A uniform classification of response to chemotherapy is essential to allow comparison of local effect and ultimate prognosis between different therapy schedules. We define a histological grading system for assessment of the response to chemotherapy in Ewing's sarcoma, based on the amount and architectural pattern of residual histologically viable-appearing tumour, the preferential sites of minimal residual tumour and the amount of tumour necrosis. Twenty-six consecutive patients with a biopsy-proven Ewing's sarcoma were treated with chemotherapy prior to surgery. The effect of chemotherapy was evaluated microscopically on the specimens obtained after surgery. Response to chemotherapy was classified as minimal or no effect (< 10% tumour necrosis), moderate effect (solid areas of remnant viable tumour), minimal residual disease, and no evidence of disease (100% tumour necrosis or well-vascularized fibrous tissue). The subperiosteal area in particular and, less frequently, soft tissues and intramedullary compartment were identified as sites of predilection for persistence of microscopic viable tumour foci, frequently depicted as pseudo-rosettes in a characteristic scattered pattern. Although it is not well known whether morphological viability of these residual clusters in Ewing's sarcoma indicates biological viability, accurate preoperative local staging, with special attention to preferential sites of residual viable tumour, is essential. The proposed grading system can be used to standardize assessment of chemotherapy in trials, and may serve as a standard for non-invasive monitoring of preoperative chemotherapy with magnetic resonance imaging.