Fumonisins, mycotoxins produced by the fungus, Fusarium moniliforme, which grows on maize, are a major worldwide agricultural problem. Consumption of contaminated maize feeds causes a wide variety of toxic effects in animals depending on the species of animal. In humans, high concentrations of fumonisins have been shown to correlate with increased incidence of oesophageal cancer (OC). Most analyses for fumonisins are done using high performance liquid chromatography (HPLC) which requires time-consuming extraction and clean-up prior to preparation of a fluorescent derivative. Enzyme-linked immunosorbent assays (ELISA), which are sensitive and specific, are a viable alternative but commercially available antibodies and kits are extremely expensive. Polyclonal antibodies against fumonisin B, (FB,) were raised in chickens and rabbits; all animals produced antibodies from week 2 onwards, the highest titre was at week 8 from one of the chickens. Cross-reactivities with FB, analogues were checked. A sensitive, quantitative competitive indirect ELISA (CI-ELISA) was developed and optimised; range 0.2 to 20 ng/ml (in buffer), detection limit 0.2 nglml (in buffer), intra-assay coefficient of variation (CV) was 5.33 % and inter-assay 7.04%. This method was adapted to analyse human plasma and urine samples. After removal of proteins by boiling, the range of recoveries of FBI were 94.7% toI12.4% at 4 ng/ml; and 94.6% to 108.7% at 8 ng/ml. Blood and urine samples from patients with OC (40 plasma, 17 urine), controls (21 plasma, 12 urine) and patients with other forms of cancer (20 plasma, 10 urine) were collected from hospitals in the Durban Metropolitan area and analysed for fumonisins. Detectable levels (>0.4 nglml) were found in 86.9% of plasma samples and 94.9% of urine samples. Statistical evaluation showed a highly significant difference between plasma results for OC and controls (p<0.000 1) but no significant difference between the urine results. Comparison of other forms of cancer and controls showed no significant differences for either the plasma or the urine samples. However, there was a highly significant difference between the OC and other forms of cancer results for both plasma (p<0.005) and urine (p<0.05) samples. Some samples (9 plasma, 8 urine) were checked by HPLC. For plasma samples there was correlation between the ELISA and HPLC methods (r = 0.656, p<0.005) but not for urine samples.