Previous work has found that feature attention can modulate electrophysiological responses to visual symmetry. In the current study, participants observed spatially overlapping clouds of black and white dots. They discriminated vertical symmetry from asymmetry in the target dots (e.g., black or white) and ignored the regularity of the distractor dots (e.g., white or black). We measured an electroencephalography component called the *sustained posterior negativity* (SPN), which is known to be generated by visual symmetry. There were five conditions with different combinations of target and distractor regularity. As well as replicating previous results, we found that an orthogonal axes of reflection in the distractor dots had no effect on SPN amplitude. We conclude that the visual system can processes reflectional symmetry in independent axis-orientation specific channels.

*sustained posterior negativity*(SPN). A typical SPN is shown in Figure 1A. In this example, SPN amplitude scales with the proportion of symmetry in the symmetry plus noise displays (Makin, Rampone, Morris, & Bertamini, 2020).

*complete Liverpool SPN catalogue*(Makin et al., 2022). The “W-load” variable on the X axis is a theoretical measure of regularity salience (van der Helm & Leeuwenberg, 1996). The more obvious the regularity, the higher the W-load. SPN amplitude increases (becomes more negative) with W. SPN is also enhanced when regularity is task relevant (blue dots). Figure 1B suggests that no task manipulations can abolish the SPN response to high W regularity (0.6 or above). There is one isolated exception (Rampone, Makin, & Bertamini, 2014), but this was not replicated in a recent unpublished study.

^{2}and black dots of 14 cd/m

^{2}.

*independent*axes of reflection and those with two

*dependent*axes of reflection. The later are sometimes known as double reflection or twofold reflection. According to the bootstrapping model of symmetry perception (Wagemans, van Gool, & D'ydewalle, 1991), the elements of a single-axis vertical reflection are unified into midpoint-collinear virtual trapezoids. The visual system finds additional correlation trapezoids by “bootstrapping” along the vertical global axis. For double reflection, with dependent horizontal and vertical axes, some of these correlation quadrangles are rectangles. Our new Ref(RefOrtho) condition could be called a twofold reflection, if we ignore color, but it does not feature correlation rectangles (Treder et al., 2011).

*M*= 7.27).

*p*value gives the probability of obtaining the observed effect (or larger), given the null. However, we are interested in the probability of the null being true given the observed effect. In other words, null hypothesis significance testing gives

*p*(D|H0), and we are interested in

*p*(H0|D). We thus used Bayesian

*t*tests to obtain the desired

*p*(H0|D). We computed Bayes factors (

*BF*01 and

*BF*10) using free JASP software (JASP Team, 2022). We used the default, uninformed prior, which assigns the null and alternative models equal prior odds. With this conventional default in place,

*BF*01 = posterior odds in favor of H0, and

*BF*10 = posterior odds in favor of H1. We can thus derive

*p*(H0|D) with the formula

*BF*01/(1 +

*BF*01). Bayesian

*t*tests also require one to set priors on parameters within the models. We used the default Cauchy prior with an

*r*scale of 0.707.

*BF*s between 1/3 and 3 are inconclusive.

*BF*01 > 3 is evidence in favor of H0.

*BF*10 > 3 is evidence in favor of the H1.

*BF*01 < 1/3, is evidence in favor of H1.

*BF*10 < 1/3 is evidence in favor of H0.

*N*= 60, we have 80% power for finding within-subjects effects of

*d*= 0.36 (

*α*= 0.05, two-tailed). Analysis of the SPN catalogue suggests relatively small 0.37 microvolt difference between conditions would typically be associated with an effect of this magnitude. This means our study is powered to find 0.37 microvolt SPNs or SPN modulations. However, the study was completed in two parts, and we observed the results of the vertical target group before running the horizontal target group. The original sample size of 30 was not selected by a priori power analysis, and the second sample of 30 was chosen for consistency with the first.

*p*correct scores conditions were negatively skewed in all conditions (Shapiro-Wilk test,

*p*< 0.01).

