The purpose of this study was to evaluate patient setup in head and neck IMRT using daily portal imaging. At our institution, orthogonal digital portal images are taken daily to check patient positioning prior to head and neck IMRT treatment. Isocenter misalignments are corrected using a couch shift (3 mm action level). Therapists also compare the DRRs and portal images looking at points more distant from the isocenter, particularly in the supraclavicular region, and re‐position the patient's shoulders in the mask if considered necessary. The daily isocenter shifts (C2 region) and frequency of patient repositioning were investigated by review of record‐and‐verify records for 15 patients. The magnitude of the shoulder repositioning was evaluated for 10 of these patients by comparing portal images and plan DRRs for a point 8 cm inferior of isocenter (T2‐T4).For all patients, pretreatment isocenter discrepancies 3 mm or smaller were recorded for a median of 92.5% of fractions (range: 71.4 – 100%). Patients were repositioned in the immobilization mask before treatment for a median of 14% of fractions (3–34%). Thirty percent of these were for shoulder shifts of 1 cm or larger. Twenty percent of patients needed shoulder shifts of 1 cm or more for more than 7/35 fractions, meaning that without setup based on daily imaging, parts of the CTV would have received less than 95% of the prescribed dose. In conclusion, with our current immobilization, isocenter positioning accuracy is excellent, while correct shoulder position is more variable, particularly for a small subset of patients. Frequent imaging of head and neck IMRT patients is essential to accurate delivery of therapy, with shoulder position an important factor.PACS number: 87.53.Oq