An aircraft equipped with cylindrical electric field mills and a chaff dispenser was used to release large numbers of 10-cm-long chaff fibers at the base of thunderclouds. The electric fields and lightning discharges were monitored during repeated flight passes at cloud base. Twenty-eight thunderstorms that met certain qualifications were seeded with chaff or left unseeded, the unseeded storms becoming 'control' cases. An analysis of the data from 10 thunderstorms seeded with chaff and 18 unseeded control storms shows that seeding thunderstorms with chaff reduced the number of observed lightnings to about one third or less of those observed in the control storms. A statistical evaluation of the occurrence of lightning discharges in the time interval before seeding and in the corresponding time interval of the control storms revealed that there was no significant difference between seeded and control storms. This indicates that both categories of storms belong to the same population. If the same test is applied to the time interval after seeding and the corresponding interval of the control storms, we find that there is a significant difference between seeded and control storms after seeding. Therefore the lightning reduction by chaff seeding is not due to chance.