Aims of the study and neurofunctional predictions

The present fMRI study aimed to investigate the neurocognitive processes underlying professional artists’ aesthetic and creative experiences in the visual domain, while also challenging the Embodied and the Perceptual Learning and Inhibitory Hypotheses. For this purpose, we recruited expert painters who performed two different fMRI tasks. In one task, they observed and mentally expressed aesthetic judgments about their own paintings and those by other artists (i.e., aesthetic judgment task). In the other task, they had to reconstruct what they thought and felt when they conceived some of their own works of art or, as a control task, when they visited now-familiar places for the first time (i.e., reconstruction by imagery task). Importantly, the selection of places, own and others’ artworks was individualized for each artist and fully matched for trivial differences (e.g., artistic style; see the Preliminary Phase below). We planned to match aesthetic judgments across the different classes of stimuli shown during fMRI (i.e., 20 own paintings, 20 others’ paintings, and 20 places). To this aim, aesthetic judgments were collected in a first behavioral session (Day 1, see below). This procedure ensured that any differences observed in neural responses to one’s own vs. others’ paintings or between the reconstruction by imagery of the creation of one’s own artworks vs. the visit to familiar places could not be attributed to variations in subjective liking. Own and others’ artworks were also matched for style, content, and dynamism.

Comparing artists’ responses to their own versus others’ paintings allowed us to investigate whether the authorship factor modulates the neurofunctional responses during the aesthetic judgment task through embodied simulation mechanisms or through perceptual learning and inhibitory processes. The Embodied Hypothesis provides a relatively direct prediction for the self/other contrast. People can recognize and discriminate the products of their past actions from those generated by others [23,24]. This so-called “authorship effect” may be driven by the perception of self-kinematic motor information embedded in static traces, rather than by visual familiarity alone [25,26] or episodic memory [27]. Accordingly, the motor system may resonate more strongly with signals stemming from prior self-produced actions [24,28]. From this perspective, self-generated paintings should elicit motor representations of gestures that artists previously enacted during their creation. Thus, professional artists should show stronger recruitment of brain regions involved in motor resonance while observing their own vs. others’ paintings.

By contrast, the Perceptual Learning and Inhibitory Hypothesis was originally formulated to explain aesthetic pleasure as emerging from perceptual learning, uncertainty reduction, and “insight” (i.e., Aha experience) on the generative model of the artwork (i.e., a model of the artwork’s life cycle, from its conception to its generation [13]). The general prediction from this model is an enhanced perceptual processing and motor inhibition for aesthetically preferred stimuli. However, using the same model, one can make predictions on a self/other authorship manipulation of the stimuli as the one adopted in this experiment. Other-authored artworks may retain greater epistemic uncertainty than self-authored ones, and therefore greater potential for prediction-updating. By contrast, artists have more precise predictions about their own works because they know the actions and intentions that gave rise to them. Consistent with evidence for enhanced prediction and sensory attenuation for self-generated actions [23,28–30], viewing others’ paintings should therefore engage perceptual and motor-inhibitory brain regions more strongly, whereas viewing one’s own paintings should place lower demands on perceptual learning. To summarize, what counts here is that the two theories make opposite functional anatomical predictions about the authorship effects.

The two frameworks also differ in their predictions for the reconstruction by imagery task. On the one hand, according to the Embodied Hypothesis, during the reconstruction of creative moments, professional artists should recruit brain regions involved in sensorimotor simulation to a greater extent than in the control condition. On the other hand, according to the Perceptual Learning and Inhibitory Hypothesis, inhibitory-control processes should play a relevant role in the reconstruction of creative moments. This prediction is grounded in the Creative Cognition literature, where inhibitory control has been proposed to support the suppression of dominant or conventional associations in order to generate novel solutions [14,15]. Thus, in the present study, we used the two tasks to examine whether artists’ aesthetic judgment and reconstruction of the creative moment were more strongly characterized by sensorimotor/motor-simulation processes or by perceptual-learning and inhibitory-control mechanisms.

Participants

Twenty-eight professional artists participated in the study’s first behavioral session (Day 1, see below). However, four of them did not join in the second fMRI session (Day 2) and other four were excluded due to areas of altered signal intensity revealed by the MRI, thus leaving a final group of twenty participants (10 females, 14 right-handed, 4 ambidextrous, age range 22–74 years, mean age 42.15 ± 18.81), all with high education (range 16–18 years). The sample size is consistent with previous fMRI studies on aesthetic appreciation and creativity: across the thirty-four experiments included in a recent meta-analysis on the same topics, the median sample size is twenty participants [16]. All the participants had normal or corrected to normal vision, and none reported a history of neurological or psychiatric disorder. All the artists were screened to exclude age-related neurocognitive decay by using the Mini-Mental State Examination (MMSE) [31]; no one showed pathological performance after correcting the raw scores for age and education (MMSE corrected score range 25.85–30). The experimental protocol was approved by the Ethics Committee of the University of Milano-Bicocca, where data collection took place (from May 2, 2022 to July 19, 2023), and the study was conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants.

