Early attempts to link solar filaments and their disappearances with geomagnetic activity1–7 were inconclusive. Joselyn and McIntosh8 have suggested that some magnetic storms are produced by disappearing filaments. However, we feel that their study included too few cases to establish this link conclusively because, of the 12 possible filament–storm associations shown in their Table 2, they acknowledge that seven are subject to doubt. Therefore we have investigated this problem using a large data sample obtained between 1974 and 1980 inclusive. We found that the average level of geomagnetic activity increases 3–6 days after the disappearance of large solar filaments. We report here that both the magnitude of the disturbance and the delay in its occurrence seem to depend on the filament size. These results support the inclusion of disappearing filaments within techniques that are used for predicting geomagnetic activity.