Narrative coherence shapes functional connectivity in default mode and frontoparietal networks

Abstract

Narratives are central to human cognition, yet the brain mechanisms supporting coherent story construction remain incompletely understood. In this study, we used a well-controlled naturalistic paradigm to isolate the neural basis of narrative coherence. Participants watched a foreign-language film with either intact or scrambled English subtitles, disrupting narrative understanding while preserving auditory and visual input. Whole-brain functional connectivity multivariate pattern analysis (fc-MVPA) revealed significant differences between conditions in the ventromedial prefrontal cortex (vmPFC), posterior cingulate cortex (PCC), lateral prefrontal cortex, and right angular gyrus—regions spanning the default mode and frontoparietal control networks. Seed-based correlation analyses showed stronger vmPFC and PCC coupling with lateral temporal and frontoparietal regions during coherent narratives, supporting semantic and contextual integration. In contrast, incoherent narratives increased connectivity between the default mode and attention networks, consistent with elevated cognitive demands. Finally, we observed a functional dissociation within the DMN: vmPFC–PCC coupling was enhanced during incoherent viewing, while PCC–precuneus coupling was stronger during coherent narratives. These results highlight dynamic reconfiguration across large-scale brain networks in support of real-world narrative comprehension.

Highlights

• •Used fc-MVPA to examine coherent narrative formation in brain.
• •Coherent narratives reconfigure default mode and frontoparietal networks.
• •Coherent narratives strengthen vmPFC and PCC links with temporal and control regions.
• •Incoherent narratives increase default to attention coupling and vmPFC to PCC links.

1. Introduction

In daily life, we seamlessly move between various simultaneous narratives, such as work-related projects and discussions, the daily and weekly routines of family life, and the episodes of our favorite TV shows. Despite these contexts being presented in interrupted and dispersed segments over long periods, they usually form complete and coherent experiences in our conscious memory. The capacity to aggregate, synthesize, and shape fragmented (and often incomplete) information implicitly over extended timescales is an important human trait, forming the basis for episodic memory. These to-be-encoded events exist within dense connections across time and space in a rich, multimodal environment, thus forming an individual narrative of events encountered in everyday life that can be retrieved involuntarily as new information is encountered.

Narrative formation in the brain has primarily been studied using naturalistic stimuli, such as audiovisual movies, podcasts, gaming environments, and virtual reality (Hasson, 2004; Sonkusare et al., 2019). Naturalistic stimuli better encapsulate the task demands of everyday life, thus providing a more accurate proxy of how the brain processes and encodes events in the real-world. Using naturalistic stimuli, research has shown that integration of multimodal information into ongoing autobiographical narratives is reliant on the default mode network (DMN), suggesting it plays a crucial role encoding, retrieving, and integrating incoming multimodal information into pre-existing memory schemas (Menon, 2023; Tylén et al., 2015). The DMN includes cortical regions such as the posterior cingulate cortex (PCC), precuneus, retrosplenial cortex, medial prefrontal cortex (mPFC), anterior temporal cortex, middle temporal gyrus (MTG) in the lateral temporal cortex, angular gyrus, and the medial temporal lobe (MTL) and subcortical nodes in the anterior and mediodorsal thalamic nuclei, medial septal nuclei, and nucleus accumbens (Menon, 2023; Smallwood et al., 2021). Further, a review by Jääskeläinen et al. (2020) highlights differential roles of DMN structures in processing narratives, whereby the ventromedial PFC (vmPFC) appears to be related to emotional regulation, the precuneus is associated with narrative interpretation, and the angular gyrus is involved with attentional engagement and suspense. Yeshurun et al. (2021) argue that the DMN acts as ‘sense-making’ network, integrating external information with prior intrinsic information to create coherent models of ongoing events. As such, the DMN plays an important role in constructing coherent representations of complex information from diverse sensory modalities, such as visual, auditory, and semantic inputs.

Previous research on understanding how the brain recalls and connects relevant information into ongoing memory has generally focused on how disrupting narrative coherence affects brain activation. For example, Kauttonen et al. (2018) used audiovisual films to identify brain regions involved in recalling and connecting relevant memories in an ongoing narrative. To do so, participants viewed two versions of an audiovisual film during functional magnetic resonance imaging (fMRI), one in reverse chronological order (the original order of the film), where key-frames serve as memory cues to recall previous scenes, and an altered version of the same film in chronological order (the control condition). Results from a multivariate event-related pattern analysis showed that, during key-frames, there was greater activation in DMN areas such as the precuneus, angular gyrus, and cingulate gyrus, as well as the lateral, superior, and middle frontal gyri. This suggests that these regions are associated with recalling relevant memories and connecting them with incoming information.

In line with this, Ren et al. (2018) used naturalistic stimuli to examine involuntary episodic memory retrieval during fMRI. While most previous research relied on voluntary recollection of abstract and static stimuli to assess episodic memory, Ren et al. (2018) argued that the majority of memory retrieval in everyday life occurs involuntarily. To assess involuntary memory retrieval, participants viewed a series of news clips under two conditions: one where the second half of the news clips were continuations of the first clips, and another where the cues were not continuations. They hypothesized that viewing the continuation of previously viewed news clips would result in involuntary memory retrieval, allowing them to probe memory processes more akin to real-life. Results from this study showed that viewing continuous film clips led to greater activation of the bilateral anterior hippocampus, precuneus, and angular gyrus compared to viewing non-continuous clips, suggesting that these regions play a key role in involuntary memory retrieval under naturalistic conditions. The researchers also found that connectivity to the precuneus was a key predictor of memory confidence, but not accuracy. Thus, this study provides important implications for understanding ongoing narrative consolidation and retrieval in contexts more akin to those in the real-world.

Several other studies have taken connectivity approaches to examining narrative and memory processing in naturalistic contexts. Van Kesteren et al. (2010) found increased functional connectivity between the hippocampus and vmPFC while watching the last 15 min of a movie when participants were previously presented with a scrambled version of the first 80 min of a movie. Further, Assouline and Mendelsohn (2023) sought to identify connectivity patterns associated with causal connections in audiovisual narratives. These researchers used functional connectivity multivariate pattern analysis (fc-MVPA; a whole-brain, data-driven approach that provides information about functional connectivity patterns of each voxel with all other voxels) to examine differences in functional connectivity when participants watched a film in correct chronological order (i.e., with causal connections intact) vs. reverse chronological order. Results showed enhanced functional connectivity between the cerebellum and frontoparietal regions for the chronological vs. reversed narrative. Finally, Türker et al. (2023) found evidence of characteristic patterns of connectivity between the DMN and other functional connectivity networks that was unique to intact movie viewing (as opposed to viewing scrambled movie viewing or resting state), including higher coherence between the DMN and frontoparietal and visual networks, and decreased coherence between the DMN and salience networks. Thus, the formation of coherent narratives in the brain appears to be associated with functional connectivity between regions of the DMN, hippocampus, and frontoparietal regions.

