Creative experiences and brain clocks

Abstract

Creative experiences may enhance brain health, yet metrics and mechanisms remain elusive. We characterized brain health using brain clocks, which capture deviations from chronological age (i.e., accelerated or delayed brain aging). We combined M/EEG functional connectivity (N = 1,240) with machine learning support vector machines, whole-brain modeling, and Neurosynth metanalyses. From this framework, we reanalyzed previously published datasets of expert and matched non-expert participants in dance, music, visual arts, and video games, along with a pre/post-learning study (N = 232). We found delayed brain age across all domains and scalable effects (expertise>learning). The higher the level of expertise and performance, the greater the delay in brain age. Age-vulnerable brain hubs showed increased connectivity linked to creativity, particularly in areas related to expertise and creative experiences. Neurosynth analysis and computational modeling revealed plasticity-driven increases in brain efficiency and biophysical coupling, in creativity-specific delayed brain aging. Findings indicate a domain‑independent link between creativity and brain health.

Fig. 1. Data characterization, preprocessing pipeline, and data analysis.

a We included M/EEG data from diverse populations (N = 1472) across 13 countries: Canada, Chile, Argentina, Cuba, Colombia, Brazil, the United Kingdom, Ireland, Italy, Greece, Turkey, Poland, and Germany. b We used a subsample of N = 1240 participants for training the support vector machines (SVMs) using EEG. These SVMs were used to predict the participants’ brain ages across all domains. c The remaining data (N = 232) was used for out-of-sample validation, and it consisted of M/EEG datasets related to different types of creative expertise and learning. Four of these groups represent creative expertise in dance (tango), music (instrumentalists and singers), visual arts (drawing), and video games (StarCraft 2) (study 1 about expertise). Additionally, we included one group with video game learning (StarCraft 2) (study 2 about pre/post-learning). d Before training the SVMs, raw M/EEG signals were preprocessed, normalized, and transformed into the source space using the AAL brain parcellation. From the source-transformed signals, we computed the functional connectivity matrices for all participants. We used data augmentation when training the SVMs to increase the model’s robustness and accuracy. From the trained SVMs, we computed the brain age gaps (BAGs) of participants as the difference between the model predictions and their chronological ages. BAGs > 0 can be interpreted as accelerated brain aging, and BAGs <0 as delayed brain aging. Points and violin plots in the figure are schematic examples. Created in BioRender. Migeot, J. (2025) https://BioRender.com/99vpcts (EEG device and brain illustrations).

Fig. 2. Brain clock model and brain age gaps (BAGs).

a The model’s performance was assessed by computing the Pearson’s correlation and the mean absolute error (MAE) between the predicted age and the real chronological age of participants (N = 1,240 participants). Red/blue colors represent accelerated/delayed aging. b Most important (informative) brain connections for predicting age. Top connections reflect the highest absolute SVR weights, indicating their importance in age prediction. c The brain age network comprises the set of most informative connections; the thickness of the edges represents the features’ importance of connections for predicting age using SVMs. d BAGs in the expertise study (N = 196 participants), i.e., tango dancers (ΔBAGs = −5.50, [−8.17, −2.84]_95%, t(194) = −4.823, p < 0.001, D = −0.69), musicians (ΔBAGs = −5.38, [−10.21, −0.56]_95%, t(58) = −2.237, p = 0.035, D = −0.60), visual artists (ΔBAGs = −6.2, [−10.79, −1.60]_95%, t(28) = −2.761, p = 0.028, D = −1.04), and gaming (ΔBAGs = −5.38, [−10.21, −0.56]_95%, t(58) = −2.237, p = 0.035, D = −0.60). e. BAGs in the pre/post-learning study (N = 24 participants) (ΔBAGs = −3.06, [−5.27, −0.85]_95%, t(23) = −2.863, p = 0.028, D = −0.46, FDR-corrected). Points in scatter plots represent participants. Box plots show the median and the first and third quartiles; whiskers mark the minimum and maximum values, and each point represents one participant. Groups were compared using t-statistics with two-sided p-values, FDR-corrected.

Fig. 3. Creative experiences relationship with brain age gaps (BAGs).

a Correlation between BAGs and scores of expertise for tango dancers, musicians, visual artists, and gaming (r = −0.306, p = 0.003, N = 105 participants, Cohen’s f ² = 0.103). Scores and BAGs were previously transformed into z-scores. b Scheme representation of the pre/post-learning study design. EEG recordings were acquired before learning (period 0) and thereafter (period 2). The in-game performance was assessed using the average actions per minute (APM) between periods 2 and 1 (post), and periods 0 and 1 (pre). c, d. Post- and pre-learning APM (N = 20 participants) (ΔAPM = 3.83, [2.45, 5.21]_95%, t(19) = 5.804, p < 0.001, D = 0.68), and correlation between changes in APM and BAGs (r = -0.508, p = 0.022, N = 20 participants, Cohen’s f ² = 0.349). Points represent participants. Box plots show the median and the first and third quartiles; whiskers mark the minimum and maximum values, and each point represents one participant. Groups were compared using t-statistics with two-sided p-values, FDR-corrected. Created in BioRender. Migeot, J. (2025) https://BioRender.com/99vpcts (EEG device and brain illustrations).

Fig. 4. Topographic patterns of connectivity associated with creative experiences.

a The anticorrelation between nodal functional connectivity and age represents age vulnerability. Associations between age vulnerability and increased brain connectivity are driven by creative experiences. The areas that have the greatest increase in connectivity are the ones with higher Cohen’s D effect sizes. The brain of the experts’ group is the average brain across dance dancers, musicians, visual artists, and gamers. Scatter plots represent the associations in expertise (r = 0.345, p < 0.001, N = 78 brain areas, Cohen’s f ² = 0.135) and learning (r = 0.326, p < 0.001, N = 78 brain areas, Cohen’s f ² = 0.119). Points in scatter plots represent brain areas. b Neurosynth associations with brain connectivity increase in creative experiences. We reported the absolute Pearson’s correlation between brain connectivity and association maps of different cognitive processes (N = 78 brain areas). FDR-corrected p-values are shown, and the thickness of the circles represents statistical significance. The p-values were computed using the Spin test up to 10,000 permutations before performing the FDR correction.

Fig. 5. General organizational and mechanistic principles associated with brain age gaps (BAGs) in creative experiences.

a Using M/EEG functional connectivity, we characterized two efficiency-based properties of brain topology, namely integration, related to general information processing, and segregation, ascribed to specialized information processing. We then used a generative model of EEG activity to test mechanisms based on global coupling modulation. b Efficiency metrics and modeling parameters in the expertise design (tango dancers, musicians, visual artists, and gaming). We reported significant correlations between BAGs and global efficiency (r = −0.247, p < 0.001, N = 195 participants, Cohen’s f ² = 0.065), local efficiency (r = −0.479, p < 0.001, N = 195 participants, Cohen’s f ² = 0.298), and global coupling (r = −0.351, p < 0.001, N = 195 participants, Cohen’s f ² = 0.105). c Efficiency metrics and modeling parameters in the pre/post-learning design. We reported significant correlations between BAGs and local efficiency (r = −0.490, p = 0.023, N = 24 participants, Cohen’s f ² = 0.316), but not with either global efficiency (r = 0.351, p = 0.111, N = 24 participants, Cohen’s f ² = 0.141), or global coupling (r = −0.148, p = 0.492, N = 24 participants, Cohen’s f ² = 0.022). Points in scatter plots represent participants. FDR-corrected p-values.