Efforts to make more precise and reliable measurements of the electromagnetic response of the Earth at periods less than a few days are baffled by the contamination of the magnetic variation data by currents which flow in the heterogeneous outer shell of the Earth. In this paper, quantitative estimates of the potential contamination are made by analysing induction in an Earth model which, in addition to the usual steep rise in conductivity at a depth of 500km, also involves a superficial conducting layer of integrated conductance 4,000S. It is concluded that (a) it may be possible, by careful selection of suitable stretches of data from observatories far from conductivity anomalies, to extend global electromagnetic response estimates to periods as short as 8 hours, but that (b) present response estimates in the period range 1-10 days may be contaminated by the currents in the outer shell.These hypotheses are tested by an investigation of the spatial and temporal variations in geomagnetic transfer functions observed at stations in Europe during a three-month period. Two events recorded during that time show distinctly different responses in the period range 1-3 days. The first yields temporally and spatially consistent values of the transfer function linking vertical and horizontal north components. The excitation is most probably of ring current origin, and the bulk of the induced current flows in the deep, spherically symmetric part of the Earth. The transfer functions generated by the second event are also consistent with time, but differ substantially from one obervatory to another. The source of this event is a current system which is spatially non-uniform. The transfer function is partly determined by the configuration of the localised external current system, and partly by the geometry of induced currents in the surface sheet. Both factors give rise to persistent relationships between the magnetic field components, which contaminate the global response estimates.