PURPOSE: Problem representation, as mediated by semantic qualifiers (SQs), has been associated with better diagnostic outcomes. The purpose of this study was to assess the effect of training medical students to use semantic abstractions as a means of building problem representations. METHODS: Sixty second-year medical students were assigned to either an intervention group (n = 20) or a control group (n = 40) during 8 months of an Essentials in Clinical Medicine course which used standardized patient-based workshops. Students were trained to transform findings into SQs and to use abstractions to compare and contrast diagnostic hypotheses. Students were assessed using a standardized patient data collection checklist, a post-encounter patient finding questionnaire (PFQ), and case summaries and write-ups. RESULTS: Experimental subjects used over twice as many SQs in their summaries as control group members (1.40 versus 0.63, P = 0.006). The correlation between checklist and PFQ scores was higher for the experimental group than for the control group (r = 0.70 versus r = 0.58, P ≤ 0.001). There was no difference between groups in either the number of SQs used in write-ups nor in diagnostic accuracy (P > 0.56). CONCLUSION: A short instructional intervention was successful in promoting the use of SQs and enabled students to recall elicited findings better. This intervention did not enhance data interpretation and diagnostic accuracy. Use of SQs may therefore be a necessary tool for efficient problem representation but one that is insufficient when used in isolation. The naturalistic setting used in this study imposed a number of limitations, implying that further research should test whether instructional efforts should also emphasize recognition of key patient findings and knowledge representation.