46o ALBERT V. G. JAMES 



thus see that in a homogeneous lava, vertical cracks arranged in 

 hexagons, slowly penetrate the solid lava from above, downward. 

 SoUdification of the lava precedes the cracking. 



The size of the columns depends chiefly on the temperature 

 and rate of cooHng. The more rapidly cooled a flow is the thinner 

 will be the columns. 



Owing to the fact that lava masses are seldom homogeneous 

 (on account of included gases, convection currents within, varia- 

 tion in chemical composition, included fragments, etc.) it seldom, 

 if ever, occurs, that only hexagonal prisms are formed. It does not 

 need the fact that olivine crystals have been found divided by 

 vertical joints, nor the preservation of flow structure in glass to 

 show us that the splitting took place in solid lava.^ It is difficult 

 to imagine liquid or even plastic lava cracking. Tension in that 

 case is relieved by fiowage, not by cracking. 



For the following reasons the writer beheves that columnar 

 structure in basalt is due to slow rather than rapid cooling: 



1. Basalt has a very low thermal conductivity and therefore 

 great thicknesses of molten basalt must necessarily cool slowly. 

 The fact that columnar basalt is not well developed where flows 

 are thin appears to be due to two factors: (a) heterogeneity of 

 the lava, (b) rapid cooling of the thin flow. 



2. Cracking takes place in the hot solid rock when the contrac- 

 tion becomes greater than the expansion under tension. The rock 

 is able to resist parting for a longer period when the tension is very 

 slowly increased by slow cooling. If tension is more rapidly 

 increased the fl'ow will break into thin columns or fragments. These 

 statements are based on physical laws. The rate of cooHng at the 

 upper surface is much more rapid than at the lower because con- 

 vection currents in the air above rapidly convey away the heat but 

 the valley floor being a poor heat conductor offers much resistance to 

 the transference of heat. In a thick flow, columns grow from all 

 cooling planes. The upper columns are found to rest on thick basal 

 columns and this fact upholds the statement that the size of 

 columns is determined by rate of coohng. 



'J. P. Iddings, Anier. Jour. Sci., XXXI (1886), p. 324; R. B. Sosman, Jour. 

 Geo/., XXIV (i9i6),p. 218. 



