K. Rayner, A. Pollatsek, D. Drieghe, T. J. Slattery, and E. D. Reichle argued that the R. Kliegl, A. Nuthmann, and R. Engbert corpus-analytic evidence for distributed processing during reading should not be accepted because (a) there might be problems of multicollinearity, (b) the distinction between content and function words and the skipping status of neighboring words was ignored, and (c) there are inconsistencies with experimental results. Reanalyses with linear mixed-effect models demonstrate that (a) regression coefficients are stable across 9 samples, (b) lexical status and skipping status (and their interactions) are highly significant but do not account for the effects of word frequency for content and for function words, and (c) there is strong evidence for lexical processing of content words while fixating function words to the left of them. A critical result about fixation durations prior to skipped words is replicated in an experiment. The distinction between correlational analyses and well-controlled experiments and questions about generalizability of results are discussed. The author argues for a complementary role of corpus analysis, computational modeling, and experiments in reading research.