Traditional approaches to the global induction problem have often determined the conductivity structure of the earth for specific modes of the external source field. In the case of “disturbed time” (or Dst) epochs, workers have tended to use the P₁⁰ term, or, in the case of “quiet time” (or Sq or L) epochs, workers have used various low-order P_n^m terms. Each of these “modes” has tended to lead to a distinct earth model which differs from those derived from other modes. Since one and only earth model should jointly “fit” any and all modes simultaneously, it is not clear what these separate models represent—we need an approach which is essentially “mode-blind”. We address this problem by applying an extension of the magnetic gradiometric technique to data from 31 standard magnetic observatories in the European sector. We employ the two years of hourly values culled by WINCH (1981) to 1) characterize the dominant statistical properties of the primary data set and reject outliers; 2) construct low order polynomial forms over the array of observatories; 3) estimate gradiometric response parameters; 4) explore model space for classes of earth models which account for all modes simultaneously, and which fit the data while minimizing the resistivity structure at depth.