However, while previous research has made great strides in understanding the neural correlates of coherent narrative formation in the brain, the majority of this research compares conditions where the narrative is intact versus conditions where the core structure or audiovisual content of the narrative is altered in some way (e.g., reordering scenes, presenting the narrative in a scrambled or reverse order, attempting to use approximately matched narrative stimuli; e.g., Assouline and Mendelsohn, 2023; Kauttonen et al., 2018). Therefore, differences in brain activation may reflect other unanticipated cognitive demands, such as increased cognitive load from processing an incoherent narrative, reduced or inconsistent engagement to the altered narrative, different memory demands, and a loss of temporal coherence. To more accurately study how coherent narratives are formed in the brain at different levels of understanding, it is essential to use appropriately matched stimuli. Thus, the current study seeks to provide a more comprehensive understanding of how narratives are formed in the brain by varying the semantic information available to the participant, while keeping the audiovisual characteristics of the stimulus constant.

In this study, we investigated differences in functional connectivity during the processing of a complex audiovisual narrative (a foreign language short-film) under two conditions: a coherent context condition (where subtitles provide semantically relevant details aligned with the ongoing narrative) and an incoherent context condition (where subtitles are presented out of order while the other audiovisual elements remain unchanged). By keeping the order of the scenes intact and manipulating only the subtitle sequence, we ensured that both conditions were matched in visual and auditory information while preserving the causal structure of the plot. This design contrasts with previous studies that disrupted coherence by altering scene order (e.g., Assouline and Mendelsohn, 2023; Kauttonen et al., 2018), allowing us to more directly disentangle functional connectivity networks associated with coherent narrative formation.

Based on previous research, we hypothesized that there would be differences in functional connectivity between the coherent narrative condition and the incoherent narrative condition to regions of the DMN, particularly areas suggested to be relevant to narrative processing, such as the vmPFC, PCC, precuneus, and angular gyrus (Jääskeläinen et al., 2020; Tylén et al., 2015). Further, in line with previous research (Menon, 2023; Smallwood et al., 2021; Türker et al., 2023), we hypothesized that the coherent narrative condition would show greater DMN connectivity than the incoherent narrative condition, as well as greater connectivity between the DMN and the frontoparietal and visual networks. Overall, this study will provide the most comprehensive analysis to date of functional connectivity analyses associated with coherent narrative formation in the brain.

2. Methods

2.1. Participants

45 healthy participants participated in this study. One participant dropped out during scanning, and data from two participants was removed prior to analysis due to the participants pressing the call button during a film, requiring the film to be restarted. This resulted in 42 participants in the final analysis (28 females; mean age 27.7, standard deviation 9.1 years). All participants had normal or corrected-to-normal vision, spoke English as their first language, did not speak Danish (the language of the foreign language film), and none of the participants were familiar with the movie presented in the study. Participants received monetary compensation of $20/hr for participation in this study ($40 per session, which included data collection for other studies not reported here). The experimental procedures were approved by the University of Alberta research ethics board in accordance with the Declaration of Helsinki (protocol number Pro00120063, date of approval April 20, 2022). All participants provided written informed consent prior to the experiment.

2.2. Stimuli and experimental design

In the MRI scanner, participants watched a modified version of the short Danish film ‘9 meter’ (Walter, 2012) twice: once with English subtitles in the properly presented order (Intact Subtitle condition), and once with the same English subtitles presented in a pseudorandom order throughout the film (Scrambled Subtitle condition). The order of film conditions was counterbalanced across participants. Participants were informed that the subtitles would differ between the two versions of the film, but were not told the nature of these differences. This film depicts the story of a 16-year-old boy who believes that he needs to continue breaking records in long jump to keep his dying mother alive, whereby the jumps continue to escalate in danger. This film was chosen based on its strong, visually driven narrative, whereby the overarching narrative is conveyed primarily visually, rather than through dialogue. The main plot points of the film can be ascertained without subtitles (i.e., the overarching goal of the protagonist), but semantic details from the subtitles are much more salient in the Intact Subtitle condition. It was shortened from its original runtime of 17 min to 12 min 14 seconds using Veed movie editing software (https://www.veed.io), removing the beginning and end credits and plot irrelevant frames, but keeping the overall plot intact. Subtitles were manually added to the bottom of the relevant scenes by the experimenter using Veed. Subtitles in the Scrambled Subtitle condition were pseudorandomly scrambled as whole lines by the experimenter to approximately match the length of the Intact Subtitle condition for each scene while disrupting the coherence of the subtitles with the ongoing visual narrative. We chose to keep the lines intact in the subtitles (i.e., we scrambled sentences rather than words) to ensure that differences in brain activation were not solely due to reading processes.

Stimuli were presented using PsychoPy version 2021.2.3 (Peirce et al., 2019)on a high-resolution 40” MRI monitor (InroomViewingDevice, Nordic Neuro Lab) placed at the front of the scanning bed, and participants viewed the screen through a mirror on the headcoil. We chose to place the monitor at the front of the room to prevent the stimulus from being obstructed by the scanner bore. Audio was presented using an Avotec Silent Scan audio system via noise attenuating MRIaudio (https://mriaudio.com) in-ear headphones. The onset of the stimulus was synchronized with the start of the fMRI scan. To ensure that participants were actively attending to each stimulus, participants were informed that they would be completing a short comprehension questionnaire immediately following the presentation of each film condition. After each version of the film, participants were presented with five multiple choice questions about the movie and were required to make a verbal response as to which item they chose. This response was recorded by the experimenter via keypress before advancing to the next question.

2.3. MRI acquisition

Whole brain imaging was performed using a 3T Skyra Siemens Magnetom MRI system (Siemens Medical Systems, Erlangen, Germany) at the U3T MRI Centre on the University of Lethbridge campus using a standard 32-channel headcoil. A T1-weighted anatomical image (TR = 2300 ms, TE = 2.26 ms, flip angle = 8°, voxel size = 1 mm isotropic, FOV = 256 mm²) using a 3D magnetization-prepared rapid acquisition with gradient echo (MPRAGE) pulse sequence was collected at the start of scanning. During movie viewing, a T2*-weighted functional scan (TR = 1500 ms, TE = 34 ms, flip angle = 90°, voxel size = 2.4 mm isotropic, FOV = 204 mm², multiband factor = 4) was collected using a gradient-echo echo-planar imaging sequence. Each movie run consisted of 497 volumes.

