Further Evidence Is Needed to Support the Perceptual Theory of Schizophrenia Authors: Cemal Demirlek, Paulo Lizano Published: 2026 In a recent article published in Biological Psychiatry: Global Open Science, Adámek et al. (1) propose the Perceptual Theory of Schizophrenia (PerTh), a reframing in which instability of retinal dopamine signaling acts as an upstream initiating mechanism that seeds aberrant salience and, through thalamocortical propagation, compensatory synaptic sprouting, and complement-mediated hyperpruning, culminating in an increased risk for schizophrenia. It examines literature demonstrating that the retina and visual system are altered in schizophrenia; assembles disparate observations into an explicit, falsifiable cascade within a developmental timeline; and insists that the retina, a noninvasively accessible part of the central nervous system, warrants a mechanistic center rather than an annotation. PerTh is an ambitious and integrative synthesis, drawing together strands of research that are usually treated in isolation in schizophrenia. It also yields concrete, testable predictions and points to specific experimental avenues, from longitudinal pattern electroretinography in at-risk youth to induced pluripotent stem cell (iPSC)–derived retinal models. The authors (1) are careful to position PerTh as complementary to, rather than a replacement for, dopaminergic, glutamatergic, neuroimmune, and neurodevelopmental accounts. It is plausible that the visual system might offer a uniquely tractable framework for psychosis research. However, PerTh needs to be more precisely mapped to distinguish between which hypotheses are ready for testing and which are not fully supported by existing data. Meta-analytic evidence for retinal alteration in schizophrenia is robust, but effect sizes are generally small-to-moderate (2) and nonspecific. Retinal changes are broadly transdiagnostic, appearing across psychiatric and neurologic disorders. Moreover, much of this retinal signal is entangled with, and partly explained by, cardiometabolic and broader biopsychosocial factors such as diabetes, hypertension, and social deprivation (3,4), making it difficult to determine whether retinal abnormalities are upstream contributors, parallel manifestations of shared liability, or downstream systemic markers, which are similar issues encountered in neuroimaging studies. On the other hand, support for a retinal origin includes enrichment of schizophrenia polygenic risk genes in retinal amacrine cells (5), indexing genetic liability expressed in a cell class common to the retina and the brain, which frames the retina as a proxy for synaptic pathology. Enrichment of shared liability in a particular cell type is not evidence that this cell type initiates the disorder, since synaptic pathology can be more widespread in schizophrenia. Thus, a mechanism that claims to be disease specific cannot rest primarily on markers that are transdiagnostic, and the PerTh could be better framed as a candidate mechanism for a psychosis-relevant dimension. This is not a failing unique to PerTh since nonspecificity is the norm across contemporary psychiatry, where most putative risk factors and biomarkers, from childhood trauma to inflammation and cognitive impairment, cut across diagnostic boundaries. The tension is that PerTh attempts to advance a disease-specific mechanism in schizophrenia while drawing on markers that behave transdiagnostically, and this is compounded by diagnostic categories themselves being porous, with psychiatric disorders sharing common genetic and biological substrates that blur the symptom-based boundaries on which any specificity claim depends. Importantly, PerTh’s central construct—chronically unstable retinal dopamine—cannot be measured directly in living humans. Retinal dopamine is inferred only indirectly, from the electroretinogram, and that inference has limitations. The same electroretinographic changes attributed to a hypodopaminergic retina are equally attributable to other causes, co-occurring with and correlating with insulin resistance and childhood trauma in early psychosis (3). Indeed, retinal dopamine content can remain unchanged even when central dopamine is markedly elevated or reduced, despite electroretinographic alteration (6). Moreover, PerTh does not simply claim that retinal dopamine is altered; it claims that the signal is unstable and noisy over time. Yet, current measures only capture the average level of these indirect signals, not their moment-to-moment variability. This distinction matters, because in PerTh, it is the instability itself, rather than the average level, that drives the disease. Thus, when a theory’s initiating variable is unmeasured, and its available proxies are confounded and point to possible central or metabolic sources, instability remains a hypothesis rather than an observation. Future work should quantify temporal variability under controlled luminance, adaptation, medication, and metabolic conditions, rather than relying only on cross-sectional mean differences. We nonetheless agree with the authors that emerging approaches, including iPSC-derived retinal cells from patients with schizophrenia, are promising avenues for moving beyond association toward causal testing of the PerTh. Additionally, PerTh leans heavily on 2 anchors. The first is the reported protection of congenital cortical blindness against schizophrenia. The authors (1) acknowledge the statistical fragility of this finding. Moreover, if degraded but present retinal signaling were pathogenic, then many of the patients with peripheral eye disease would show elevated, not comparable, rates of schizophrenia, which is not observed. Real-world data demonstrate that the linkage of ophthalmologic disorders to psychosis is again transdiagnostic or tied to brief psychotic disorders rather than being schizophrenia specific (7). The second anchor is the fact that auditory hallucinations are far more common than visual ones in schizophrenia. To explain this, PerTh must add 2 additional assumptions: The brain adapts to the visual problem so patients never notice it, and auditory hallucinations arise from a separate self-monitoring failure. A central-noise model explains the same pattern more simply, without these add-ons, because brain-based noise is already strongest in the auditory system, and overly strong prior expectations most easily distort speech, which is rich in information (8). Even if we were granting a genuine retinal contribution, its explanatory scope is likely narrow, given that schizophrenia is biologically and clinically heterogeneous and its cognitive impairments precede psychosis and are arguably among its earliest and most disabling features. Also, brain structural signatures can be global with great heterogeneity rather than being confined to a specific pathway. A model built on peripheral sensory lesions struggles to account for disorganized thought and language—capacities that are human specific and not obviously downstream of retinal gain control. For these reasons, PerTh may be best understood not as a theory of schizophrenia per se but as a candidate contributor to a subset of patients with psychosis spectrum or psychotic-like experiences. Finally, 3 of the proposed observations favor a central, rather than peripheral, origin of psychotic symptoms. First, Mendelian randomization findings in schizophrenia challenge a simple 1-way retina-to-brain account and are at least compatible with central-to-retinal effects, shared liability, or bidirectional causation (9). Second, although the evidence is preliminary and heterogeneous, the earliest detectable retinal changes in clinical high risk and first-episode cohorts are increases (or no differences) in thickness that covary with cognition and antipsychotic dose (10), rather than a consistent structural correlate for PerTh’s proposed functional retinal instability. Third, and most directly, across large prodromal and early-psychosis samples, attenuated delusions reliably emerge before attenuated hallucinations, and delusion severity falls as hallucinations appear (8), which sits uneasily with a visual lesion–first model. Overall, we agree with the authors that the retina is an accessible readout of shared central nervous system liability, providing a strong rationale for longitudinal, multimodal, adequately powered studies incorporating prospective retinal imaging across the lifespan, harmonized multicenter protocols, and study designs capable of determining whether retinal changes precede, parallel, or follow brain changes. PerTh’s lasting service may be to compel the measurements that can finally distinguish an upstream trigger from an accessible marker that could be used to predict which individuals will respond to which particular therapy. For now, the retina is best regarded as a window onto neuropsychiatric disorders and their comorbidities rather than their origin, emphasizing the need for high-quality longitudinal, multimodal, and treatment studies before its role can be interpreted as causal. Acknowledgments and Disclosures PL serves as a consultant to Bristol-Myers Squibb. CD reports no biomedical financial interests or potential conflicts of interest. References