The prominent felsic granulites in the southern part of the Bohemian Massif (Gföhl Unit, Moldanubian Zone), with the Variscan (∼340 Ma) high-pressure and high-temperature assemblage garnet+quartz+hypersolvus feldspar ± kyanite, correspond geochemically to slightly peraluminous, fractionated granitic rocks. Compared to the average upper crust and most granites, the U, Th and Cs concentrations are strongly depleted, probably because of the fluid and/or slight melt loss during the high-grade metamorphism (900–1050°C, 1·5–2·0 GPa). However, the rest of the trace-element contents and variation trends, such as decreasing Sr, Ba, Eu, LREE and Zr with increasing SiO2 and Rb, can be explained by fractional crystallisation of a granitic magma. Low Zr and LREE contents yield ∼750°C zircon and monazite saturation temperatures and suggest relatively low-temperature crystallisation. The granulites contain radiogenic Sr (87Sr/86Sr340 = 0·7106–0·7706) and unradiogenic Nd ( = − 4·2 to − 7·5), indicating derivation from an old crustal source. The whole-rock Rb–Sr isotopic system preserves the memory of an earlier, probably Ordovician, isotopic equilibrium.

Contrary to previous studies, the bulk of felsic Moldanubian granulites do not appear to represent separated, syn-metamorphic Variscan HP–HT melts. Instead, they are interpreted as metamorphosed (partly anatectic) equivalents of older, probably high-level granites subducted to continental roots during the Variscan collision. Protolith formation may have occurred within an Early Palaeozoic rift setting, which is documented throughout the Variscan Zone in Europe.

The high-dT/dP-type Hidaka Metamorphic Belt in Hokkaido, northern Japan, represents a tilted crustal section of a magmatic arc of Tertiary age. The highest metamorphic grades reached are granulite facies, and the syn-metamorphic granitic rocks are widely distributed in this metamorphic terrane. The granitic rocks are mainly tonalitic and granodioritic in composition, and are classified into peraluminous (S-type) and metaluminous (I-type) granitoids. A large amount of pyroxene-bearing S-type tonalites (garnet-orthopyroxene tonalite) is distributed in the Niikappu river region in the northern part of the Hidaka Metamorphic Belt. Pyroxene-bearing I-type tonalite (two-pyroxene hornblende tonalite) bodies are also distributed in this area.

The pyroxene-bearing tonalites are classified into several sub-types on the basis of their field occurrence, texture, mineral assemblage and geochemical features. Homogeneous IH- and SH-type tonalite are thought to represent original magmas, i.e. those which have been generated by partial melting of mafic metamorphic rocks and pelitic-psammitic metamorphic rocks, respectively. Model calculations assuming batch partial melting indicate that possible restites are garnet-two-pyroxene mafic granulite for IH-type and garnet-orthopyroxene aluminous granulite for SH-type. The unexposed lowermost crust of the ‘Hidaka crust’ is thought to be composed of garnet-two-pyroxene mafic granulite, garnet-orthopyroxene aluminous granulite and metagabbros.

The exposed Precambrian cratonic crust in South Korea is divided into two massifs – the Gyeonggi massif to the north and the Yeongnam massif to the south. Mesozoic granites intruded into both massifs and are mostly I-types. The Jurassic granites form extensive deep-seated batholiths, the Triassic granites are deep-seated stocks and the Cretaceous granites occur as volcanic-plutonic complexes.

