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

2 Models, Mechanisms, and Cognitive Theories

2 Models, Mechanisms, and Cognitive Theories 2.1 Cognitive Autonomy: Models and Multiple Realizability The previous chapter suggested that despite the arguments of Meaning Rationalists and Neurobiological Eliminativists, personal -level psychology is not an autonomous explanatory realm. In fact, there must be an explanatory relationship between neuroscience and folk psychology (as the everyday practice of personal -level explanation is sometimes called). To take an obvious case, amnesia following traumatic brain injury drastically compromises personal-level psychology. The person with amnesia will experience and react to the world quite differently after her injury. Insofar as her beliefs and ability to deploy them in reasoning depend on memories, she will also think about the world differently. Memory loss caused by brain damage seems the most plausible hypothesis here. An important step in completing our understanding of disorders like amnesia involves producing the correct account of the way the brain encodes information acquired in experience and then reconstructs representations of that information when subsequently cued. This is the type of project I’m describing with the phrase cognitive theorizing. The most serious objection to this approach comes not from Neurobiological Eliminativists or Meaning Rationalists, but 22 Chapter 2 from cognitive psychologists who regard cognitive explanation as theoretically autonomous. If cognitive explanation is a genuinely autonomous realm, then it cannot bridge personal and neurobiological explanation. The argument for the autonomy of cognitive explanation is correct in one (and fortunately only one) sense: namely, when applied to cognitive models considered independently of their physical implementation. For example, if we succeed in describing the algorithms involved in parsing sentences or mapping retinal information to representations of the visual scene caused by the retinal array, we can run and test them on a laptop. Thus, in principle, computational processes can be realized in both brains and laptops (Lewis 1980). Not only that, but the same computational problem can be solved in different ways, even in the same machinery, as anyone who has switched word-processing programs can attest. Consequently, individual differences between human brains and the computational solutions they have developed should make us wary of assuming that two people who behave in the same way in response to a stimulus are processing information identically (Gerrans 2003). This is especially so for belief, the paradigmatic, personal-level state. The same belief can be the product of testimony, theoretical reasoning, or bitter personal experience, and in each case the information processing and neural machinery involved will differ. It follows that there is no necessary connection between the computational structure of a problem, its algorithmic solution, and the nature of the mechanisms that implement it (Coltheart 2010; Plaut and Patterson 2010). This line of thinking leads to the idea that cognitive theorists produce models that operate at a level of abstraction that makes it impossible either in principle or in practice to draw Models, Mechanisms, and Cognitive Theories 23 any conclusions about the role played in cognition by neural mechanisms. All that cognitive theorizing can add to our understanding are models that help with the process of correlation but not explanation (Craver 2001). For example, we might learn that faces can be recognized by features or configuration and that damage to different areas of the brain correlates with impairments to the different types of process. But that takes us no closer to understanding how the brain implements feature-based processing and why, therefore, damage to particular cells affects it in different ways (Patterson and Plaut 2009; Arbib 2003). 2.2 Causal Relevance and the Personal Level In order to avoid this problem we need a methodology in which cognitive theories play a different role, namely, pointing toward the way human brains actually perform cognitive processes. One such methodology is that advocated by Carl Craver in Explaining the Brain (2007; see also Craver 2002). Craver’s (2002) work is a sustained argument that explanation in neuroscience involves the identification and description of relations of causal relevance between levels of systemic organization.1 Causal relevance is defined in terms of manipulability and intervention (Woodward 2008a,b). For example, unless we understand the mechanisms of memory at all levels, we cannot understand the effects on memory of interventions at levels as different as phosphorylation within the neuron, calcium channels, temporal lobe lesion, or such paradigmatically personal encounters with the world as armed combat or traumatic early childhood (Lupien et al. 2005, 2007). Events at each of these levels are causally...