We introduce a novel scheme that enables natural silicic glasses to be proj
ected into the synthetic system Qz-Ab-Or-H2O in order to relate variations
in volcanic glass chemistry to changing pressure (P) and temperature (T) co
nditions in the sub-volcanic magma system. By this means an important disti
nction can be made between ascent-driven and cooling-driven crystallisation
under water-saturated or undersaturated conditions. In samples containing
feldspar and a silica phase (quartz or tridymite), quantitative P-T estimat
es of the conditions of last equilibrium between crystals and melt can be m
ade. Formation of highly silicic melts (i.e. >77 wt% SiO2) is a simple cons
equence of the contraction of the silica phase volume with decreasing press
ure, such that high silica glasses can only form by crystallisation at low
pressure. Resorption of quartz crystals appears to be a further diagnostic
feature of decompression crystallisation. Groundmass and inclusion glasses
in dacites from the 1980-1986 eruption of Mount St Helens volcano (WA) span
a wide range in SiO2 (68-80 wt%, anhydrous). The compositions of the least
evolved (SiO2-poor) inclusions in amphibole phenocrysts record entrapment
of silicic liquids with less than or equal to 5.4 wt% water, corresponding
to a water saturation pressure of similar to 200 MPa at 900 degreesC. The c
ompositions of more evolved (higher SiO2) plagioclase-hosted inclusions and
groundmass glasses are consistent with extensive ascent-driven fractional
crystallisation of plagioclase, oxide and orthopyroxene phenocrysts and mic
rolites to low pressures. During this polybaric crystallisation, plagioclas
e phenocrysts trapped melts with a wide range of dissolved water contents (
3.5-5.7 wt%). Magmas erupted during the Plinian phase of the 18 May 1980 er
uption were derived from a large reservoir at depths of greater than or equ
al to6 km. Subsequent magmas ascended to varying depths within the sub-volc
anic system prior to extraction. From glass chemistry and groundmass textur
e two arrest levels have been identified, at depths of 0.5-1 and 2-4 km. A
single dome sample from February 1983 contains groundmass plagioclase, trid
ymite and quartz, testifying to temperatures of at least 885 degreesC at 11
MPa. These shallow storage conditions are comparable to those in the crypt
odome formed during spring 1980. The corresponding thermal gradient, less t
han or equal to 0.2 degreesC MPa-1, is consistent with near-adiabatic magma
ascent from similar to8 km. We argue that the crystallisation history of M
ount St Helens dacite magma was largely a consequence of decompression crys
tallisation of hot magma beyond the point of water saturation. This challen
ges the conventional view that phenocryst crystallisation occurred by cooli
ng in a large magma chamber prior to the 1980-1986 eruption. Because the cr
ystallisation process is both polybaric and fractional, it cannot be simula
ted directly using isobaric equilibrium crystallisation experiments. Howeve
r, calculation of the phase proportions in water-saturated 910+/-15 degrees
C experiments by Rutherford et al. (1985) over the pressure range 220-125 M
Pa reproduces the crystallisation sequence and phenocryst modes of Mount St
Helens dacites from 18 May 1980. By allowing for the effects of fractional
versus equilibrium crystallisation, entrained residual source material, an
d small temperature differences between nature and experiment, phase compos
itions can also be matched to the natural samples. We conclude that decompr
ession of water-saturated magma may be the dominant driving force for cryst
allisation at many other silicic volcanic centres.