Inevitably, our experiment did not rely on comparisons with a control group for several principled reasons:

1. The core task in this study involves artists viewing and reflecting on artworks they created themselves. This kind of self-referential engagement requires authorship, which, by definition, non-artists lack, together with a broad body of artworks usable as stimuli.
2. Asking non-artists to reflect on “their own artworks” would require fabricated or trivial creations, which would lack the motor, emotional, and aesthetic depth central to the phenomenon under study.
3. Not all questions require control groups: control groups are essential when isolating variables in general cognitive mechanisms, but this study addresses a specific expert population and experience. The aim was not to determine what all humans do, but to uncover how expertise and authorship uniquely shape neurocognitive responses in experienced artists.

Procedure

The study was divided into three phases: a preliminary phase, a first behavioral session (Day 1), and a second fMRI session (Day 2).

Preliminary phase. Participants were professional artists selected by an artist and professor of the Academy of Fine Arts of Brera (the author S.P.). They were recruited via e-mail or phone call, during which they were informed about the purposes of the study. If they consented to participate, they were asked to send forty images of their own pictorial artworks and forty pictures of places they knew. For each participant, S.P. provided a list of renowned artists with the same pictorial style, and two independent experimenters (the authors G.T. and M.A.M.) selected a “control” painting for each artwork (i.e., a one-to-one matching with the forty self-selected works of art by each artist) according to the following criteria: i) abstract/figurative; ii) living/non-living; iii) static/dynamic; iv) canvas coverage (very/little covered); v) very/little colored; vi) pictorial technique. For each artwork of each participant, a third independent researcher (the author L.M.S.) selected the best “control” painting out of the two options.

Day 1. In the first session, the artists performed two behavioral judgment tasks. In the first one, they saw their own forty and other artists’ forty paintings and had to provide liking judgments on a Visual Analogue Scale (VAS) ranging from 0 to 100. In the second task, they evaluated the grade of appreciation of the forty places using the same VAS (see S1 Fig in the Supporting Information for the trial-timeline). For both tasks, each image was presented twice to obtain a more robust evaluation. During the same session, the artists were also screened for contraindications to fMRI. For each subject and class of stimuli (i.e., own paintings, others’ paintings, and places), we calculated the mean score from the two evaluations of each image. We then selected twenty items for each category by including the ten least and ten most liked items, while also ensuring that there were no significant differences in ratings across classes of stimuli (own mean: 63.68 ± 24.53; others mean: 63.69 ± 24.57; places mean: 64 ± 26.43). The twofold aim of this procedure was (i) to ensure that any differences across classes of stimuli were not due to differences in liking, and (ii) to evaluate in a further analysis whether the artists’ brain reacts differently to paintings that they like the most compared to those they liked the least. For each artist, their own and other artists’ selected artworks were also matched for aspects related to the content (i.e., number of figurative and abstract artworks, number of paintings carrying body-related cues) and the dynamism (i.e., number of static and dynamic artworks). Moreover, although familiarity was not formally assessed, artists informally reported strong familiarity with most of the other artists’ pieces of art. The sixty selected images (twenty own paintings, twenty other artists’ paintings, and twenty places) were used in two fMRI tasks.

Day 2. In the second session, participants completed two fMRI tasks in two different runs: an aesthetic judgment task and a reconstruction by imagery task. The order of the tasks was counterbalanced between participants. Prior to the scanning session, the artists completed a brief training using images that were not selected to familiarize themselves with the tasks.

During the aesthetic judgment task, participants observed both their own and others’ artworks and were instructed to mentally formulate an aesthetic judgment. No motor response was required to avoid confounding motor-related activations associated with response preparation and execution, which could, in principle, amplify motor signals and artificially bias the data toward an embodied interpretation. However, to ensure participants remained focused and engaged in the task, and keep sufficient control on their performance, a question randomly appeared on the screen, prompting them to press a button with either their index finger or thumb depending on whether the artwork they had just seen belonged to another artist (index finger) or if it was their own (thumb; see Fig 1a for the trial timeline). The exact question was: “Was this your own artwork or an artwork by another artist?”

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Trial-timeline of the experimental tasks.

