Functional Connectivity Differences in the Perception of Abstract and Figurative Paintings

Abstract

The goal of neuroaesthetic research is to understand the neural mechanisms underpinning the perception and appreciation of art. The human brain has the remarkable ability to rapidly recognize different artistic styles. Using functional connectivity, this study investigates whether there are differences in connectivity networks formed during the processing of abstract and figurative paintings. Eighty paintings (forty abstract and forty figurative) were presented in a random order for eight seconds to each of the 29 participants. High-density EEG recordings were taken, from which functional connectivity networks were extracted at several time points (−300, 100, 300 and 500 ms). The debiased weighted phase lag index (dwPLI) was used to extract the connectivity networks for the abstract and figurative conditions across multiple frequency bands. Significant connectivity differences were detected for both conditions at each time point and in each frequency band: delta (p < 0.0273), theta (p < 0.0292), alpha (p < 0.0299), beta (p < 0.0275) and gamma (p < 0.0266). The topology of the connectivity networks also varied over time and frequency, indicating the multi-scale dynamics of art style perception. The method used in this study has the ability to identify not only brain regions but their interaction (communication) patterns and their dynamics at distinct time points, in contrast to average ERP waveforms and potential distributions. Our findings suggest that the early perception stage of visual art involves complex, distributed networks that vary with the style of the artwork. The difference between the abstract and figurative connectivity network patterns indicates the difference between the underlying style-related perceptual and cognitive processes.

1. Introduction

One of the most extraordinary human abilities is to create and appreciate art. The earliest evidence of this is cave paintings that can be traced back to tens of thousands of years ago [1]. Until recently, the study of art and aesthetics has been the focus of interest of researchers in the humanities. The emergence of neuroaesthetics [2] as a research field has opened up this area to scientists from other fields, enabling them to investigate the neural mechanisms underpinning the perception and appreciation of art. Neuroaesthetics combines neuroscience, psychology and aesthetics research to investigate how the human brain processes and responds to artistic stimuli. It seeks to understand how our neural architecture enables the rapid recognition of and emotional response to artistic styles. Studies [3,4,5,6,7] have shown that beyond the visual cortical areas, other specific brain regions, such as the orbitofrontal cortex, the anterior insula and posterior parietal cortex, are also involved in the aesthetic appreciation of visual art, highlighting the intricate relationships among perception, cognition and emotion. Our response to art is also influenced by its visual and contextual qualities. For instance, grouping, symmetry, order, color, contrast, complexity, content and style all play a key role in how art is visually perceived and emotionally evaluated.

Advanced neuroimaging techniques, particularly functional magnetic resonance imaging (fMRI) and electroencephalography (EEG), have opened up new opportunities for exploring the neural basis of aesthetic processing. EEG, due to its excellent temporal resolution, allows researchers to track brain activity as it unfolds in milliseconds, which is not possible with fMRI. Most existing EEG-based neuroaesthetics research has focused on event-related potential (ERP) analysis that relies on trial-averaged EEG signal waveforms, or time–frequency analysis that examines the brain’s oscillatory activity in different frequency bands [8,9,10,11]. These methods provide information about individual electrodes or regions only and are not suitable for detecting cooperation between different brain regions. Consequently, there is growing interest in brain connectivity networks that have the potential to offer a more comprehensive view of how different brain regions interact during the perception and appreciation of art.

Brain connectivity can be categorized into structural, functional and effective connectivity [12,13,14,15,16,17]. Structural connectivity pertains to the physical wiring of the brain, such as white matter tracts connecting different regions [17,18], which can be studied using techniques like diffusion tensor imaging (DTI). Functional connectivity refers to the temporal correlations between neural activities in different brain areas in the resting state or during task execution [14,15,16,17], and is computed using techniques like M/EEG and fMRI. Effective connectivity, on the other hand, involves extracting the causal relationships between brain regions and how one area influences another’s activity [14,15,17,18].

