the structural details present in the eyes of many different organisms.
Each design on the route shows some simple change from the previ- ous one, and at each stage spatial resolution improves, by having each separate receptor cell exposed to a different field of view. The heart of the explanation is the demonstration that there is a series of gradual changes in morphology that take us from a light sensitive patch to a fo- cused-lens eye. Models of this kind analyze complex biological mecha- nisms into their components, and show how small changes in these components and their interactions can enhance the mechanism of the performance as a whole. Thus they map adaptively possible trajectories through phenotype space (see Calcott [forthcoming] for a detailed ac- count of these explanations and of their evolutionary importance). Thus, with such models we can represent many possible biological forms, including those that are very different from actual biological forms. Even so, there are still many forms that elude the modeler’s skill. There is, for example, no model that describes anything as complex as a mammal (for a good discussion of current limitations and the prospects for future advances in this methodology, see McGhee 1999, 282–87). So even with contemporary computational techniques, there are limits on the complexity of systems whose differences can be compared. Within these limits, theoretical morphology has focused strongly on spatial rep- resentations of biological possibility. The process by which such spaces are developed is simple. Parameters (length, height, rate of coiling, angle of branching, number of iterations, and so on) are deployed as dimensions of a space. These spaces thus have a dimensionality equal to the number of developmental or morphological characters studied. While, in theory, such hyperspaces might have very high dimensionality, in practice, theo- retical morphologists have become adept at representing biological forms using a relatively small number of parameters. Having determined the geometry of the morphospace, it can then be used to “describe the total spectrum of physically possible forms” (Raup 1966, 1178). It makes sense to think of a physical object being “as wide as it is deep” because width and depth can be measured in the same units. In contrast, most of the dimensions of morphospaces are only distantly related to our familiar three spatial dimensions, and their dimensions are rarely commensurable. Usually, the relationship between the di- mensions is more like the relationship between the standard spatial dimensions and time. Many of the “distances” in such spaces are not even magnitudes. They are instead values within a set of discrete pos- sible states, as in Thomas and Reif’s skeleton space. Yet in some ways these morphospaces are indeed spatial. We can “lo- cate” actual and merely possible organisms within them. Morphospatial