During the past two years the teaching collection of rocks in the Geological Department of the Imperial College has been underre-arrangement, and in the course of the work the writer has had occasion to consider very carefully the principles on which igneous rocks should be classified. He is aware that in the present state of our knowledge any such classification must be tentative and experimental, and should be judged according to its general convenience, both for teaching purposes and for understanding the various problems that arise from petrological studies. The following article summarizes the conclusions arrived at, and while the writer must be held entirely responsible for all expression of opinion, he wishes to acknowledge his gratitude and indebtedness to Professor Watts and Dr. Evans, both of whom, in numerous discussions, have freely offered suggestions and criticism which have proved to be of the greatest value.