Biological evolution is, as has often been noted, both fact and theory. It is a fact that all extant organisms came to exist in their current forms through a process of descent with modification from ancestral forms. The overwhelming evidence for this empirical claim was recognized relatively soon after Darwin published On the Origin of Species in 1859, and support for it has grown to the point where it is as well established as any historical claim might be. In this sense, biological evolution is no more a theory than it is a “theory” that Napoleon Bonaparte commanded the French army in the late eighteenth century. Of course, the details of how extant and extinct organisms are related to one another, and of what descended from what and when, are still being worked out, and will probably never be known in their entirety. The same is true of the details of Napoleon’s life and military campaigns. However, this lack of complete knowledge certainly does not alter the fundamental nature of the claims made, either by historians or by evolutionary biologists. (Pigliucci et al. 2006: 1)

On the other hand, evolutionary biology is also a rich patchwork of theories seeking to explain the patterns observed in the changes in populations of organisms over time. These theories range in scope form “natural selection,” which is evoked extensively at many different levels, to finer-grained explanations involving particular mechanisms (e.g., reproductive isolation induced by geographic barriers leading to speciation events). (Pigliucci et al. 2006: 1)

(….) There are a number of different ways in which these questions have been addressed, and a number of different accounts of these areas of evolutionary biology. These different accounts, we will maintain, are not always compatible, either with one another or with other accepted practices in evolutionary biology. (Pigliucci et al. 2006: 1)

(….) Because we will be making some potentially controversial claims throughout this volume, it is crucial for the reader to understand two basic ideas underlying most of what we say, as well as exactly what we think are some implications of our views for the general theory of evolutionary quantitative genetics, which we discuss repeatedly in critical fashion. (Pigliucci et al. 2006: 2)

(….) The first central idea we wish to put forth as part of the framework of this book will be readily familiar to biologists, although some of its consequences may not be. The idea can be expressed by the use of a metaphor proposed by Bill Shipley (2000) …. the shadow theater popular in Southeast Asia. In one form, the wayang golek of Bali and other parts of Indonesia, three-dimensional wooden puppets are used to project two-dimensional shadows on a screen, where the action is presented to the spectator. Shipley’s idea is that quantitative biologists find themselves very much in the position of wayang golek’s spectators: we have access to only the “statistical shadows” projected by a set of underlying causal factors. Unlike the wayang golek’s patrons, however, biologists want to peek around the screen and infer the position of the light source as well as the actual three-dimensional shapes of the puppets. This, of course, is the familiar problem of the relationship between causation and correlation, and, as any undergraduate science major soon learns, correlation is not causation (although a popular joke among scientists is that the two are nevertheless often correlated). (Pigliucci et al. 2006: 2)

The loose relationship between causation and correlation has two consequences that are crucial…. On the one hand, there is the problem that, strictly speaking, it makes no sense to attempt to infer mechanisms directly from patterns…. On the other hand, as Shipley elegantly show in his book, there is an alternative route that gets (most of) the job done, albeit in a more circuitous route and painful way. What one can do is to produce a series of alternative hypotheses about the causal pathways underlying a given set of observations; these hypotheses can then be used to “project” the expected statistical shadows, which can be compared with the observed one. If the projected and actual shadows do not match, one can discard the corresponding causal hypothesis and move on to the next one; if the two shadows do match (within statistical margins of error, of course), then one had identified at least one causal explanation compatible with the observations. As any philosopher or scientist worth her salt knows, of course, this cannot be the end of the process, for more than one causal model may be compatible with the observations, which means that one needs additional observations or refinements of the causal models to be able to discard more wrong explanations and continue to narrow the field. A crucial point here is that the causal models to be tested against the observed statistical shadow can be suggested by the observations themselves, especially if coupled with further knowledge about the system under study (such as details of the ecology, developmental biology, genetics, or past evolutionary history of the populations in question). But the statistical shadows cannot be used as direct supporting evidence for any particular causal model. (Pigliucci et al. 2006: 4)

The second central idea … has been best articulated by John Dupré (1993), and it deals with the proper way to think about reductionism. The term “reductionism” has a complex history, and it evokes strong feelings in both scientists and philosophers (often, though not always, with scientists hailing reductionism as fundamental to the success of science and some philosophers dismissing it as a hopeless epistemic dream). Dupré introduces a useful distinction that acknowledges the power of reductionism in science while at the same time sharply curtailing its scope. His idea is summarized … as two possible scenarios: In one case, reductionism allows one to explain and predict higher-level phenomena (say, development in living organisms) entirely in terms of lower-level processes (say, genetic switches throughout development). In the most extreme case, one can also infer the details of the lower-level processes from the higher-level patterns produced (something we have just seen is highly unlikely in the case of any complex biological phenomenon because of Shipley’s “statistical shadow” effect). This form of “greedy” reductionism … is bound to fail in most (though not all) cases for two reasons. The first is that the relationships between levels of manifestation of reality (e.g., genetic machinery vs. development, or population genetics vs. evolutionary pathways) are many-to-many (again, as pointed out above in our discussion of the shadow theater). The second is the genuine existence of “emergent properties” (i.e., properties of higher-level phenomena that arise from the nonadditive interaction among lower-level processes). It is, for example, currently impossible to predict the physicochemical properties of water from the simple properties of individual atoms of hydrogen and oxygen, or, for that matter, from the properties of H₂0 molecules and the smattering of necessary impurities. (Pigliucci et al. 2006: 4-5)