Recent advances in neuroimaging and electrophysiological techniques have provided new insights into the neural mechanisms underlying symmetry perception (Beck et al., 2005; Carr et al., 2025). Functional magnetic resonance imaging (fMRI) and electroencephalography (EEG) studies have identified a distributed network of brain regions involved in processing symmetrical patterns. These include early visual areas such as the primary visual cortex (V1) and extrastriate areas (V2, V4), as well as higher-order regions like the lateral occipital complex (LOC) and the intraparietal sulcus (IPS) (Bertamini et al., 2018; Makin et al., 2021). These regions work in concert to encode local symmetrical features and integrate them into global perceptual representations. For example, Tyler et al. (2005) used fMRI to investigate the LOC exhibits strong neural responses to symmetry patterns (such as mirror and rotational symmetry), supporting its role in symmetry grouping and perception (Tyler et al., 2005). This suggests that the LOC plays a critical role in the perceptual grouping of symmetrical patterns. Similarly, Bertamini et al. (2018) demonstrated that symmetry processing involves both feedforward and feedback mechanisms, with early visual areas encoding local features and higher-order regions integrating these features into coherent perceptual objects (Bertamini et al., 2018).

While neural mechanisms provide the foundation for symmetry detection, cognitive factors such as attention, expectation, and learning also play a significant role (Elena et al., 2021; Makin et al., 2023). Recent studies have shown that symmetry perception can be modulated by top-down processes, such as attentional focus and task demands. For instance, research by Bertamini and Makin (2014) demonstrated that attention can enhance the perception of symmetry, particularly when participants are explicitly instructed to focus on symmetrical patterns (Bertamini and Makin, 2014). This suggests that attention can amplify the neural responses to symmetry, making it more salient in the visual field. Moreover, understanding hemispheric specialization in symmetry perception is critical for cognitive neuroscience because it reveals fundamental principles of how the brain achieves perceptual organization (Rabbito et al., 2023). The two hemispheres exhibit distinct processing biases - with the left hemisphere preferentially analyzing local features and the right hemisphere specializing in global integration (Fink et al., 1997; Makin et al., 2021; Paulraj et al., 2018). As symmetry detection requires both local element processing and global configuration analysis, it provides an ideal paradigm to investigate how divided hemispheric computations are coordinated through callosal connections.

In this study, we aim to further explore the mechanisms underlying symmetry perception, with a particular focus on the role of the two cortical hemispheres in the detection of reflection symmetric patterns. Specifically, does there exist hemisphere specialization for the detection of reflection symmetric patterns? The callosal hypothesis posits that the anatomical symmetry of the human visual system underlies the efficiency of vertical symmetry detection. However, recent evidence suggests that symmetry perception may involve more complex interactions between the two hemispheres. We aim to examine whether there is a functional specialization of the left and right hemispheres for processing different orientations of reflection symmetry.

To address these issues, we conducted a series of psychophysical and electrophysiological experiments in which participants were presented with two types of symmetrical patterns: cycle and right-left quarter-circle. These patterns were presented in the right, left, or both visual fields to assess the role of hemispheric processing in symmetry detection. By combining behavioral measures with EEG recordings, we aim to provide a comprehensive understanding of the neural and cognitive mechanisms underlying symmetry perception.

Fifteen students (10 males; mean age = 26.3; SD = 3.4) with normal or corrected-to-normal vision from Okayama University volunteered for the experiments. They provided written informed consent for their participation in this study, which was previously approved by the ethics committee of Okayama University.

We used MATLAB (v 14; the MathWorks, Okayama, Japan) and GERT, the Grouping Elements Rendering Toolbox (Demeyer and Machilsen, 2012), to construct arrays of non-overlapping Gabor elements on a uniform gray background (Figure 1). The arrays comprised 496 × 496 pixels. Each Gabor element was defined as the product of a sine wave luminance grating (frequency of 3 cycles per degree of visual angle and presented at 100% Michelson contrast) and a circular Gaussian (standard deviation of 3 arc min). There were 512 Gabor elements, interior 62 elements, exterior 410 elements and outline 40 elements. The number of elements inside and outside the shape outline was adjusted to ensure a homogeneous spacing between the Gabor elements. The orientations of the inside and outside Gabors were randomly and unchanged. A subset of 40 Gabor elements were positioned along the contour outline of an artificial circle shape. The shape outline was generated by plotting the sum of 5 radial frequency components in polar coordinates. The orientation of Gabor on contour was defined by stimulus type and stimulus intensity. There were two kinds of stimulus type: circle, right-left quarter-circle, see Figure 1A. The elements had orientations parallel to the local tangent of the shape outline. For example, the orientation of left 20 Gabors were local tangent of the shape outline but right 20 Gabors were randomly when stimulus was left semicircle. In addition, we created three different noise levels by altering the orientations of the contour elements. As shown in Figure 1B, Stimuli noise 0.0 (SN0.0): The orientations of the Gabors are aligned with the local tangents of the shape contours; Stimuli noise 0.4 (SN0.4): Randomly change the orientations of the 8 Gabors; Stimuli noise 0.8 (SN0.8): Randomly change the orientations of the16 Gabors.

The subjects were instructed to perform the experiment in a dimly lit, electrically shielded and sound-attenuated room (laboratory room, Okayama University, Japan) with their head positioned on a chin rest. Stimulus presentation and response collection were conducted using E-prime 1.1 software (Psychology Software Tolls, Inc., Pittsburgh, PA, USA). Figure 2 showed the experiment consisted of three sessions of 120 trials each, all stimuli were presented randomly. For each stimulus condition, subjects were presented with a central fixation cross for a randomized duration between 1200 and 1800 ms, followed by the stimulus display for 150 ms. During the experiment, participants were instructed to press the “1” key on the keyboard when a “Cycle” stimulus was displayed, the “2” key for a “left quarter-circle” stimulus, and the “3” key for a “right quarter-circle” stimulus.