In this study, we report on whether and how sensor space functional connectivity analysis can help in uncovering and eventually understanding how different brain regions work together to process artistic stimuli. Our analysis is based on phase-based connectivity techniques that are closely related to brain oscillations (rhythmic electrical activity in the brain) [19,20,21,22], and offer distinct advantages over traditional methods such as amplitude correlation-based connectivity. Oscillations play a critical role in inter-regional communication within the brain, with different frequency bands associated with various cognitive functions [23,24,25]. Phase-based methods rely on phase synchronization information across time or trials, and express the presence of connections with measures such as the phase locking value (PLV) [19] or the phase lag index (PLI) [26] that are less influenced by volume conduction, a common issue in EEG where signals from different regions can interfere with each other due to their overlapping electromagnetic fields [19,22,26,27,28,29,30,31].

This exploratory research aims to identify differences in brain connectivity when subjects engage with abstract versus figurative paintings. Using phase-based connectivity analyses with the debiased weighted phase lag index (dwPLI) [28] as the calculation method, this study seeks to uncover the neural communication patterns linked to the perception, recognition and processing of different artistic styles. The distinction between figurative art, that depicts recognizable real-world objects or scenes, and abstract art, that focuses on shapes, colors and forms, often without specific references to real objects, provides a unique opportunity to investigate how the brain processes different types of visual artistic stimuli.

2. Materials and Methods

2.1. Participants

Twenty-nine right-handed university students (14 male, 15 female) with no formal art education background volunteered to participate in this study. Their age ranged from 19 to 35 years (mean age = 23.14 ± SD = 3.81). Each participant had normal or corrected vision, and had no history of neurological or psychiatric disorders. The participant’s lateral dominance was determined using the Edinburgh Handedness Inventory [32], and to exclude potential confounding handedness effects on connectivity, only right-handed participants were selected for the study. A customized version of the Art Expertise Assessment [33] was used to evaluate the participant’s level of experience and knowledge in the field of visual arts. On the art expertise scale, participants scored an average of 4.66 (SD = 4.30) out of a maximum possible score of 45, indicating minimal exposure to art and art-related expertise. Prior to the EEG experiment, each participant was asked to complete a Positive and Negative Affect Schedule (PANAS) [34] questionnaire to measure the participant’s actual emotional state and verify that they were able to perform the given experiment.

Participants were also briefed about the nature of the study, the risks and benefits, and their right to withdraw at any time. Written informed consent was obtained from every participant confirming their participation and permission for the data to be used for research and publication purposes. The study was approved by the local ethics committee of the university.

2.2. Experiment

The stimuli consisted of 80 digital color images of paintings selected from the less familiar works of famous Hungarian artists arranged into two experimental conditions: 40 abstract and 40 figurative paintings. The paintings collected for the figurative condition belong to the representational art category. The paintings show easily recognizable real-world figurative objects, such as people (faces, portrait and groups), animals, still life and various scenes (landscapes, townscapes, etc.). The abstract condition group consists of non-representational (showing non-recognizable shapes, colors or forms) and highly abstract figurative paintings. For each condition, paintings were selected carefully in terms of content, complexity and visual features to create a balanced, non-biased set of stimuli. Thumbnails of the paintings can be found in Supplementary Materials Figures S1 and S2.

The study was performed in the Bioelectric Brain Imaging Laboratory of the Faculty of Information Technology at the University of Pannonia. Participants were seated in a dimly lit room in front of the computer monitor used to display the paintings at a comfortable viewing distance (80–90 cm). The paintings were scaled to fit the display height of the monitor (1080 pixels). The sequence and timing of the experiment is as follows (also shown in Figure 1). Each painting was displayed for 8 s after a 1 s cue mark showing the remaining number of paintings in the experiment. This was introduced to maintain subjects’ attention and help them focus on the center of the screen. Paintings were presented in a random order unknown to the subjects. After the 8 s presentation, a blank black screen appeared for 4 s during which the subject had to press either the left or right mouse button with their right hand’s index or middle finger, respectively, to indicate a ‘Like’ or ‘Dislike’ response. The measurements took place in a single session for each subject without breaks. The presentation and control of the experiment were conducted by a custom script executed in the Psychtoolbox-3 (Psychophysics Toolbox Version 3) software package [35].

