The plasma of fifteen patients with malignant primary intracranial central nervous system (CNS) neoplasms was examined using proton nuclear magnetic resonance (NMR) and lipid level analysis. Twenty-three benign intracranial neoplasms and twelve non-CNS, non-malignancy patients served as control. At 300 MHz the NMR mean line width of plasma with malignant tumours was (+/- 2 SE) 37.2 +/- 3.0 Hz and for benign tumours 33.5 +/- 3.0 Hz (P = 0.09). The mean plasma triglyceride level for the malignancy patients was 100 +/- 25 mg/decilitre and 173 +/- 74 mg/decilitre for the patients with benign tumours (P = 0.07). The mean plasma cholesterol was 182 +/- 42 mg/decilitre for the malignancy patients and for the patients with benign tumours 241 +/- 29 mg/decilitre (P = 0.03). Unfortunately overlaps of the values of the different groups, in spite of statistically significant differences, detracts from the clinical usefulness of these criteria. Attempts to combine these values and correct for tumour volume, as calculated from computerized tomography, did not improve the differentiation between these two groups. Although it has been reported that plasma NMR and lipid levels can differentiate between malignant and benign tumours in general and in the central nervous system, these criteria are not adequately sensitive and specific to replace histology for the definite diagnosis of central nervous system tumours.