Limitations and suggestions for reporting interactions with
spatial frequency
A limitation is that we do not know why groups differed in acuity; thus,
it is not yet clear whether it is appropriate to match groups on acuity
in the way we have done. On the one hand, at least one study has found
that people with schizophrenia less often visit an optometrist (Viertiö
et al., 2007; see also, Silverstein & Rosen, 2015), suggesting that
poor contrast sensitivity at higher spatial frequencies may arise from
not having appropriate eyewear (Zemon et al., 2020; Keri et al., 2002).
This possibility should be taken seriously because optical blur within
the “normal” 20/20 range can diminish sensitivity to Gabor elements
with a frequency as low as four cycles/degree (Keane et al., 2022). On
the other hand, people with anti-NMDA receptor encephalitis– a
condition that symptomatically resembles schizophrenia and that attacks
the same receptor that is commonly implicated in schizophrenia (Beck et
al., 2020; Singh et al., 2022)–have worse acuity than matched controls,
especially for more severe bouts of the infection (Brandt et al., 2016).
Thus, either uncorrected refractive error, neural factors, or some
combination could worsen contrast sensitivity deficits at higher spatial
frequencies in schizophrenia. The same conclusion may hold for other
special populations. For example, individuals of advanced age may have
impaired acuity due to a combination of neural factors and optical
under-correction (La Fleur & Salthouse, 2014; Liou et al., 1999).
Interactions with spatial frequency can be properly reported in a few
ways. First, as may already be obvious, log-transforms can approximate
homoscedastic, normal distributions, and generalized estimating
equations may provide an even better way to model such data (Prekár &
Brabec, 2018; Feng et al., 2014). Boxplots or histograms could reveal
unexpected data distributions (such as those in Fig. 1B). To avoid
confounds with optical blur, an optometrist could measure and correct
acuity beforehand so that all subjects have their best corrected visual
acuity at the time of testing (BCVA). Note that some investigators
mistakenly use the term “BCVA” to refer to habitual acuity rather than
optimal acuity (Elliot, 2016). However, only the latter can remove
confounds associated with refractive error since many individuals with
contacts or glasses will have out-of-date prescriptions. If groups
cannot be matched on BCVA, this would be informative as it would
indicate a neural origin to poor acuity and argue against any further
matching based on acuity. If providing optimal correction to every
subject is impractical, refractive error could instead be quantified
with an auto-refractor. Portable auto-refractors generate spherical
equivalent refractive error estimates that are similar to those of
subjective refraction and retinoscopy with or without cycloplegia
(Ciuffreda & Rosenfeld, 2015). In this approach, subjects with
excessive refractive error could be excluded (e.g., >0.5
diopters, roughly equivalent to 20/30 vision), and subject groups could
then be matched on refractive error in a post-hoc analysis, if not in
the overall sample. Either way, refractive error must be considered
before interpreting interactions with spatial
frequency.