Peralkaline rhyolitic lava-like lobes showing direct evidence for a spatter-fed origin are associated with eruptive centres that document a transition from explosive to effusive fountain-fed activity. Owing to the preservation of spatter clasts, clast dimensions can be used to model the eruptive dynamics of these deposits. We conducted an integrated field, experimental and theoretical study of the Cala di Tramontana centre, Pantelleria, Italy, with the aim to better understand its eruptive behaviour. This centre is composed of four units and a close association of fall lapilli, tack-welded agglutinate, and spatter-fed lava-like lobes. These units were fed initially from a fissure vent, and later, from a circular vent. For the last-erupted lobe, field evidence indicates that fall-out accumulated rapidly, coalescing into a 10 m thick deposit that flowed downslope and became lava-like. High eruption rates and high heat flows are interpreted to promote welding of spatter proximal to the vent. Overburden pressures and thermal insulation from the accumulating spatter caused complete homogenisation of spatter within the interior of the flow. Laboratory vesiculation experiments show that peralkaline obsidian can vesiculate at temperatures as low as 620 °C (over a 4 hour period), and indicate that airborne agglutinate can deform and weld rapidly within minutes of landing. We measured viscosity on a natural Pantellerian sample using the parallel-plate method over a temperature range of 560–600 °C. Viscosities of 1010–1011 Pa s, and an activation energy of 311 kJ/mol were obtained. Based on undeformed clast sizes and densities, eruption dynamics modelling indicates that gas eruption velocities ranged from ˜ 90 to ˜ 130 m s−1. Erupted mass fluxes were ˜ 200 kg s−1 m−1 and ˜ 2 × 104 kg s−1 from the fissure and localised circular vents, respectively. Magmatic viscosities of between 105 and 107 Pa s (for T = 750–850 °C) precluded intermittent Strombolian activity due to gas bubble coalescence. In addition, the amount of magmatic water available to drive the eruption fountains ranged from 0.4 to 0.6 wt.%—significantly less than the estimated ˜ 1.5 wt.% of exsolved volatiles—implying that significant passive degassing occurred.