We present a new palaeomagnetic investigation of the two sites from the Steens Mountain (Oregon) volcanic record of a Miocene polarity reversal which were supposed to record very fast changes of the geomagnetic field or impulses (Mankinen et al. 1985; Prtvot et al. 1985a,b). Approximately 130 cores were first drilled from the two initial sites, belonging to sections A and B, in order to obtain at least one detailed and complete vertical sampling of each lava flow. Thermal analyses of natural remanent magnetization, complemented by some alternating field treatments, low-field thermomagnetic curves, microscopic observations and electron probe analyses of key magnetic phases, lead us to somewhat different conclusions for the first and the second impulses. At site B (first impulse), we find that the dependence of the remanence direction on the sample vertical position in flow B51 does not imply a directional field change during flow cooling, but is better explained by a thermochemical overprinting due to the overlying B50 flow. However, this conclusion does not challenge the existence of the first impulse because this field change seems to be recorded some 25 m away in flow B51 (Coe & Prevot 1989), at a place where it is thick enough for this record not to have been erased by the baking due to B50. Regarding the second impulse, restudied at site A, our new findings are more comprehensively explained by a change in the field direction during cooling of flow A41-2 than by some overprinting. Using a simple model of flow cooling, the angular rate of change of the field is estimated to have been of the order of 2 "-3 O or 250-350 nT per day during the impulse. This figure is similar to that previously obtained from site D, some 250m away. However, the directional paths describing the field change are somewhat different at the two sites. New investigations are planned to try to understand the origin of this discrepancy.