330 
PACIFIC SCIENCE, Vol. XIX, July 1965 
The failure of magma to rise along the 
caldera-boundary fractures itself calls for an ex- 
planation. The erupting basaltic magma has a 
specific gravity of about 2.7, and it appears 
likely that the magma even at the depth of 
the magma chamber has a density of only about 
2.73 (Macdonald, 1963:1076). Compared with 
this, the gross density of the caldera-filling rocks 
is at least 2.8, and probably is between 2.9 and 
3.0. If a mass of this density is underlain by a 
magma body of equal or greater horizontal di- 
mensions and lower density, why does not the 
caldera block sink completely into the magma? 
The answer probably lies in the wedge form 
of the sinking block, bounded by fractures that 
dip inward instead of outward. Reynolds 
( 1956) has pointed out that a downward con- 
vergence of the boundary fractures is implied 
by the up-bending of the edges of the lava beds 
filling many cauldron subsidences, including 
that of Glen Coe. The margins of the older, 
eroded Hawaiian calderas are seldom well 
enough exposed to reveal whether or not the 
edges of the beds are bent upward. However, 
Stearns (1940: Fig. 7) has described a basin- 
ing of the lavas in the Koolau caldera, Oahu, 
and at least in one sector the beds filling the 
Kauai caldera are dragged slightly upward 
against the caldera boundary ( Macdonald, Davis, 
and Cox, 1960:36). There is a definite implica- 
tion that the boundary faults converge down- 
ward. Since sinking of a wedge-shaped block 
would tend to keep the boundary fractures 
tightly closed, this would also help to explain 
the failure of magma to rise to the surface 
along them and the lack of eruptive vents on 
the caldera boundary. 
Why do the fractures converge downward, 
instead of diverging in the ring-dike manner 
deduced mathematically by Anderson (1936)? 
Perhaps the answer lies in the fact that the 
fractures were first established as a result of 
upthrust of magma beneath a relatively small 
portion of the mountain top, resulting in up- 
ward-divergent fractures of cone-sheet type, 
and that once established these fractures served 
as the surfaces on which the caldera block later 
sank. The tumescence frequently observed at 
Kilauea shows that magmatic pressure is great 
enough to push up the top of the mountain, 
and the piston-like rise of the caldera floor in 
the 1850’s shows that at times the elevation 
takes place by displacement of a fault-bounded 
block rather than by quasi-plastic arching. 
Where upward pressure continues long enough 
cone sheets may form, like the numerous con- 
centric inward-dipping dikes that surround the 
caldera of the Ofu-Olosega volcano in Samoa 
(McCoy, 1965); and concentric lines of spat- 
ter and cinder cones may form by surficial erup- 
tion on these fractures, as on some of the vol- 
canoes of the Galapagos Islands (H. Williams, 
personal communication, 1964). In Hawaii, 
however, only a few dikes with the attitudes 
of cone sheets have been found, on Oahu and 
Kauai. Even fewer examples are known of erup- 
tion on concentric fractures, but one such line 
of cones lies just southwest of - Kilauea caldera. 
For some reason, in Hawaii magmatic pressure 
has usually resulted in distension of the volcanic 
structure by upward bending of the summit 
followed by splitting open of the rift zones 
instead of lifting of the apex of the volcano on 
inward-dipping conical fractures. 
A factor that must be explained before we 
can accept Williams’ Kilauean mechanism is 
the very considerable discrepancy that exists 
between the volumes of some of the subsidences 
in the caldera during historic times and those 
of the simultaneously-erupted lava flows. The 
volumes of historic subsidences are listed in 
Table 1, the figures being taken from papers 
by Finch ( 1940, 1941) except the one for 1955, 
which is from Macdonald and Eaton (1964). 
The volumes for 1924 and 1955 include both 
marked collapses at Halemaumau and a general 
sinking of the whole mountaintop over a radius 
of 15 km or more, but this was undetectable 
without instrumental measurements in the ear- 
lier episodes of collapse. In these the figure 
given is only for the conspicuous sinking that 
took place in the caldera. Undoubtedly, how- 
ever, a wider-spread general subsidence, like 
those in 1924 and 1955, also took place during 
each of the earlier episodes of collapse, and 
the volumes of those were accordingly greater 
than shown. The volumes of the lava flows are 
taken from Stearns and Macdonald (1946), 
again except for that of 1955. They include 
estimates of the volume of subaerial flows that 
