An attempt was made to identify all patients presenting with amyotrophic lateral sclerosis (ALS) during a 10-year period in Lancashire and South Cumbria, England. An analysis of their place of residence was made using the postal code. This was used to allocate each patient to an electoral ward. Using such small areal units, it was found that more cases of ALS had arisen than would have been expected by chance in several wards during the study period. Although the actual number of wards showing this effect was probably not increased, the significance levels suggest that the disease may not show a random geographical distribution. No evidence of clustering of year of birth was found. These findings indicate the need for further work seeking geographical clusters of ALS and are discussed in relation to current etiological hypotheses of the disease.