the right vlPFC/insula, medial occipitotemporal cortex and inferior pre- and postcentral cortex, as well as in the left temporoparietal junction. Finally, we found a positive correlation between AUT efficiency and GMV in both the bilateral rostrolateral PFC and the left rostromedial PFC, as well as in the bilateral OFC, dorsomedial PFC (dmPFC) and insula, the right vlPFC and dlPFC, and in the left precentral PFC.

To ensure that these results were not driven by group differences and represent generalizable brain-behaviour relationships, we ran the same analyses after covarying out participants’ clinical status (bvFTD or controls). These control analyses showed similar results to the initial analyses, with lower statistical power (Supplementary Fig. 1 and Supplementary Table 4). Nonetheless, all clusters identified in the initial analyses were also observed in the control analyses, except for the right vlPFC cluster related to CAT efficiency.

Functional connectivity gradients analyses

Principal gradient of the rostral PFC

The principal functional gradient of the rostral PFC, i.e. the first of the top ten gradients explaining functional variation within the rostral PFC, computed at the group level in the 24 healthy controls, accounted for 45% of variance (Supplementary Fig. 2).

We plotted principal gradient values on the brain to explore the topographical pattern of functional variation across the rostral PFC (Fig. 4). We showed that the rostral PFC principal gradient is organized along a mediolateral axis, with the rostromedial PFC on one end of the gradient, and the bilateral rostrolateral PFC on the other end.

Figure 4.

Group-level rostral prefrontal cortex principal resting-state functional connectivity gradient in healthy subjects. Regions depicted with colours that are closer together on the colour spectrum exhibit greater similarity in their correlation over time during resting-state functional MRI compared to regions depicted with colours further apart on the spectrum. The plus and minus symbols represent the two opposite poles of the principal gradient, but the assignment of a particular end to either pole is arbitrary.

Rostral PFC principal gradient external validity

We ran the same gradient analysis in two independent datasets of 93 (Ovando-Tellez et al.¹²⁶) and 36 (Moreno-Rodriguez et al.⁵⁰) young healthy subjects. The results (Supplementary Fig. 3) successfully replicated the group-level mediolateral principal functional gradient of the rostral PFC shown in Fig. 4.

Principal gradient functional characterization

In the reference healthy controls group (n = 24), the algebraic connectivity of the group-level rostral PFC principal gradient was 0.77, indicating a highly progressive functional gradation from the rostromedial PFC to the rostrolateral PFC.

Finally, to characterize the functional implications of the mediolateral gradient in the normal brain, we explored which cerebral regions were preferentially connected to each end of the rostral PFC principal gradient (Fig. 5 and Supplementary Table 5) by contrasting their seed-based connectivity patterns in healthy controls (n = 24). The rostromedial third of the gradient was preferentially connected to the bilateral vmPFC, OFC, precuneus, anterior and posterior cingulate cortex, and medial and anterolateral temporal cortex. These regions mainly overlapped with the DMN (55%) and, to a lesser extent, the limbic network (18%). On the other hand, the rostrolateral third of the gradient was preferentially connected to the bilateral dlPFC and cerebellum (Crus I and lobule VI), the left vlPFC and the left PPC. These regions fall mainly within the ECN (52%) and, to a lesser extent, the SN-VAN (29%).

Figure 5.

Seed-based intrinsic connectivity networks of each end of the group-level rostral prefrontal cortex principal gradient extracted in healthy subjects. We compared the resting-state functional connectivity with the rest of the brain of the two ends of the rostral prefrontal cortex (PFC) principal gradient (regions corresponding to the lower and higher thirds of the gradient values) using whole-brain non-parametric paired two-sample t-tests, in healthy controls (n = 24). Top: Significant clusters that are significantly more functionally connected to the rostromedial than the rostrolateral PFC. Bottom: Reverse comparison. Statistical analyses were thresholded for significance at P_FWE-corrected < 0.05 at the voxel level, with an additional spatial cluster extent threshold of 10 contiguous voxels. Lower and higher thirds of the gradient value in rostral PFC that were used as seeds are represented in black. The proportions of overlap of the resulting functional networks with Yeo’s seven intrinsic functional networks^130-132 are reported on radar plots for each extreme third. DAN = dorsal attention network; DMN = default mode network; ECN = executive control network; FWE = family-wise error; SMN = somatomotor network; SN-VAN = salience network-ventral attention network.

