Ontogenetic development, says Roux, is the production of a _visible_ manifoldness. It cannot be said that this cautious description of the developmental process has been apprehended by those who expound the dogmas of mechanistic biology. Development is indeed the production of a diversity, but this diversity is only a phase of a preceding diversity, a rearrangement of spatially extended pre-existing elements. How else could the developing embryo and its material environment be regarded as a system of physico-chemical elements, capable of study by the methods of experimental and mathematical physics, except by regarding it as a system passing through phases each of which is a necessary consequence of the preceding one, and each of which contained the same elements separated from each other in space? Let us think of water occupying a vessel at a high temperature and continually cooling. The states of this system are (1) the gaseous state in which the molecules of the water are moving at a high velocity and are a relatively considerable distance apart, and in which they are incessantly colliding with each other and with the walls of the vessel; (2) the state of the system consisting of the separate phases, liquid water and gaseous steam in contact with it; and (3) the solid phase, in which the molecular motions almost, or quite, cease. Here the progress of the system through its phases leads to physical diversity and then again to physical homogeneity. But the diversity of the different phases is in a sense an apparent one only: any single phase, or at least those which involve the passage of the system from the gaseous to the liquid phases, and _vice versa_, can be represented by van der Waal’s general equation, RT = (_p_ + _a_/_v_^2) (_v_ - _b_). Does anything in modern biological investigation, except, of course, the speculations of non-physical physiologists, suggest that an ontogenetic process can be represented in such a manner?