A special-purpose trigger unit [36] was used to generate event triggers for the EEG device. Stimulus onset events were created by displaying a 1 × 1 cm white square on the right edge of the display simultaneously with the painting, which was then registered by a display-mounted light sensor. The light sensor and the mouse micro-switches were connected to the trigger unit to generate hardware trigger signals for the stimulus and response, respectively, that were then sent to the EEG device for digitization.

2.3. EEG Data Collection and Analysis

EEG activity was recorded using a 128-channel Biosemi ActiveTwo (BioSemi B.V. WG-Plein 129 1054SC Amsterdam, The Netherlands) measurement device and Ag/AgCl active electrodes arranged in the Biosemi equiradial ‘ABC’ electrode layout [37] held in place by an elastic cap. Conductive saline gel (Signa) was applied to ensure good electrical contact between the electrodes and the scalp. DC electrode offset values were kept stable below the recommended 40 μV amplitude level. The analog input EEG signal was low-pass-filtered (fc = 410 Hz, −3 dB) by the EEG device before being digitized at a sampling rate of 2048 Hz. Data were saved in BDF EEG file format for later offline processing.

2.3.1. Pre-Processing

The EEG data processing was performed using the EEGLAB Toolbox [38] in three key steps: (i) pre-processing, (ii) removing unwanted noise and artifacts, and (iii) segmenting the data into epochs. The raw EEG signals were filtered with 1 Hz and 40 Hz cut-off frequency high- and low-pass zero-phase FIR filters (EEGLAB eegfiltnew filter), respectively. Next, the data were down-sampled to a 512 Hz sampling rate. The filtered EEG datasets were inspected for bad channels (having extremely high or low amplitudes or excessive noise) and were interpolated in EEGLAB using spherical spline interpolation (16 subjects, average number of bad channels: 3). After interpolation, the dataset was re-referenced to the average reference, and Independent Component Analysis (ICA) [39] was performed to separate eye, muscle and channel noise artifacts from the neural signals. The resulting independent components (ICs) were labeled as potential artifacts using the IC Label EEGLAB plugin [40]. Eye blinks, muscle and channel noise artifact components with a confidence level above 80% were removed from the component set to reconstruct the artifact-free signals. Datasets still showing artifacts after cleaning were further inspected and the remaining artifactual independent components with a sub-80% confidence level were removed manually.

2.3.2. Data Epoching

After pre-processing, the EEG data were segmented to stimulus-locked 2 s epochs (with −1 s pre-stimulus and 1 s post-stimulus intervals). Following the baseline (−500 to 0 ms) correction, the trials were labeled as ‘figurative’ or ‘abstract’ depending on the style of the stimulus, creating datasets for the two conditions. After epoch averaging, the ERP waveforms (Figure 2) showed peaks at approx. 120 ms, 250 ms and 450 ms. A well-known effect in event-related potential studies is the increasing temporal blur of data as we move further away from the stimulus. We presented the paintings for 8 s but, as shown in Figure 2, the ERP waveform after 1 s post-stimulus is hardly distinguishable from the baseline period. After approximately one second, the probability of phase locking and synchronization across trials sharply decreases, resulting in strong signal cancelation and a low signal-to-noise ratio after epoch averaging. Figure 3 illustrates the spatial distribution of the ERPs from 0 to 1100 ms. These maps show that the potential distribution is dominated by occipital area activations, with hardly any detectable activations in other areas. For these reasons, we chose to use functional connectivity analysis to extract the interactions of different brain regions and focused our analysis on the 0–1 s post-stimulus interval only. By displaying the painting for 8 s, however, we can separate the decision making and response planning/execution processes from the perception and recognition of the paintings.

