Thinking Skills has, since its conception as a teachable subject, been difficult to assess due to lack of a universal definition of Thinking Skills. McMahon (1999) warns that the chosen strategy for assessing Thinking Skills must itself promote and reward thinking skills. Self- and Peer- assessment are the methods of assessment proposed by Costa and Kallick (2000) for the assessment of their Thinking Skills programme, Habits of Mind. This study investigates the diversity of perceptions which can be generated through self-, peer- and teacher-assessment, and how a balance can be struck between them through incorporating dialogue into assessment. 12 Grade 8 learners who study Habits of Mind as a school subject in a South African all-girls' private school participated in 4 assessment tasks. For each task, participants were assessed by a peer, a teacher and themselves. After each task, all participants reflected on the rubrics from all three assessors. Participants in this study demonstrated more extensive learning across Anderson's (2010) Dimensions of Growth when they engaged in reflective dialogue compared to when they engaged in written reflections. Through dialogue, learners were empowered as role-players in their own assessment and became able to shift their own perspective to include the perspectives of others. Dialogical Assessment also facilitated the development of meta-cognition in participants.