Effectively teaching low-achieving and learning disabled students is challenging. Concept-based instruction is recognized as a particularly effective technique for helping students in these populations attain high levels of achievement (Tindal, Nolet, & Blake, 1992). The maze (a reading selection where certain words have been deleted and students are given choices of words to use in each blank) is well established as a classroom-based reading measure (Parker, Hasbrouck, & Tindal, 1992; Fuchs, Hamlett, & Fuchs, 1990; Cranney, 1972-73; Guthrie, 1973; Kingston & Weaver, 1970) that can provide reliable assessments of student improvement in reading. Technology enables individuals to improve on the maze, delivering it via the internet, and thereby making it possible for students to receive almost instant feedback on their performance. To isolate the components of the MAZE that influence student learning, the authors designed their study in two parts. In Part One, they tested the effect of regular use of concept-based versus random mazes without feedback. In Part Two, they shifted their focus. This study was conducted at a public high school of approximately 750 students in a small town in the Pacific Northwest. This study advances individuals' understanding of how the MAZE affects student learning. Part I of the study suggests that the simple act of completing mazes without receiving any feedback does not positively affect student ability to identify or apply concepts in language arts. Part II of the study begins to offer some insights into the type of feedback that is most effective in promoting students' ability to identify and apply concepts in language arts. It should be noted, however, that although more frequent interactions with concept MAZEs are correlated with better performance on one of the post test measures, this improvement in performance does not generalize to the other measures. Also, although the difference in performance is statistically significant, it is not overwhelming. This study provides information that will be useful in identifying areas for future research. This study highlights the need to refine MAZE tasks to be sure that they are reinforcing the concepts and attributes the teacher has decided are most significant for each unit. One of the insights gleaned from this study is the realization that concept-based MAZE activities might be improved by changing the format in such a way that students interact more with the actual concepts and attributes being studied as opposed to examples of those concepts. An additional area for further exploration involves the degree to which interaction with MAZE activities, which require relatively low level cognitive processes, promotes improvement in low level cognitive processes as opposed to improvement in higher level, more cognitively demanding tasks. Although students in this study did, indeed, improve their performance on tasks that required them to identify concepts and attributes, there was no improvement on tasks that necessitated application of concepts and attributes in a novel situation. If the goal of instruction is to improve students' ability to perform on cognitively-demanding tasks, then additional structures that would reinforce such learning should also be included in future studies. (Contains 12 figures and 9 tables.)