2.4. MRI preprocessing

Preprocessing of neuroimaging data took place in fMRIPrep 21.0.1 (see Supplementary Materials B for full preprocessing details; Esteban et al., 2019). Briefly, fMRIprep preprocessing of T1-weighted images included intensity normalization, skull-stripping, brain tissue segmentation, and volume-based spatial normalization to the ICBN 152 Nonlinear Asymmetrical template version 2009c standard space (Fonov et al., 2009) via nonlinear registration. Preprocessing of the functional scans included skull stripping, motion correction, slice timing correction, affine registration to the T1-weighted image, and resampling into standard space.

The preprocessed data from fMRIprep was imported into CONN (Whitfield-Gabrieli and Nieto-Castanon, 2012) release 22.a (Nieto-Castanon and Whitfield-Gabrieli, 2022) and SPM release 12.7771 (Penny et al., 2011) for further preprocessing and analysis. Functional data were denoised using a standard denoising pipeline, including the regression of potential confounding effects characterized by white matter timeseries (5 CompCor noise components), CSF timeseries (5 CompCor noise components, motion linear and quadratic parameters and their first order derivatives [24 factors], outlier scans, session and task effects and their first order derivatives [4 factors], and linear trends [2 factors]) within each functional run, followed by bandpass filtering of the BOLD timeseries between .008 Hz and .09 Hz. Importantly, task effects were modeled as variables of non-interest during denoising to reduce the contribution of task-evoked co-activation to the functional connectivity estimates (see Cole et al., 2019). This increases confidence that the observed differences reflect changes in connectivity profiles rather than simple co-activation differences between conditions, although it does not eliminate all possible condition-related differences in processes such as eye movements, error monitoring, or working memory demand.

2.5. Functional connectivity analysis

In order to detect brain regions that differed in functional connectivity between the Intact- and Scrambled-Subtitle conditions, the following steps were implemented: 1) fc-MVPA of pairwise connections between all measured voxels; 2) second-level between-condition analyses testing for differences in functional connectivity between the conditions using an F test; and 3) post-hoc seed-based correlation analysis (SBC) on each of the identified regions of interest (ROIs) from the second-level results.

• (1)First-level analysis: Functional connectivity multivariate pattern analyses (fc-MVPA; Nieto-Castanon, 2022) were performed to estimate the first ten eigenpatterns characterizing the principal axes of heterogeneity in functional connectivity across participants and conditions. The first four eigenpatterns were then chosen for analysis to follow the convention of one component per every ten subjects (Nieto-Castanon, 2022; Vittinghoff and McCulloch, 2007). The four components implies that the 4-dimensional representation of spatial pattern of connectivity of each voxel with all other brain voxels. Eigenpatterns and eigenpattern-scores were computed separately for each individual seed voxel and the rest of the brain (a matrix with one row per target voxel, and one column per participant and condition). Individual functional connectivity values were computed from the matrices of bivariate correlation coefficients between the BOLD timeseries from each pair of voxels, estimated using a singular value decomposition of the z-score normalized BOLD signal (subject-level SVD) with 64 components separately for each participant and condition.
• (2)Second-level analysis: Group-level analyses were performed using a General Linear Model (GLM). For each individual voxel, a separate GLM was estimated, with first-level connectivity measures at this voxel as dependent variables (one independent sample per participant and one measurement per task or experimental condition), and condition as an independent variable. Voxel-level hypotheses were evaluated using multivariate parametric statistics with random-effects across subjects and sample covariance estimation across multiple measurements. Inferences were performed at the level of individual clusters (groups of contiguous voxels). Cluster-level inferences were based on nonparametric statistics from randomization/permutation analyses (Bullmore et al., 1999; Nieto-Castanon, 2020), with 1000 residual-randomization iterations. Results were thresholded using a cluster-forming p < 0.01 voxel-level threshold and a familywise corrected p-FWE <.05 cluster-mass threshold (Chumbley et al., 2010).
• (3)Post-hoc seed-based correlation analysis on the identified regions of interest. SBC were estimated to characterize the spatial pattern of functional connectivity during the Intact vs. Scrambled Subtitle contrast with each seed area, whereby the seed regions included the ROIs identified by the group level analysis in step 2. Functional connectivity was represented by Fisher-transformed bivariate correlation coefficients from a weighted GLM, estimated separately for each seed area and target voxel, modeling the association between their BOLD signal timeseries. Group-level analyses were performed using a GLM. For each individual voxel, a separate GLM was estimated, with first-level connectivity measures at this voxel as dependent variables. Voxel-level hypotheses were evaluated using multivariate parametric statistics with random-effects across subjects and sample covariance estimation across multiple measurements. Cluster level inferences were based on nonparametric statistics from randomization/permutation analyses (Bullmore et al., 1999; Nieto-Castanon, 2020), with 1000 residual-randomization iterations. Results were thresholded using a combination of a cluster-forming p < 0.01 voxel-level threshold and an adjusted cluster-mass corrected p-FWE < .05/5 = .01. To further understand the nature of these effects in terms of coupling and anticorrelations in each condition, we conducted SBC analyses for each fc-MVPA ROI for the Intact Subtitle and Scrambled Subtitle conditions separately and used significant voxels from the SBC analyses from the Intact vs. Scrambled Subtitle contrast as masks.

3. Results

3.1. Movie questionnaire

Behavioral results from the paired-samples t-test examining differences in accuracy on the movie comprehension questionnaire showed the participants scored significantly higher for the Intact Subtitle condition (M = 3.95, SD = .94) than the Scrambled Subtitle condition (M = 3.19, SD = .99; t(41) = 3.20, p = 0.003). For participants who viewed the Intact Subtitle film first and the Scrambled Subtitle film second, participants showed significantly higher comprehension scores in the Intact Subtitle condition (M = 4.24, SD = .83) than the Scrambled Subtitle condition (M = 2.81, SD = .79; t(20) = 4.26, p < 0.001). Notably, we did not find significant differences in comprehension scores between the Intact Subtitle condition (M = 3.67, SD = .97) and the Scrambled Subtitle condition (M = 3.57, SD = .75; t(20) = .35, p = 0.733) for the participants who viewed the Scrambled Subtitle film before the Intact Subtitle film.

