BACKGROUND: Little attention has been paid to the metacognitive ability of medical students. AIM: We used confidence marking to explore certainty of knowledge and ignorance. METHODS: One hundred and twenty-seven of 169 general practice trainees took part. Students sat a written multiple choice question (MCQ) test. Each answer was followed by a degree of certainty judgement. Answers attributed with a high degree of certainty were used to compute overall usable knowledge, hazardous ignorance, proportions of knowledge that is usable and of ignorance that is hazardous. The former variables were analysed according to MCQ score, year of training and gender. RESULTS: At a group level, the mean amount of usable knowledge on the MCQ was 21.13%, mean amount of hazardous ignorance on the MCQ was 5.21%, mean proportion of knowledge that was usable was 36.57%, mean proportion of ignorance that was hazardous was 14.32%. There were neither significant differences between highest and lowest quartiles of MCQ score, nor according to year of training. Men had higher levels of ignorance that is hazardous. CONCLUSION: A third of trainees' knowledge was partial. A sixth of their ignorance was hazardous. Confidence marking can aid formative assessment and could potentially be implemented into summative assessments.