The systematic variation of εNd and SrI in the Korean Mesozoic granites could result from the mixing of two components in different proportions to produce the source of the granites. Although most Mesozoic I-type granites were apparently derived from more juvenile crust, the old evolved crustal components seem to have been incorporated in the magmas in various proportions. Mantle–crust mixing can account for the generation of the source of the Triassic and Cretaceous granites in the Gyeongsang basin. On the other hand, crust–crust mixing can feasibly produce the source of the Triassic and Jurassic granites in the Yeongnam massif, the Jurassic granites in the Gyeonggi massif, and the Cretaceous granites in the Yeongdong–Gwangju basin and the Okcheon belt. However, some Jurassic granites in the Yeongnam massif and Cretaceous granites in the Yeongdong–Gwangju basin can be also explained by the mantle–crust mixing. Combined geochemical and isotopic signatures indicate that a simple binary mixing model is inadequate to explain both the geochemical and isotopic data. The chemistry of the granites is considered likely to reflect the composition of the igneous protolith that derived from depleted mantle, which explains why most Mesozoic granites in South Korea are represented by I-types, regardless of their temporal and spatial position. Nd-Sr isotopic signatures of the Mesozoic granites and basement rocks indicate that the continental crust beneath the Korean peninsular is vertically structured by the successive underplating of mantle-derived materials. It is postulated that the crust is vertically stratified from the surface to the lowermost crust with late Archean to early Proterozoic, early to middle Proterozoic (ca. 1·9 Ga), middle Proterozoic (ca. 1·5 Ga), and late Proterozioc (younger than 1·5 Ga) components.

A strong correlation exists between the Li isotopic compositions of CarboniferousTriassic granites from the New England Batholith, and the previously inferred involvement of sedimentary and mantle/infracrustal source components. Isotopically (Nd and Sr) juvenile, low-K, Cordilleran I-type granites of the Clarence River supersuite have δ7 Li= +2·2 to +8‰ similar to those of arc magmas, the inferred source of these granites (Bryant et al. 1997). Isotopic variability within this supersuite probably arises from heterogeneity within primary mantle-derived magmas, combined with subsequent modifications through interactions with crustal materials. Oxidised, high-K granites of the Moonbi Supersuite have more homogenous and slightly lighter Li isotopic compositions (δ7 Li= +1·9 to +4·2‰). The observed range of values lies within the range of arc magmas, and is consistent with partial melting of arc shoshonites within the crust (cf. Chappell 1978) or the involvement of high-K mantle-derived magmas (cf. Shaw & Flood 1981; Landenberger & Collins 1998). S-type granites of the Bundarra (δ7 Li= −0·1 to +2·1‰; average= +1˙3‰; n=6) and Hillgrove supersuites (δ7 Li= +0·4 to +1·7‰; average= +0·8‰) define a narrow range of isotopic compositions which are, overall, lower than those observed in NEB I-type granites or generally observed in primary arc magmas. Their isotopic compositions are equivalent to those typically observed in shales (primarily δ7 Li= −3·2 to +2·0‰; Moriguti & Nakamura 1998; Teng et al. 2004). No difference is evident in the isotopic compositions of the two S-type supersuites despite inferred differences in the degree of weathering experienced by the sedimentary protolith, or differences in mineralogy of the granites. Granites of the Uralla Supersuite, which have been have formed from mixtures of local meta-igneous and meta-sedimentary components, span a broad range of values (δ7 Li= −1·3 to +3·9‰) which overlap with both the sediment-poor New England Batholith I-type intrusions of the Clarence River and Moonbi supersuites, and the S-type granites of the Bundarra and Hillgrove supersuites. Lower δ7 Li values primarily occur in lower-K plutons from the northern portion of the Uralla Supersuite.

Overall, anatexis and magma differentiation do not appear to contribute to significant fractionation of Li isotopes relative to the inferred source components. However, subtly lower δ7 Li values, evident in the three leucogranites analysed herein, imply that subtle Li isotopic fractionation may occur in association with the exsolution of an aqueous fluid. Like most isotopic systems, the Li isotopic composition of rocks is not a definitive guide to source rock compositions, but given the results herein, the present authors suggest that it may play a very useful role in understanding crustal processes.

