Concurrent chemo-radiotherapy (CT-RT) has been shown to be superior to sequential CT-RT for stage III non-small cell lung cancer (NSCLC). Pre-chemotherapy gross tumor volumes (GTV) are commonly contoured for sequential CT-RT and, as significant inter-clinician variability exists in defining GTV's for lung cancer, we postulated that the poorer local control observed with sequential CT-RT may partly be due to the larger errors in defining GTV after chemotherapy-induced tumor regression. Pre-and post-chemotherapy CT scans for RT planning (RTP) were performed in ten patients who received induction chemotherapy for NSCLC. Image registration of pre- and post-chemotherapy RTP scans was performed for all patients. GTV's were first contoured in the conventional manner by two clinicians, i.e. by visual reconstruction from hard copies of the pre-chemotherapy diagnostic CT scans ('GTV-visual'). A 'GTV-match' was then contoured after image-registration, and the 'gold standard' volume was considered to be the overlap of the 'GTV-match' generated by both clinicians. The 'GTV-match' was on average 31-40% larger than 'GTV-visual'. The mean percentage of the 'gold standard', which was not covered by the 'GTV-visual' was similar for both clinicians, i.e. 26.3+/-12.5 and 28.0+/-15.0%. The inter-clinician agreement in contouring improved after image registration. These data suggest that conventional visual contouring of pre-chemotherapy GTV's may fail to treat the actual pre-chemotherapy tumor volume, and thus confound studies evaluating optimal sequencing of chemo-radiotherapy in NSCLC.