Our lab investigates the relation of language and thought, in computational terms.

How is the structure of the human mind reflected in the world's languages? Are there concepts that are universally available to speakers of all languages? Or does language create and shape our thoughts, such that speakers of different languages think about the world in fundamentally different ways? We probe questions such as these, and do so in an integrative fashion that seeks to move beyond simple nature vs. nurture conceptions of the mind. We seek to understand which aspects of the mind shape language, and which are shaped by it - and what general principles govern the interaction of our language and our thoughts.

Computational modeling is an important means to this end. Thinking in formal computational terms provides a level of generality, and simultaneously a level of detail, that can accommodate empirically complex data in a conceptually straightforward way. We also draw on behavioral experiments, and on data from many languages. Our work is broadly collaborative, and we maintain active working relationships with researchers at many different institutions. Below we describe several specific streams of research we have been pursuing.

Semantic typology of the lexicon

Words in different languages often partition the world differently. For example, some languages use a single color term to cover both green and blue, and some languages have a kinship term that can be used, by a woman, to refer to her maternal grandmother or to her daughter's children of either sex. Yet this variation is constrained, and unrelated languages often have words with the same meaning. Why do words in the world's languages carry the particular variety of meanings they do? One major line of our research suggests a simple answer: across languages, words have meanings that support efficient communication. We have tested this idea computationally against cross-language data in the semantic domains of color, space, kinship, and number, and have found that it accounts for universal tendencies in these domains, as well as for substantial cross-language variation.

The Sapir-Whorf hypothesis

The Sapir-Whorf hypothesis holds that our perception and thoughts are shaped by our native language. Our research has found evidence that is consistent with this view - including evidence that the Sapir-Whorf hypothesis may be supported in the right visual field more than the left, perhaps as a result of the functional organization of the brain. At the same time, we have also found evidence consistent with the opposing view that there are universals of cognition. We sometimes find evidence for both views in the same set of data - suggesting an interesting interplay between universal and language-specific forces in cognition. We have proposed that some of the controversy surrounding this hypothesis may be resolved by considering it through the lens of probabilistic inference.

Language learning

An influential view in cognitive science holds that some knowledge of language is innate - and that it must be innate because the child's input is too limited to allow that knowledge to be learned. We have been pursuing a different possibility. We argue that some such linguistic knowledge can in fact be learned from the input, by relying on general cognitive forces such as indirect evidence and a preference for simplicity.