3.2. Functional connectivity multivariate pattern analysis

Results from the fc-MVPA group-level GLM identified 5 significant clusters of differential functional connectivity between the Intact and Scrambled Subtitle conditions using a cluster-forming p < 0.01 voxel-level threshold and a familywise corrected p-FWE <.05 cluster-mass threshold using nonparametric statistics with 1000 random permutations (see Fig. 1 and Table 1). These clusters included the bilateral vmPFC (Yeo 17 network: Default A), left middle/inferior frontal gyri (MFG/IFG; Yeo 17 network: Control A, extending into Default B and Control B), right angular/supramarginal gyri (AG/SMG; Yeo 17 network: Control B), bilateral posterior cingulate gyrus (PCC; Yeo 17 network: Default A, extending into Control C and Somatomotor A), and an additional cluster in the left middle frontal gyrus (Dorsal Attention B, extending into Salience/Ventral Attention A and Control A).

Fig. 1.

Results from fc-MVPA omnibus analysis

Note. Fc-MVPA identified five significant clusters with significant differences in connectivity profiles between the coherent and incoherent narrative conditions, which included the bilateral vmPFC, bilateral PCC, left MFG/IFG, left MFG/precentral gyrus, and right angular/supramarginal gyrus. Inset images show the unthresholded whole-brain difference in functional connectivity between each significant cluster and all other voxels in the brain, included for descriptive visualization of the distributed connectivity patterns contributing to the multivariate fc-MVPA effect rather than as independent inferential results. Intact > Scrambled is shown in red, and Scrambled > Intact is shown in blue.

Table 1.

Results from fc-MVPA omnibus analysis.

Cluster Location	Max Voxel Coordinates (x, y, z)			Cluster size	F-value	Mass p-FWE
Bilateral vmPFC	06	44	06	184	13.52	.009
L IFG pars opercularis	−36	14	26	130	8.23	.011
R angular gyrus	60	−50	38	93	14.74	.017
Bilateral PCC	00	−26	46	78	14.73	.019
L MFG/precentral gyrus	−40	00	44	59	10.91	.041

Note. Coordinates are in MNI space.

To evaluate whether order could account for the observed fc-MVPA results, we conducted three supplementary fc-MVPA analyses (see Supplementary Materials A). First, we repeated the Intact > Scrambled omnibus fc-MVPA separately within each counterbalancing group (n = 21 per group). When participants viewed the Intact version first, fc-MVPA identified 14 significant clusters (mass pFWE < .05), including the bilateral MFG/IFG, bilateral vmPFC, bilateral PCC, and multiple temporoparietal clusters involving angular gyrus/supramarginal regions, as well as additional occipital and cerebellar clusters (Supplementary Fig. 1; Supplementary Table 1). When participants viewed the Scrambled version first, fc-MVPA identified 4 significant clusters (mass pFWE < .05), including the precuneus, right angular gyrus, right MFG, and a right occipital/LOC cluster (Supplementary Fig. 2; Supplementary Table 2). Importantly, these order-stratified analyses did not show a reversal of the condition effect; instead, they indicated that the coherence-related connectivity differences are detectable in both orders, with differences in the spatial extent of significant clusters across sequences.

Second, to further reduce any influence of second-run factors (e.g., familiarity, fatigue), we analyzed first presentations only, comparing Intact vs Scrambled when each was viewed first. This analysis identified two clusters in the left and right IFG/MFG (voxel p < 0.01 uncorrected; cluster pFWE < .05, cluster-size corrected; parametric; Supplementary Fig. 3; Supplementary Table 3). Together, these supplementary analyses argue against the interpretation that the primary findings are solely driven by order effects. Rather, they suggest that order may modulate the magnitude and detectability of the coherence effect (consistent with expected carryover in repeated viewing), while the directionally consistent Intact > Scrambled connectivity-profile differences persist across order-stratified and first-presentation analyses.

3.3. Post-hoc seed-based correlation analysis on the identified regions of interest

Post-hoc SBC analyses were completed for the five ROIs identified from fc-MVPA (i.e., bilateral vmPFC, bilateral PCC, right angular/supramarginal gyrus, left MFG/precentral gyrus, and left MFG/IFG). Our findings are summarized below.

3.3.1. VMPFC

The SBC analysis comparing the Intact Subtitle condition to the Scrambled Subtitle condition using the vmPFC ROI as a seed for whole brain connectivity identified 12 regions that showed significant differences in connectivity. These included the bilateral middle and superior temporal gyri, crus II of the cerebellum, precuneus, IFG, and PCC, the left MFG/precentral gyrus, LOC/dorsal precuneus, and MFG, and the right supramarginal gyrus (SMG). The vmPFC showed greater connectivity for the Intact Subtitle condition than the Scrambled Subtitle condition in bilateral middle and superior temporal gyrus, precuneus, crus II of the cerebellum, orbitofrontal IFG, and the left MFG/precentral gyrus (see Fig. 2 and Table 2). The vmPFC showed greater connectivity for the Scrambled Subtitle vs. Intact Subtitle condition in the left MFG, right angular/supramarginal gyrus, left LOC/precuneus, and bilateral PCC.

Fig. 2.

Results from seed-based correlation analysis

Note. ROIs with significant differences in connectivity between the Intact and Scrambled Subtitle from the fc-MVPA in the bilateral vmPFC, bilateral PCC, right angular/supramarginal gyrus, left MFG/IFG, and left MFG/precentral gyrus were used as seed regions for a seed-to-voxel connectivity analysis. Seed regions are shown on the left side of the panel, and significant regions of connectivity are shown on the left side of the panel. In the left panel, regions shown in red showed greater connectivity with the seed region for the Intact Subtitle condition than the Scrambled Subtitle condition, whereas regions in blue show greater connectivity for the Scrambled Subtitle than the Intact Subtitle condition. Results were thresholded using a combination of a cluster-forming p < 0.01 voxel-level threshold and an adjusted cluster-mass corrected p-FWE < .05/5 = .01.

Table 2.

Results from seed-based correlation analysis.

Seed Region	Contrast	Cluster Location	Max Voxel Coordinates (x,y,z)			Cluster Size	t-value	Mass p-FWE
vmPFC	Intact > Scrambled	L pMTG/STG	−52	−14	−10	1692	7.04	<.001
		R pMTG/STG	52	−18	−06	1126	7.64	<.001
		R Crus II	22	−76	−38	368	5.77	<.001
		Bilateral Precuneus	−02	−54	38	255	5.98	.002
		L MFG/Precentral	−48	−02	54	233	6.29	.002
		L Crus II	−24	−74	−40	229	6.13	.002
		L Frontal Orbital	−42	32	−04	159	5.81	.003
		R IFG	56	28	06	113	4.98	.009
	Scrambled > Intact	Bilateral PCC	00	−34	44	251	−6.11	.002
		L LOC/Precuneus	−12	−72	50	185	−4.62	.002
		R SMG	62	−36	44	129	−4.29	.005
		L IFG	−42	34	14	123	−4.49	.004
PCC	Intact > Scrambled	L pSTG	−60	−24	−02	1423	6.18	<.001
		R pSTG	48	−34	00	889	5.92	<.001
		L IFG/MFG	−46	18	24	837	6.02	<.001
		R IFG/MFG	48	18	26	383	6.89	<.001
		Bilateral PCC	−04	−50	38	278	5.90	<.001
		R Crus II	26	−76	−38	181	5.39	<.001
		L SMA (SFG)	−04	04	62	177	4.52	<.001
		R Sup LOC	34	−68	48	92	4.69	.014
		R Crus II/VIIb	34	−72	−52	61	4.61	.065
	Scrambled > Intact	Bilateral vmPFC	00	44	−04	187	−5.52	<.001
		L IFG	−36	32	14	62	−4.56	.057
Right Angular Gyrus	Intact > Scrambled	L pMTG/STG	−54	−34	00	718	6.15	<.001
		L Angular Gyrus	−54	−58	30	323	5.45	.001
		R Crus II	16	−82	−46	160	5.50	.004
		R Temporal Pole	60	06	−14	148	4.94	.004
		L Precuneus	−02	−52	38	147	5.03	.004
		L MFG/Precentral	−42	06	50	144	6.64	.003
	Scrambled > Intact	R Occipital pole	22	−92	30	286	−5.29	.001
		L Precuneus	−12	−48	58	158	−4.69	.004
		R Precuneus	04	−42	58	110	−5.01	.009
Left MFG/IFG	Intact > Scrambled	Bilateral PCC/precuneus	00	−34	44	351	6.00	<.001
		R vmPFC	10	34	00	177	5.14	.001
		L LOC	−12	−80	50	155	5.07	.001
		R Lingual/Parahippocampal	32	−38	−12	154	5.96	.001
		R SMG	64	−20	42	116	3.93	.009
		Bilateral vmPFC	04	34	22	112	5.37	.004
	Scrambled > Intact	L pSTG	−62	−34	06	998	−7.00	<.001
		R pMTG/STG	50	−26	−04	146	−4.59	.002
		R Crus II	20	−84	−34	141	−6.31	.001
		Bilateral precuneus	−02	−54	42	125	−5.44	.001
		L Crus II	−16	−76	−36	104	−4.38	.007
Left MFG/precentral	Intact > Scrambled	Bilateral vmPFC	00	40	−02	337	5.14	<.001
		R Angular Gyrus	58	−52	46	249	4.97	<.001
		R vmPFC	14	34	20	202	5.61	<.001
		Bilateral PCC	04	−28	38	136	5.49	.003
	Scrambled > Intact	L pSTG	−54	−30	00	184	−5.35	.001

Note. Coordinates are in MNI space.

To further understand the nature of these effects in terms of coupling and anticorrelations in each condition, we conducted SBC analyses for each fc-MVPA ROI for the Intact Subtitle and Scrambled Subtitle conditions separately and used significant voxels from the SBC analyses from the Intact vs. Scrambled Subtitle contrast as masks. Results from this analysis showed that, in the Intact Subtitle condition there was coupling (i.e., positive correlations) between the vmPFC and the bilateral: MTG/STG, orbitofrontal IFG, precuneus, and PCC, whereas there were anticorrelations between vmPFC and the: left MFG/precentral gyrus, left MFG, left LOC/precuneus, and right supramarginal gyrus (see Fig. 3). In contrast, in the Scrambled Subtitle condition, there were anticorrelations between the vmPFC and the: bilateral MTG/STG, bilateral orbitofrontal IFG, and left MFG/precentral gyrus, and coupling between vmPFC and the: right supramarginal gyrus, bilateral precuneus, left LOC/dorsal precuneus, bilateral PCC, and left MFG.

Fig. 3.

Seed-to-voxel correlation results for the Intact and Scrambled Subtitle conditions masked with the significant voxels from the Intact vs. Scrambled Subtitle contrast

Note. Seed-based correlation for the Intact and Scrambled Subtitle conditions masked with significant voxels from the Intact vs. Scrambled Subtitle contrast. Seed regions are shown on the left side of the panel. Areas shown in red represent positive correlations, whereas areas in blue represent negative correlations.

3.3.2. Bilateral PCC

The SBC analysis using the bilateral PCC ROI as a seed for whole brain connectivity for the Intact Subtitle vs. Scrambled Subtitle conditions identified 11 regions with significantly different connectivity patterns. These included the bilateral pSTG, bilateral IFG/MFG, bilateral PCC, bilateral vmPFC, right crus II of the cerebellum, left superior frontal gyrus, right superior LOC, and right crus II/VIIb of the cerebellum. The PCC showed greater connectivity for the Intact subtitle condition than the Scrambled subtitle condition in all regions except regions in the left MFG and the bilateral vmPFC, which showed greater connectivity in the Scrambled subtitle condition (see Fig. 2 and Table 2). When examining the direction of correlations for each contrast, in the Intact Subtitle condition there was coupling (i.e., positive correlations) between the PCC and the bilateral precuneus, vmPFC, and portions of the IFG and MFG, whereas there were anticorrelations between the PCC and all other regions (see Fig. 3). The Scrambled Subtitle condition showed a similar pattern of result to the Intact Subtitle condition, with the exception of the bilateral precuneus, which showed anticorrelations with the PCC.

3.3.3. Right AG/SMG

The SBC analysis using the AG/SMG ROI as a seed for whole brain connectivity for the Intact Subtitle vs. Scrambled Subtitle conditions identified 9 regions with significantly different connectivity patterns. These included the left MTG/STG, left AG, right occipital pole, right crus II of the cerebellum, left precuneus (three clusters), right temporal pole, and left MFG/precentral gyrus. The right AG/SMG showed greater connectivity for the Intact subtitle condition than the Scrambled subtitle condition in the left MTG/STG, left AG, left precuneus, right temporal pole, and left MFG/precentral gyrus (see Fig. 2 and Table 2). The right AG/SMG showed greater connectivity for the Scrambled Subtitle vs. Intact Subtitle condition in the right occipital pole and left precuneus. When examining the direction of correlations for each contrast, in the Intact Subtitle condition there was coupling (i.e., positive correlations) between the right AG/SMG and the left MTG/STG, left AG, right temporal pole, left precuneus, and left MFG/precentral gyrus, whereas there were anticorrelations between the bilateral precuneus and right LOC (see Fig. 3). In the Scrambled Subtitle condition, there was coupling (i.e., positive correlations) between the right AG/SMG, and the MTG, and portions of the bilateral precuneus, whereas there were anticorrelations between the left MTG/STG, left AG, right LOC, right temporal pole, and left precentral gyrus.

