Three hundred and twenty-seven patients with inoperable squamous cell carcinomas of the head and neck were entered on a randomized study comparing a mixture of neutron and photon ("mixed beam") radiation therapy with photon/electron radiation therapy. Neutron treatment was delivered with fixed-beam, physics-laboratory-based equipment. Patients with histologically proven tumors of T-stage T2, T3, or T4 and any N-stage were eligible for randomization. Primary tumor sites were limited to cancers originating in the oral cavity, oropharynx, supraglottic larynx, or hypopharynx. Patients entered on this study now have a minimum at-risk follow-up period of 6 years. Study results reveal no significant differences in overall loco-regional tumor control rates or survival. Subgroup analysis reveals significant differences based on whether or not patients presented with positive lymph nodes. Loco-regional tumor control rates for patients presenting with positive lymph nodes were 30% for mixed-beam-treated patients versus 18% for photon-treated patients (p = 0.05). Loco-regional tumor control rates for patients presenting without positive lymph nodes were 64% for photon-treated patients and 33% for mixed-beam-treated patients (p = 0.004). Control of tumor located in the nodal sites favored mixed beam over photons by a margin of 45% (49/109) to 26% (23/87) with a significance of p = 0.004. Possible explanations for these contradictory findings are discussed.