1] Lavas of Pliocene-Pleistocene age were sampled in two regions in Kenya: Mount Kenya on the equator and the Loiyangalani region, east of Lake Turkana, at about 3°N. We sampled 100 sites distributed around the Mount Kenya Massif and to the northeast along the Nyambini Range. The equator bisects Mount Kenya, and all sites were sampled within 40′ of the equator. Thirty-two sites were sampled in the Loiyangalani area, making a total of 132 sites. Many sites from the Mount Kenya study were severely affected by lightning; however, after progressive AF demagnetization 69 sites yielded directions with α95 equal to or less than 10°. Normal polarity sites dominate (N = 58 and a mean of declination (dec) = 1.2°, inclination (inc) = −0.7°, and α95 = 3.6°) with only 11 reverse polarity sites (mean of dec = 182.3°, inc = 0.6°, and α95 = 7.2°); no transitional directions were identified. Inverting the reverse sites yields a combined mean direction of dec = l.4°, inc = −0.7°, and α95 = 3.2°. This result is not significantly different from what is expected from the geocentric axial dipole for the mean locality (dec = 0° and inc = 0°); a quadrupole component was not resolved. The samples from the Loiyangalani region were not seriously affected by lightning, and all 32 sites gave satisfactory data with α95 less than 10° (17 reverse sites, dec = 183.4°, inc = 0.8°, and α95 = 6.7°; 15 normal sites, dec = 358.6°, inc = −1.1°, and α95 = 4.7°); after inverting the reverse sites the combined mean was dec = 1.1°, inc = −1.0°, and α95 = 4.1°. Altogether, we had a total of 101 successful sites. A virtual geomagnetic pole (VGP) was calculated from each site mean; the VGP dispersion is low, with Sb = 10.9° for Mount Kenya and 9.8° for the Loiyangalani region. This dispersion agrees with updated Model G of McElhinny and McFadden (1997) and model TK03 of Tauxe and Kent (2004) that was tuned to the compilation of McElhinny and McFadden (1997) but disagrees with the higher dispersion near the equator and the smaller latitudinal gradient in dispersion estimated by Johnson et al. (2008). A new database is presented, and the included studies support a systematic decrease of dispersion from high to low latitudes.