Plinian/ignimbrite activity stopped briefly and abruptly 16 and 45 h after commencement of the 1912 Novarupta eruption defining three episodes of explosive volcanism before finally giving way after 60 h to effusion of lava domes. We focus here on the processes leading to the termination of the second and third of these three episodes. Early erupted pumice from both episodes show a very similar range in bulk vesicularity, but the modal values markedly decrease and the vesicularity range widens toward the end of Episode III. Clasts erupted at the end of each episode represent textural extremes; at the end of Episode II, clasts have very thin glass walls and a predominance of large bubbles, whereas at the end of Episode III, clasts have thick interstices and more small bubbles. Quantitatively, all clasts have very similar vesicle size distributions which show a division in the bubble population at 30 ??m vesicle diameter and cumulative number densities ranging from 107-109 cm-3. Patterns seen in histograms of volume fraction and the trends in the vesicle size data can be explained by coalescence signatures superimposed on an interval of prolonged nucleation and free growth of bubbles. Compared to experimental data for bubble growth in silicic melts, the high 1912 number densities suggest homogeneous nucleation was a significant if not dominant mechanism of bubble nucleation in the dacitic magma. The most distinct clast populations occurred toward the end of Plinian activity preceding effusive dome growth. Distributions skewed toward small sizes, thick walls, and teardrop vesicle shapes are indicative of bubble wall collapse marking maturation of the melt and onset of processes of outgassing. The data suggest that the superficially similar pauses in the 1912 eruption which marked the ends of episodes II and III had very different causes. Through Episode III, the trend in vesicle size data reflects a progressive shift in the degassing process from rapid magma ascent and coupled gas exsolution to slower ascent with partial open-system outgassing as a precursor to effusive dome growth. No such trend is visible in the Episode II clast assemblages; we suggest that external changes involving failure of the conduit/vent walls are more likely to have effected the break in explosive activity at 45 h. ?? Springer-Verlag 2006.