Learning about language and learners from computer programs

Making Nation’s text analysis software accessible via the World Wide Web has opened up an exploration of how his learning principles can best be realized in practice. This paper discusses 3 representative episodes in the ongoing exploration. The first concerns an examination of the assumptions behind modeling what texts look like to learners with different levels of lexical knowledge; the second concerns approaches to handling proper nouns in text profiling within an international context; and the third involves the future of the Academic Word List as new frequency information appears to undermine its utility. Underlying these explorations is an argument that writing computer programs is a useful way to investigate language and language learning.