Several granites and basement gneisses along the southern margin of the Siberian Craton were sampled to obtain precise age constrains and compositional isotopic data. The analysed granite plutons are interpreted to have been emplaced between 1880 and 1850 Ma, and are related to Palaeoproterozoic collisional and post-collisional events. Pb-Pb whole data of a granulite (1884 ± 26 Ma) and a two mica granite (1821 ± 29 Ma) constrain these U-Pb single zircon ages. Sr, Nd and Pb-Pb isotope data reveal the crustal origin of the investigated rocks and the reworking of Archaean material. Nevertheless, a minor influence of a mantle component is still visible in the Pb isotopes. Geodynamically, the magmatic and metamorphic ages in the Kitoy area are linked to several Early Proterozoic events along the southern margin of the Siberian Craton. Since these events are all older than the assembly of Rodinia (about 1·4 Ga ago) the collisional processes are linked to the consolidation of the Siberian Craton itself in Early Proterozoic times.

Late Cretaceous (90–100 Ma) A-type granites are widespread in the coastal area of the Zhejiang and Fujian Provinces, SE China. According to mineralogical and geochemical characteristics, the A-type granites in this belt can be further divided into aluminous and peralkaline subgroups. The aluminous subgroup often contains aluminous-rich minerals (e.g. spessartine and Mn-rich muscovite), while the peralkaline subgroup usually contains riebeckite, arfvedsonite and aegirine. Geochemically, the aluminous A-type granites show lower Nb, Zr, Ga, Y and REE abundances, and lower FeO*/MgO and Ga/Al than the peralkaline subgroup. When they occur in the same area, the two subgroups of A-type granites display quite similar initial Nd isotopic compositions, which are indicative of mixing of ancient basement crustal rocks with variable amounts of mantle materials. Integrated geological and geochemical investigations indicate that both the aluminous and the peralkaline magmas are highly evolved and reflect the residual liquids left after high degrees of fractional crystallisation in a deep magma chamber. The present authors suggest that the mineralogical and geochemical differences between the aluminous and peralkaline subgroups are likely to have been generated via different differentiation paths controlled by varying fluorine contents of the parent magmas.

Late Yanshanian magmatism in SE China includes three stages of thermal event induced by the interaction between the continental margin of Eurasia and the paleo-Pacific plate during the Cretaceous period. Products of syn-orogenic magmatism (130–110 Ma) include high-Al gabbros (HAG), and gneissic tonalite, trondhjemite and granodiorite (TTG), which intruded into the deep basement (18–24 km). Rocks of the post- and an-orogenic magmatism are shallow-level (6–8 km) I-type granitoids (110–99 Ma), and miarolitic A-type granites plus rhyolite-dominate bimodal volcanics (94–81 Ma), respectively. Geochemically, HAG and TTG belong to the medium-K calc-alkaline affinity with high Sr/Y, whereas other granitoids are mainly high-K calc-alkaline to shoshonitic rocks with low Sr/Y. Sr and Nd isotope compositions suggest different sources of HAG and TTG from other rocks. Progressive depletions of Ba, Sr, Eu and P from I- to A-type granites reflect partial melting of felsic granulites from hydrous to dry conditions, whereas high Sr/Y in HAG and TTG are compatible with dehydration melting of amphibolites. Tectonic models which accommodate HAG and TTG may involve thickening of the lithosphere to convert the pre-existing lower-crust basic rocks into amphibolites. It was followed by basaltic underplating which is attributed to delamination of the thickened lithosphere and led to triggering of crust melting under exten

Mafic rocks in the Ryoke belt, the Cretaceous granitic province in Southwest Japan, occur in two modes: (1) as mafic dykes and pillow-shaped enclaves, and (2) as isolated kilometresized bodies of gabbroic cumulate. The dykes and pillows have fine-grained textures with thin radiating plagioclase laths, indicative of quenching. The gabbroic cumulates are predominantly coarse-grained and commonly lithologically layered.

