The remanent magnetization of basaltic flows in the contact zones of two feeder dikes of the Columbia Plateau series was investigated in detail by af and thermal methods with the aim of estimating the depth of the sampling levels at the time of dike intrusion. In these cases these depths are known from geological and geochemical correlation of the dikes and certain flows stratigraphically higher in the series. In both cases the host rock direction differs strongly from the younger dike direction. In the Fields Spring contact zone, one sample in the profile across the contact displays a hybrid direction between that of the flow and that of the dike. Both remanence components show distinct blocking temperature spectra with an interface at 540 ¿C. This implies that the maximum temperature to which this sample (taken at 60 cm from the contact with the 17.7-m-thick dike) was heated was 540 ¿C. Applying heat conduction theory, the maximum temperature increase due to the cooling dike at 60 cm from the contact is calculated to be 614 ¿C, which would imply a significant negative ambient temperature in the host rock prior to the intrusion. Although complicated by the absence of a hybrid sample in the profile across the Almota contact and the occurrence of spontaneous oxidation during heating of the samples, a comparable negative ambient temperature is derived. No evidence for multiple intrusions is observed, and the Almota dike is only 1.8 m thick. Furthermore, substantial heat leakage to the surface can be ruled out because of the relatively large depths of the sampling levels (250 m for Almota and 850 m for Fields Spring) with respect to the short distances to the contacts. For this reason a model is set up based on heat conduction theory allowing for 5% porosity in the flows. Assuming stagnant water in the flows, a reduction of about 9 ¿C of the temperature increase is calculated. Consequently, if the discrepancy is to be explained by evaporation of water, high permeability would be required to allow perhaps 10 times the volume of the stagnant water by the high-density jointing in the lavas and the present dry condition of the surface. In this case the conclusion is that the groundwater layer extended from a minimum depth of 250 m (Almota sampling level) to at least 850 m (Fields Spring sampling level) at the time of dike intrusion about 15 m.y. ago. A semiquantitative model for convecting groundwater is also presented and, using published data for basaltic flows, is shown to represent a useful model to account for the derived temperatures in the hybrid zone.