Introduction The aim was to determine the diagnostic accuracy and additional value of diffusion-weighted imaging for detection of malignant lymph nodes in head and neck squamous cell carcinoma.Methods Two hundred nineteen lymph nodes, predominantly smaller than 10 mm (95.4%), in 16 consecutive patients were evaluated at 1.5 T. Lymph nodes were evaluated for maximum short axial diameter, morphological criteria, and apparent diffusion coefficient (ADC) values (b=0 and b=1,000 s/mm 2 ). Sensitivity, specificity, positive and negative predictive values as well as diagnostic odds ratios (DORs) and areas under the curves (AUCs) of ROC curves were calculated for the various magnetic resonance imaging (MRI) criteria individually and in combination. Histological examination of lymph nodes in the neck dissection specimen was the gold standard to determine malignant involvement. Results The optimal ADC threshold was 1.0×10 −3 mm 2 /s. Using this cutoff point, sensitivity and specificity were 92.3% and 83.9%, respectively. When used in combination with size and morphological criteria, ADC value <1.0× 10 −3 mm 2 /s was the strongest predictor of presence of metastasis (DOR=97.6). A model which added ADC values to the other MRI criteria performed significantly better than a model without ADC values: AUC=0.98 versus AUC=0.91 (p=0.036). Conclusion In this study, with predominantly small lymph nodes, the ADC criterion is the strongest independent predictor of presence of metastasis. The use of ADC values in combination with the other MRI criteria significantly improves the discrimination between malignant and benign lymph nodes.