Background and purpose: In cancer treatment precise definition of the tumor volume is essential, but despite development in imaging modalities, this remains a challenge. Here, pathological tumor volumes from the surgical specimens were obtained and compared to tumor volumes defined from modern PET/MRI hybrid imaging. The purpose is to evaluate mismatch between the volumes defined from imaging and pathology was estimated and potential clinical impact. Methods and Materials: Twenty-five patients with head and neck squamous cell carcinoma were scanned on an integrated PET/MRI system prior to surgery. Three gross tumor volumes (GTVs) from the primary tumor site were delineated defined from MRI (GTV_MRI), PET (GTV_PET) and one by utilizing both anatomical images and clinical information (GTV_ONCO). Twenty-five primary tumor specimens were extracted en bloc, scanned with PET/MRI and co-registered to the patient images. Each specimen was sectioned in blocks, sliced and stained with haematoxylin and eosin. All slices were digitalized and tumor delineated by a head and neck pathologist. The pathological tumor areas in all slices were interpolated yielding a pathological 3D tumor volume (GTV_PATO). GTV_PATO was compared with the imaging GTV's and potential mismatch was estimated. Results: Thirteen patients were included. The mean volume of GTV_ONCO was larger than the GTV's defined from PET or MRI. The mean mismatch of the GTV_PATO compared to the GTV_PET, GTV_MRI and GTV_ONCO was 31.9 %, 54.5 % and 27.9 % respectively, and the entire GTV_PATO was only fully encompassed in GTV_ONCO in 1 of 13 patients. However, after the addition of a clinical 5 mm margin the GTV_PATO was fully encompassed in GTV_ONCO in 11 out of 13 patients. Conclusions: Despite modern hybrid imaging modalities, a mismatch between imaging and pathological defined tumor volumes was observed in all patients. A 5 mm clinical margin was sufficient to ensure inclusion of the entire pathological volume in 11 out of 13 patients.