In this study, I record and describe dialectal and register-based variation observed in the production of undergraduate students of Arabic as a foreign language at a major U.S. research university. As a point of curricular design, these students receive explicit instruction in multiple, discretely identifiable varieties of their target language, and the present research investigates the ways in which this diversity of input is reflected and reshaped as an element of student output. In so doing, it is posed to fill a gap in the theoretical and applied linguistic literatures between studies of the untutored acquisition of second language variability (e.g., Friesner & Dinkin, 2006) and those of native second dialect learning in a classroom context (see Siegel, 2010), and in addition addresses a pressing issue of curricular design and implementation in 21st century Arabic education (cf. Al-Batal, 2017). Utilizing material drawn from over 30 hours of class observation and recorded individual interviews, I evaluate student production data on the basis of both type and frequency of language mixing behaviors attested and further draw on observed classroom practice and metalinguistic commentary to identify the motivations and avenues of actuation underlying these behaviors. The results show a striking correspondence to theoretically postulated outcomes of language and dialect contact in a “natural” setting, reflecting widely recognized aspects of code-switching typology and new dialect formation: particular characteristics evidenced include the observation of a three way distinction between alternation, insertion and congruent lexicalization in mixing (Muysken, 1997) and pressure toward interdialectal leveling and fudged forms (Trudgill 1986). These traits lead to the analysis of the data as reflecting four readily definable stages of mixing practices which vary across student experience levels. These stages sketch a progressive trajectory of evolution in the diverse linguistic output of Arabic learners, and argue strongly against fears of fossilization of non-native-like mixing patterns that may temporarily dominate individual phases of learning. The marked parallels with language/dialect contact phenomena in non-classroom contexts show the relevant developmental patterns to represent predictable and explicable phases in the acquisition of linguistic variability, which have the potential to eventually lead to native-like patterns of diversity in target language production. Knowledge of this progression has clear implications for Arabic language educators, ranging from assessment to classroom management to curricular design, and additionally serves to inform the work of theoretically oriented scholars of linguistic variation in noncanonical contexts.