Relationship between principal functional gradient and bvFTD pathology

In all patients and controls, we computed individual-level rostral PFC gradients aligned to the reference group-level gradients (see the ‘Materials and methods’ section). We then explored the range of individual-level principal gradients. We confirmed our hypothesis of a decreased (U = 120.00; P = 0.043) principal gradient range in bvFTD patients (mean = 16.80; SD = 5.56) as compared to controls (mean = 20.02; SD = 5.31), indicating reduced differentiation in the connectivity patterns of both ends of the mediolateral gradient.

To characterize the impact of bvFTD on the rostral PFC principal gradient, we explored whether GMV could account for decreases in gradient range across patients and controls (n = 39, 15 patients and 24 controls). We found that whole-brain GMV positively related to gradient range variations [β = 0.049, standard error (SE) = 0.020, t(32) = 2.41, P = 0.022, standardized β = 0.50], and GMV in BA 10 showed a similar positive association with gradient range [β = 1.26, SE = 0.59, t(32) = 2.12, P = 0.041, standardized β = 0.40]. Partial regression plots are provided in Supplementary Fig. 4.

Relationship between principal functional gradient and creativity tasks

To understand the implications of gradient range modifications on different aspects of creativity, we tested whether the rostral PFC principal gradient range contributed to creative efficiency in the CAT, FGAT and AUT using linear regressions combining patients and controls (n = 39, 15 patients and 24 controls). While gradient range variations did not statistically explain variations in generating [FGAT efficiency; β = 0.038, SE = 0.035, t(32) = 1.07, P = 0.291] or combining [CAT efficiency; β = 0.021, SE = 0.049, t(32) = 0.43, P = 0.672] remote semantic associations, they did contribute to individual creative abilities as assessed by the AUT efficiency [β = 0.065, SE = 0.025, t(30) = 2.62, P = 0.014, standardized β = 0.41]. Partial regression plots for each task are provided in Supplementary Fig. 5.

Control analyses in additional Brodmann areas

To test the specificity of gradient mapping findings in the rostral PFC (BA 10), we ran the same gradient analysis in BAs 8, 9, 11, 20, 23, 26, 44, 45 and 47, where the DMN and ECN co-exist (each network representing over 10% of the BA’s volume; Supplementary Table 6). Results showed that only BAs 8, 9 and 20 displayed a principal RSFC gradient segregating the DMN and ECN (Supplementary Fig. 6) However, for these three BAs, no significant association was found between the range of the principal gradient and FGAT efficiency [BA 8: β = 0.012, SE = 0.048, t(32) = 0.26, P = 0.798; BA 9: β = 0.003, SE = 0.052, t(32) = 0.056, P = 0.955; BA20: β = 0.071, SE = 0.058, t(32) = 1.22, P = 0.233], CAT efficiency [BA 8: β = −0.006, SE = 0.065, t(32) = −0.087, P = 0.931; BA 9: β = 0.010, SE = 0.071, t(32) = 0.14, P = 0.890; BA20: β = −0.027, SE = 0.081, t(32) = −0.33, P = 0.744] or AUT efficiency [BA 8: β = 0.052, SE = 0.032, t(32) = 1.59, P = 0.122; BA 9: β = 0.070, SE = 0.037, t(32) = 1.88, P = 0.070; BA20: β = 0.009, SE = 0.042, t(32) = 0.22, P = 0.831]. Partial regression plots are provided in Supplementary Fig. 7.

Discussion

This study provides new insights into creative cognition, how it relates to PFC organization, and how it is affected in bvFTD. It reveals the distinct critical roles of the rostromedial and rostrolateral PFC in the generation and combination of remote semantic associates, respectively, but also their joint involvement in individual creative abilities. Furthermore, this study takes advantage of the recent functional connectivity gradient mapping methods to better characterize the poorly understood role of the rostral PFC in human cognition. This approach allowed us to unveil the importance to creative abilities of the graded functional segregation between rostromedial and rostrolateral PFC, respectively connected to the DMN and ECN.

