Over the last decades, linguistic research has become more diverse both theoretically and methodologically. With regard to the former, after a long period in which “theoretical linguistics” was synonymous with “generative linguistics”, now a wider variety of approaches have emerged; for this talk of interest are cognitive/usage-based approaches, which assume a less-than-modular linguistic system that is ‘governed’ to a large extent by domain-general mechanisms such as frequency, contingency, recency, context etc. With regard to the latter, linguists are now routinely using a wider range of data and it is corpus data that have seen a particular increase.Against this background, I will discuss in this talk ways in which corpus-based work can help explore the lexicon/construction in ways that properly operationalize the above domain-general determinants of processing and learning: frequency, contingency, recency, context. I will discuss two brief case studies – one on phonological similarity within lexical units (involving
frequency), one on multi-word identification (adding contingency) – before I turn to a broader discussion of how to involve recency and context properly to our corpus-linguistic toolkit.