*χ*

^{2}(5) = 104.054,

*p*< 0.001). The Ref(Ref), Ref(RefOrtho) and Ref(Rand) conditions are most interesting. Here the correct answer was “symmetry.” Median

*p*correct was near ceiling in the Ref(Ref) condition (0.98), but reduced in the Ref(Ortho) condition (0.88) and the Ref(Rand) condition (0.90). The ∼10% performance reductions in Ref(Rand) and Ref(RefOrtho) were significant (

*p*< 0.001, Wilcoxon signed ranks tests). Apparently, these task irrelevant distractor dots could not be completely suppressed by feature attention. Averaged across conditions, performance was similar in the vertical and horizontal target groups (0.88 vs. 0.91,

*p*= 0.801, Mann-Whitney

*U*test).

*p*= 0.756, Mauchly's test). A mixed analysis of variance confirmed that there was significant difference among the five conditions (

*F*(4, 232) = 39.798,

*p*< 0.001, partial

*η*

^{2}= 0.407). This did not interact with the between subject's factor Target orientation (

*F*(4, 232) = 1.837,

*p*= 0.123, partial

*η*

^{2}= 0.031), and there was no main effect of Target orientation (

*F*(1, 58) < 1,

*NS*).

*t*(59) −3.561,

*p*< 0.001,

*d*

_{z}= −0.460) and the difference between Ref(Ref) and Ref(Rand) (

*t*(59) = −3.110,

*p*= 0.003,

*d*

_{z}= −0.402). All three conditions where the correct answer was “symmetry” all produced a significantly larger SPN than either of the two conditions where the correct answer was “asymmetry” (smallest effect

*t*(59) = 6.271,

*p*< 0.001,

*d*

_{z}= 0.810).

*BF*01 wave, where the BF01 from successive Bayesian one sample

*t*tests is plotted over time on a log scale. When

*BF*01 rises above 3, there is unlikely to be a difference from Rand(Rand). When this falls below 1/3, there is likely to be a difference from Rand(Rand). This suggests no SPN in the Rand(Ref) and Rand(RefOrtho) conditions.

*BF*01 = 7.065,

*p*(H0|D) = 0.876). This again implies that the task irrelevant horizontal reflection was not registered by the visual system when participants were attending to vertical reflection and vice versa.

*BF*01 over time. Figure 6B shows prior and posterior plot from one Bayesian

*t*test on amplitude averaged over the 300 to 1000 ms window. Figure 6C shows sequential analysis from this Bayesian

*t*test. Figure 6C indicates that

*BF*01 drifted further above 1 as more participants were added to the analysis. In summary, Bayesian analysis supports the conclusion that Ref(RefOrtho) and Ref(Rand) waves were the same.

*against*preattentive symmetry processing. First, reflectional symmetry does not pop out in visual search tasks (Olivers & van der Helm, 1998). Second, reflected contours do not attract spatial attention to regions of crowded displays (Kimchi, Yeshurun, Spehar, & Pirkner, 2016). Third, reflectional symmetry does not produce priming effects in the absence of conscious awareness (Devyatko & Kimchi, 2020). There is also evidence

*for*preattentive processing. First, unconscious symmetry processing happens in hemispatial neglect patients, who are subjectively blind the to the left side of objects (Driver, Baylis, & Rafal, 1992). Second, symmetry in task irrelevant outer contours biases judgements about symmetry in task relevant inner contours (van der Helm & Treder, 2009). Third, concentric rings and radial symmetries activate V4 cells in anaesthetized monkeys (Gallant, Connor, Rakshit, Lewis, & van Essen, 1996).

*is*processed automatically. When participants are classifying patterns as vertical reflection or random, horizontal plus vertical reflection is classified more rapidly than a single vertical reflection (Palmer & Hemenway, 1978). The current results suggest that the orthogonal axis of two independent axes is ignored, but this does not mean the horizontal axis of a double horizontal plus vertical reflection is routinely ignored.

*Cerebral Cortex,*32, 2277–2290, https://doi.org/10.1093/cercor/bhab358. [CrossRef]

*Vision Research,*19, 783–793, https://doi.org/10.1016/0042-6989(79)90154-8. [CrossRef] [PubMed]

*Annals of the New York Academy of Sciences,*1426, 111–126, https://doi.org/10.1111/nyas.13667. [CrossRef]

*Acta Psychologica,*95, 119–140, https://doi.org/10.1016/s0001-6918(96)00038-8. [CrossRef] [PubMed]

*Vision Research,*177, 68–75, https://doi.org/10.1016/j.visres.2020.09.003. [CrossRef] [PubMed]