3.3.4. Left MFG/IFG

The SBC analysis using the left MFG/IFG ROI as a seed for whole brain connectivity for the Intact Subtitle vs. Scrambled Subtitle conditions identified 11 regions with significantly different connectivity patterns. These included the bilateral PCC, bilateral vmPFC (two clusters), left LOC, right lingual/parahippocampal gyrus, bilateral MTG/STG, bilateral crus II of the cerebellum, bilateral precuneus, and right supramarginal gyrus. The left MFG/IFG showed greater connectivity for the Intact subtitle condition than the Scrambled subtitle condition in the bilateral PCC, bilateral vmPFC, bilateral crus II of the cerebellum, right lingual/parahippocampal gyrus, left LOC, and right supramarginal gyrus (see Fig. 2 and Table 2). The left MFG/IFG showed greater connectivity for the Scrambled Subtitle vs. Intact Subtitle condition in the bilateral MTG/STG and bilateral precuneus.

When examining the direction of correlations for each contrast, in the Intact Subtitle condition there was coupling (i.e., positive correlations) between the left MFG/IFG and the bilateral MTG/STG, left precuneus, and a portion of the bilateral PCC, whereas there were anticorrelations between the bilateral vmPFC, portions of the bilateral PCC, left LOC, right lingual/parahippocampal gyrus, and right angular gyrus. Similarly, in the Scrambled Subtitle condition, there was coupling (i.e., positive correlations) between the left MFG/IFG and the bilateral MTG/STG, precuneus, and a portion of the bilateral PCC, whereas there were anticorrelations between the bilateral vmPFC, portions of the bilateral PCC, left LOC, right lingual/parahippocampal gyrus, and right angular gyrus.

3.3.5. Left MFG/precentral

The SBC analysis using the left MFG/precentral ROI as a seed for whole brain connectivity for the Intact Subtitle vs. Scrambled Subtitle conditions identified 5 regions with significantly different connectivity patterns. These included the bilateral vmPFC, right angular gyrus, left pSTG, and bilateral PCC. The left MFG/precentral showed greater connectivity for the Intact subtitle condition than the Scrambled subtitle condition with all regions except the left pSTG, which showed greater connectivity in the Scrambled subtitle condition (see Fig. 2 and Table 2). When examining the direction of correlations for each contrast, in the Intact Subtitle condition there was coupling (i.e., positive correlations) between the left MFG/precentral and the left pSTG, and portions of the right angular gyrus, vmPFC, and PCC, whereas there were anticorrelations between the left MFG/precentral and portions of the vmPFC, PCC, and right angular gyrus (see Fig. 3). The Scrambled Subtitle condition showed a similar pattern of results to the Intact Subtitle condition, however there were greater anticorrelations between the left MFG/precentral and the right angular gyrus, vmPFC, and PCC.

4. Discussion

In this study, we sought to identify patterns of functional connectivity associated with coherent narrative formation using well-matched naturalistic stimuli. To do so, we used two versions of the same audiovisual foreign language film, one with subtitles in the correct order (the coherent narrative) and one with subtitles in a scrambled order (as full lines, resulting in a less coherent narrative). The stimuli were identical other than the order of the subtitles, thus controlling for visual and auditory information, and leaving the overall plot of the film intact. We hypothesized that we would see differences in functional connectivity between the coherent and incoherent versions of the film in regions of the default mode and frontoparietal networks. In line with our hypotheses, results from this study show that coherent narrative formation is related to the functional connectivity of regions within the core DMN and frontoparietal network. Our main findings are summarized below.

Results from our fc-MVPA identified five regions of the brain with connectivity patterns that were significantly different between the coherent and incoherent narrative conditions, including the bilateral vmPFC, bilateral PCC, left inferior and middle frontal gyrus, left precentral/middle frontal gyrus, and the right angular/supramarginal gyrus. This suggests that, in line with previous research (e.g., Assouline and Mendelsohn, 2023; Kauttonen et al., 2018; Ren et al., 2018), functional connectivity to core DMN structures such as the vmPFC and PCC is associated with successful formation of coherent narratives. We also found that areas in the frontoparietal control network, including the lateral prefrontal cortex and right angular gyrus, show differential connectivity patterns between the Intact and Scrambled narrative conditions. By taking a whole-brain connectivity approach, we can identify fine-grained differences in connectivity across regions without the need to define a priori ROIs. Importantly, by modeling task effects as a variable of non-interest during preprocessing (see Cole et al., 2019), it is unlikely that these results are due to task differences between the two conditions (e.g., eye movements, error processing, working memory, etc.). Together, these results suggest that fine-grained connectivity profiles to areas in the core DMN (vmPFC and PCC) and control network (right angular gyrus, left inferior and middle frontal gyrus/precentral gyrus) are important for coherent narrative formation.

4.1. Brain connectivity associated with coherent narrative formation

To further understand the differential patterns of connectivity between the Intact and Scrambled Subtitle conditions, we also performed SBC analyses using each of the significant fc-MVPA clusters as a ROI for a whole-brain seed-to-voxel connectivity analysis. For the Intact > Scrambled Subtitle contrast, the vmPFC showed significantly greater functional connectivity with the bilateral precuneus, bilateral orbital portion of the inferior frontal gyrus, left middle frontal/precentral gyrus, and bilateral lateral temporal cortex. Significantly greater connectivity for the Intact Subtitle condition between the bilateral lateral temporal cortex and left MFG/precentral gyrus and the vmPFC appears to be due to anticorrelations between these regions in the incoherent narrative condition. This heightened vmPFC-lateral temporal connectivity during the coherent narrative may reflect increased semantic integration and schema-construction processes (Baldassano et al., 2018; Simony et al., 2016). Thus, coherent narrative processing is related to heightened connectivity between the vmPFC and language and semantic networks.

