A controversy exists as to whether the injection of particles into the ring current is caused by the induced electric field associated with the substorm expansive phase or by the enhanced convection electric field that is directly driven by the solar wind. To investigate the causal storm-substorm relationship, artificial neural networks are used to examine the correlation between the westward auroral electrojet index (AL) and the ring current index (Dst). We use the hourly solar wind number density n, the solar wind velocity V, the interplanetary magnetic field (IMF) magnitude B, and the southward component of the IMF, Bs, to predict separately the hourly Dst and AL indices. For the first time we show, using the prediction residuals, the significant UT dependent effects of the substorm current wedge on Dst, as a consequence of the sparse distribution of only four stations that are used to compute the Dst index. When none of the Dst stations is found beneath a typical substorm current wedge, during the main phase of a magnetic storm, we find no correlation of Dst residuals with AL residuals. However, if a Dst station is found beneath a substorm current wedge, during the main phase of a magnetic storm, then a correlation of Dst residuals with AL residuals is observed. Moreover, the mean correlation of Dst residuals with AL residuals over all phases of a magnetic storm is only 2.4%. We therefore suggest that the expected effect of substorm expansive phase activity, in the development of the storm-time ring current, is not detectable in the Dst index, but that the Dst index is contaminated whenever a Dst station passes beneath the substorm current wedge. ¿ 2000 American Geophysical Union