2.3.3. Functional Connectivity Computation

As shown in Figure 3, ERPs show few temporal variations in the first second after the stimulus onset. We hypothesize that functional connectivity based on phase synchronization measures can provide more insights into the cortical processes in this time interval. Connectivity computation and analysis was performed using functions of the Fieldtrip [41] software (fieldtrip-20201002) in custom scripts. We chose the debiased weighted phase locking index (dwPLI) as the basis of the connectivity network calculation. This measure has the advantage of not being sensitive to phase angle differences of 0 and 180 degrees, thus mitigating spurious connections generated by volume conduction effects. Functional connectivity can be computed in the sensor space (i.e., on electrodes) or in source space (on cortical regions or source locations). We chose to use sensor-based connectivity, since, in most cognitive experiments, it is difficult to obtain subject-specific MRI images, without which it is impossible to perform source localization and subsequent source-based connectivity calculation. While sensor space connectivity may not be able to correctly localize activation sources, it can be a valuable tool to find experimental condition differences, hence providing an effective method for testing condition effects that might otherwise remain undetectable.

First, a time–frequency spectral computation was performed (ft_freqanalysis) for the [−1, 1] s time interval and 1–40 Hz frequency range by wavelet transformation using 3-cycle Morlet wavelet windows. The output was than divided into standard EEG frequency bands (delta: 1–4 Hz, theta: 4.5–8 Hz, alpha: 8.5–13 Hz, beta: 13.5–30 Hz and gamma: 30.5–100 Hz) and fed into the ft_connectivityanalysis() function to compute the connectivity matrices containing the dwPLI values for each electrode pair for each frequency band. After averaging the frequency entries for each band and the full band, we generated a single connectivity matrix for each band and the full frequency band for the abstract and figurative conditions.

The full-band connectivity matrix was used to compute the strength (the weighted version of the node degree) of each electrode. The strength is then the sum of the weights of all edges connecting to a given electrode.

For visualization purposes, we also computed the modularity and node betweenness network measures using the Brain Connectivity Toolbox (BCT) [42] functions modularity_und() and betweenness_wei().

2.3.4. Statistical Analysis

Since we were interested in the differences between two conditions (abstract and figurative) executed for each subject, we employed a within-subject design with two conditions. A non-parametric Monte Carlo permutation test was performed using the ft_freqstatistics() function. A dependent samples one-tailed t-test was performed for each connection edge–time point pair with a randomization level of 1000. The alpha level was set to 0.05 and FDR multiple comparison correction was used.

The statistical analysis was performed in two steps, first for figurative > abstract, then for abstract > figurative to detect significant increases in connectivity (network edges) during abstract and figurative painting perception, respectively.

The Fieldtrip connectivity analysis function expects a binary input network, in which an edge is either present or missing, i.e., without a weight. This requires a preceding thresholding operation that can be problematic since one can use, e.g., the absolute threshold, which may result in widely varying network densities across conditions, or the proportional threshold, where the top x % strongest edges are retained [18]. This latter method can make very weak weights as important as strong weights in another network. To avoid this problem, we modified the statistical calculation to use a weighted input network and perform the statistics on the non-thresholded network. The resulting significant edges (p < 0.05) were retained and saved as a mask for the next step, in which the original networks were proportionally thresholded, keeping 5% of the strongest edges, and then masked with the significant edge mask to keep only those edges from that set that represent true connection increases in each condition.

3. Results

3.1. Node Strengths

First, we show the changes in the node (electrode) strength values (group grand average) depicted as a scalp topography time series (Figure 4: figurative condition, Figure 5: abstract condition). The node strengths were computed for the entire group and the full frequency band (2–40 Hz). A strong increase in the node strength occurs first at 100 ms post-stimulus for both conditions in the occipital area as an indication of the initial visual response. This is followed by a short frontal increase at 150 ms, again, in both conditions but with slightly different magnitude. A third prominent increase occurs from 300 to 400 ms for the figurative and from 350 to 550 ms for the abstract condition, respectively.