The figure illustrates the trial-time line of the a) aesthetic judgment task, and the b) reconstruction by imagery task. In the aesthetic judgment task, artists viewed 20 self-authored and 20 other-authored paintings and mentally formulated an aesthetic judgment. The main contrast of interest was self-authored versus other-authored paintings. In the reconstruction by imagery task, artists mentally reconstructed either the moment in which they conceived one of their own artworks or the moment in which they first visited a now-familiar place. This task included 20 self-authored paintings and 20 familiar places, and the main contrast of interest was reconstruction of art creation versus reconstruction of places. For each stimulus category, the 20 selected items included 10 more-liked and 10 less-liked stimuli, based on the Day 1 ratings. Red frames indicate the onset of the events of interest in the fMRI analyses. For illustrative purposes, the figure includes two paintings by the artist Stefano Pizzi.

https://doi.org/10.1371/journal.pone.0355737.g001

During the reconstruction by imagery task, artists were instructed to close their eyes and mentally evoke the moment when they conceived their own artworks, when one of their paintings appeared on the screen, or the moment when they visited the places for the first time, when an image of a place appeared (control condition). The exact instructions were: “Imagine as vividly as possible what appeared in your mind and what you felt when you i) conceived the artwork that will be presented to you or ii) visited the place that will be presented to you for the first time”. They were asked to press a button either with the index or thumb finger when the recall was vivid (see Fig 1b for the trial-timeline). We designed the control condition in order to isolate the neurofunctional underpinnings of artistic production from those more generally involved in episodic memory retrieval.

Once outside the scanner, we asked the artists to qualitatively describe how they performed the aesthetic judgment and reconstruction by imagery tasks. Then, i) they rated the imaginability of each own artwork and place (i.e., how well they imagined and evoked the moments during the reconstruction by imagery task) on a VAS, and ii) they rated their emotional involvement with each stimulus (own and others’ paintings and places) also using a VAS.

Stimuli and trial timeline

During the fMRI session, the visual and auditory stimuli were delivered through MRI-compatible devices (NordicNeuroLab InRoom Viewing Device and Audio System). The aspect ratio of the visual stimuli was preserved by adjusting the image size to a maximum dimension of 1920 x 1080 pixels. The responses were recorded through an MRI-compatible response-box (NordicNeurolab ResponseGrip) placed under the right hand of the participants.

The aesthetic judgment task comprised ten mini-blocks of six trials each (for a total duration of ~ 10 min). Each mini-block comprised four paintings (i.e., two of their own paintings and two of others), an attention-check question (4,000 ms), and an implicit baseline (i.e., black screen displayed for 12,000 ms) presented in random order. The trials started with a fixation-cross displayed on the screen (variable duration ranging from 500 to 2,000 ms). The image of an artwork was then presented (8,000 ms), followed by its scrambled version (2,000 ms, see Fig 1a). Once per mini-block, the attention-check question was presented, for a total of 10 times during the aesthetic judgment run. The question remained on screen for up to 4 s. If participants responded before the end of this interval, a black screen was presented so that the total duration of the question-related event remained fixed at 5 s.

The reconstruction by imagery task comprised ten mini-blocks of five trials each (for a total duration of ~ 14 min). Each mini-block comprised four images (i.e., two of their own paintings and two places) and an implicit baseline (i.e., black screen displayed for 12,000 ms) presented in random order. Each trial started with a fixation-cross (duration ranging from 500 to 1,500 ms), followed by the image of one of their paintings or a place (3,000 ms). Then, simultaneously, the scrambled version of the image appeared, and participants heard “close” (1,000 ms). After they had closed their eyes, the participants had a maximum of 12,000 ms to evoke what appeared in their mind and what they felt when they conceived their own artworks or when they visited places for the first time. If they took a shorter time, they were instructed to press a button with their index or thumb finger when the recall was vivid. Anyway, they had to remain with their eyes closed for the entire duration (12,000 ms), until they heard “open” (1,000 ms), i.e., the instruction to open their eyes and interrupt the mental reconstruction (see Fig 1b). This procedure allowed us to collect the same number of scans across participants, although participants took different durations to complete the trials. The software E-Prime 3 (Psychology Software Tools, Inc.) was used both for stimuli presentation and randomization.

Behavioral Analyses

All behavioral data are available in the Open Science Framework at the following address: https://osf.io/e4fdb/.

Day 1. Based on the liking judgments provided by the artists on Day 1, we selected the stimuli for the fMRI experiments (i.e., artists’ own paintings, other artists’ paintings, and places). The stimuli were matched for liking ratings across categories, as ensured by the results of a three-way (own artworks/others’ artworks/places) within-subject non-parametric ANOVA (i.e., Friedman test) performed for each artist (all p values were greater than 0.063). Additionally, for each artist, we checked that aspects related to content and dynamism were matched between the artists’ own and others’ artworks. We indeed classified each artwork based on whether i) its content was figurative or abstract, ii) it carried body-related cues or not, and ii) its configuration was dynamic or static, and we checked that the frequencies of these features did not differ significantly between own and others’ paintings, as ensured by the results of a Chi-Squared test performed for each artist (all p values were greater than 0.058).

Day 2. At the behavioral level, we wanted to check that there were no significant differences in the duration of the mental reconstruction during the fMRI task, nor in imaginability between own paintings and places to avoid entering confounds in the analysis that could explain some of the differences in the brain activations. From the behavioral analysis on duration of mental reconstruction, we excluded one subject because, although she attempted to press the button, no response was recorded from the response box. For this participant, duration was set equal to 0 for the fMRI analyses in all trials. Then, we explored whether there were differences in emotional involvement across the three classes of stimuli. All these analyses were performed using linear mixed effects models with stimuli as fixed effects. By-subjects random intercepts and slopes were included to account for variability across subjects. Data were analyzed using the statistical software Jamovi (retrieved from https://www.jamovi.org.) and the GAMLj module (retrieved from https://gamlj.github.io/). Degrees of freedom were estimated using the Satterthwaite method. Post hoc test applied the Bonferroni correction when needed, and all inferential tests of significance were based upon an α level of .05.

