Crespi & Badcock's (C&B's) autism-psychosis opposition has analogous diametrical differences in the syndromes' self-monitoring and inner private speech. In autism, self-talk for cognitive regulation is absent or reduced, but in psychosis, it is overt, yet ineffectual. These differences gain detail and explanatory power if understood in the context of cognitive-computational control.

Processing architectures distinguish logic (representations per se) from control (management of dataflow) (Frawley Reference Frawley1997; Kowalski Reference Kowalski2005a; Reference Kowalski2005b; Pylyshyn Reference Pylyshyn1985). Control is clearly seen in cross-domain thinking, either across whole knowledge domains (Carruthers Reference Carruthers2002, Frawley Reference Frawley2002b; e.g., coordination of visual and motor representations) or across levels within domains (Frawley Reference Frawley2002a; e.g., at the phonology-syntax interface in language). Inner speech is an instrument of the former and surfaces as private speech to facilitate and regulate cognition and behavior during complex, online thinking. The structure and effectiveness of private speech correlate with task challenge and performance (though, see Fernyhough & Fradley Reference Fernyhough and Fradley2005).

The issue for the autism-psychosis account is exactly how these syndromes and their regulatory disruptions add to the control mix. Self-regulatory talk exhibits two major control failures:

1. Failure to surface at all – or to do so in reduced form. This is a potentially very serious disorder of cognitive control because the lack of self-talk prevents external speech-scaffolding for online thinking.

2. Surfacing with deficits in structure and so functionally ineffective. Such private speech cannot perform appropriate regulatory functions and hence will run parallel to behavior instead of guiding it; it cannot be developmentally internalized (or “go underground,” as Vygotsky said); nor can it dynamically manifest itself and re-surface appropriately when it is needed for cognitive and behavioral regulation.

C&B cite research to the effect that in autism, inner speech is absent or reduced. But under my reading, the autistic individuals in question did indeed have inner speech, but less private speech. Without inner speech, individuals would lack an internal language for cross-domain thinking, and it is not clear that anyone lacks that, whatever the severity of the disorder. Evidence for the existence of inner, but not private, speech in autism can be seen in the results of Williams et al. Reference Williams, Happé and Jarrold(2008), who have found that, on a direct phonological encoding task, individuals with autism-spectrum disorder deployed inner speech as much as individuals of comparable mental age without autism.

But reduced private speech is another story. Whitehouse et al. Reference Whitehouse, Maybery and Durkin(2006) have found that autistic individuals have truncated articulations in overt labeling tasks. But more curiously, articulatory suppression did not affect autistic individuals' performance in task-switching scenarios (a difficult cognitive requirement), yet it did affect the performance of normal controls. Why?

Normal individuals alternate implicit and explicit self-speech to scaffold behavior under duress. Autistic individuals have inner speech, but are unable to use it overtly in self-guiding tasks. With its repetitive and self-directed behaviors, autism might be expected to encourage private speech. But as a disorder of overactive input filters, autistics use these behaviors to block input, which readily reaches excessive intake in autism. Private speech does not function to block input. So while normal individuals naturally allow inner speech to surface as private speech because they have no blocking problem, autistic individuals naturally suppress it. Additional support for this argument can be found in C&B's observation that neurological studies of autistics reveal abnormal lack of deactivation of default executive functions when attention is redirected: This suggests that persistence of internal executive control is the normal state for the autistic end of the spectrum when it should otherwise deactivate to refocus attention on new input. Put simply, in autistics, inner speech stays inner. A task which enhances private speech suppression should therefore have no effect on the performance of autistic subjects, as was found, because suppression ratifies the default state of autistic inner speech. It would be surprising if suppression of speech did have an effect on autistic individuals and surprising if it did not for controls.

These findings may in fact relate closely to the severity of the disorder. Low-functioning autistics should have unexpressed inner speech, but high-functioning autistics – those who fall along the cline toward “normal,” from Asperger's syndrome to Angelman's syndrome to pervasive developmental disorder not otherwise specified (PDD-NOS) – ought to have private speech, but it should be functionally defective. I know of no study of this predicted difference, but my informal observations of high-functioning autistics, including my own son, bear this out. PDD-NOS individuals' private speech has deficits in metapragmatics and discourse markers, the very features that index cross-domain control because they encode the line between linguistic and nonlinguistic knowledge. Problems with these structural elements are typical in ineffective private speech, whatever its cause (Frawley Reference Frawley1997).

Thus, as autistic severity lessens and the autistic spectrum approaches normal, private speech should also surface in difficult tasks but be functionally ineffective and have predictable structural deficits. Systematic study of this prediction remains to be done. But the larger point is that the autism spectrum has an analogous private speech spectrum, and the effectuality of private speech might be a new measure of the severity of the autism.

In contrast, psychosis has diametrically opposite private speech manifestations, though with similar functional results. C&B do not illustrate psychotic private speech, but it is well known that such speech is elaborate and ostensibly social, yet regulatorily ineffective. Jones and Fernyhough Reference Jones and Fernyhough(2007) observe that verbal self-mediation in psychosis manifests problems in distinguishing self from other and often takes the form of commands. Psychotic individuals hear their own private speech as orders from others.

As with autistic private speech, problems in psychotic private speech evidence deficits in metapragmatics and discourse forms. Normal private speech metapragmatically signals that the speech is self-derived, as can be seen in the prevalence of self-directed questions (“Now what did I do wrong?”). Moreover, the illocutionary force of normal private speech utterances can be other than a command (directive). Self-directed “Blue, blue” in a task could mean “There's the blue one” (presentative), “I need a blue one” (desiderative), and so forth.

Private speech in individuals with psychosis is overt to match their over-social cognition, but relatively undifferentiated in the way the speech encodes and executes control. As a consequence, this speech for thinking cannot “go underground” to serve inner functions. Just as speech for thinking in severe autism keeps the mind inside itself, so the same speech for thinking in severe psychosis keeps the mind away from itself. A detailed linguistic analysis of the private speech in psychotic syndromes would likely show deficits in the same metapragmatic and discourse markers as in autism, but with different functional effects. A similar diagnostic cline to “normal” should also present itself, with movement from severe to less severe, as measured by private speech and signaled by the gradual reduction and internalization of private speech and concomitant increase of functional cognitive control.

Autism and psychosis are diametric control syndromes – true private speech opposites. Each moves toward normal by the other's failures – autism via increased explicitness and psychosis via increased implicitness.