To evaluate diffusion-weighted (DWI) magnetic resonance imaging (MRI) for treatment prediction during chemoradiotherapy (CRT) of head and neck squamous cell carcinoma (HNC). Thirty patients with HNC underwent echo-planar DWI and anatomical MRI before and 2 and 4 weeks into CRT. Patient follow-up lasted 2 years post-CRT. Tumour ADC (DeltaADC) and volume changes (DeltaV) between baseline, and 2 and 4 weeks' follow-up were compared for lesions with recurrence versus complete remission (CR) using a Mann-Whitney U test. The predictive value of the DeltaADC and DeltaV for locoregional control (LRC) was examined with the Kaplan-Meier method. The study was approved by the local ethics committee. All patients gave written informed consent. The DeltaADC in primary tumours and nodal metastases, 2 and 4 weeks after the start of CRT, was significantly lower in lesions with post-CRT recurrence than in lesions with CR (DeltaADC(2 weeks) and DeltaADC(4 weeks) for primary tumours, relative to nodal metastases: p < 0.0001). The DeltaV only showed a significant difference for primary tumours at 2 weeks (DeltaV(2 weeks): p = 0.03). The DeltaADC correlated significantly with 2-year LRC (p < 0.001); the DeltaV did not (p > 0.05). DWI during CRT for HNC allows more accurate response prediction than anatomical imaging, correlating significantly with 2-year LRC.