An EEG system (BrainAmp MR plus, Gilching, Germany) was used to record EEG signals through 32 electrodes mounted on an electrode cap (Easy Cap, Herrsching-Breitbrunn, Germany). Horizontal eye movements were measured by deriving the electrooculogram (EOG) from one electrode placed about 1cm from the outer canthi of the left eye. Vertical eye movements and eye blinks were detected by deriving an EOG from an electrode placed approximately 1.5 cm below the subject’s left eye. All signals were referenced to left and right earlobe, and the impedance was maintained below 5 kΩ. Raw signals were acquired at a sample rate of 500 Hz and stored for off-line analysis.

The ERPs elicited by target stimuli were analyzed by using the Brain Vision Analyzer software (version 1.05, Brain Products GmbH, Munich, Germany). The data were band-pass filtered from 0.01 to 30 Hz. Then, the data were divided into epochs, from −100 ms before stimulus onset to 500 ms after stimulus onset, and baseline corrections were made to the data from −100 ms to stimulus onset. Epochs contaminated by artifacts (i.e., eye movements, eye blinks, amplifier blocking) were rejected based on a threshold of ±100 μV in all channels before averaging. All averaged ERP waveforms were obtained across all participants for each stimulus type in each electrode.

Therefore, our analysis was performed across electrode regions of interest focused at left occipital region (O1), right occipital region (O2), left parietal region (P3 and P7), right parietal region (P4 and P8), see Figure 3.

Murray et al. (2006) found that both correct and incorrect responses did not affect the contour integration effect or shape detection within the first 300 milliseconds (Murray et al., 2006). Therefore, four main components before 300 ms were statistically analyzed. The first component was the P/N100 complex (peaking around 100 ms after stimulus onset); the second component was the P/N130 complex (peaking around 130 ms after stimulus onset); the third component was the N190 (peaking around 190 ms after stimulus onset); the last component was P260 (peaking around 260 ms after stimulus onset). Amplitudes of these component respectively measured as an average of 20 ms duration around the peak. ANOVA (Greenhouse-Geisser corrections with corrected degrees of freedom).

Table 1 showed the proportion correct for the 6 stimuli. Analysis of the proportion correct using 2 (stimulus type) × 3 (noise level) repeated measures ANOVA revealed a significant main effect in noise level [F (2, 28) = 190.008, p < 0.001, η_p² = 0.931], showing that the identification ability of participants was decreased with noise enhancement. Whereas, no significant difference was found in stimulus type [F (1, 14) = 2.663, p = 0.125, η_p² = 0.160]. The interaction between stimulus type and noise level was significant [F (2, 28) = 6.705, p = 0.012, η_p² = 0.324], indicated that contour effect was different in varying noise level. The pairwise comparisons showed that for the SN 0.0 condition, the responses to cycle stimuli were significantly higher than the responses to quarter stimuli (p < 0.05). All reported statistics reflect Green-Geisser corrections at a significance level of 0.05.

Figure 3 showed the ground-averaged ERPs across all participants for 6 stimuli. All stimuli evoked similar P100, N130, N190, P260 components at occipital electrode, and evoked N100, P130, N190, P260 components at parietal electrode. Analysis of the peak latency with 4 components revealed there was no significant difference in 6 stimuli across both hemispheres and electrodes (all p > 0.05). Analysis of the mean amplitude of 4 components (P/N100, P/N130, N190, P260) × 2 electrode (occipital, parietal) × 2 hemisphere (left, right) × 2 stimuli type (cycle, quarter) × 3 noise level (SN0.0, SN0.4, SN0.8) repeated measures ANOVA revealed a significant four-way component × electrode × hemisphere × stimuli type × noise level interaction [F (6, 84) = 3.937, p = 0.015, η_p² = 0.219]. The results suggested different processing patterns for different noise conditions and for left hemisphere and right hemisphere. Therefore, we analyzed these differences in detail as follows.

One of the key questions addressed in this study was whether there is hemispheric specialization for the detection of reflection symmetry. Our ERP results provide partial support for this idea, with significant hemisphere effects observed at multiple stages of processing. For example, the P/N100 component showed more negative amplitudes in the left hemisphere at parietal electrodes, while the N190 and P260 components revealed significant interactions between hemisphere, stimulus type, and noise level. These findings suggest that the two hemispheres may play distinct roles in symmetry perception, with the left hemisphere potentially specializing in the processing of local features and the right hemisphere in the integration of global shapes (Fink et al., 1997; Sasaki et al., 2005). However, further research is needed to clarify the specific contributions of each hemisphere to symmetry perception.

The callosal hypothesis posits that the anatomical symmetry of the human visual system, particularly the role of the corpus callosum in interhemispheric communication, underlies the efficiency of vertical symmetry detection (Wright, 1972). While our findings do not directly contradict this hypothesis, they suggest that symmetry perception involves more complex interactions between the two hemispheres than previously thought. For example, the significant hemisphere effects observed in our study indicate that symmetry processing is not solely determined by the anatomical midline but is also influenced by functional specialization and task demands. This aligns with recent reviews suggesting that the strong version of the callosal hypothesis is unlikely to fully account for the observed phenomena in symmetry perception (Bertamini et al., 2018; Treder, 2010).