3.2. Functional Connectivity Networks

Here, we present the functional connectivity network results for the five bands and two conditions. The results are presented as an increase in connectivity of one condition over the other (one group subtracted from the other), i.e., ‘figurative’ condition: figurative minus abstract, and ‘abstract’ condition: abstract minus figurative networks, respectively. For each band, we show the network changes at four selected time points (−300 ms/baseline/, and 100, 300 and 500 ms post-stimulus). The size of the nodes represent the relative node strength, whereas the node colors depict the network modularity (nodes with identical colors belong to the same module).

The delta band results for abstract and figurative conditions across time points are shown in Figure 6. The figurative condition shows a significant increase in node strength (size) and interconnections at 100 ms (p = 0.0259), particularly in the left occipital and parietal regions. The node strengths further increase in the 300 ms (p = 0.0273) to 500 ms (p = 0.0247) interval, showing robust connectivity across various regions (frontal, temporal and parietal areas). Conversely, the abstract condition displays increased, dense and widespread connections in the frontal and posterior areas (temporal and occipital) from 100 ms (p = 0.0239) to 300 ms (p = 0.0233), which intensifies over time.

In the theta band (Figure 7), we observe the highest figurative connection increase at 500 ms (p = 0.0234), initially with frontoparietal involvement at 300 ms (p = 0.0288) that becomes more right-hemisphere-dominant with stronger frontal connections by 500 ms. In comparison, the abstract condition reveals dense, increased connectivity, earlier than that seen in the figurative condition, at 100 ms (p = 0.0223) in the frontal, right parietal and occipital regions. This connectivity density then decreases from 300 ms (p = 0.0270) to 500 ms (p = 0.0306), with a noticeable shift towards the left temporoparietal region at 300 ms, and central and left occipital areas at 500 ms.

The alpha band connectivity patterns (Figure 8) also progressed differently for both the abstract and figurative conditions. The figurative condition exhibits focused connectivity in the bilateral central regions, particularly in the left temporal area at 100 ms (p = 0.0264), progressing to denser connections in the fronto-occipital network by 300 ms (p = 0.0267), and shifting to the right frontal and posterior areas at 500 ms (p = 0.0268). The abstract condition, however, shows increased connectivity at 100 ms (p = 0.0259), with connections spanning from the occipital to frontal regions in both hemispheres. The connectivity pattern then shifts to the right hemisphere at 300 ms (p = 0.0243), involving the frontal, right temporal and left parietal areas, before becoming more distributed across the frontal, left temporal and occipital areas at 500 ms (p = 0.0269).

The beta band connectivity (Figure 9) reveals significant connectivity differences across time points for both conditions. For the figurative condition, increased connectivity can be observed in the frontal and posterior regions, particularly at 100 ms (p = 0.0232) and 300 ms (p = 0.0232), with a higher beta connectivity in the frontoparietal network at 500 ms (p = 0.0260). In contrast, the abstract condition shows stronger connectivity, focusing in the occipital and parietal regions at 100 ms (p = 0.0252) and 300 ms (p = 0.0275), with frontal involvement at 300 ms. By 500 ms (p = 0.0256), the abstract condition displays more widespread beta connectivity across the scalp (occipital, parietal and central areas).

