I am broadly interested in understanding how cognitive pressures in individual minds lead to law-like patterns in social systems, like language. Focusing on language, I have studied how cognitive and pragmatic pressures can lead to patterns in behavior at the level of individuals, and over time, how these pressures can lead to regularities at the level of words. I am particularly interested in the role of developmental (learning) pressures in these dynamics. Below are more detailed descriptions about four areas of my research (see papers for full publication list).

Measuring and describing variability in cross-cultural semantic space

In my postdoc, I have been working on two projects exploring cross-cultural semantics. In one project with James Evans, I have been exploring word-embeddings from English text written by speakers of non-English first languages. With Gary Lupyan, I have been examining the Google Draw dataset to uncover cross-cultural variability in meaning representations (Lewis & Lupyan, 2018).

Left: The prototypical “bread” drawing by country, calculated as the drawing with the smallest average pairwise distance to other drawings from the same country.

Complexity Bias

As a case study of how linguistic regularities emerge, I have studied the bias for languages and people to tend to use longer words to refer to more complex concepts, or a "complexity bias" (Lewis & Frank, 2016a; Lewis & Frank, 2016b; Lewis & Frank, 2016c; Lewis & Frank, 2015; Lewis & Frank, 2014).

Left: (a) Schema of word mapping task (b) Artificial “geon” stimuli. (c, d) Experimental results from a task in which participants were asked to map a novel word of varying length to one of two possible referents.

Strong sampling and the suspicious coincidence effect

How does the source and quantity of information about a concept affect what we infer (Lewis & Frank, 2016; Lewis & Frank, in press)?

Left: Effect sizes for Xu and Tenenbaum (2007a) and five replication attempts.

Word learning

How do children identify the referent of a word in context, and how do they generalize that meaning to new situations (Lewis & Frank, 2013b)? I have been particularly interested in understanding the bias for children to assume a novel word maps onto to a novel concept ("mutual exclusivity" bias; Lewis & Frank, in prep.; Lewis & Frank, 2013a). In another project (Metalab), I have explored how we can use meta-analytic data to develop a meta-theory of language acquistion (Lewis, Braginsky, Tsuji, Bergmann, Piccinini, Cristia, & Frank, M., under review; Bergmann, Tsuji, Piccinini, Lewis, Braginsky, Frank, & Cristia, in press).

Left: Three left panels show three possible systems-level theories of language acquisition. Right panel shows the empirical trajectory of different language skills, as measured through meta-analysis.