Critical role of the rostromedial prefrontal cortex in generating remote semantic associations

Using whole-brain VBM, we found a critical role of rostromedial PFC integrity in the intentional generation of remote semantic associations (FGAT efficiency), which was impaired in patients with bvFTD. The involvement of the rostromedial PFC in the generation of creative semantic associations has previously been shown in healthy subjects^26,42,53,137 and patients. Indeed, our results replicate, in a different pathological population, previous findings obtained with the same FGAT task in patients with focal frontal lesions.¹⁶ The rostromedial PFC is hypothesized to relate to memory retrieval processes involved in self-driven creative idea generation, and is a critical region for remote thinking.¹² This cortical region is a key node of the core DMN,^33,138 whose central role in the self-generation of mental contents^33,139,140 and in creativity^{16,30,31,141,142} has been well established. The rostromedial PFC appears essential for processing semantic and episodic memory information^143-145 that are instrumental to creative thinking.^3,5,146 FGAT efficiency also relied on the integrity of the right vlPFC and insula, regions that were also found to be critical for CAT efficiency and AUT efficiency. These cortical regions, classically related to inhibition processes,¹⁴⁷ may be involved in the suppression of inadequate candidate ideas, a key aspect of creative cognition.^{12,48,148-152}

Critical role of the rostrolateral prefrontal cortex in combining remote semantic associations

The capacity to combine semantic associates to solve a creative problem (CAT efficiency) was also impaired in patients with bvFTD. Here also, VBM results in bvFTD patients replicated previous findings observed in healthy controls or in other pathological conditions using the CAT, demonstrating the critical role of the rostrolateral PFC for this task.^16,27,79 In addition to associative processes^4,8,153 which are also assumed to be involved in the FGAT, the CAT is hypothesized to involve high-level control processes such as relational integration and combination,^16,22,26-28 constrained strategic memory search,^3,19 or implementing, monitoring and coordinating multiple subtasks.¹⁶ Hence, our results are consistent with the established role of the rostrolateral PFC in processing relational information between words,^{23,24,26,42,53,117,154} multitasking^59,69,155 or branching.^60,156 As part of the ECN,¹³⁰ the rostrolateral PFC is indeed central to several controlled processes supporting the combination and relational integration of multiple semantic elements, a key aspect of creative cognition.¹⁶ Interestingly, the CAT also related to the GMV in regions of the visual (right medial occipitotemporal cortex) and somatomotor (right inferior frontoparietal cortex and left temporoparietal junction) networks, whose role in creativity has been highlighted recently.^21,157

Overall, our current results replicate in another pathological model the results from Bendetowicz and colleagues’ study¹⁶ showing the critical role of the rostromedial and rostrolateral PFC for generative and combinatory processes of creative cognition, respectively.

Critical role of both the rostromedial and rostrolateral prefrontal cortices in individual creative abilities

After extending the results from Bendetowicz et al.¹⁶ to another pathology, we explored how these regions are critical for creative potential, as assessed with the AUT. VBM results for AUT efficiency confirmed the critical roles of the rostromedial and rostrolateral PFC for creative abilities. This finding aligns with prior fMRI literature⁴² but crucially adds the first direct evidence that creative abilities depend on the integrity of both regions. These results support the conceptual framework of creativity as relying on the complementary interaction between associative and executive processes.^5,8 In this view, the rostromedial PFC, located at the core of the DMN, is likely to play a central role in the associative processes involved in the generation of novel, uncommon, semantic associations that allow original ideas to emerge, while the rostrolateral PFC, connected to the ECN, appears central for controlled processes required to efficiently combine and integrate multiple associative elements in a way that satisfies the goal and constraints of a given situation. Hence, our results position the rostral PFC as a critical hub in the interplay between the DMN and ECN, which respectively support the associative and controlled components of creative cognition. Yet, our results are at odds with a recent work suggesting that lesions to the right rostral PFC or its connected regions may enhance creativity.¹⁵⁸ First, aligned with our results, this study identified the right frontal pole as a key region for creativity based on seed-based connectivity from meta-analytic activation peaks across 36 fMRI studies in healthy participants. However, by comparing this network to brain damage patterns in various diseases through large-scale meta-analysis, the authors suggested that focal or neurodegenerative lesions affecting the frontal pole or a network positively connected to it enhance creativity. Despite the elegant and large-scale nature of these analyses, their conclusions rest on assumptions that remain debated and untested: first, that focal frontal pole lesions promote creativity, an inference not statistically supported by the cited work¹⁵⁹; and second, that creativity increases in bvFTD, a notion based on isolated case reports but contradicted by findings from controlled patient studies.^{54,90,160,161} Moreover, this study did not statistically test the relationship between individual lesion location and creative performance, limiting causal interpretation. In contrast, the current study demonstrates creativity impairment in bvFTD and provides direct empirical evidence linking rostral PFC integrity with creative abilities. In addition, we demonstrated through functional connectivity analyses that the rostral PFC's connectivity organization underpins its role at the DMN-ECN interface, is disrupted in patients with bvFTD, and predicts creative abilities, as discussed later.

