Language is inherently variable, and learner language is particularly variable. The variationist paradigm considers learner language a heterogeneously variable yet inherently rule-governed system. Specifically, learners’ alternation between native-like and nonnative-like variants of a variable or invariable target native speaker (NS) form constitutes learner language variation. Variation is also viewed as an indication of a transitional phase towards acquisition (e.g., Regan, 2013; Tagliamonte, 2011). With a particular concentration on second language (L2) morphosyntactic variation, this dissertation explored inter-learner variation and intra-learner variability together with interlanguage development by analyzing Japanese L2 learners’ oral performances in English oral proficiency interviews. The research observed and studied the variation pattern in the interview data and identified the linguistic, paralinguistic, and nonlinguistic factors and factor groups which may give rise to Japanese L2 learners’ repeated exercise of their interlanguage grammar for four morphosyntactic features: preposition/particle, article, object pronoun-dropping, and modal auxiliary verb. The data were analyzed by using classification trees, random forests, and mixed-effects variable rule methods which together identified a hierarchy of variable importance among potential factors and factor groups and the influential factor levels within each significant factor group. With modern mixed models, the dissertation concluded that the observed morphosyntactic variation is subject to inter-lingual and intra-learner factors. Additionally, learners may also have individualized baselines and grammar. More importantly, the findings of the current research have provided important theoretical and empirical justification on whether and how individual patterns mirror the interlanguage patterns and hence an inter-lingual developmental understanding of L2 morphosyntactic competence.