This report describes an in-depth analysis of the errors made by users of SOLO, a programming language written for Open University students studying cognitive psychology. The study was designed to (1) determine the effectiveness of SOLO's current error-handling routines by evaluating how often SOLO produced "sensible" messages or automatic corrections; (2) examine the context surrounding errors and the kinds of conceptual misunderstandings which characterized their true causes; (3) suggest improvements to error-handling routines and other user aids; and (4) estimate the overheads incurred by such improvements and calculate the number of erroneous lines which could be successfully "trapped" or "pre-empted" by such improvements. An introduction describes the constraints from which SOLO evolved, features that "friendly systems" should incorporate, and features of SOLO. Data were collected by recording SOLO user interactions using the DEC-20 "Photo" program, which stores all the user input lines and SOLO response lines into a file in order of occurrence. Results are presented in relation to recommended changes in the system's user aids, specific type-related changes, and general changes. Appendices include a summary of all the changes, their requirements and benefits, and a breakdown of the errors that occurred by message and by types. (Author/LMM)