We study language and language-like systems both from a theoretical and from an experimental perspective. The theoretical perspective is driven by the construction of models that make explicit and testable predictions. The experimental perspective helps us evaluate these models. For many years, formal linguistics has heavily relied on data based on introspection. While this method has the advantage of providing at little cost rich arrays of data to constrain complex theories, there are several situations that require psycholinguistic methods. These are of four types:

Subtle data As formal models have become more sophisticated, the crucial data required to adjudicate among them have become increasingly subtle and sometimes unstable. Empirical controversies have arisen as a result;  in these cases, psycholinguistic methods must be used to establish the facts.

New types of data Our models aim to describe cognitive faculties, and as such they capture data that are inaccessible to introspection. For instance, we developed and applied psycholinguistic methods to gather information about the processing aspects of several phenomena of interest; in this case, psycholinguistic methods have brought to the fore data that would have been unavailable without them.

Special populations Our work with some special populations (e.g. children) call for the development of indirect measures of linguistic abilities.

Variability We wish to systematically investigate cross-linguistic variability and stability (most prominently between spoken and sign languages, or across sign languages). Our methods must thus minimize irrelevant differences at the level of data collection. At the other extreme, we aim to investigate the interplay between subjects’ abilities at the intra-individual level. For instance, one may expect to find correlated skills in two given linguistic domains (e.g. NPI licensing and inferential judgments), or between domains (e.g. musical and phonological skills). The tools to assess such correlations require precise, quantifiable data.