Among the many excursions reported in the late Brunhes, the Skalamaelifell excursion, recorded in 22 flows in three hills of assumed late Wisconsin basalts in the active rift zone of the Reykjanes peninsula, is one of the most reliable. It is characterized by a single, well‐defined transitional direction (D = 260, I = −16). Paleointensity (PI) measurements, using the Thellier method as modified by Coe in a vacuum with partial thermoremanent magnetization checks, were made on six samples from the excursion flows. For comparison, 10 samples from the adjacent and probably nearly contemporaneous normally magnetized flows were also measured. Because of sample scarcity, four of the excursion samples and seven of the normal samples had been already af demagnetized to 20 mT in Iceland. The samples, however, were chosen because of their high coercivity and lack of viscous remanent magnetization. They retained 70–95% of their natural remanent magnetization (NRM) after 20 mT demagnetization, and their remanence directions were stable during the 20 mT demagnetization and during 2–6 years of storage following it. Saturation magnetization (Js) measurements suggest that the reason for the low NRM of the excursion samples was a paleofield one fourth to one tenth that of the normal basalts. The mean PI from four excursion samples is 4.3±0.6 μT, whereas the mean PI of the five normal samples considered reliable is 30±9 μT. The quality of the NRM‐thermoremanent magnetization (TRM) data correlated well with the reversibility of the Js‐T curves. The PI differences agree, in general, with those predicted by the NRM/Js data, especially when comparing the means of the two groups, which average the effects of different grain sizes. A prior, partial alternating field (af) demagnetization of one sample in each of two pairs from the same cores caused the PI to be underestimated by only 10%. Continuous af demagnetization during the NRM‐TRM measurements of one sample in another pair that had nonlinear TRM‐TRM plots, even in the undemagnetized sample, caused the PI to be overestimated by 40%. These results, combined with other studies, suggest that the effect of partial af demagnetization on high‐coercivity samples depends on when the demagnetization is done, i.e., only prior to or during the NRM‐TRM measurements, and, probably more importantly, the thermochemical stability of the sample. If correct, this increases the usefulness for PI determinations of old, hard to recollect, partially demagnetized collections. The very low PI of 4 μT is similar to that found for excursions and transitions in Iceland and other places. However, o its virtual geomagnetic pole lies 90° from a north–south vertical plane through the sites, which rules out a purely axisymmetric model for the excursion. The low normal intensities are similar to those found in the Chaine des Puys (France) and elsewhere for the period 10,0000–50,000 years B.P. Given its very low PI and a direction only 70° from complete reversal, the Skalamaelifell excursion is most likely the result of nondipole fields and an unusually weak dipole field, such as that observed in France for the 10,000‐year interval, 35,000–45,000 years B.P., surrounding the Laschamp excursion.