MRI data acquisition and analyses

Neurosynth decoding

To further decipher the functional role of the brain regions resulting from our analyses, we used the decoder function implemented on the website Neurosynth (https://neurosynth.org, [34]. It permits quantifying the functional association of a given brain map with functional/psychological terms, after interrogation of 14,371 functional MRI studies and their semantics. We conducted two decoding analyses: in one, we entered as input image the brain map resulting from the direct comparison between the appreciation of one’s own and others’ artworks (own > others) masked with the simple effect related to the appreciation of one’s own paintings. In the other one, we entered as input image the brain map resulting from the reconstruction of art creation compared with the control task (own paintings > places) masked with the simple effect related to the reconstruction of own paintings’ creation.

Behavioral results

Concerning the imaginability, we did not find significant differences between their own paintings and places (F (1,19) = 1.02, p = 0.326; Estimate: 2.46, 95%CI from −2.32 to 7.25; M_OwnPaintings = 76.2, SE_OwnPaintings = 3.05 and M_Places = 78.7, SE_Places = 2.23. See Supporting Information S2a Fig). Similarly, participants did not take different durations in recalling the creative moments compared to the ones when they visited the places (F(1,17.5) = 0.323, p = 0.577. Estimate: 0.109, 95%CI from −0.266 to 0.483; M_OwnPaintings  = 9.68, SE_OwnPaintings  = 0.387; M_Places  = 9.58, SE_Places = 0.4. See Supporting Information S2b Fig). Finally, regarding emotional involvement, we found a significant main effect of the class of stimuli (F(2,19) = 26.8, p < 0.001). Post-hoc comparisons revealed that participants were less emotionally engaged with others’ paintings than with their own paintings (Estimate: 20.5, 95%CI from 15 to 26, t(19) = 7.284, p_corr < 0.001) and places (Estimate: 19.7, 95%CI from 12.9 to 26.6, t(19) = 5.648, p_corr < 0.001; M_others = 50, SE = 3.33; M_own = 70.5, SE = 2.28; M_places = 69.8, SE = 2.26). Instead, no differences emerged between own paintings and familiar places (t(19) = 0.284, p_corr > 0.99. See Supporting Information S2c Fig).

fMRI results

The simple effects (task versus implicit baseline) and the common effects in the aesthetic judgment task and the reconstruction by imagery task are reported in the Supporting Information (see S3 Fig, S1-S3 Tables and S4 Fig, S4-S6 Tables respectively).

Aesthetic judgment task.

The direct comparison of the two appreciation conditions (for self-authored and for artworks made by others) revealed that in professional artists the judgment of their own paintings was associated with stronger recruitment of the right superior parietal lobule (SPL) and anterior intraparietal sulcus (AIP), the left lateral occipitotemporal cortex (LOTc), and the dorso-central sector of the right insula (Fig 2a and Table 1). A left superior parietal cluster also showed a similar trend (k = 395, p_uncorr = 0.019 cluster-level; local maximum at MNI −18, −60, 44; z-score = 3.8; p_uncorr = 0.00007 voxel level).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Neurofunctional results of the aesthetic judgment task (“own> others” masked by the simple effect of own artworks).

https://doi.org/10.1371/journal.pone.0355737.t001

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. FMRI results.

The figure illustrates a) the brain regions that show a greater engagement during the formulation of aesthetic judgments of one’s own compared with others’ artworks, and b) the cerebral regions that exhibit a greater involvement during the reconstruction of the creative moments compared with the control task. All the analyses were conducted at the whole-brain level, and the data are reported by applying the same statistical threshold discussed in the text (p_uncorr < .001 at the voxel level and p_FWER-corr < 0.05 at the cluster or voxel level). For illustrative purposes, we also report the left superior parietal cluster (own > others, aesthetic judgment task), and the left ventral premotor cluster (own paintings > places, reconstruction by imagery task), though they do not survive the correction for multiple comparisons. LOTc = lateral occipito-temporal cortex; SPL = superior parietal lobule; AIP = anterior intraparietal sulcus; DCI = dorso-central insula; IFG = inferior frontal gyrus; vPM = ventral premotor cortex; aMCCv = anterior-ventral midcingulate.

https://doi.org/10.1371/journal.pone.0355737.g002

Instead, the contrast in the opposite direction (“others> own”) did not show any significant effect.

