We aimed to detect new cases of konzo and monitor cyanogen exposure from cassava flour in communities previously affected by konzo epidemics in Nampula Province, northern Mozambique. Other objectives were to detect subclinical upper motor neuron damage in schoolchildren and test a new kit to measure urinary thiocyanate concentration. In 1999 and 2000, we carried out active and passive case detection for konzo in Memba and Mogincual Districts. In July and October, 1999, we collected cassava flour from 30 houses in three communities and measured cyanogen concentrations with a picrate kit. In October 1999, we examined all schoolchildren in three communities for ankle clonus and measured urinary thiocyanate concentration in thirty schoolchildren in each of five communities with a picrate kit. We found 27 new cases of konzo in Mogincual District. Mean total cyanogen concentrations in cassava flour varied between both seasons and years, but were always high, ranging from 26 to 186 ppm. Very high mean levels at three sites in November 1998 and July 1999 were probably due to low rainfall in the 1997-1998 season. The proportion of schoolchildren with ankle clonus varied from 8 to 17%. The new picrate kit for urinary thiocyanate worked well; mean concentrations in schoolchildren ranged from 225 to 384 micromol x l(-1). Konzo and sub-clinical upper motor neuron damage persist in poor rural communities in northern Mozambique, associated with high cyanogen concentrations in cassava flour and high urinary thiocyanate concentrations in schoolchildren.