Jake Spinella

Logically competent agents routinely fail to draw obvious consequences of what they believe, and routinely fail to notice that they hold contradictory beliefs. These problems of deduction have seemed to many to tell against the received view of belief as a unified system, and to favor a fragmentational view, on which belief is a three-place relation between an agent, a proposition, and an index. On this view, beliefs stored in different fragments are inaccessible to one another in token inferences, which explains how problems of deduction are possible. In this paper, I argue that problems of deduction do not motivate fragmentational accounts of belief. This is because there can be intrafragmental problems of deduction, and this leaves the fragmentationist with a dilemma. If fragments are deductively closed, then intrafragmental failures are impossible and the fragmentationist has changed the topic rather than explained the phenomena that motivate the view in the first place. If fragments are not closed, the fragmentationist must appeal to some further notion of inferential availability, a notion which the unified theorist is equally entitled to. I then develop an alternative, the Costly Inference View, on which belief is a two-place relation over a unified store, and failures of deduction are explained by the computational costs of search, retrieval, and evaluation. On this view, belief is a unified system that is very, very large, and the cost of inference over so large a store explains how fragmentation-like phenomena emerge without any architectural commitment to fragments proper.

In this dissertation I argue that the mind has an architecture, and that perception and cognition are functionally distinct systems within it. What makes them distinct is a single property: the belief-independence of perception. A mental process is belief-independent when cognitive states—beliefs, desires, and thoughts—are dispensable to its computational explanation, which is to say that its outputs cannot be semantically traced back to cognitive contents in the way the conclusion of a genuine inference can be traced back to its premises. In Chapter 1, I make this notion more precise. I separate out the functional and representational senses in which cognitive states might be dispensable to a process, and I argue that inferential isolation from cognition is the mark of belief-independence. In Chapter 2, I ask how a psychological system could come to be isolated in this way, and I argue that informational encapsulation is sufficient for it, though not necessary. Defending the claim that perception is encapsulated means taking on the cognitive penetration literature. I go through the cases one at a time and argue that in each, either a purely perceptual explanation of the phenomenon is at least as well supported as the penetration explanation, or the perceptual outputs in question cannot be semantically traced to the cognitive contents said to affect them, so that whatever relation holds between them falls short of the inferential relation that would threaten the architectural distinction. In Chapters 3 and 4, I argue that belief-independence explains features of perception, and so is more than a criterion for locating the perception–cognition boundary. In Chapter 3, I argue that encapsulation makes perception computationally tractable. On the standard view, encapsulation does this by shrinking a hypothesis space specified in advance. I argue instead that it works by leaving hypotheses outside that space out of consideration entirely, and that this same feature makes perception more accurate as well as faster. In Chapter 4, I argue that belief-independence is also necessary for perception to anchor thought to the world. Following a line of thought from Russell and Strawson, I argue that successful singular thought requires acquaintance relations of the kind perception provides, and that the FINST-and-object-file structure underlying visual tracking is one such relation. A perceptual system whose contents were fixed in part by belief would face the same reduplication problems that acquaintance is supposed to solve. Taken together, the four chapters argue that perception and cognition are distinct stems of the mind's architecture, that the belief-independence of perception marks the line between them, and that it is because perception is belief-independent that it can make cognition both tractable and answerable to the world.

Peter Lamarque, in chapter 4 of his 2010 book Work and Object, argues that certain artworks, like musical scores and literary texts, are such that there can be no forgeries of them that purport to be of an actually existing work—what Lamarque calls “referential forgeries”. Lamarque motivates this claim via appeal to another distinction, first made by Goodman, between “allographic” and “autographic” artworks. This article will evaluate Lamarque’s argument that allographic literary works are unable to be referentially forged and will find them wanting. In so doing, the distinction between allographic and autographic artworks (and therefore artforms) will be called into question. In section I, I characterize referential forgery and Lamarque’s definition of allographic and autographic artforms. Section II critically examines Lamarque’s argument against the possibility of referential forgery in allographic artforms. Section III offers a case where it appears that a putatively allographic text’s type membership is sensitive to facts about its causal-intentional provenance. This case serves as pretext for Section IV’s identification of this causal-intentional relation with the sanctioning relation as formulated by Sherri Irvin. On the basis of considerations treated in sections I through IV, section V questions the tenability of the allographic/autographic distinction.

I argue, contra traditional interpretations of James's emotion theory like Antonio Damasio and alternative interpretations like Phoebe Ellsworth and Lisa Barrett, that James is best classified as a functionalist regarding emotion categories. In arguing for this point, I will make four textual claims: (1) James was an important precursor to Basic Emotion Theory (BET) and his theory is best identified as a flavor of BET; (2) James's theory individuates emotion categories by their evolutionary, functional roles; (3) The only necessary condition on something being an emotion is that it is a bodily feeling; and, (4) Contrary to Barrett and Ellsworth, James was loath to offer a definitive list of basic emotions not because he loathed taxonomy but rather because he thought psychology was not yet a natural science with well-defined theoretical categories. I close by arguing that a proper understanding of James's emotion theory defangs some critiques of BET and of Neo-Jamesian theory.

A problem with recent work about the relationship between metaphysics and science, especially in the theorizing of those who identify as “naturalized metaphysicians”, is the spotty, metaphorical characterization of what it means for science to “constrain” metaphysics. The most robust account of scientific constraint on metaphysical theorizing is advanced by James Ladyman and Don Ross in their 2007 book Every Thing Must Go. Ladyman & Ross claim that the only legitimate metaphysical hypotheses are those that unify two previously disparate scientific explanations. I will critique Ladyman & Ross’ account of naturalized metaphysics, and offer an alternative view of naturalized metaphysics as a practice of constructing physically possible models of reality. This account yields a different view of science’s constraint on metaphysics, specifically, that models must be physically possible in order to be of methodological and heuristic use to scientists.