We also assessed that the neurofunctional differences between the appreciation of their own paintings and paintings made by other artists were not driven by the behavioral differences in the emotional involvement, performing a control analysis with the delta of the emotional involvement (own – others) as a covariate of a one-sample t-test in which we entered the contrast “own> others” calculated at the first level of analysis. These results confirmed that differences in neural activations were not due to differences in emotional engagement (see Supporting Information S7 Table).

Moreover, we were interested in evaluating whether the brains of our artists reacted differently to artworks they liked more. When collapsing across one’s own and others’ paintings, neither the main effect of liking nor the painting-by-liking interaction effect yielded clusters that survived correction for multiple comparisons. We then examined the simple effects of liked > less-liked paintings separately for self-authored and other-authored artworks. For other-authored paintings, no liking-related anatomical effect was present. On the other hand, for self-authored paintings, the contrast own liked > own less-liked revealed a trend in the ventral anterior cingulate cortex (vACC, k = 306, p_uncorr = 0.044 cluster-level. Local maxima: MNI −4, 28, −2; z-score = 4.28; p_uncorr = 0.000000009 at the voxel-level, and MNI 0, 38, −8; z-score = 3.39; p_uncorr = 0.000347 voxel-level; see S5 Fig), within a medial frontal region compatible with the so-called field A1, previously associated with the experience of beauty [35–37]. This activation did not survive whole-brain FWER correction. However, given the functional relevance of field A1 for aesthetic experience of beauty [35–37], we applied a small volume correction (SVC) using an independently defined spherical region of interest centered on the coordinates reported by Ishizu and Zeki [35] for field A1 (MNI: −3, 41, −8; radius: 8 mm). Within this search volume, the effect survived FWER correction (cluster size: k = 34 voxels; p_FWER-SVC = 0.009 at the cluster level; local maximum: MNI: 0, 36, −4; z-score = 3.53; p_FWER-SVC = 0.005 at the voxel level. Additional local maxima within the same cluster were observed at MNI: −4, 34, −6; z-score = 3.46; p_FWER-SVC = 0.006 at the voxel level and MNI: 0, 38, −8; z-score = 3.39; p_FWER-SVC = 0.007 at the voxel level; see Fig 3a). We also found a trend within the left nucleus accumbens (k = 95, p_uncorr = 0.239 cluster-level. Local maximum: MNI −10, 4, −10; z-score = 3.92; p_uncorr = 0.00004 voxel-level; see S5 Fig), but this effect did not survive correction for multiple comparisons at the whole brain level, nor did it belong to the volume of interest identified by Ishizu and Zeki (2011) and is therefore reported only as an exploratory observation.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Relationship between Day 1 liking ratings and brain activity during the aesthetic judgment task and the reconstruction by imagery task.

The figure illustrates (a) the anterior cingulate cluster that survived small volume correction (SVC) for the contrast own liked > own less-liked (aesthetic judgment task), reflecting differential engagement during the observation of self-authored artworks that were more appreciated compared with self-authored artworks that were less appreciated. This cluster corresponds to the portion of the broader uncorrected effect (see S5 Fig) that survived FWER correction after restricting the search volume to the independently defined field A1, a region of the medial frontal cortex related to aesthetic experiences of beauty. The SVC was applied using a spherical region of interest centered on the Montreal Neurological Institute (MNI) coordinates reported for field A1 (−3, 41, −8), with a radius of 8 mm. The figure also shows (b.) the right dorsal premotor/precentral region whose activity positively covaried with the mean Day 1 liking ratings for own artworks during the reconstruction by imagery task. Individual contrast images for the comparison between reconstruction of art creation and reconstruction of familiar places (own paintings > places) were entered into a second-level regression analysis, with each artist’s mean Day 1 liking rating for own paintings included as a covariate. This analysis was conducted at the whole-brain level, applying the same statistical threshold reported in the text (p_FWER < 0.05 at the voxel level).

https://doi.org/10.1371/journal.pone.0355737.g003

Reconstruction by imagery task.

Mental reconstruction of the creative moments compared with the control task was associated with greater recruitment of the left SPL and AIP, the left LOTc, the pars triangularis of the left inferior frontal gyrus, and the ventral portion of the anterior midcingulate (aMCC) (Fig 2b and Table 2). A cluster within the left ventral premotor cortex also showed a similar trend, though it did not survive the correction for multiple comparisons (k = 244, p_uncorr = 0.045 cluster-level; local maximum at MNI −46, 8, 22; z-score = 3.82; p_uncorr = 0.00006 at voxel-level). As could be anticipated from the literature on spatial and topographic cognition, the comparison of recall of places versus the recall of one’s own paintings revealed greater activation in specific regions, including the para-hippocampus, a region recently labelled as the para-hippocampal place area [38] (see S6 Fig and S8 Table in the Supporting Information). Comparing the present findings with the ones of the aesthetic judgment task in terms of macroscopic distribution of the brain activations, we remark that both reconstructing the creation of one’s own paintings and the appreciation of the same artworks are associated with the recruitment of the left LOTc and the left SPL.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Neurofunctional results of the reconstruction by imagery task (“own paintings> places” masked by the simple effect of own artworks).

