This study was conducted in accordance with the ethical standards set forth in the Declaration of Helsinki and approved by the Institutional Review Board (IRB) of Binghamton University. Informed consent was obtained via Qualtrics from all participants prior to their inclusion in the study.

Control participants were recruited from the Binghamton University research participation pool for credits as part of a degree requirement and participants with VSS were recruited from an advertisement on the Visual Snow Initiative website (5). If a participant reported psychoactive or psychedelic drug use in the past 12 months and had persistent hallucinogenic effects, they were removed from analyses to eliminate overlap with possible Hallucinogen Persisting Perception Disorder (HPPD), since HPPD can mimic VSS (21, 6, 7). Additionally, if a participant declared themself as having VSS, but stated they did not experience visual static, they were removed from the study (as this should not be possible). Note that control participants were informed that the study was directed at VSS; this may have affected recruitment, as participants within a pool self-selected into the study.

The study was administered asynchronously through Qualtrics. Before taking the CVSS, all participants read the ICHD-3 diagnostic criteria, as well as a rendering of visual static, and were asked to self-report if they believed they had VSS. If self-reported VSS, they were additionally asked to self-report if they had been formally diagnosed by a physician. Participants who answered “yes” to this question became our VSS diagnosed (VSS-D) sample, while participants who answered “no” became our VSS undiagnosed (VSS-UD) sample. All other participants were labeled as controls.

Participants were instructed to complete the study alone, in a quiet setting, without distractions. All participants were given a unique, randomly generated URL to the survey. Experimenters were not given access to participant identities beyond the random string generated with each URL. Data was stored on a password-protected university computer. All code and anonymized data is available at Open Science Framework: https://osf.io/x5sru/.

We sought to identify (1) which symptoms show differences between self-reported severity and impairment, (2) the consistency of responses for each symptom, (3) the symptoms that are most useful for identifying VSS presence.

To investigate which items on the CVSS best predict group membership (i.e., VSS as compared to not VSS), we conducted a logistic regression and a principal component analysis (PCA).

A total of 144 participants were recruited, 92 were female (63.9%), with 78 control participants and 66 participants with VSS. We recruited every potential VSS participant who responded to our advertisement posted on the Visual Snow Initiative website (5) if they met the criteria listed above. The control sample was sized to match, but with the caveat that the overall sample should exceed 100. No power analysis was conducted, as the size of the sample was seen as likely to be limited by the recruitment of VSS population participants and little prior work on which to base an estimate of effect size exists, but consideration was given to the approach outlined by Machin et al. (11) in determining the number. One goal of this study was to demonstrate effective differences and provide evidence of effect sizes for future research. The VSS population was divided into two subgroups: those with a self-diagnosis of VSS (“undiagnosed”; VSS-UD) and those with self-report of a VSS diagnosis by a physician (“diagnosed”; VSS-D). Nine participants from the university recruitment pool were included in the VSS sample and placed into the appropriate subgroup as they self-reported visual snow syndrome diagnosis (or self-diagnosis). Eight participants were removed from the sample completely for declaring themselves as having VSS but not experiencing visual static at all, because this symptom is considered to be critical for the diagnosis (per ICHD-3 criteria). Detailed demographics for the full sample are displayed in Table 2.

The average Likert score across all metrics for all 10 symptoms for the total VSS group was 2.27 (SD = 0.79); the average for the control group was 0.74 (SD = 0.43). Figure 1 contains the score density (i.e., frequency of response/total responses), across symptoms. Symptom prevalences are listed in Table 3.

The average CVSS scores for 36 out of 40 metrics were significantly greater for the VSS total sample compared to the control group, p < 0.05/40, with 32 of 40 metrics withstanding significance at p < 0.001/40 (Supplementary Table S1, metric means in Figure 2). The 4 metric average for each symptom demonstrated significant differences in all 10 symptoms between the total VSS sample and the control sample, p < 0.05/10, while 0/10 significant differences between the VSS diagnosed and undiagnosed sample (Figure 3). Our diagnosed and undiagnosed VSS samples did not significantly differ for any metric, and will be treated as one sample moving forward.

