It has been shown that the optimal linear prediction filter relating the solar wind electric field and the geomagnetic activity, as measured by the AL index, is both bimodal and dependent on the level of activity in the magnetosphere. Further studies truncated the prediction filter to a five-parameter model containing two low-pass-filtered delta functions of arbitrary amplitude and delay time. The present study elaborates on the nature of the bimodal response by using the five-parameter model to quantify the effects of the level of geomagnetic activity on each of the modes of the filter individually. We find that at all levels of activity, the second mode, occurring at ~1 hour, is relatively unchanged. The first mode, however, has a necessary one-parameter dependence on the level of activity in the magnetosphere. The amplitude of the first mode is shown to increase with respect to activity, and this dependence is sufficient in characterizing the changing properties of the magnetosphere with respect to activity. ¿ 1998 American Geophysical Union