A rostral prefrontal mediolateral gradient segregates default mode and executive control networks

We used RSFC gradient mapping techniques to characterize the functional connectivity profile within the rostral PFC and explore whether medial and lateral PFC are segregated by abrupt boundaries or graded transitions. This approach allowed us to characterize the principal connectivity gradient within the rostral PFC, explaining 45% of the rostral PFC’s functional variation in healthy individuals. The rostral PFC’s principal gradient, which was replicated in two independent datasets, was organized along a mediolateral axis, with the rostromedial and rostrolateral PFC at each end of the gradient. Hence, the main variation in the functional connectivity profile within the rostral PFC, spanning from the rostromedial to the rostrolateral PFC, echoes our VBM findings, which highlight distinct cognitive roles for these regions. We further showed that the functional segregation between the rostromedial and rostrolateral PFC along this gradient is characterized by a high level of progressive gradation rather than a sharp delimitation between subregions. These results extend previous literature on the rostral PFC organization by using, for the first time, the potential of functional gradient mapping to reveal a finer-grained characterization of its functional connectivity variation.¹⁶² Indeed, while our results are supported by previous connectivity studies showing structural and functional differentiation along a mediolateral axis,^73-76 these past studies used hard parcellation techniques that resulted in various and debated numbers of parcels inside the rostral PFC.⁷³ By using a data-driven exploration of functional variation without a priori, we were able to show that the principal functional organization of the rostral PFC does not follow a hard parcellation but rather a highly graded pattern of functional variation along the mediolateral axis.

Using seed-based intrinsic connectivity, we then showed that the rostromedial third of the gradient mostly connected with regions of the DMN and, to a lesser extent, of the limbic network, itself highly connected with the DMN.¹⁶³ Conversely, the rostrolateral third of the gradient mostly connected with regions of the ECN and, to a lesser extent, of the salience network. These results confirm that the rostromedial and rostrolateral PFC are rooted in the DMN and ECN, respectively,^73,75,130 as previously proposed.^73,75 Furthermore, these findings align with our VBM results, which link generative aspects of remote thinking, related to the DMN, to the rostromedial PFC, and controlled combinatory aspects, related to the ECN,^5,7,8,30 to the rostrolateral PFC.

Overall, this graded connectivity between the rostromedial and rostrolateral PFC, and thus between the DMN and ECN, has theoretical implications for the understanding of the cerebral bases of creativity. Indeed, the graded mediolateral functional connectivity of the rostral PFC places this region in a position to act as a DMN-ECN connecting hub, facilitating the efficient coordination between these two networks and thus between several specialized cognitive processes supported by them. Combined with our VBM results, the current findings begin to clarify how the functional organization of the rostral PFC supports the generation and combination of remote semantic associations, and possibly their integration for creativity.¹⁶⁴

Functional distance between rostromedial and rostrolateral prefrontal cortices predicts creative abilities

Interestingly, we also found that the range of the mediolateral principal gradient, i.e. the dissimilarity in functional connectivity between the rostromedial and rostrolateral PFC, positively predicted individual creative abilities (AUT efficiency). However, while creative abilities are thought to involve both generative and combinatory processes, the rostral PFC’s principal gradient range did not predict the generation (FGAT efficiency) or combination (CAT efficiency) of remote semantic associates. Hence, greater functional segregation between rostromedial and rostrolateral PFC may not specifically underpin one distinct process of creative cognition supported by these regions, but instead contributes to broader creative abilities that emerge from the interaction between these processes. The rostral PFC gradient may provide the functional architecture that allows the complementary generative and combinatory processes to be flexibly engaged, independently or jointly, depending on task demands. Greater functional segregation between these two interconnected regions and their two affiliated networks, the DMN and ECN, could allow distinct types of cognitive processes to work independently when needed, without hindering their ability to cooperate during certain phases, favouring fluid and efficient dynamic interactions. Indeed, generating remote semantic associations might result in more original ideas when the DMN is working independently from the constraints of control processes supported by the ECN, while coordinated recruitment of both networks could be needed when combining several semantic associations to meet specific constraints.^{31,142,165,166} Further work will be required to clarify how these generative and combinatorial components of creative cognition, and the corresponding medial and lateral rostral PFC subregions, cooperate to support overall creative abilities. Our findings nonetheless align with recent results indicating that hub regions, which connect the DMN and ECN and show the highest diversity in their functional connectivity, may facilitate interactions between functionally dissimilar cortical regions, supporting integrated cognitive functions such as creativity.¹⁶⁴ Furthermore, the control analysis applied in other BAs showed that the existence of a principal connectivity gradient with a range of DMN-ECN segregation associated with creative abilities was specific to the rostral PFC.

