The data documenting evolution tend to fall into one of two categories: either they are observations made on living organisms over the course of a few years, or they are observations made by comparing species whose differences have evolved over millions of years. These two categories are often referred to respectively as microevolution and macroevolution. While both types of data unquestionably demonstrate evolutionary processes, major questions remain about how microevolution is translated into macroevolution. Are macroevolutionary differences simply the result of microevolutionary processes spun out over millions of years? Are the small evolutionary steps that we observe in living populations consistent with the larger differences that we observe among species and in the fossil record?

Computer simulations are an important tool for determining whether the things we think are key to evolution over short timescales are adequate to explain differences that have evolved over long time periods. Simulations have been important for studying the evolution of simple biological properties (e.g., the average body weight of species or the average length of bones and teeth) because the evolution of these single value measurements is easy to model. Structures, such as mollusc shells, have also been successfully simulated, but more complicated structures (e.g., dinosaur skulls, mammal teeth, echinoderm skeletons) have been more difficult to model because of the interactions between their parts.

This paper explains one procedure for running simulations of complex structures. The interactions between parts are measured directly from real populations and used to construct a mathematical space in which the simulation can be run efficiently. By representing the structures as geometric points or lines, its evolving shape can be pictured for comparison to real data. The evolution of the teeth of the Common shrew, Sorex araneus, was simulated using four different types of evolution that have been observed over short timescales: (1) randomly changing selection, in which the fittest shape changes with each generation in response to changing conditions; (2) directional selection, in which the fittest shape never changes; (3) stabilizing selection, in which the fittest shape is the one that exists now; and (4) genetic drift, in which any shape is as fit as another. Comparison of the results with the teeth of real shrew species suggests that the latter have evolved in response to randomly changing selection.