My research focuses on the philosophy of language and foundations of cognitive science. My main focus is the nature of linguistic meaning, the operations that produce complex meanings and thoughts, and their interaction with context, memory and general reasoning. Those topics overlap with central topics in philosophy of mind, including questions about the structure of concepts and their role in judgment and decision-making. I also work on the foundations of cognitive science and neuroscience, focusing on reductionism, multiple-realizability, and the nature of scientific inference and inter-level explanation. I briefly describe my current and future work in each of these topics below, and include some selected papers. 

language, meaning, and Natural logic  

A unique capacity of human cognition is that it can take simple meanings and concepts and combine them to form novel thoughts of unbounded variety and complexity. A fundamental questions in philosophy and the cognitive sciences concerns the nature of the combinatorial systems that underlie the productivity of human language and thought. Focusing on the language system, key related questions include to what extent this system is like a formal computational system, and how it interfaces with general knowledge and reasoning. My work in this area focuses on the ways in which the language system is like a logical language. I have argued that human languages are strictly compositional. I have also argued, more radically, that they include a largely classical natural logic or inferential system which automatically filters out many expressions that are strictly well formed but are informationally useless.  

Selected papers: 

MEANING, CONTEXT, and multidimensional semantics

Traditionally, work in semantics and philosophy of language models the meaning of expressions as that which determines their extension. The meaning of complex expressions is compositionally determined from that of their parts. Recently, I have developed a multidimensional theory according to which meaning includes more than just extension-determining components. These additional dimensions include information about the extension of terms, but unlike Fregean senses they do not directly determine their extension. These additional dimensions are still part of linguistic meaning because they can determine the compositional contribution of terms to complex expressions of which they are parts. I argue that my multidimensional semantics can help us understand how linguistic meaning systematically interacts with context. I also argue that it helps us understand how linguistic meaning interacts with perception and inference. I show that this approach can be implemented in fully compositional and type-driven formal semantic theories, and give accounts of some puzzling kinds of complex expressions. 

Selected papers: 


Multidimensional theories (as described above) can inform the study of concepts and social cognition. For example, research on implicit bias assumes that concepts encode salient or statistical associations between features and categories, e.g., stripped is encoded as salient and typical of tigers. Similar structures are thought underlie stereotypes, such as when the negative feature lazy is used to represent entire social groups. However, concepts also encode additional information, esp., how conceptual features ‘depend on each other’, which determines their degree of centrality. In a series of collaborative papers, we have argued that these non-associative, dependency networks can encode significant biases. These biases are undetectable by measures of saliency/typicality, and have been largely ignored by research in social psychology and the philosophical accounts of implicit bias that rest on them. 

Selected papers: 


The scientific study of cognition---incl., language, memory and reasoning---is being increasingly pursued with neuroscientific tools and sophisticated inferential techniques (e.g., neuroimaging + machine learning). Some argue that this approach rests on objectionable forms of mind-brain reductionism. Critics even labeled the approach a `new Phrenology’, arguing that it assumes that most cognitive functions are implemented in particular locations of the brain. Clearly, we cannot yet reduce most complex cognitive processes to their neural implementations. Under what conditions can neuroscience advance our understanding of cognition? Most of my research in this area focuses on this question, arguing that the kinds of inferences currently used to draw conclusions about the mind from data about the brain do not depend on objectionable forms of reductionism or functional locationism. Indeed, recent techniques—esp., machine learning algorithms---can extract information about cognitive function from widely distributed patterns of activation. In forthcoming work, I explore the limits and prospects of the increasingly influential use of machine learning tools to draw inferences from neural data to cognitive functions.

Selected papers: