Embodied Simulation and Emotion Recognition Deficits in Schizophrenia — Epoche C2
What the deficit is, and how big Asked to name the emotion on a photographed face, people carrying a diagnosis of schizophrenia are wrong more often than demographically matched controls by close to a full standard deviation; this essay asks what that failure is a failure of . Kohler and colleagues (2010) pooled the facial-emotion perception literature and reported a large group difference, of the order of $d \approx 0.9$ in the pooled standard deviation metric. It is worth translating that figure, because "large effect" is a label rather than a fact. If the two distributions are approximately normal with equal variance, the probability that a randomly drawn patient scores below a randomly drawn control is $\Phi(d/\sqrt{2}) = \Phi(0.636) \approx 0.74$. So in roughly three comparisons out of four the patient does worse — a substantial and clinically real shift, and also one that leaves a quarter of comparisons running the other way. Any account of the deficit has to be an account of a distributional shift, not of an abolished capacity. Two further features of the empirical picture constrain what can be said. The deficit is not confined to static photographs: it appears for dynamic displays and for vocal prosody. And it is not confined to schizophrenia: comparable impairments are reported in bipolar disorder, in autism, and after bilateral amygdala damage. That last point bears directly on the thesis defended below, and it is the reason the phrase "schizophrenia-specific", used without qualification in the earlier version of this essay, has to be withdrawn and replaced with a claim that can actually be tested. The representationalist reading, and what it commits one to The account this essay opposes treats emotion recognition as a two-stage inference. Stage one extracts perceptual features from the face — the configuration of brow, eyelid and mouth. Stage two matches that feature vector against stored, amodal categories: representations of anger , fear and the rest that are not themselves in any sensory or motor format, and whose content is fixed by their inferential role. A deficit is then located in one of the two stages, or in the mapping between them. Set out this way, the account makes a definite prediction, and it is the prediction that gives the dispute empirical teeth: whatever is wrong should be dissociable from the state of the observer's own body. Storage and retrieval of an amodal category has no more to do with the observer's viscera than storing the category prime number does. A patient with intact perceptual encoding, intact stored categories and an unresponsive body should recognise emotions normally. The embodied account denies exactly this. Whose theory embodied simulation is, and a correction The earlier version of this essay attributed embodied emotion simulation to Shaun Gallagher, and specifically to How the Body Shapes the Mind (2005). That attribution inverts his position and has to be corrected plainly, because the essay's argument is stronger once the two figures are kept apart. The theory and the term are Vittorio Gallese's. On his account (Gallese 2005), observing another's action, emotion or sensation automatically reactivates, in the observer, a subset of the motor, visceromotor and somatosensory circuitry that would be active were the observer performing the action or undergoing the state. The reuse is sub-personal, fast and non-inferential; what it yields is not a proposition about the other's state but an experiential acquaintance with it. Gallese's evidence base is the mirror-neuron literature in macaque premotor and parietal cortex extended, by analogy and by human imaging, to the anterior insula for disgust and to somatosensory cortex for touch. Gallagher, by contrast, is one of simulation theory's most persistent critics. How the Body Shapes the Mind devotes its treatment of social cognition to arguing that neither theory-theory nor simulation theory is needed: on what he calls interaction theory, we perceive emotion directly in expressive bodily movement, and no intervening replication of the other's state is required, because the movement is not evidence for a hidden inner state but is itself the visible shape of the emotion. Gallagher and Gallese are both embodied theorists; they disagree about whether simulation is the mechanism. Treating them as a single position loses the only live alternative to the view being defended. Gallagher's book nevertheless supplies something the argument genuinely needs, and it is a different thing: the distinction between body image and body schema . The body image is a system of perceptions, beliefs and attitudes in which one's own body figures as an intentional object — what one notices and thinks about one's body. The body schema is a system of sensorimotor capacities that operates below the threshold of perceptual monitoring, keeping posture and movement organised without being attended to. The distinction is doing real work below. One more correction to the framing. The earlier text opposed "embodied simulation" to "representational" accounts as though the contrast were between representation and its absence. Gallese's mechanism is a reuse of neural representations; it is representational in any sense broad enough to be uncontroversial. The contrast that carries empirical content is between amodal representation — a format shared by no sensory or motor system — and modality-specific reuse , in which the vehicle of the understanding is the observer's own bodily machinery. Stated that way, the dispute makes differential predictions instead of trading in a metaphysical contrast that no experiment could settle. What goes wrong with the body in schizophrenia The embodied hypothesis needs an independently characterised bodily anomaly to point at, or it is merely the negation of the representationalist story. The best-developed candidate is not a motor deficit but a disturbance of self-experience, and Sass and Parnas (20