The linguistics of schizophreniathought disturbance as language pathology across positive symptoms

For over a century, psychiatry has treated schizophrenia as a disease of thought. The language patients produce — the fractured speech, the voices, the fixed impossible beliefs — has been seen as a window onto something broken beneath the surface. A readout. A symptom. Never the broken thing itself. Wolfram Hinzen's argument flips that frame entirely: language isn't displaying the pathology; language is the pathology. This reversal has consequences for everything — how we understand each symptom, what brain imaging is actually showing us, and where treatment should aim. The standard view has a name. Hinzen calls it the Cartesian conception of language: the idea that language is a vehicle for communicating pre-formed thoughts, that thought and language are separable, and that the real action happens upstream of words. According to this framing, a patient who believes they are Jesus Christ has a disorder of thought, not of language. Frith made this explicit — "Thinking is a private matter, whereas language is arguably the most important method we have for communicating with others." Bleuler said something similar: schizophrenia's primary symptoms find expression in language, but the abnormality lies not in the language itself but in what it has to say. The Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition — DSM-5 — criteria reflect this inheritance. Three positive symptoms qualify for diagnosis: hallucinations, delusions, and disorganized speech. The received interpretation is that disorganized speech results from disorganized thought, not that it constitutes it. The problem with this framing is that it leaves a conceptual hole. If thought is the broken thing in schizophrenia, what exactly is thought made of? Hinzen's answer, developed with Joana Rosselló and drawing on earlier work with Sheehan, is that human thought — the kind that can be true or false, shared between people, embedded inside other thoughts — is not separable from grammar. It is grammar. This is the un-Cartesian claim, and it has a precise empirical content: there is a one-to-one correlation between human-specific meaning and forms of grammatical organization. Words in isolation do not carry truth value. Only when grammar assembles them into sentences do propositions emerge — contents that can be true or false, denied, questioned, and reasoned about. Grammar doesn't simply dress thought up; it generates the kinds of thought only humans have. Two dimensions of meaning do the heavy lifting in this model. Referential meaning is the machinery that transforms a conceptual item — like the concept HOUSE — into a definite referent, something you can point to with a pronoun, a name, or a deictic expression like "this one." Propositional meaning is what grammar builds when it combines a subject and a predicate: a claim about the world that can be evaluated. Running through both is what Hinzen and Rosselló call the deictic frame — the grammatical apparatus of person. Every utterance is implicitly anchored in a triangle: I, the speaker; you, the addressee; and the world, the third-person domain of things talked about. That triangle is not decorative. It is the structure that makes language a shared human practice rather than private noise. With that scaffold in place, the three positive symptoms of schizophrenia fall into three distinct corners. Hinzen and Rosselló model language as a triangle of speech perception, speech production, and semantic content — and argue that each major symptom corresponds to a breakdown at one vertex. Auditory verbal hallucinations are a disorder of speech perception. In schizophrenia, these are "perception-like experiences that occur without an external stimulus" — most commonly, voices speaking in words or short sentences. A patient hears their own inner speech but does not recognize it as their own. Hinzen and Rosselló explain this as a deictic shift: a first-person thought like "I am weak" gets represented as though someone else is saying it, moving from first person to second or third. The patient loses ownership of their own voice because the grammatical person anchoring has failed. Baethge and colleagues confirmed in a sample of four thousand nine hundred seventy-two hospitalized patients that auditory experiences are overwhelmingly the modal form of hallucination. The subtypes documented by McCarthy-Jones and colleagues map onto linguistic distinctions — commanding voices use imperatives, some voices speak in the hearer's own first person, others replay memorized speech — which is exactly what you would expect if the disruption is grammatical rather than generically perceptual. Formal Thought Disorder sits at the production corner. Speech runs without its normal feedback control. Hinzen and Rosselló's review illustrates this with clinical examples drawn from McKenna and Oh: answers that never address the question, sentences that derail from topic to topic, invented words like "plausity," a patient describing a lamp as a "night-illuminating object," or producing the sentence "A conclusion was a French professor." The surface grammar can be intact. What's missing is the propositional point — the discourse-level organization that gives speech a goal. Morice and Ingram achieved a differential diagnostic accuracy of ninety-five percent for schizophrenia versus mania and controls based purely on reduced syntactic complexity and syntactic or semantic errors in speech samples. That number matters. It means a linguistically trained listener, armed with the right measures, can distinguish these patients from the outside at near-perfect accuracy — which is not what you would expect if language were merely a secondary surface effect. Delusions occupy the content corner. They are linguistically articulated propositions held with fixed, incontrovertible certainty — and they show the deictic frame failure in the opposite direction from hallucinations. Where hallucinations push first-person content outward to a perceived other, delusions pull third-person predicates inward: "I am Jesus." Hinzen and Rosselló note that such delusional sentences are characteristically non-embedded, copular, and present tense. An embedded formulation — "I think I am Jesus" — is not a delusion in the technical psychiatric sense, because embedding signals uncertainty, a recognizable gap between the speaker's perspective and reality. Schneiderian first-rank symptoms like thought insertion and withdrawal fit the same pattern: the patient loses the grammatical boundary between their own first-person domain and others'. The neuroimaging fits this architecture. Sans-Sansa and colleagues found that Formal Thought Disorder is associated with gray-matter reductions in Broca's area and the superior temporal gyrus — Wernicke's area — together with the ventromedial prefrontal cortex and orbitofrontal regions. For auditory verbal hallucinations, imaging shows increased activity in temporal-lobe neocortex, often bilaterally. Plaze and colleagues reported that patients hearing hallucinations show decreased activation in the left superior temporal gyrus when listening to external speech — and concluded that "auditory hallucinations compete with normal speech for processing sites within the temporal cortex." That competition is exactly what the perception-vertex model would predict: the same neural territory handling external speech is being occupied by internally generated voice. The overlap with Theory of Mind — often treated as a competing explanation — turns out to be predicted rather than problematic. The ventromedial prefrontal cortex is implicated in both Formal Thought Disorder and classical Theory of Mind tasks. Ferstl and colleagues found strong overlap between what they called the extended language network and regions implicated in Theory of Mind processing. Hinzen and Sheehan treat this as a corollary, not a complication: the grammatical machinery of person and propositional embedding is precisely the machinery used to model other minds. They are not separate systems occupying the same neural real estate by coincidence — they are, on this account, the same system. The model also explains a persistent puzzle in the neuropsychology literature. Dibben and colleagues' meta-analysis found that the pooled correlation between executive-function impairment and positive "reality distortion" symptoms is close to zero. Standard neuropsychological batteries keep missing something. Hinzen and Sheehan's answer is that delusions and the other positive symptoms target propositional content and referential structure, not the processing components those batteries are built to measure. The solution lies in grammar-based tests: probes targeting deictic and definite noun phrases, pronoun tracking across discourse, and clausal embedding. These should outperform conventional measures at revealing the language-cognition link in schizophrenia. If the model holds, the clinical stakes are real. Language circuitry — Broca's area, the superior temporal gyrus, the extended language network — becomes a primary neurocognitive target, not an afterthought. Cognitive remediation programs that specifically address phonological and referential-discursive components might do more than generic executive-function training. The question of what is broken, at the most basic level, gets a different answer than the one psychiatry has held for a century. Not thought, and then language. Language itself — the grammatical machinery that makes human thought possible — and the thought disorders as consequences. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.