*Cerebral Cortex,*17, 1423–1432, https://doi.org/10.1093/cercor/bhl054. [CrossRef]

*Symmetry,*12, 1872, https://doi.org/10.3390/sym12111872. [CrossRef]

*Nature,*360, 73–75, https://doi.org/10.1038/360073a0. [CrossRef] [PubMed]

*Journal of Neurophysiology,*76, 2718–2739, https://doi.org/10.1152/jn.1996.76.4.2718. [CrossRef] [PubMed]

*Scientific Reports,*6, 29287, https://doi.org/10.1038/srep29287. [CrossRef] [PubMed]

*International Journal of Psychophysiology,*65, 20–31, https://doi.org/10.1016/j.ijpsycho.2007.02.007.

*Cognitive Affective & Behavioral Neuroscience,*3, 289–299, https://doi.org/10.3758/CABN.3.4.289. [PubMed]

*JASP (Version 0.16.4)*.

*Human Brain Mapping,*39, 3813–3826, https://doi.org/10.1002/hbm.24211. [PubMed]

*Vision Research,*126, 34–51, https://doi.org/10.1016/j.visres.2015.07.008. [PubMed]

*The Journal of Neuroscience,*36, 714–729, https://doi.org/10.1523/JNEUROSCI.2962-15.2016.

*The analysis of sensations and the relation of the physical to the psychical*. Mineola, NY: Dover Publications.

*Journal of Vision,*9, 1–11, https://doi.org/10.1167/9.12.11. [PubMed]

*Journal of Cognitive Neuroscience,*32, 353–366, https://doi.org/10.1162/jocn_a_01485. [PubMed]

*Psychophysiology,*50, 1045–1055, https://doi.org/10.1111/psyp.12082. [PubMed]

*PLoS ONE,*16, e0254361, https://doi.org/10.1371/journal.pone.0254361. [PubMed]

*ELife,*11, e66388. [PubMed]

*Neuropsychologia,*50, 3250–3261, https://doi.org/10.1016/j.neuropsychologia.2012.10.003. [PubMed]

*Cerebral Cortex,*26, 4416–4434, https://doi.org/10.1093/cercor/bhw255.

*Perception,*36, 1305–1319, https://doi.org/10.1068/p5794. [PubMed]

*Perception & Psychophysics,*60, 1101–1116, https://doi.org/10.3758/bf03206161. [PubMed]

*Journal of Experimental Psychology-Human Perception and Performance,*4, 691–702, https://doi.org/10.1037//0096-1523.4.4.691. [PubMed]

*Journal of Neuroscience Methods,*162, 8–13, https://doi.org/10.1016/j.jneumeth.2006.11.017. [PubMed]

*Vision Research,*40, 2621–2644, https://doi.org/10.1016/S0042-6989(00)00110-3. [PubMed]

*Neuropsychologia,*58, 107–117, https://doi.org/10.1016/j.neuropsychologia.2014.04.005. [PubMed]

*Proceedings of the National Academy of Sciences of the United States of America,*102, 3159–3163, https://doi.org/10.1073/pnas.0500319102. [PubMed]

*Symmetry,*2, 1510–1543, https://doi.org/10.3390/sym2031510.

*Attention, Perception, & Psychophysics,*73, 1487–1502, https://doi.org/10.3758/s13414-011-0115-9. [PubMed]

*NeuroImage,*24, 306–314, https://doi.org/10.1016/j.neuroimage.2004.09.018. [PubMed]

*Symmetry,*3, 54–71, https://doi.org/10.3390/sym3010054.

*Psychological Review,*103, 429–456, https://doi.org/10.1037/0033-295x.103.3.429. [PubMed]

*Vision Research,*49, 2754–2763, http://dx.doi.org/10.1016/j.visres.2009.08.015. [PubMed]

*Current Biology,*8, 889–892, https://doi.org/10.1016/S0960-9822(07)00353-3.

*Spatial Vision,*9, 9–32, https://doi.org/10.1163/156856895X00098. [PubMed]

*Psychological Bulletin,*138, 1172–1217, https://doi.org/10.1037/a0029333. [PubMed]

*Perception & Psychophysics,*50, 413–427, https://doi.org/10.3758/BF03205058. [PubMed]

*Perception,*23, 221–236, https://doi.org/10.1068/p230221. [PubMed]