SHRIMP zircon U-Pb ages of both types of mafic rocks are in the range 71–86 Ma, late Cretaceous. The mafic rocks become younger eastwards, matching the along-arc age trend of the associated Cretaceous granites (Nakajima et al. 1990). Both types of mafic rocks were apparently generated during the same magmatic event that produced the Ryoke/San-yo granites. The mafic dykes and pillows are aphyric basaltic-andesites to andesites (SiO2 54–60 wt.%), with microphenocrysts of biotite and hornblende. They have a composition which is similar to mafic rocks from the northern Sierra Nevada, and also to medium-K calc-alkaline rocks from present-day arc volcanics. The gabbroic cumulates are mostly pyroxene-hornblende gabbros (SiO2 43–52 wt.%). Their bulk-rock chemical compositions are mostly unlike any magma compositions.

Both types of mafic rocks from the Ryoke belt have relatively high 87Sr/86Sr initial ratios (SrI), 0·7071–0·7097, which are similar to those of the associated granites. The granites were formed either by fractional crystallisation of the mafic magmas, or by partial melting of newly formed mafic rocks at depth. The high SrI indicates that the mafic magmas were derived from enriched mantle or mixed with enriched crustal materials. Even if the mixing occurred between primitive basaltic magma and metasedimentary rocks, then the basaltic andesite–andesite magmas must have contained more than 60% mantle-derived components. The Cretaceous magmatism in Southwest Japan represents a major episode of crustal growth by additions from the upper mantle in an arc setting.

Felsic and related veins within mantle-derived peridotite xenoliths from Tallante, Spain, were examined in order to understand the mantle-wedge processes related to the behaviour of Si-rich melt. The thickest part of the vein has a quartz diorite lithology, and is composed mainly of quartz and plagioclase, with pyroxenes, hydrous mineral, apatite, zircon and rutile present as minor phases. The thinner parts are free of quartz and predominantly composed of plagioclase. Orthopyroxene always intervenes between the internal part (plagioclase ± quartz) and host peridotite, indicating that it is a product of interaction between silica-oversaturated melt and olivine. This indicates that a sufficiently high melt/wall rock ratio enabled the melt to retain its silicaoversaturated character.

The quartz diorite part has adakite-like geochemical signatures, except for negative Ba, Rb Eu and Sr anomalies, and positive Th and U anomalies. These negative anomalies indicate that fractionation of plagioclase and hydrous minerals was achieved between the upper most mantle and the slab melting zone. The shape of the rare-earth element (REE) pattern of clinopyroxene in quartz diorite is strikingly similar to that of clinopyroxene phenocrysts from Aleutian adakites. However, the former has one order higher REE contents than the latter, except for Eu which shows a prominent negative spike. This feature was caused by the precipitation of large amounts of plagioclase and small amounts of clinopyroxene from a fractionated adakitic melt before and during solidification. This adakitic melt was produced by partial melting of a detached and sinking slab beneath the Betic area in the Tertiary.

The mid-Miocene Aztec Wash pluton is divisible into a relatively homogeneous portion entirely comprising granites (the G zone, or GZ), and an extremely heterogeneous zone (HZ) that includes the products of the mingling, mixing and fractional crystallisation of mafic and felsic magmas. Though far less variable than the HZ, the GZ nonetheless records a dynamic history characterised by cyclic deposition of the solidifying products of the felsic portion of a recharging, open-system magma chamber.

Tilting has exposed a 5-km section through the GZ and adjacent portions of the HZ. A porphyry is interpreted as a remnant of a chilled roof zone that marks the first stage of felsic GZ intrusion. Subsequent recharging by felsic and mafic magma, reflected by repeated cycles of crystal accumulation and melt segregation in the GZ and emplacement of mafic flows in the HZ, rejuvenated and maintained the chamber. Kilometre-scale lobes of mafic HZ material were deposited as prograding tongues into the GZ during periods of increased mafic input. Thus, they are lateral equivalents of the cumulate GZ granites with which they interfinger. Conglomerate-like units comprising rounded, matrix-supported intermediate clasts in cumulate granite are located immediately above the lobes. These ‘conglomerates’ appear to represent debris flows shed from sloping upper surfaces of the lobes. Thus, the GZ can be viewed as comprising distal facies, remote from the site of mafic recharging in the HZ, and the HZ as comprising proximal facies.

