Formal Semantics and the Psychology of Reasoning: Building new bridges and investigating interactions

Research in linguistic semantics in the past forty years has produced sophisticated mathematical models that represent the meanings of natural language utterances and explain how meanings relate to one another to form entailment patterns. At the same time, research on human reasoning within psychology has discovered a wealth of fallacious inference patterns, establishing that human reasoning is fallible in highly predictable ways. The psychological study of reasoning has been characterized by extensive experimental work analyzed in terms of often informal theories focusing on the processes of reasoning. Although data on reasoning are almost always collected by means of linguistically presented stimuli, psychology has largely disregarded issues having to do with the interpretation of the linguistic signal. On the other hand, linguistic semantics has a longstanding tradition of focusing on normatively sanctioned inference patterns, and it has so far largely ignored fallacies. The two domains of research overlap significantly, but they have progressed almost completely in parallel, with next to no interaction. This dissertation contributes to furthering the connection between the two fields. The work herein makes contributions of two kinds toward this goal.;Mapping sentences to mental representations.;At least since the work of Davidson (1967), most of the research done in philosophy of language and in linguistic semantics has focused on investigating mappings between linguistic expressions and truth conditions (or constitutive elements of truth conditions in the case of sub-clausal expressions). This program has been successful in producing interesting and insightful research on language. However, a different program for semantics can be envisaged that will potentially help bridge the gap between semantics and psychology. Instead of mapping expressions to truth, this program maps expressions to mental representations, whose properties can be independently discovered using the methods of psychology. The past fifty years of research in psychology have shown that mental representations do not capture truth entirely faithfully, and that the mind does not put at our disposal perfectly truth-preserving operations on mental representations. There is no reason to expect, and no reason to assume, that the mental structures in charge of representing the world should do so completely faithfully. The first part of this dissertation pursues this program, by building on the fact that certain non-truth- conditional semantics (inquisitive semantics and some of its relatives) share interesting properties with an entirely independently proposed theory of human reasoning from psychology (mental model theory). Inquisitive semantics and mental model theory both posit non-classical, non-truth-conditional representations for sentences containing disjunctions, and this dissertation combines the two approaches in one unified theory. In the process, independent and reciprocal justifications appear for inquisitive semantics and mental model theory. The success of mental model theory in predicting observed inference-making behavior is predicated on a theory of mental representations and the relation between sentences of natural languages and those representations. This theory of mental representations bears striking connections to the models of inquisitive semantics, which has independent and strictly linguistic motivations. This work demonstrates how non-trivial convergences between linguistics and psychology are not only possible in principle, but can be attained in practice by looking at the space of theories already on the market in both fields, and seeing what theories can be combined in simple and insightful ways.;Exploring the line between reasoning and interpretation.;The second part of this dissertation explores a less original point of contact between semantics and psychology, but an absolutely indispensable and vastly understudied one. The psychology of reasoning overwhelmingly collects its data by means of experiments with linguistically presented stimuli. Experimental subjects then have two tasks to perform. First they must decode the linguistic signal using their faculty of language, then they must manipulate the representations they arrive at using their faculty for reasoning. The two processes are distinct, and most likely also different.;In the second part of this dissertation I explore two cases of (superficially) fallacious inferential behavior for which there exist interpretation-based explanations. First, I look at illusory inferences from disjunction and formulate an interpretive account of these (apparent) fallacies. I also present a puzzle for the interpretive theories used earlier, arguing that a particular technical aspect of those theories is at odds with commonly accepted facts about psychology. The last chapter of this dissertation focuses on one of the most well-known instances of fallacious reasoning, the conjunction fallacy. I argue that the conjunction fallacy also has an alternative explanation in terms of interpretation. I conclude with first steps toward a paradigm that can separate the predictions of reasoning-based and interpretation-based accounts of the conjunction fallacy.