To measure internal consistency, we computed the Cronbach’s alpha for each symptom within each group (Table 4), with our optimal range defined between 0.7–0.9. The mean Cronbach’s alpha averaged across groups and all symptoms was α = 0.84 (SD = 0.10). In our VSS sample, the metrics pertaining to afterimages, trails, floaters, and anxiety, and sadness symptoms all fell above our optimal range, which indicated these questions may be too similar to one another.

To assess which metrics best predict group membership, we computed a logistic regression using each individual’s average score for 9 symptoms from Part 1 of the CVSS (visual static was omitted, see section 2.3.1) as predictors. Group membership (control vs. VSS) was the predicted variable; status as -D or -UD was not included. We treated our predictor variables as numerics (as opposed to ordered factors).

Ultimately, we found our model–even with visual static omitted–demonstrated a strong fit for our data, McFadden’s pseudo R² = 0.79. We display below in Table 5 the odds ratios and 95% confidence intervals of each included symptom. In brief, the odds ratio reported indicates how many times more likely someone is to have VSS if they display a given symptom, compared to someone who does not display the given symptom.

For the 40 total metrics (i.e., 4 metrics on intensity, interference questions, and reduction of daily activities for each of the 10 symptoms), we sought to identify which metrics shared similar patterns of responses through the use of a principal component analysis (PCA). The model generated from the PCA indicated that 56.9% of the variance was from two principal components. Factor 1 was most associated with the metrics from the symptoms: visual static, afterimages, blue field entoptic phenomenon, and night vision impairment, while Factor 2 was most associated with anxiety, sadness, and depersonalization/derealization. This pattern continues even when visual static is removed from the analysis (Supplementary Table S2). For visual clarity, we averaged each symptom’s four metric factor loadings in Table 6.

Even in the earliest discoveries into visual snow syndrome, depression and anxiety were found to be common comorbidities (1). More recently, researchers studying the psychiatric profile of VSS found VSS patients displayed heightened rates of anxiety, depression, and depersonalization, relative to neurotypical controls (6). While our results cannot speak to depression, our results do demonstrate (1) higher prevalence rates of depersonalization, anxiety, and sadness in our VSS sample compared to our control sample (Table 3), a (2) much higher Likert scores on all four sadness/depersonalization metrics (intensity, interference with socializing, interference with wellness, and reduction of daily activity) in our VSS sample compared to our control sample (Supplementary Table S1).

Our lack of statistical difference between one sadness and three anxiety metric scores between our VSS sample and control sample may be due to the fact that our VSS sample was significantly older than our control sample, as our control sample was composed primarily of university students, a population known to experience higher self-reported scores of depression than the general population (18). In other words, we compared our VSS sample to a sample that may already experience depression with a higher severity ratings, relative to a true general population. Therefore, we do not discount the possibility that a VSS sample could report higher metric scores of depression and/or anxiety relative to a true general population.

In any case, both our results and previous results speak to a broader mental health concern associated with VSS, and in particular, depersonalization and derealization. We echo the concerns of Solly et al. (6): it is unclear if depersonalization arises out of the same mechanisms that induce VSS, or if the primary symptoms of VSS (i.e., visual static, etc.) give rise to depersonalization. In either case, psychological symptoms–and depersonalization/derealization in particular–deserve additional emphasis in treatment of VSS in order to improve patient quality of life.

Our VSS sample is similar to other VSS samples in the literature in terms of demographics and symptom presence, except the VSS group in this study had a higher prevalence of tinnitus than the group studied by Thompson et al. (7). All the other symptoms were within 10% of the prevalence noted in other studies (1, 2, 7).

One limitation for our study is the use of self-reported physician VSS diagnosis and self-diagnosis of VSS and the age differences between the VSS group and the control group. It is possible that members of our VSS-D sample do not actually have a physician diagnosis of VSS, or that people in our VSS sample more broadly do not have VSS.

A second limitation is the stark age difference in ages between our VSS samples (M = 37.1, SD = 12.7) and control sample (M = 18.8, SD = 0.89). In an ideal study, controls would be age-matched to VSS participants. However, it is worth noting that a cursory series of zero-order Spearman correlations revealed that only afterimage scores and sadness/LOI intensity scores correlated with age (Supplementary Table S3). Still, we acknowledge that we cannot rule out the influence of age on our results.