Elemental chemistry suggests that the GZ cumulate granites represent a second-stage accumulation from an already evolved melt, and that coarse, more mafic, feldspar+biotite+accessory mineral ± hornblende rocks trapped between mafic sheets in the HZ are the initial cumulates. Fractionated melt accumulated roofward and laterally, and was the direct parent of the ‘evolved’ GZ cumulates. The most highly fractionated, fluid-rich melts accumulated at the roof.

The Kenogamissi complex represents a large exposure of folded Late Archaean crystalline crust exposed within the Abitibi Subprovince, Ontario, Canada. It is composed of an heterogeneous amphibolite-grade orthogneiss unit, and several generations of batholiths and plutons of tonalite, granodiorite and granite composition. Together, the various units represent granitic magmatism during the period from 2740 Ma to 2660 Ma. Structural mapping and petrographic studies were focused on the orthogneiss unit (2723 Ma), on the newly defined Roblin tonalitegranodiorite batholith (ca. 2713 Ma) and on the highly strained metavolcanic rocks within the deformation aureole that surrounds the Kenogamissi complex. Structural analysis indicates that the Kenogamissi complex was emplaced into the greenstones as a dome that caused severe flattening and recumbent F2 refolding of earlier F1 folds in the deformation aureole. Doming is interpreted to be caused by the emplacement and inflation of tonalite-granodiorite batholiths, such as the Roblin Batholith, into the actively folding Swayze greenstone belt. Continued regional folding resulted in F3 refolding of F1 and F2 in the deformation aureole. Continued regional folding also deformed and folded the Kenogamissi complex and resulted in further uplift and emplacement of the complex into the greenstone belt. The early-formed magmatic foliation and compositional layering in the Roblin Batholith were folded by F3 while the batholith was still a crystal mush, and an F3 axial-surface magmatic foliation was locally formed. Folding of the partially molten Roblin Batholith also resulted in the remobilisation of fractionated liquids into shear zones which formed on the limbs of the F3 magmatic folds. Similar structures are present in the orthogneiss unit and are interpreted to represent remobilisation of melts which intruded the orthogneiss at the time of emplacement of the Roblin Batholith. The formation of the dykes on sheared fold limbs may be attributed to increased dilatancy during localised shearing of the crystal mush. Deformation-assisted remobilisation and extraction of fractionated liquids, and the possible transport of the fractionated liquids to higher levels in the crystallising Roblin Batholith, may have played a role in its magmatic differentiation.

Combined field and geophysical data show that plutons from the Bega Batholith are elongate, meridional, wedge-shaped bodies which intruded during a period of regional east–west extension in the Palaeozoic eastern Lachlan orogen, eastern Australia. Plutons within the core of the batholith have intruded coeval, syn-rift sediments and co-magmatic volcanics. The batholith is bound by high-temperature, dip-slip faults, and contains several major NE-trending transtensional faults which were active during batholith construction. In the central part of the batholith, the Kameruka pluton is an asymmetric, eastward-thickening, wedge-shaped body with the base exposed as the western contact, which is characterised by abundant, shallow-dipping schlieren migmatites which contain recumbent folds and extensional shear bands. A shallow (<30°), east-dipping, primary magmatic layering in the Kameruka pluton steepens progressively westward, where it becomes conformable to the east-dipping basal migmatites. The systematic steepening of the layering is comparable to sedimentary units formed during floor depression in syn-rift settings. The present authors suggest that the wedge-shaped plutons of the Bega Batholith are the deeper, plutonic expression of a hot, active rift. The batholith was fed and sustained by injection of magma through sub-vertical dykes. Displacement along syn-magmatic, NE-trending faults suggests up to 25 km of arc-perpendicular extension during batholith construction. The inferred tectonic setting for batholith emplacement is a continental back-arc, where modern half-extension rates of 20–40 mm yr−1 are not unusual, and are sufficient to emplace the entire batholith in ∼1 Ma. This structural model provides a mechanism for the emplacement of some wedge-shaped plutons and is one solution to the ‘room problem’ of batholith emplace

