Numerous models have been developed to simulate the reaction of magmas to changes of thermodynamic variables, such as pressure, temperature, oxygen fugacity, and water activity. However, the extensive experimental database still lacks information on the distinct effect of small amounts of H2O on olivine + plagioclase + clinopyroxene cotectic crystallization in tholeiitic basalt. We present an experimental study addressing the effects of pressure (at 100, 200, 400, and 700 MPa) and small amounts of H2O on phase relations and liquid lines of descent in three tholeiitic basalts representing different evolutionary stages of the Shatsky Rise oceanic plateau magmatic system (compositions AH6, AH3, and AH5 with 8·6, 8·0, and 6·4 wt % MgO, respectively). Two experimental approaches (dry and low H2O) are designed to maintain contrasting H2O activities during crystallization using (1) graphite–platinum double capsules to perform nearly anhydrous experiments (<0·15 wt % H2O in the melt) and (2) Fe pre-saturated Au20Pd80 capsules to obtain low melt H2O contents ranging from 0·4 to 1·1 wt % H2O. Under dry conditions, at lower pressures (≤400 MPa), the crystallization in the MgO-rich AH6 and intermediate AH3 basalts follows the typical sequence of tholeiitic differentiation with olivine crystallization at the liquidus followed by olivine + plagioclase and olivine + plagioclase + clinopyroxene. Both basalts are close to multiple saturation at pressures between 400 and 700 MPa. At high pressure (700 MPa) the crystallization sequence is reversed, starting with clinopyroxene at the liquidus. Under low-H2O conditions, AH6 and AH3 are very close to multiple saturation, even at the low pressures of 100 and 200 MPa, and the reversed crystallization sequence (clinopyroxene, plagioclase + clinopyroxene, olivine + plagioclase + clinopyroxene) is observed already at 400 MPa. In contrast to the two more MgO-rich basalts, in the most evolved AH5 basalt, clinopyroxene is the liquidus phase at all investigated pressures and under both dry and low-H2O conditions, followed by crystallization of plagioclase + clinopyroxene and olivine + plagioclase + clinopyroxene. The most striking observation in our experiments is that the stability of clinopyroxene increases not only with pressure increase but also in the presence of small amounts of H2O (when compared with dry counterparts at similar pressures). Small amounts of H2O increase the proportion of clinopyroxene in the olivine + plagioclase + clinopyroxene phase assemblage. Our experiments clearly show that the effect of adding 0·4 wt % H2O to cotectic melt compositions (e.g. CaO/Al2O3 ratio at a given MgO) is similar to that caused by an increase of pressure from 100 to ∼ 300 MPa. This implies that small amounts of H2O can lead to significant overestimation of cotectic crystallization pressures (by up to 300 MPa) and that H2O contents need to be taken into account in geobarometric models. Our new experiments emphasize the role of low melt H2O contents in stabilizing clinopyroxene and provide some new insights into the problem of the ‘pyroxene paradox’. The apparent mantle pressures obtained for some mid-ocean ridge basalts using ‘dry’ geobarometric approaches can actually represent depths within the lower crust, if small amounts of H2O are present. The application of our experimental data to natural Shatsky Rise basalts implies that the magmas record partial crystallization processes occurring mainly at low pressure (100 MPa), corresponding to depths of ∼3 km beneath the former spreading center, although the more primitive lavas show evidence of differentiation in a deeper reservoir at ∼14 km depth (400 MPa).