This study of 52 undergraduates of English as a foreign language (EFL) involves an empirical assessment of the pedagogical suitability of data-driven learning (DDL) in three Taiwanese grammar classes. One class (16 students) was taught using a traditional deductive approach (TDA), and the others (one of 17 and one of 19 students) were taught using blends of DDL and TDA. The participants' performance in grammar and their judgments of the teaching effects of DDL were both collected for analysis. Using a covariance analysis, the study results indicate no significant differences between the three classes in grammar proficiency, although paired-sample t-tests reveal significant gains for each class. However, the results of quantifying participants' perceptions of the treatments over time show clear changes as the experiment proceeded; there was a growing preference for DDL-integrated treatments but a disinclination towards the TDA. Although it seems premature to claim DDL's pedagogical suitability here, the overall results lend support to the legitimacy of practicing DDL in different educational areas. This is particularly notable for Taiwan's EFL context, given that most of its grammar classrooms are still employing conventional approaches, including the Grammar Translation method, even if they are not inclined towards them. The article concludes with a discussion of DDL's effects on future EFL grammar classes and possible avenues for further studies.