Felsic magmatism has occurred over a large region of East Asia since Jurassic times and has provided important mineral resources such as tin, tungsten, base metals and gold. The circum-Japan Sea region preserves various geological records of active continental margins, including Jurassic to Early Tertiary magmatic arcs and subduction zones and pre-Jurassic continental basements, which were separated by the opening of the Japan Sea during the Miocene. The felsic magmatism in this region shows a wide variation in terms of redox state and related mineralisation, encompassing east–west contrasts around the Pacific Ocean. A review of granitoids and associated ore deposits in this region indicates that the character of the crust, sedimentary versus igneous, is an essential factor to control the redox state, and a tectonic setting may be an additional factor in some cases.

The reduced-type granitoids, characterised by tin mineralisation, were generated in carbonbearing sedimentary crust which was composed mainly of accretionary complex material and not influenced by previous magmatism. Involvement of sedimentary materials is corroborated by oxygen, sulphur and strontium isotope data. The oxidised-type granitoids, characterised by gold or molybdenum mineralisation, were generated in igneous crust which was depleted in reducing agents as a result of previous magmatism. Granitoid magmatism in a given area tends to become more oxidised with time.

Jurassic accretionary complexes in East Asia are thought to have been largely displaced from the original place of accretion and stacked up against the northeastern margin in the Khingan and Sikhote–Alin Mountains. This region, dominated by sedimentary crust, was subsequently subjected to Cretaceous felsic magmatism and converted to a large province of reduced-type granitoids and tin–tungsten mineralisation. Diverse geodynamic processes, including the change of the arc-trench system, the creation and collapse of the back-arc basin and the collision of continents, may have prepared many favourable sites for the generation of reduced-type granitoids in northeast Asia. These processes may have resulted in a remarkable contrast with the Pacific margin of North America, where repeated arc magmatism during the Mesozoic formed granitoid batholiths of the oxidised-type.

The granitoid types may also be controlled by the tectonic setting and mode of magma emplacement. In the northern Kitakami area of Northeast Japan, Early Cretaceous episodic magmatism occurred in a Jurassic accretionary complex, and formed the oxidised-type granitoids accompanied by submarine bimodal volcanism associated with kuroko mineralisation. Granitoids of fissure-filling type emplaced under extensional environments may be oxidised, irrespective of basement geology, because of insignificant crustal input.

The Tombstone, Mayo and Tungsten plutonic suites of granitic intrusions, collectively termed the Tombstone-Tungsten Belt, form three geographically, mineralogically, geochemically and metallogenically distinct plutonic suites. The granites (sensu lato) intruded the ancient North American continental margin of the northern Canadian Cordillera as part of a single magmatic episode in the mid-Cretaceous (96–90 Ma). The Tombstone Suite is alkalic, variably fractionated, slightly oxidised, contains magnetite and titanite, and has primary, but no xenocrystic, zircon. The Mayo Suite is sub-alkalic, metaluminous to weakly peraluminous, fractionated, but with early felsic and late mafic phases, moderately reduced with titanite dominant, and has xenocrystic zircon. The Tungsten Suite is peraluminous, entirely felsic, more highly fractionated, reduced with ilmenite dominant, and has abundant xenocrystic zircon. Each suite has a distinctive petrogenesis. The Tombstone Suite was derived from an enriched, previously depleted lithospheric mantle, the Tungsten Suite is from the continental crust including, but not dominated by, carbonaceous pelitic rocks, and the Mayo Suite is from a similar sedimentary crustal source, but is mixed with a distinct mafic component from an enriched mantle source.

