Sixty target words were selected from two sets of graded readers. One set contained three readers with the same title, "The Railway Children," and the other set, three books about Sherlock Holmes. Students chose one of the two sets to read and were given a pretest, an immediate posttest, and a five-week delayed posttest on their acquisition of spelling, aural meaning, written meaning, and use. Five fixed factors (time, frequency of word occurrence, glossing, word frequency levels, and four dimensions of vocabulary knowledge) in vocabulary learning and a random variable (the participants) were analyzed with generalized linear mixed models. The results show that the odds of improvements in the knowledge of written and aural meanings were significantly better than those for the knowledge of spelling and use. Significant interaction effects were found between time and other fixed factors, except for glossing. Pedagogical implications of the results are discussed.