https://doi.org/10.1371/journal.pone.0355737.t002

Moreover, with regard to inter-individual differences, mean Day 1 liking ratings for self-authored artworks positively covaried with activity in the right dorsal premotor/precentral cortex during creative reconstruction (own paintings > places; MNI: 58, 0, 50; z-score = 4.82; p_FWER = 0.023 at the voxel-level; see Fig 3b). There were also similar trends in the right anterior midcingulate (MNI:14, 22, 38; z-score = 3.7; p_uncorr = 0.0001 at the voxel-level) and in the right lingual gyrus (MNI: 22, −44, −2; z-score = 3.51; p_uncorr = 0.0002 at the voxel-level; see S5 Fig), though these effects did not survive correction for multiple comparisons at the whole brain level.

Neurosynth decoding

We used the Neurosynth.org decoder to qualify further the functional role of the brain regions resulting from the direct comparison between i) the aesthetic judgment of one’s own and others’ artworks, and ii) the reconstruction of art creation compared with the control task. Concerning the appreciation of own works of art, besides self-referential anatomical terms like “intraparietal sulcus” (r = 0.303) and “posterior parietal” (r = 0.293), the Neurosynth decoder identified associations with more cognitive-related terms like “attentional” (r = 0.238), “spatial” (r = 0.213), “calculation” (r = 0.191), “working memory” (r = 0.126), “visual” (r = 0.126), and also “manipulation” (r = 0.1). Concerning the reconstruction of art creation, besides self-referential anatomical terms like “intraparietal” (r = 0.266), the Neurosynth decoder identified associations with cognitive terms like “calculation” (r = 0.228), “working memory” (r = 0.194), and also “grasping” (r = 0.146).

Aesthetic appreciation of own and others’ artistic production

Before discussing the results in detail, it is worth recalling that several important low-level and stimulus-related factors were controlled, including the subject – abstract paintings vs. figurative carrying body-related cues – as well as the dynamism or the subjective liking. Accordingly, these factors should not explain the effects discussed below. On the other hand, the self/other distinction was a crucial factor of interest addressed in our study yet entangled with the realm of aesthetic appreciation, rather than based on a generic distinction, to represent how artists make aesthetic judgements on their own work. One could argue that the distinction between self versus other could be confounded with factors like emotional attachment, or autobiographical memory in our artists: however, as far as the aesthetic evaluation was concerned, these hypotheses were made more unlikely by the results of the fMRI investigation (see the discussion below). Having made these preliminary clarifications, we believe that, for aesthetic appreciation, our results are in line with the Embodied Hypothesis. The appreciation of one’s own compared with others’ paintings was indeed associated with greater recruitment of brain regions belonging to the so-called action observation network (AON) [39,40] most of which have documented mirror-like properties [40]. These regions included the right superior parietal lobule (SPL), with a strong trend in the left-sided region (which, however, did not survive correction for multiple comparisons), the right anterior intraparietal sulcus (AIP), the right dorso-central insula, and the left lateral occipitotemporal cortex (LOTc). These areas are systematically recruited in tasks requiring the execution, observation, or imagination of arm/hand actions [39–41] and tool manipulation [42,43]. Previous fMRI studies reported the involvement of parietal regions (both AIP and SPL) and LOTc both during the actual manipulation of everyday objects to perform different goal-directed actions [42], as well as during the mental simulation of their use [43]. Additionally, a direct neuron-recording study has also provided evidence of the involvement of the posterior parietal cortex (i.e., AIP and SPL) in reaching behaviors [44]. The results of the analysis with Neurosynth further support our interpretations. As much as no brain area, with perhaps very few exceptions, is associated with a single function, it is worth emphasizing that the brain map from the aesthetic judgment task correlated with the term “manipulation”. Other correlating terms were “attentional” and “spatial” and “working memory”. Our results also revealed the recruitment of the dorso-central sector of the right insula. Although the insular cortex fulfills different functional roles (i.e., sensorimotor, socio-emotional, olfactory-gustatory, and cognitive), meta-analytic evidence indicates that its dorso-central portion is reliably recruited by sensorimotor tasks [45]. This region is anatomically connected with the main nodes of the parieto-frontal circuit involved in reaching/grasping upper limbs movements execution and observation (e.g., with the AIP and the ventral premotor cortex) [46]. Interestingly, the dorso-central sector of the right insula has a key role in expressing, perceiving, and imagining the affective component of arm gestures [41]. Regardless of their goal, actions can be endowed with different “vitality forms” [41,47]. Vitality forms could be defined as the “prosody” of movements, that is, how the gestures are performed (i.e., the style), reflecting the affective state of the agent. The individuals’ internal states modulate the kinematics of actions: the same action can thus be performed in different ways (e.g., gently, vigorously, or neutral) according to the emotional states [41].

