Understanding the processes by which volcanic eruptions are triggered is crucial for volcanic hazard prediction and assessment. In intermediate and silicic magmas, where water is the dominant volatile phase(1,2),’second boiling’ provides an effective eruption trigger(1)-as the magma cools and crystallizes, volatiles are increasingly concentrated in the residual melt where they eventually become saturated and are exsolved, thereby raising the magma pressure to the point of eruption(1,2). In contrast, in basaltic magma chambers in which carbon dioxide is the dominant volatile phase, the effect of second boiling is to decrease the chamber pressure. But basaltic magmas are also relatively inviscid, so volatile bubbles can separate from the turbulently convecting melt to produce a foam layer at the chamber roof(3,4) : decompression of the rising bubbles can increase the chamber pressure and so trigger an eruption(5,6). Here we model the complex competition between the processes of magma cooling and bubble ascent. We find that in basalt of very low viscosity (1-10 Pas), bubble separation dominates and can increase the reservoir pressure by as much as 10 MPa (sufficient to trigger an eruption(7)) on timescales of 1-10 years, comparable to the duration of eruptive episodes at Kilauea volcano, Hawaii(4,8).