Research has described the key role of formulaic language use in both written and spoken communication (Schmitt, 2004; Wray, 2002), as well as in relation to L2 learning (Ellis, Simpson--Vlach, & Maynard, 2008). Relatively few studies have examined related fixed and semi-fixed multi-word units (MWUs), which comprise fixed parts with the potential for flexible slots in L2 learners' language use over time. Building upon the corpus linguistic methodology of concgramming and usage-based linguistic research on a construction-based developmental sequence in language acquisition, this article describes learners' multi-word inventories through the analysis of a learner corpus of out-of-class instant messaging and blog use in a U.S. high school advanced placement Spanish language course. We examined this corpus using concgram analysis of phraseological tendencies to document related MWUs used in fixed and schematic fashion. Patterns in student language use are described in terms of meaningful permutations of constituency variation over time. Related patterns used by different learners are analyzed and discussed in light of methodological considerations in tracking second language use and the pedagogical implications of the concgramming approach to MWUs. (Contains 2 figures and 1 note.)