We report magnetic properties of submarine basalts 3.5 to 16 Ma in age recovered from depths as great as 530 m in layer 2 near the Mid-Atlantic Ridge at 37° N during Leg 37 of the Deep Sea Drilling Project. The rocks are classified as type-I if they have reversible in-vacuum thermomagnetic curves and as type-Il if they are thermomagnetically irreversible and develop a high-Curie-point phase (believed to be magnetite) when heated. Initial Curie points are low: 140–200 °C in type-I rocks, 250–300 °C in type-II rocks. The phases responsible are thought to be stoichiometric and cation-deficient (oxidized) titanomagnetite, Fe2.4Ti0.6O4, respectively. Only the 3.5 Ma basalts contain any type-I material; the older basalts are completely oxidized.Viscous magnetization is uniformly strong in type-I rocks, weaker and variable in type-II rocks. Hysteresis properties explain this difference. It is not due primarily to the chemical difference between stoichiometric and oxidized titanomagnetites, but to a difference in grain size. Type-I rocks are magnetically very soft: the coercive force (Hc) is 15–90 Oe (1194–7162 A/m), the median demagnetizing field of natural remanent magnetization (NRM) is 35–135 Oe (2785–10743 A/m), the ratio between saturation remanence Jrs and saturation induced magnetization Js is generally and the ratio of remanent coercive force, HR, to Hc is . These results all indicate multidomain grains of titanomagnetite ≥ 40 μm in size. Opaques of this size are seen in polished thin sections. Type-II rocks have Hc > 150 Oe (11937 A/m), (27 853 A/m), and HR/Hc generally < 2, indicating single-domain or pseudo-single-domain behaviour in micron- or submicron-size grains. The small magnetic grain size in type-II rocks could result from preferential oxidation of fine grains and/or subdivision of larger grains by inhomogeneous oxidation. The pronounced viscous magnetization of type-I rocks is therefore thought to be due to coarse, unoxidized multidomain grains of titanomagnetite.Long-term viscous magnetization is simulated by measuring viscous decay curves at temperatures up to 200 °C. Relaxation times are strongly temperature dependent: relaxation times as long as 106 yr can be activated in laboratory experiments at 75 °C if a low-Curie-point phase like Fe2.4Ti0.6O4 carries the viscous magnetization, or at 200 °C if Fe3O4 is the carrier. Viscous remanent magnetization (VRM) over 106 years seems to be no more than a factor 2 or 3 times the VRM estimated by extrapolating room-temperature data determined over a laboratory time scale. Even in type-I rocks, long term VRM is insufficient to completely erase the NRM.