The current study applied the person-fit statistic, l[subscript z], to data from a Canadian provincial achievement test to explore the usefulness of conducting person-fit analysis on large-scale assessments. Item parameter estimates were compared before and after the misfitting student responses, as identified by l[subscript z], were removed. The changes of item parameter estimates were found to be noticeable for some items. In addition, analyses were conducted to identify student and class factors associated with misfitting responses. Hierarchical linear modeling was used due to the hierarchical structure of the data. Although student-level and class-level variables were found as statistically significant predictors of the degree of person-fit as indicated by the values of l[subscript z], the percentage of variability accounted for by these variables was considerably small--only 3.65%.