These results suggest that, in professional artists, observing their paintings automatically triggers the internal mental representation of the gestures they performed to realize the graphic traces on the canvas. These embodied simulation mechanisms may also extend to the vitality forms and affective states that accompanied and modulated those movements, leaving their imprint in the lines, shapes, and colors of the artwork [48].

Motor simulation processes might also be elicited while appreciating others’ artworks. Indeed, we found the recruitment of bilateral premotor cortices during the observation of both own and other paintings. Yet, embodied mechanisms seem more strongly involved in one’s own works of art. Indeed, own artworks are characterized by a “motor familiarity”, that is, by higher availability and access to the more vivid motor representations of the gestures performed while creating one’s own piece of art. Importantly, these results were independent and therefore not attributable to the different emotional involvement between one’s own and others’ paintings, as confirmed in our control analysis (S7 Table in the Supporting Information). In addition, these observations are also independent of the specific content or stylistic approach of the works of art observed, one’s own or others, as own and others’ artworks were matched for content.

Similar findings are documented by experiments derived from performing arts. Expert dancers show stronger recruitment of mirror areas (e.g., premotor cortex, intraparietal sulcus, and SPL) when they observe dance movements that belong to their motor repertoire compared with kinematically comparable dance movements that do not [49]. While expert ballet dancers greatly activate the fronto-parietal mirror system when they see ballet moves compared with capoeira ones, the opposite effect is exhibited by expert capoeira dancers [49], suggesting that these brain responses depend on motor familiarity (i.e., the previous motor experience of performing those actions) and not on visual familiarity [50]. In the same vein and more specifically, Calvo-Merino and colleagues [50] performed a fMRI study on male and female classical ballet dancers, showing that the mirror system was more engaged when dancers viewed ballet movements they could represent motorically (i.e., the movements specific to their gender) compared with movements for which they had visual but not motor familiarity (i.e., the ballet movements specific to the other gender, that were visually familiar since female and male dancers train and exhibit together).

Finally, we explored whether the neural activity was modulated by the extent of appreciation of artists’ own and others’ paintings. We did not find a main effect of liking. However, we identified a trend within the ventral anterior cingulate cortex and the left nucleus accumbens: these regions showed differential activations, based on liking judgments, only when artists observed their artworks, with deactivation when they evaluated their own less-liked paintings. Although these effects did not survive FWER correction at the whole-brain level, the activation within the anterior cingulate cortex survived small volume correction within the so-called field A1, a medial frontal region previously associated with the experience of beauty across the visual and musical domains [35–37]. We also observed a cluster in the left nucleus accumbens, a key node of the mesolimbic reward circuit, which is reliably and directly linked to the experience of pleasure, including in the music domain [51–54]. However, this latter effect did not survive any correction for multiple comparisons and should therefore be considered as exploratory. By contrast, no liking-related response was present during the observation of others’ artworks.

Art-making as explored by the reconstruction by imagery task

When participants were asked to mentally reconstruct their past artworks – recalling what appeared in their minds and how they felt during the creative process – a distinct neurofunctional pattern emerged, consistent with the Embodied Hypothesis. Brain activations included the left superior parietal lobule (SPL), anterior intraparietal sulcus (AIP), lateral occipitotemporal cortex (LOTc), the pars triangularis of the inferior frontal gyrus (IFG), and the ventral portion of the anterior midcingulate cortex (aMCC). The same trend was also observed in the left ventral premotor cortex.

As previously noted, the SPL, AIP, and LOTc are regions with mirror-like properties involved in executing, observing, and imagining arm and hand movements [40]. The pars triangularis of the IFG is likewise engaged during tasks involving the observation or imitation of grasping actions [55,56]. Anatomically, this region connects with the SPL and AIP via both the superior longitudinal fasciculus (the dorsal stream) and the extreme capsule fasciculus (the ventral stream) [56].

Finally, the ventral aMCC also plays a central motor role [57]. In a large-scale intracortical stimulation study involving over 1,700 sites across the cingulate cortex in 329 patients, Caruana and colleagues provided a comprehensive functional map detailing the specific contributions of each cingulate subregion to behavioral, affective, and sensory functions. [57]. Their findings revealed that the stimulation of the aMCC, and particularly its ventral portion, significantly elicited goal-directed behaviors, ranging from simple movements of hands and arms to more complex integrated patterns of reaching and grasping actions performed in a smooth natural way [57]. The aMCC also shows mirror-like properties, being involved not only in action execution but also in the observation of actions, especially those expressed through different vitality forms [58]. Notably, it is anatomically connected with the dorso-central insula, forming part of a “vitality form” circuit for hand actions [58].

Supporting this interpretation, comparing the brain map generated during art creation imagery with the Neurosynth.org database revealed the association with the term “grasping”.