The PCC showed significantly greater connectivity in the Intact > Scrambled subtitle contrast with the bilateral precuneus, bilateral IFG/MFG, left SMA, bilateral lateral temporal cortex, and right intraparietal sulcus. In contrast to the vmPFC, connectivity between the PCC and both the IFG/MFG and lateral temporal cortex (part of the language network; Bottenhorn et al., 2019; Fedorenko et al., 2024) was negative for both the Intact and Scrambled Subtitle conditions, which may reflect functional antagonism between the posterior DMN and language and semantic processing. We also found positive connectivity in the Intact Subtitle condition between the PCC and the precuneus. Further, we found increased connectivity between the orbital IFG and less anticorrelation between the PCC and the intraparietal sulcus in the intact vs. scrambled condition, suggesting that functional segregation was decreased between these regions in coherent narrative condition.

The right AG/SMG showed significantly greater connectivity in the Intact Subtitle condition with the left precuneus, left MFG/precentral gyrus, left lateral temporal cortex, and right temporal pole. These positive correlations suggest greater coupling between the right AG/SMG and the left precuneus, left MFG/precentral gyrus, left lateral temporal cortex, and right temporal pole during coherent narrative formation. The left IFG ROI showed significantly greater connectivity in the Intact Subtitle condition with the bilateral PCC, right angular gyrus, bilateral anterior cingulate cortex, and the right parahippocampal cortex. Connectivity between the left IFG and each of these regions was negative, suggesting that there was less anticorrelation between the left IFG and these areas in the Intact Subtitle condition than the Scrambled Subtitle condition. Finally, the left MFG ROI showed significantly greater connectivity with the bilateral vmPFC and right angular/supramarginal gyrus in the Intact Subtitle vs. Scrambled Subtitle contrast. This appears to be primarily due to stronger anticorrelations between these regions in the Scrambled Subtitle condition, rather than increased coupling in the Intact Subtitle condition.

In sum, coherent narrative formation may be reliant on integrated functional connectivity between default mode, lateral temporal, and frontoparietal networks. These results are in line with previous research showing between network synchronization when processing naturalistic stimuli (Aboud et al., 2019; Bailey et al., 2018), which may be indicative of coordinated information transfer between networks, particularly semantic and language information.

4.2. Brain connectivity associated with incoherent narrative formation

For the Scrambled Subtitle > Intact Subtitle contrast, the vmPFC showed significantly greater functional connectivity with the right supramarginal gyrus, left MFG, bilateral PCC, and left LOC/dorsal precuneus. Significantly greater connectivity for the Scrambled Subtitle condition between these regions appears to be due to increased coupling, suggesting that, under more difficult narrative encoding conditions, there is greater functional connectivity between the vmPFC and regions of the core DMN (i.e., PCC), frontoparietal control (i.e., the right supramarginal gyrus, left MFG) and dorsal attention (i.e., left LOC/precuneus) networks. This highlights the potential role of increased neural engagement in processing incoherent narrative information. In line with this, the PCC showed significantly greater positive connectivity in the Scrambled > Intact Subtitle contrast with the vmPFC and left MFG. It also showed stronger anticorrelations between the PCC and regions of the language and semantic networks, which may reflect shifting cognitive strategies when processing the incoherent narrative.

The right AG showed significantly greater connectivity in the Scrambled > Intact subtitle contrast with the right LOC and bilateral anterior precuneus (part of the ventral attention network). Correlations in these regions were generally positive, suggesting that there may be heightened reliance on visual and attention networks when processing the incoherent narrative. The left IFG ROI showed significantly greater connectivity in the Scrambled > Intact Subtitle condition with the right precuneus and bilateral MTG/STG. Heightened connectivity between control network regions and lateral temporal language and semantic networks may indicate compensatory processing for understanding the incoherent narrative (e.g., Whitney et al., 2011). This theory, and our results, align well with the semantic control network described by Jackson (2021), which implicates the left IFG and MTG in semantic control processing (see Fig. 4). The connectivity changes we observe (heightened connectivity in the Scrambled condition between IFG and MTG), support Jackson's (2021) and Noonan et al.'s (2013) proposal that semantic control processes should be regulating conceptual representations in a task-appropriate manner, and this should be done via interactions (i.e., task-dependent shifts in connectivity) between the left IFG and MTG (and other semantic control network regions).

Fig. 4.

Overlap between semantic control network and inferior frontal gyrus results

Note. The left MTG cluster (yellow, right) exhibits heightened connectivity with the left IFG (orange, left) in the Scrambled > Intact contrast. Both clusters overlap the semantic control network (blue) identified in a previous meta-analysis (Jackson, 2021). Note that the colours differentiate binary masks and do not reflect statistical values.

However, our results do not purely isolate the semantic control network. Given the absence of differential connectivity between the left IFG and dorsomedial PFC in our results (which is present in the Jackson's semantic control network), and involvement of default-mode vmPFC regions, it is most likely the case that our paradigm is promoting integration between the semantic control, general frontoparietal control, and narratively sensitive default mode networks. It is important to note that Jackson's (2021) meta-analysis of semantic control focused primarily on traditional trial-based task designs, which may isolate semantic control at the expense of missing out on richer narrative and emotional processing which occurs in naturalistic semantic contexts (such as our present study). Future research will be necessary to fully evaluate the involvement of semantic control during naturalistic narrative processing.

4.3. Functional dissociations in narrative processing

Interestingly, we found a functional dissociation in connectivity between core DMN regions (vmPFC, PCC, and precuneus) between the coherent and incoherent narratives. Notably, connectivity between the vmPFC and PCC was greater in the Scrambled Subtitle condition than the Intact Subtitle condition, specifically through stronger positive correlations between regions. This is in line with previous research showing that decoupling of the vmPFC from the rest of the DMN facilitates schema memory (Müller et al., 2020). Further, the PCC and precuneus showed opposing patterns of connectivity between conditions. While the PCC showed stronger connectivity with the precuneus in the Intact Subtitle condition, this stronger connectivity was due to anticorrelations between the PCC and precuneus in the Scrambled Subtitle condition that were not present in the Intact Subtitle condition. This suggests that the PCC and precuneus operate more independently during effortful narrative processing and highlights the need to examine fine-grained differences in connectivity. This finding is in line with research showing that the precuneus acts as a hub within the DMN, interacting with the DMN and control networks (Utevsky et al., 2014). Thus, this study provides novel evidence to support theories of a functional dissociation between core DMN structures, particularly the precuneus and PCC.

