Here be some data

Tokenized Brown Corpus ("representative" 1-million word corpus)
ICNALE_500 Corpus (500 learner texts)