Taken together, our results suggest that motor simulation processes of the kind recruited during motor imagery are involved in artistic production and its recall. In support of this speculation, from the post-scanning debriefing, we learned that more than half of the artists, when reconstructing their moments of creation, visualized themselves in the act of physical creation, rather than in a moment of conceptual conception. Indeed, in addition to remembering the emotions they were experiencing (an aspect common to the re-enactment of the moments when they visited the now familiar places) and the motivations that drove them to realize each artwork or to choose specific colors, twelve out of twenty artists recalled flashes of moments in which they were spreading the color on the canvas, cutting the paper, or sewing fabrics. As some of them indicated, during the art creation process there would not be a sharp distinction between a phase of conception and another of physical implementation, and artworks may arise by implementing motor acts. Coherently, further analyses (see S1 Text in the Supporting Information) confirmed that the regions of the action observation mirror network previously described were recruited by the subgroup of painters who post-scanning declared to employ such motor imagery strategies, while this was not true for the non-motor imagery subgroup (see S7 Fig, S9 and S10 Tables).

The recruitment of sensorimotor regions in visual artists is also consistent with previous results in the musical domain. During musical improvisation, professional musicians show deactivation of prefrontal brain regions involved in executive control [59,60] and a greater involvement of premotor and parietal regions [61]. Artistic training could thus lead to lower demands for cognitive control [60,62] and higher recruitment of action simulation processes [61] when creating (and reconstructing the creation of) something new.

Our data underscore the existence of a main creative-reconstruction network that distinguishes artistic-creation retrieval from familiar-place retrieval independently of appreciation. However, besides this left-sided system, a right-sided system was modulated by individual differences in the degree of appreciation of one’s own art. Specifically, we found a positive relationship between the aesthetic value attributed to one’s own artworks and the recruitment of the right dorsal premotor cortex, accompanied by similar trends within the right anterior midcingulate and lingual cortices. These findings suggest that artists who assign higher aesthetic value to their own artworks may engage in the creative-reconstruction process by more strongly recruiting sensorimotor representations associated with the creative gesture, possibly extending to affective and visual-imagery components.

Comparison with previous results and conclusion

While the present results hint that both art-viewing and art-making in professional painters rely on embodied simulation mechanisms, different results (i.e., better aligning with the Perceptual Learning and Inhibitory Hypothesis) were found in a recent meta-analysis on the neurofunctional underpinnings of the aesthetic and creative experiences in the visual domain [16]. However, several crucial aspects differentiate the present study from previous ones. Only participants naïve to the history or practice of visual art were included in the meta-analysis, and almost all the neuroimaging studies so far applied a “perceptual approach”: artworks were presented for very brief durations [18], and particular care was taken to avoid too easily recognizable pieces of art [16]. To bypass the effect of artistic education, experimental subjects were required to evaluate a large number of paintings as beautiful/non-beautiful, independently of previous knowledge. In these instances, the results of the meta-analysis suggested that predictive and inhibitory processes may play an important role in enabling enhanced perceptual processing in “non-expert” participants observing little-known paintings [16]. However, expertise influences the emotional and cognitive evaluation of artworks [63,64]. Individuals who have been extensively exposed to artistic experiences may have acquired diverse aesthetic sensibility and developed different pathways for appreciating art, which, according to our results, strongly rely also on embodied simulation. Likewise, Alluri et al. [65] demonstrated that, while musicians process the auditory stimuli via an action-based approach during music listening, the non-musicians’ approach is perception-based.

Concerning visual art production, no previous attempt has been made to investigate the neural activities associated with “big creativity”, i.e., artistic production, in professional artists [16], but see [62]. Most fMRI studies on visual creativity investigated the neural foundations of what is known as “little creativity” and divergent thinking in naïve participants, asking subjects to find original solutions to open-ended problems under strict time constraints [16,18]. Thus, similarly to other domains, different processes and strategies could be employed according to the degree of expertise. While naïve subjects would need to employ inhibitory control over more automatic/common associations, professional artists may be more prone to think out-of-the-box and require such inhibition to a lesser extent. In this view, creative gestures are not the product of artistic creation but can themselves be considered entailed in the creative process, as the artists create-by-doing. As a result, embodied simulation should be seen as the fingerprints of expertise both in visual artistic appreciation and production. In line with the classical dichotomy between the Apollonian (rational) and the Dionysian (corporal) views of art, our results seem to suggest that while artistically naïve individuals may seek conceptual learning while approaching artworks, the corporal dimension is intrinsic to how artists experience and create their own art. Given the absence of a direct comparison between the data of expert artists and naïve individuals performing similar art-related tasks, this interpretation remains speculative, though we hope it may inspire future research.

The generalizability of our results is subject to some inevitable limitations for the specific population studied: it would be interesting for the future to study aesthetic appreciation in art historians and critics, as they maximally resemble artists for aesthetic expertise while not possessing the pragmatics of an artist. The prediction is made, pending an empirical verification, that they might resemble naïve people than artists while watching works of art, as they should not possess the motoric repertoire needed in art-making. On the other hand, the study of amateur less proficient artists could also be instructive to assess whether the level of proficiency in art making counts in determining patterns of brain activity during aesthetic appreciation.