Purpose Tumour cell hypoxia is a common feature in solid tumours adversely affecting radiosensitivity and chemosensitivity in head and neck squamous cell carcinomas. Positron emission tomography (PET) using the tracer 18 Ffluoromisonidazole ( 18 FFMISO) is most frequently used for non-invasive evaluation of hypoxia in human tumours. A series of ten human head and neck xenograft tumour lines was used to validate 18 FFMISO as hypoxia marker at the microregional level. Methods Autoradiography after injection of 18 FFMISO was compared with immunohistochemical staining for the hypoxic cell marker pimonidazole in the same tumour sections of ten different human head and neck xenograft tumour lines. The methods were compared: first, qualitatively considering the microarchitecture; second, by obtaining a pixel-by-pixel correlation of both markers at the microregional level; third, by measuring the signal intensity of both images; and fourth, by calculating the hypoxic fractions by pimonidazole labelling. Results The pattern of 18 FFMISO signal was dependent on the distribution of hypoxia at the microregional level. The comparison of 18 FFMISO autoradiography and pimonidazole immunohistochemistry by pixel-by-pixel analysis revealed moderate correlations. In five tumour lines, a significant correlation between the mean 18 FFMISO and pimonidazole signal intensity was found (range, r 2  = 0.91 to r 2  = 0.99). Comparison of the tumour lines with respect to the microregional distribution pattern of hypoxia revealed that the correlation between the mean signal intensities strongly depended on the microarchitecture. Overall, a weak but significant correlation between hypoxic fractions based on pimonidazole labeling and the mean 18 FFMISO signal intensity was observed ( r 2  = 0.18, p  = 0.02). For the three tumour models with a ribbon-like microregional distribution pattern of hypoxia, the correlation between the hypoxic fraction and the mean 18 FFMISO signal intensity was much stronger and more significant ( r 2  = 0.73, p  < 0.001) than for the tumours with a more homogenous, patchy, microregional distribution pattern of hypoxia. Conclusion Different patterns of 18 FFMISO accumulation dependent on the underlying microregional distribution of hypoxia were found in ten head and neck xenograft tumours. A weak albeit significant correlation was found between the mean 18 FFMISO signal intensity and the hypoxic fraction of the tumours. In larger clinical tumours, 18 FFMISO–PET provides information on the tumour oxygenation status on a global level, facilitating dose painting in radiation treatment planning. However, caution must be taken when studying small tumour subvolumes as accumulation of the tracer depends on the presence of hypoxia and on the tumour microarchitecture.