The gamma band shows distinct temporal dynamics for both the abstract and figurative connections (Figure 10). Initially, at 100 ms, the figurative condition (p = 0.0245) exhibits increased connectivity with prominent nodes in the frontal and temporal regions, while for the abstract condition (p = 0.0252), the connectivity is mainly in the right frontal and left parieto-occipital regions, with relatively larger node strengths appearing in those areas. As processing continues to 300 ms, the figurative condition (p = 0.0254) displays extensive connections across the cortex, with larger nodes appearing in both the right frontal and occipital areas. The connections in the abstract condition (p = 0.0253) are more distributed in the frontal and occipital regions, similar to those seen at earlier time points but shifting to the right hemisphere. By 500 ms, the figurative condition (p = 0.0248) presents a more centralized connectivity whereas the abstract condition (p = 0.0245) shows widespread connectivity, particularly in the frontal and left temporal regions.

Table 1. Mean p-values of the significantly different edges for the figurative and abstract conditions across five frequency bands (delta, theta, alpha, beta, gamma) at different time points (−300 ms, 100 ms, 300 ms, 500 ms).

	Mean p-Values of the Significant Edge Weight Differences
	−300 ms (Baseline)	100 ms	300 ms	500 ms
Figurative > Abstract
Delta	0.0248	0.0259	0.0273	0.0247
Theta	0.0275	0.0292	0.0288	0.0234
Alpha	0.0299	0.0264	0.0267	0.0268
Beta	0.0263	0.0232	0.0232	0.0260
Gamma	0.0266	0.0245	0.0254	0.0248
Abstract > Figurative
Delta	0.0258	0.0239	0.0233	0.0204
Theta	0.0250	0.0223	0.0270	0.0306
Alpha	0.0273	0.0259	0.0243	0.0269
Beta	0.0256	0.0252	0.0275	0.0256
Gamma	0.0246	0.0251	0.0253	0.0245

highlights the strong statistical evidence for the increased connectivity differences between the abstract and figurative conditions, spanning the time interval from the baseline (−300 ms) to 500 ms post-stimulus across multiple frequency bands. All the connectivity differences in Figure 6, Figure 7, Figure 8, Figure 9 and Figure 10 are highly significant with the p-values being (p < 0.05) for both the figurative > abstract and abstract > figurative comparisons across all the frequency bands and time points. Table 1 lists the mean p-values of the significant network edges.

4. Discussion

In this paper, we explored the application of functional connectivity calculations and analysis in investigating human responses to different styles of art. Our results indicate that functional connectivity provides a more sensitive method that can potentially uncover the temporal and spatial details of brain activity that remain hidden in traditional ERP or time–frequency analysis.

Several M/EEG-ERP studies have focused on the time course and temporal order of processing activities during art perception and identified several key stages [43,44]. The initial stage starts at around 100 ms post-stimulus with early visual processing involving visual areas, such as V1, V2 and V4. Perceptual grouping, symmetry analysis, etc. provide input to the initial impression formation stage occuring at around 300 ms. From around 300 to 600 ms, the recognition and memory networks related to memory retrieval and processing familiarity engage with the artworks. The final stage, beyond approximately 600 ms, is assumed to be the classification stage, which involves the recognition of the content and style of the artworks. This is known as the early—automatic—aesthetic evaluation, which involves extensive right hemispheric activations. Activities after the first second fall into the deliberate, contemplative aesthetic processing phase that is believed to evoke aesthetic emotions.

Our functional connectivity study focused on the first 1 s of aesthetic processing and the time course of activities manifested in the connectivity networks. Our exploratory analysis on network node strengths revealed more activation details than traditional ERP analysis. The extracted node strength map series (see Figure 4 and Figure 5) demonstrated that for both the abstract and figurative conditions, this method could detect more detailed activation patterns in the frontal regions than would be inferred from the ERP topoplots alone and showed that occipital activation is neither constant nor symmetric over time (Figure 3).

For the connectivity network analysis, we selected three key time points (100, 300 and 500 ms), aligning with past studies [43,45,46,47] and with the main peaks found in our ERP waveform (Figure 2). The analysis based on the edge weights of the functional connectivity networks found significant differences between the abstract and figurative conditions across all the time points and frequency bands. A significant difference in connectivity represents an increase in network edge weight that in turn represents an increase in phase synchronization (stronger communication) between brain regions.

