An empirical analysis of solar wind-magnetosphere energy coupling functions is reported. Using the technique of linear prediction filtering with 2.5 minute data, we examine the relationship of auroral zone geomagnetic activity to solar wind power input functions which depend on the solar wind quantities VB2, VBs, or V2Bs. In this analysis a least squares prediction filter or impulse response function which relates a solar wind power function to an auroral zone geomagnetic index is designed directly from the data. We find that the computed impluse response functions have the characteristics of low pass filter with a time delay which may be dependent of the strength of the energy input. While the AL index is reasonably well related to the solar wind energy function, the AU index shows a substantially poorer relationship. In addition, high frequency variations of the auroral indices and some substorm expansions are not predictable with solar wind information alone, suggesting that internal magnetospheric processes partially control the AL index. We also find that the &egr; parameter which depends on VB2 in the solar wind has a poorer relationship to auroral zone geomagnetic activity than a power parameter having a VB solar wind dependence.