Overall, functional differentiation along the mediolateral rostral PFC gradient could allow a broader scope of possible interactions in the DMN-ECN interplay and along the cognitive continuum between generative and combinatory processes, leading to greater creativity.⁸¹ This interpretation, supported by recent findings involving macroscale gradients,¹⁶⁷ implies that the rostral PFC plays a cornerstone role in the segregated yet collaborative operations of complementary processes and networks. Extending recent results showing the shrinkage of macroscale gradients in bvFTD,¹⁰³ we also demonstrated a positive relationship between both brain-wide and BA10 atrophy and a decreased range of the mediolateral rostral PFC gradient in bvFTD, as well as a significant shrinkage of this gradient in patients with bvFTD as compared to controls. These results indicate that brain atrophy due to bvFTD affects functional connectivity resulting in reduced functional specialization of the rostromedial and rostrolateral PFC. This loss of differentiation between the DMN and ECN may in turn contribute to the impairment of creative abilities observed in patients.

Limitations

A few methodological limitations of this study should be considered. First, given the rarity of bvFTD and our strict inclusion criteria, our sample size is rather limited. However, VBM results replicated previous results using the same creativity tasks in other populations and experimental paradigms, supporting their reliability, as did the replication of the rostral PFC’s principal gradient in two independent datasets. Nonetheless, our multiple regression analyses linking mediolateral rostral PFC gradient range to creative behaviour are exploratory and should be replicated in larger samples to confirm our findings. Second, psychotropic medications commonly prescribed in bvFTD or affective and behavioural symptoms such as anxiety, depression, or apathy may have contributed to cognitive variability in the patient group. However, the strong collinearity of these parameters with bvFTD diagnosis in a rather limited sample size prevented formal modelling of their respective contributions to creative cognition. Thus, larger, stratified samples, including unmedicated patients and/or medicated controls (excluded from the current study to avoid confounding psychiatric pathologies), will be needed to specifically determine how psychotropic exposure or affective or behavioural disturbances may modulate creative cognition and its neural correlates in bvFTD and in other populations. Nonetheless, our findings reflect the cognitive profile of bvFTD patients as they are naturally treated in clinical settings, preserving their ecological validity. Third, our results on rostral PFC’s principal gradient were obtained using resting-state fMRI data, and their applicability to task-related fMRI data remains to be investigated. Indeed, previous studies have shown that the connectivity between the DMN and ECN is highly dynamic, with diverse patterns of coupling and decoupling depending on ongoing creativity tasks.^{30,31,80,141,165} Finally, while our results link gradient range and creative abilities, the specificities of our population should be kept in mind. Indeed, our results are likely to be driven by the gradient range decrease in patients with bvFTD, in parallel to the decrease in creative abilities. However, the gradient range related to creative abilities measured by the AUT efficiency, but not to FGAT efficiency or CAT efficiency, although they were also impaired in patients, suggesting that the association with AUT efficiency is not merely driven by bvFTD versus control differences. Nonetheless, replicating these results in other populations, particularly in those with an increased gradient range compared to healthy controls, would be of great value for generalizing the link between the rostral PFC functional gradient and creative abilities. In parallel, it is worth noting that although most controlled studies converge on impaired creativity in frontotemporal dementia, rare cases of emergent artistic production have been reported, particularly in patients with temporal involvement, such as in semantic dementia.^168-170 Interestingly, Friedberg and colleagues¹⁶⁸ have shown that left anterior temporal atrophy may lead to increased dorsomedial occipital activity, which in turn could support emergent visual creativity. While patients with bvFTD are less frequently concerned by visual artistic enhancement and temporopolar atrophy,¹⁶⁸ recent work using gradient mapping suggests that salience network dysfunction, a common and early feature of bvFTD,⁹⁴ may also enhance connectivity of the visual network located at the opposite pole of a salience-visual gradient.¹⁷¹ Although our study did not include patients with artistic emergence, future applications of gradient mapping to such cases could provide valuable insights into the neural mechanisms underlying these rare but intriguing manifestations.

