The Measure of Madness: Philosophy of Mind, Cognitive Neuroscience, and Delusional Thought

8 The Sense of Agency, Lost and Found: Experience and Thought in Schizophrenic Delusion

8 The Sense of Agency, Lost and Found: Experience and Thought in Schizophrenic Delusion A book on delusions would be incomplete without a discussion of schizophrenic delusions—the most common, clinically significant , and, in some ways, most baffling forms of delusion. (How can a person genuinely experience episodes of thinking that originate in someone else’s mind?) Previous chapters touched on aspects of schizophrenia in discussions of neurobiological eliminativism, the default system, doxastic and narrative theories of delusion, and dopaminergic mechanisms of salience. However, they did not attempt to provide a systematic explanation of schizophrenic delusion. That is the aim of this chapter, which focuses on delusions of alien control and the (I will argue, related) phenomenon of delusions of thought insertion. The standard doxastic explanation of these delusions is that they are beliefs generated to explain the loss of “sense of agency,” a subtle form of bodily awareness that arises in the process of controlling action. When this sense of agency is absent, the subject performs an action but does not experience it as self-initiated. This can be true even in cases where she explicitly intends the action. For example, in one experiment, subjects asked to move a joystick did so, which indicates that they explicitly intended the movement, but reported vivid sensations 164 Chapter 8 of not being the agent of the action (Mlakar, Jensterle, and Frith 1994; Spence et al. 1997). Delusions of control are correlated with abnormally high levels of activation in the right inferior parietal cortex (Behrendt 2004; Blakemore et al. 2000; Spence et al. 1997; Ganesan, Hunter, and Spence 2005; Langland-Hassan 2008; Spence 2002). Spence has observed ruefully that recent high-tech studies establishing this correlation merely confirm the prediction made by Angyal in 1936 that the inferior parietal lobe would be implicated in passivity experience in schizophrenia (Angyal 1936). The initial aim of the chapter is to explain this correlation using a predictive coding model of motor control. The ultimate aim is to transform that model to a neurocognitive theory by considering evidence that the neural circuitry implicated in delusions of control actually implements a version of that model. The essential idea is that sensorimotor systems “predict” sensory consequences of movements initiated by the organism. This prediction allows an organism to correct errors by comparing predicted to actual sensory inputs. Crucial dimensions of comparison for action control are between predicted consequences of motor commands, proprioceptive (the sense of bodily location ), and visual feedback (Blakemore, Wolpert, and Frith 2002; Hartmann et al. 1984; Synofzik et al. 2010). The predictive coding framework generates a model cognitive architecture for motor control, which can be represented in figure 8.1 (the forward model, as it is known). The basic idea is that a goal-directed system computes the difference between the goal state and its current state (the inverse model). This inverse model generates an intention to produce a movement that gets the system from the current state to the goal state. That intention generates, via a downward cascade of cognitive processes, a The Sense of Agency, Lost and Found 165 motor instruction whose execution realizes the intention. Execution of the motor instruction produces feedback in the form of sensory inputs. That feedback is “compared” to a “prediction” of the sensory consequences of the motor instruction, and any discrepancy generates an error signal, which is then used to generate a new set of motor commands to correct the movement. The process iterates until the error is canceled. Importantly, the system is hierarchical, with each level in the hierarchy using the predictive coding strategy. Error signals and corrections propagate up and down the hierarchy. My intention to park the car, for example, is Figure 8.1 Movement actual state sensory feedback Desired state Estimated actual state Controllers (perception to movement) Goal Predictors (movement to perception) Predicted state Affordances 166 Chapter 8 ultimately realized at low levels by minute adjustments of pressure on the steering wheel and accelerator (Pacherie 2000). Failures at low levels that cannot be automatically corrected at those levels produce error signals that propagate up the hierarchy. Different layers in the control hierarchy exploit predictive information in different ways. Automatic sensorimotor loops couple the organism with the environment by comparing actual to predicted proprioceptive feedback consequent on motor instructions . If the movement is unsuccessful, the mismatch creates an error signal that is sustained until the movement is corrected. This process of comparison...