debate within evolutionary theory. In a revised form, this mechanism forms the basis for modern views of the relationship between ontogeny and phylogeny.

Evolution and the Mechanics of Recapitulation

The fact of evolution recast Agassiz's threefold parallelism as a mirror image of its former self: the fossil sequence that he had added as a third illustration of divine wisdom became the primary cause for the other two series. Evolution also provoked a profound change in the mechanics of recapitulation. The single developmental tendency of Naturphilosophie no longer sufficed. Evolution implied a true, physical continuity of forms through time. This raised a host of questions that had never occurred to the Naturphilosophen. The need to solve them provided comparative embryology with an entirely new subject. Many evolutionists, Cope and Hyatt in particular, spent a major part of their careers trying to solve these problems and exploring the consequences of their solutions.

To Oken and Serres, embryonic stages of higher forms had "represented," "stood for," or been in some way "symbolic of" living, lower adults. No physical tie connected a fish to a human fetus with gill slits. But after 1859, recapitulationists had to view embryonic stages of descendants as the actual, physical remnants of previous ancestors. How had this remnant been transferred from an ancestral adult of large size (where it developed late in life as a permanent stage) to a tiny embryo (where it appeared early and endured but a short time)? What, in other words, is the mechanism of recapitulation? There is only one way to make recapitulation work under a theory of evolution by physical continuity. Every recapitulationist, from the staunchest Darwinian (Weismann) to the most militant neo-Lamarckist (Cope and Hyatt), upheld this mechanism; there is no other (Fig. 7). It involves two assumptions.

1. Evolutionary change occurs by the successive addition of stages to the end of an unaltered, ancestral ontogeny.3 This assumption provokes two problems. First, since many lineages involve thousands of steps, ontogenies will become impossibly long if each step is a simple additon to a previous ontogeny. Second, embryonic stages usually occur much earlier in time and at much smaller sizes than the ancestral adult stage they represent. There must be some force continually operating to shorten ancestral ontogenies, thereby keeping the descendant's period of development within reasonable limits.

2. The length of an ancestral ontogeny must be continuously shortened during the subsequent evolution of its lineage.