A comparison of the end-of-study versus months-to-years later levels of language proficiency of 49 students of Japanese at Cornell University, from both intensive and nonintensive programs, is reported. In addition to academic records and performance on the Japanese Proficiency Test, data obtained from a comprehensive background questionnaire covering language training and contact, relevant travel, language attitudes, motivation, learning style preferences, self-evaluation of language skills, other related information, and results of several interviews and case studies were analyzed. It was found that: (1) the questionnaire developed for this study was highly appropriate for obtaining student-based information concerning language-learning background, skill levels, and other data relevant to language attrition; (2) attriting and nonattriting students differed little in initial language ability or formal language training; (3) decreases in proficiency over time were inversely related to degree of language use; (4) attrition may be skill-specific; (5) case studies corroborated group data analyses; and (6) re-acquisition of skills might be faster with familiar rather than unfamiliar language learning materials. Results suggest that studies of changes in student performance use a test battery providing both general assessment and a more highly diagnostic, item-by-item probing of language elements, and that future studies include analyses of attrition patterns for subgroups at several proficiency levels as well as whole-group analyses. The questionnaire is appended. (MSE)