This paper reports on the impact of computer-mediated input, output and feedback on the development of second language (L2) word recognition from speech (WRS). A quasi-experimental pre-test/treatment/post-test research design was used involving three intact tertiary level English as a Second Language (ESL) classes. Classes were either assigned to a control group (n = 31) or to one of two alternative treatment levels which used a web-based computer application enabling self-determined opportunities to repeatedly listen to and reconstruct spoken target text into its written form. Treatment group one (n = 30) received text feedback after each of their efforts at target text reconstruction, whereas treatment group two (n = 35) did not. Results indicated that word recognition gain scores of those who used the application, regardless of treatment level, were significantly higher than those of the control group. The relationship between the quantity of self-determined exposure to input and word recognition improvements was moderate but not linear, with those choosing moderate levels of speech input deriving the greatest measurable improvement. Neither increased levels of modified output nor the provision of text feedback were associated with significant improvements in word recognition gain scores. Implications for computer-mediated approaches for the development of L2 WRS are described and areas for future empirical research are suggested.