Each suite has a distinctive metallogeny that is related to the source and redox characteristics of the magma. The Tombstone Suite has a Au-Cu-Bi association that is characteristic of most oxidised and alkalic magmas, but also has associated, and enigmatic, U-Th-F mineralisation. The reduced Tungsten Suite intrusions are characterised by world-class tungsten skarn deposits with less significant Cu, Zn, Sn and Mo anomalies. The Mayo Suite intrusions are characteristically gold-enriched, with associated As, Bi, Te and W associations. All suites also have associated, but distal and lower temperature Ag-Pb- and Sb-rich mineral occurrences. Although processes such as fractionation, volatile enrichment and phase separation are ultimately required to produce economic concentrations of ore elements from crystallising magmas, the nature of the source materials and their redox state play an important role in determining which elements are effectively concentrated by magmatic processes

The northern Colorado River extensional corridor (NCREC, USA) provides an excellent record of coeval volcanic and mid- to upper-crustal (<13 km) plutonic suites. The NCREC is a 50–100-km-wide zone that records late Tertiary lithospheric extension, volcanism, continental sedimentation and plutonism. Compilation of published studies of NCREC magmatic rocks permits an assessment of volcanic–plutonic links, magma sources and magmatic processes. The volcanic sections provide an excellent record of magma compositions (basalt, trachyandesite, trachyte and rhyolite) which span a 9-million-year period in the Miocene age (20–11 Ma).

Contemporaneous Miocene plutons span a similar compositional range (gabbro, diorite, quartz monzonite and granite) and were emplaced during a 4·5-million-year interval from 17 to 12·5 Ma. Geochemical and isotopic compositions and compositional trends allow direct correlation between plutonic and volcanic suites across the entire compositional range. Petrogenetic models demonstrate that intermediate magmas formed by a combination of magma mixing and fractional crystallisation involving mantle-derived mafic with crustal-derived felsic end-member magmas. Plutons exhibit a variety of features which suggest magma chamber processes, including (1) mafic cumulate sequences, (2) felsic cumulate sequences, and (3) magma mingling and advanced stages of magma mixing. Thus, the NCREC plutonic-volcanic record provides a link between magmatic processes recorded in pluton magma chambers and magmatic products in the form of extrusive igneous rocks. The NCREC plutons represent upper crustal magma chambers which connected volcanic eruptive centres to deeper-level magma chambers, and ultimately, to zones of mantle and crustal mel

The Bishop Tuff (USA) is a large-volume, high-silica pyroclastic rhyolite. Five pumice clasts from three early stratigraphic units were studied. Size distributions were obtained using three approaches: (1) crushing, sieving and winnowing (reliable for crystals >100 μm); (2) microscopy of ∼1 mm3 fragments (preferable for crystals <100 μm); and (3) computerised X-ray microtomography of ∼1 cm3 pumice pieces.

Phenocryst fragments coated with glass are common, and the size distributions for all crystals are concave-upward, indicating that crystal fragmentation is an important magmatic process.

Three groups are recognised, characterised by: (1) high-density (0·759–0·902 g cm−3), high-crystal content (14·4–15·3 wt.%) and abundant large crystals (>800 μm); concave-downward size distributions for whole crystals indicate late-stage growth with limited nucleation, compatible with the slow cooling of a large, gas-saturated, stably stratified magma body; (2) low-density (0·499 g cm −3), low-crystal content (6·63 wt.%) and few large crystals; the approximately linear size distribution reveals that nucleation was locally important, perhaps close to the walls; and (3) intermediate characteristics in all respects.

The volumetric fraction of bubbles inversely correlates with the number of large crystals. This is incompatible with isobaric closed-system crystallisation, but can be explained by sinking of large crystals and rise of bubbles in the magma