This study examines university learners' self-assessment and self-feedback on performance as captured in audio files from a foreign language speaking test. The learners' were guided to listen, transcribe and analyse their own speaking samples, as well as propose future actions for improvement. Content of learners' self-feedback was scrutinised against a feedback model, with data coded into various feedback categories as stipulated in the model for analysis. Results indicated that learners' self-feedback was far reaching and multifaceted. Through self-feedback, learners' identified discrepancies, answered feed up, feedback and feed forward questions, and inspected performance at task, process, self-regulation, and self-levels. Much of the feedback involved reflections on past learning history, other areas of learning, deviation of performance from preparation and learner personality traits. The self-feedback went largely beyond most teachers' feedback capacity and bore great potential for learning and instruction. In particular, contrary to theoretical presumptions, self-level feedback seemed quite enlightening. Whether the observed quality self-feedback could actually help learners improve their performance, however, was not clear. It was suggested that some teachers' time and effort be directed to the endeavour of facilitating learners' self-assessment and self-feedback. Learners' self-feedback capability should also be explored further in the development of relevant pedagogies and theories.