We also identified functional dissociations within the DMN subsystems associated with differing contextual richness. In the Intact Subtitle condition, the medial prefrontal cortex, bilateral PCC, and right angular gyrus showed stronger connectivity with other core nodes of the DMN, including the bilateral inferior frontal gyri, precuneus, dorsomedial prefrontal cortex, and inferior frontal gyrus. In line with Ren et al. (2018), we found differential connectivity patterns between regions such as the angular gyrus, precuneus, and medial prefrontal cortex, which may support continuous memory retrieval. In particular, connectivity between the angular gyrus and precuneus was greater in the coherent narrative than the incoherent narrative, suggesting that interactions between the precuneus and angular gyrus may be important for supporting ongoing narrative formation. However, the precuneus was not identified as having a significantly different functional connectivity profile between the Intact and Scrambled Subtitle conditions in our fc-MVPA, suggesting that when experimental stimuli are well matched (i.e., virtually identical visual and auditory information), the precuneus does not appear to serve as a key processing hub, rather it supports the processes of our seed regions.

Our five seed regions integrate functions from multiple networks to process and integrate complex narratives. The results support the findings of Kauttonen et al. (2018), which implicated the precuneus, AG, cingulate gyrus, and frontal gyri in the integration of narratives into memory. We see the same regions implicated in our study, and our connectivity analysis helps better characterize how the regions work together during narrative formation. As evidenced by the consistency in their connectivity profiles (see Fig. 2, Fig. 3), traditional default mode regions (vmPFC, PCC, and AG) interface with language regions (IFG and MTG) to support narrative formation. This is in line with research by Jääskeläinen et al. (2020), in which the vmPFC processed emotional components of narrative stimuli and the AG handles attentional engagement. In contrast, the language semantic control regions (IFG/MFG and MTG) interface more strongly with each other during incoherent narratives in order to resolve the semantic conflict experienced by participants, consistent with semantic control theories (Jackson, 2021; Noonan et al., 2013). Additionally, our results support the notion that the DMN consists of multiple interacting subsystems, with the PCC and precuneus potentially indexing distinct aspects of narrative context monitoring and scene construction (Utevsky et al., 2014).

4.4. Limitations and future directions

Although the aim of this study was to use stimuli that were closely matched, it is possible that some differences between the conditions may be due, in part, to differences in reading between the two conditions. Specifically, participants in the Scrambled Subtitle condition may have ignored the subtitles once it became apparent that they were not as relevant to the plot. We attempted to minimize this effect by scrambling whole sentences from the same film, which would still allow the participant to derive some meaning from the subtitles as they constructed the narrative. However, we believe that these potential effects are minimal, as we did not see differences in functional connectivity associated with reading areas in the brain, including the left inferior temporo-occipital cortices (i.e., the putative ‘visual word form area’; Price, 2012) or other visual regions. Instead, significant differences in functional connectivity were isolated to frontoparietal and DMN areas. Thus, we believe these results are primarily due to differences in ongoing narrative construction between the Intact and Scrambled Subtitle conditions, rather than reading processes.

It is also important to note that although our supplementary analyses of order argue against a simple carryover-account of the results, order effects cannot be ruled out entirely. It could be the case that the broader connectivity patterns present in the Intact-first group are reflective of a coherent narrative-specific amplification of DMN and association-network involvement. However, this alternative explanation must be interpreted cautiously, as reduced statistical power reduces the sensitivity to detect the multivariate connectivity effects that are the focus of our study. Future research could design a paradigm to specifically test whether carry-over effects are present in narrative processing studies. Despite this limitation, condition-related (Intact vs. Scrambled) connectivity profile differences were observed regardless of the order of presentation, and frontal effects persisted even when only first presentations were analyzed, supporting our theory of order-independent mechanisms driving the relationship between narrative processing and functional connectivity. In addition, while we did include a short questionnaire to ensure participant engagement with the films, we did not collect detailed information about what they recalled from each film and thus we are unable to determine exactly what was encoded for each participant. Results from the comprehension questionnaire should be interpreted with caution, as the questions were limited and do not reflect a comprehensive memory test of the film. Future research should incorporate free and/or cued recall into the experiment to probe these effects. More broadly, the functional dissociations we identified within the DMN suggest that disruptions in these connectivity patterns may underlie difficulties in constructing coherent life narratives, a hallmark of disorders such as Alzheimer's disease, schizophrenia, and depression. Further, our findings provide a foundation for future research to explore how these networks change over the lifespan and how interventions, such as cognitive training or neuromodulation, might enhance narrative comprehension in clinical populations.

5. Conclusions

Results from this study provide novel evidence supporting the relationship between the default mode and frontoparietal networks and the processing of ongoing naturalistic narratives. Using fc-MVPA, we showed that processing coherent narratives is associated with distinct functional connectivity patterns to the default mode and frontoparietal networks, particularly the vmPFC, PCC, right angular gyrus, and left precentral, middle, and frontal gyri. These findings provide compelling evidence that narrative integrity significantly alters functional connectivity in key cortical networks associated with executive function, semantic integration, and attentional control, ultimately supporting deeper engagement with the narrative content. More broadly, these results contribute to our understanding of how the brain processes complex, real-world information and may have significant implications for conditions that disrupt narrative cognition, such as aging, neurodegenerative diseases, and psychiatric disorders. Together, our findings bring us closer to understanding the networks that underlie our innate ability to aggregate, synthesize, and shape fragmented information into a coherent ongoing narrative of our everyday lives, ultimately promoting more coherent life storytelling and autobiographical continuity.

Data and code Availability statement

The data that support the finding of this study are available from the corresponding author upon reasonable request.

Funding

This work was supported by a NSERC Discovery Grant and NSERC Discovery Launch Supplement to C.E. (RGPIN-2021-03568 and DGECR-2021-00297). The funding source had no involvement in the study design; in the collection, analysis and interpretation of data; in the writing of the report; and in the decision to submit the article for publication.

CRediT authorship contribution statement

Chelsea Ekstrand: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Visualization, Writing – original draft, Writing – review & editing. Christina Haines: Conceptualization, Data curation, Methodology, Project administration, Writing – review & editing. Keva Klamer: Data curation, Project administration, Writing – review & editing. Joshua Craig: Data curation, Project administration, Writing – review & editing. KiAnna Sullivan: Data curation, Project administration, Writing – review & editing. Peter Seres: Data curation, Project administration, Writing – review & editing. Shaylyn Kress: Writing – review & editing.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:Chelsea Ekstrand reports financial support was provided by Natural Sciences and Engineering Research Council of Canada. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article.

Data availability

Data will be made available on request.