Using a panel of seven cell markers, we studied the value of immunocytochemical labelling in the histological diagnosis of desmoplastic malignant melanoma. Sections from routine formalin-fixed tissue of 45 surgical specimens were obtained from 25 cases of malignant melanoma that showed well-marked desmoplastic or neurotropic features. Routinely stained sections (Haematoxylin-and -eosin and melanin stains) were compared with the following panel of seven antibodies: S-100, neuron-specific enolase (NSE), vimentin, factor XIIIa (FXIIIa), desmin and the newer, supposedly more specific anti-melanoma antibodies HMB45 and NKIC3. S-100 and NSE were the most sensitive antibodies for desmoplastic malignant melanoma with strong labelling of spindle cells in most cases. In contrast, results for NKIC3 were more variable; results were negative in nearly half the tumours, but strong labelling was seen in six cases (27%). Positive labelling for HMB45 was noted in five tumours (22%); it was mostly confined to small groups of cells in the superficial part of the lesions. Tumour spindle cells were negative for FXIIIa in all cases; there was no increase in the number of positive dermal dendritic cells compared to conventional and spindle cell melanoma. All tumours were desmin-negative, but most were vimentin-positive. Our findings indicate that immunocytochemistry is of less value in the diagnosis of desmoplastic malignant melanoma that it is with other types of malignant melanoma. However, positive or negative labelling for S-100 protein and NSE is useful for suggesting or excluding a diagnosis of desmoplastic malignant melanoma; neither marker is specific and, in particular, positive labelling is also found in most neurofibromas and benign cellular naevi. NKIC3 and HMB45 label desmoplastic malignant melanoma much less consistently than other conventional malignant melanomas.