Self-directed learning, despite its growing popularity in education, has challenged conventional assessment practice which often foregrounds the presentation of identical conditions to ensure reliability. This article discusses the results of a case study of university academic English teachers' perceptions and reported practices of assessing self-directed learning with a view to enhancing its reliability and validity. Twenty-two teachers were interviewed about four main areas of self-directed learning: its relevance to English enhancement, the necessity for and challenges in assessing it and the evidence for its success. The study finds teachers' varying degree of understanding of self-directed learning that influence their assessment practices. The reported challenges include difficulties in monitoring students' progress and reification of autonomy. The interviewees emphasised cognitive development as a major indicator of success in self-directed learning. Pedagogical implications are identified to make the assessment more robust such as strengthening teacher-training and co-construction of criteria with students.