Conclusion

In summary, we show the critical brain regions for the generation and combination of remote associative thinking in creativity and how they are part of distinct brain networks that involve the rostral PFC. Our results place the rostral PFC as a critical cortical hub for the interplay between these functional networks, in particular the DMN and ECN. Specifically, we provide evidence for a functional organization of the rostral PFC along a graded mediolateral axis segregating the rostromedial PFC, a core node of the DMN critical for the generation of remote semantic associations, and the rostrolateral PFC, an essential node of the ECN critical for the goal-directed combination of multiple semantic associates. Furthermore, we establish the role of this functional connectivity gradient in creativity by showing that its shrinkage in bvFTD patients predicts overall creative abilities, rather than the specific generative or combinatory components of creative performance, which it may help integrate. Our findings extend existing knowledge of the proposed role of the DMN-ECN cooperation for creativity by showing that the rostral PFC plays a critical role in it, and maybe more surprisingly, that higher functional segregation within this region is conducive to higher creative abilities.

Appendix 1

Members of the ECOCAPTURE study group contributed to this study: Shany Aflalo, Victor Altmayer, Carole Azuar, Vanessa Batista-Brochard, David Bendetowicz, Théophile Bieth, Anaïs Blau, Armelle Blossier, Louis Bonnefous, Stéphanie Bombois, Mathilde Boucly, Arabella Bouzigues, Bénédicte Batrancourt, Blandine Cabrol-Douat, Guilhem Carle, Céline Chamayou, Fanny Colin, Emmanuel Cognat, Sophie Coriou, Jean-Christophe Corvol, Bruno Dubois, Johan Ferrand-Verdejo, Aurélie Funkiewiez, Paolo Fulcheri, Manon Gallouedec, Cécile Gaudebout, Marie-Alexandrine Glachant, Valérie Godefroy, Vincent Guillemot, Hortense Guitton, Sarah Hachemi, Estelle Janvier, Lucie Jeanne, Lars Jorgensen, Aline Jourdain, Mathilde Joulié, Elena Karpinski, Tatiana Landre, Isabelle Le Ber, Vincent Le Du, François-Xavier Lejeune, Richard Levy, Yongjian Liu, Karen Lecouturier, Graziella Mangone, Louise-Laure Mariani, Frédéric Marin, Constance Marque, Carla Matos Araujo, Raffaella Migliaccio, Aymen Mousli, Pierre Mullot, Raphaëlle Odobez, Alizée Paquiry, Caroline Peltier, Fanny Pineau, Flore Pochan-Leva, Myriam Provost, Armelle Rametti-Lacroux, Anaïs Raud, Rozenn Renaud, Nassim Saidi, Dario Saracino, Idil Sezer, Julie Socha, Camille Simard, Delphine Tanguy, Daphné Tessereau-Barbot, Antoni Valero-Cabre, Giulia Visentin, Hélène Vinatea, Emmanuelle Volle.

Data availability

We report all methods in the main manuscript or Supplementary material. Data and analysis codes are openly available at https://osf.io/rtnzw/overview?view_only=f6c54494c28349b083f4eec0bded9562, except for structural and RSFC brain imaging scans, which are not provided due to ethical protocol constraints related to the bvFTD clinical population. Task materials are available upon reasonable request.

Funding

The research was supported by the ‘Agence Nationale de la Recherche’ grant number ANR-19-CE37-0001-01 and received infrastructure funding from the French programs ‘Investissements d'avenir’ ANR-11-INBS-0006 and ANR-10-IAIHU-06. V.A. was supported by the ‘Fondation pour la Recherche Médicale’ (FRM), grants number M2R202003011383 and FDM202206015358. M.O.T. was funded by Becas-Chile of Agencia Nacional de Investigación y Desarrollo-Comisión Nacional de Investigación Científica y Tecnológica (ANID/CONICYT). A.L.P. was supported by the European Union’s Horizon 2020 Research and Innovation Program under the Marie Sklodowska-Curie grant agreement number 101026191. The ECOCAPTURE protocol was funded by: the program ‘Investissements d’avenir’, ANR-10-IAIHU-06, France; the ‘Fondation pour la recherche médicale’ (FRM), DEQ20150331725, France; the ENEDIS company (ERDF), France, 2015-2017; the Malakoff Humanis company, France; and by the French Federation for Research on the Brain (Fédération pour la Recherche sur le Cerveau), FRC 2020.

Competing interests

The authors report no competing interests.

Supplementary material

Supplementary material is available at Brain online.