The present study investigated problem-solving behaviour in statistics by documenting differences between successful and unsuccessful students. Two methodological approaches were used. The first, based on studies by Chi, Glaser and Rees (1982), required students to identify key structural features of common problems. The second, based on further work by Chi, Bassok, Lewis, Reimann & Glaser (1989), used verbal protocol procedures to examine the processes employed when 'Good' and 'Poor' statistics students solve an ANOVA problem. It was expected that findings obtained by these earlier researchers in the domain of Physics would also apply in Statistics. Two groups of students in an undergraduate Psychology course were selected for the study. The first group (N = 8) had performed well in a statistics unit, the second group (N---6) had failed that unit. In the first part of the study, no differences were found between groups on the task of identifying key structural features of common problems. In the second part of the study, protocol analyses showed that the 'Good' students engaged in more concept-oriented solution behaviour, displayed more monitoring/judging behaviour, and asked more questions about the problem itself. These findings are similar to those obtained in other abstract knowledge domains such as Physics and Mathematics and indicate quite fundamental differences in the way in which competent and incompetent Statisics students solve Statistics problems.