The increased connectivity in our first condition (figurative > abstract) showed a sustained and focused network, initially involving the occipital and parietal regions then progressing to the frontal and other areas as processing continues. This temporal progression aligns with previous literature suggesting that art processing begins with perceptual analyses before advancing to higher-order semantic processing and context integration [43,48,49]. Also, the spatial pattern of the figurative connectivity network showed a pronounced right hemisphere dominance at 500 ms in the theta, alpha and beta bands, matching results reported in past studies [43,44].

In comparison, our other (abstract > figurative) condition showed more widespread connectivity across different brain areas, with more prominent occipital region [50] activations in the theta, alpha and beta bands at 500 ms, which may reflect the stronger reliance on purely visual elements and the lack of visual memory and recognition tasks that might characterize abstract painting processing.

The novelty of our work is the demonstration that the presented functional connectivity analysis is able to detect significant differences and changes in connectivity patterns at distinct time points instead of temporal averages. Our study has also demonstrated that phase synchronization methods in general, and dwPLI in particular is a reliable way to extract functional connectivity networks. Using dwPLI removes the need to apply additional spatial filtering techniques to mitigate volume conduction effects to avoid the detection of spurious connections.

The work presented here has some limitations. We used distinct selected time points in the analysis instead of looking at continuous temporal changes in connectivity. The reasons for this were (i) partly inherent in the method since dwPLI uses spectrum-based calculation that requires window-based operations that make it problematic to obtain instantaneous phase information, and (ii) it would have generated severe computational issues (i.e., very long runtimes and an extreme amount of memory space requirements). For similar computational reasons, we have not looked into the calculation of effective connectivity, which could provide additional causality information on connections, which obviously would be very important for understanding the intricacies of the underlying cortical mechanisms.

In our study, we used two rather general art categories. In the future, relying on the demonstrated sensitivity of our method, one could create more conditions and then perform comparisons between more specific groups of paintings (e.g., portraits, landscapes, still life, abstract figurative and fully non-representational) to allow for a fine-grained analysis on how different artistic elements influence neural connectivity. In addition, further analyses of the effects of visual properties such as color, contrast, complexity and symmetry on the neural processing of art could be another area to investigate. Symmetry, for instance, is often associated with positive aesthetic judgments and emotional responses that contribute to the perception of beauty in both visual and auditory art forms [51].

An additional area left unexplored in this study is the effect of art expertise on the neural correlates of art perception. It has been shown [52,53,54,55] that art novices generally have a preference towards representative art. Since the subjects in our study did not have formal art education, the connectivity differences might have been influenced by the lower level of art expertise. A future study should compare the connectivity network differences between artists and non-artists to study the effect of art expertise on cognitive processing and to shed light on how they engage with abstract and figurative paintings.

Finally, in our current study, we have only looked at the edge weight and node strength parameters of the connectivity networks. However, there are many other graph measures that can be used for a more detailed comparison of the detected networks, e.g., node centrality, modularity, efficiency measures, etc. Future studies should focus on these as well. Additionally, the node strength seems to be a more sensitive measure of node activation than event-related potential or power, and consequently could be explored as a potential alternative to traditional ERP analysis methods.

5. Conclusions

This study used functional connectivity to investigate whether our brain responds differently to different styles of art paintings. We focused on the first second of viewing to find perceptual differences when viewing abstract and figurative paintings. The results indicated that there were significant differences in the connectivity networks in all frequency bands and at each selected time point for the two conditions. Our findings show that the early perception stage of viewing visual art involves complex, distributed networks that vary with the style of the artwork. The difference in the abstract and figurative connectivity network patterns imply that there are differences in the underlying style-related perceptual and cognitive processes. We also demonstrated that functional connectivity provides a sensitive method that can extract important temporal and spatial details. In the future, comparing differences between artists and non-artists could provide insight into how art education influences brain connectivity and the activation of different brain regions.