SUMMARY A palaeomagnetic and rock-magnetic investigation has been carried out on a Pliocene lava flow sequence in the Djavakheti Highland in the central Lesser Caucasus in the Republic of Georgia. In addition, a 40 Ar/ 39 Ar dating and electronic microscopic studies were performed on samples of this sequence, named the Saro section, which consists of 39 successive lava flows of doleritic basalts. A characteristic magnetization could be isolated in all studied 39 flows,yielding reverse-polarity directions inall cases,amean directionD =202.2 ◦ ,I =–60.6 ◦ (N = 39, α95 = 2.0 ◦ , k = 138) being obtained. Thermomagnetic experiments (strong-field versus temperature curves) suggested low-Ti titanomagnetites and low Curie-temperature titanomagnetites with a rather high titanium content (x ≈ 0.5–0.7) as the main carriers of remanence. Their domain structure is characterized by a mixture of single- and multidomain grains. 40 Ar/ 39 Ar dating yielded an age of 1.73 ± 0.03 Ma, interpreted as the eruption age of the uppermost lava flow of the sequence. Analysis of palaeomagnetic results and radiometric data from the present and a previous study allows two different explanations about the time of emplacement of the section: (i) The lower 36 flows of the sequence might have been emitted between the normal-polarity Reunion and Olduvai chrons, and the upper three flows after the Olduvai chron, with a long hiatus in volcanic activity of more than 150 kyr or (ii) The whole sequence has been emitted between 1.778 and 1.73 ± 0.03 Ma, after the Olduvai chron. Comparison of the palaeomagnetic results obtained in this study with the expected direction shows that while inclination values agree well, declination shows an eastward deviation of 19.2 ◦ ± 5.8 ◦ . This discrepancy can be explained with a clockwise vertical-axis rotation of the sequence, which might have been produced by extensional structures with strike-slip component, which can be found in the study area. Virtual geomagnetic pole (VGP) scatter recorded in the sequence is somewhat lower than expected. To analyse this behaviour, VGP scatter of the sequence was first calculated with respect to the palaeomagnetic pole obtained from the Fisher mean of all VGPs. Then it was calculated with respect to the appropriate Eurasian palaeomagnetic pole from VGPs obtained after applying a 19.2 ◦ counter-clockwise rotation of the sequence (i.e. undoing the observed vertical axis rotation). The observed scatter with both independent methods appeared to be very similar, allowing the deduction that