In this article we report on an 8-year process that included three successive iterations of the following cycle: (a) development of instructional technology, (b) formative evaluation, and (c) modification of the technology. From the first formative evaluation to the last, our students told us that they found heavily contextualized language difficult to learn and frequently requested more decontextualized language for textbook presentations and for practice. With the aid of formative evaluation data (e.g., performance data based on think aloud protocols, attitudinal data, retrospective interviews, and course surveys), we tried to strike a balance between what students said they wanted (i.e., more decontextualized language input and practice) and what we as language teachers and curriculum developers believed that they needed (i.e., more contextualized language input and practice). Three theoretical constructs proved particularly relevant in helping us interpret the data: activity theory (Lantolf & Appel, 1994; Lantolf, 2000); the naive lexical hypothesis (Bland, Noblitt, Armington, & Gay, 1990); and the lexical approach (Lewis, 1993), also known as the lexical syllabus (Willis, 1990). We argue that when formative evaluation becomes a central part of the development of instructional technology, the results are a more learner-centered curriculum with more user-friendly technology. (Contains 6 notes and 5 figures.)