PROBLEMS WITH APPLYING THE TRADITIONAL APPROACH OF SCIENTIFIC METAPHYSICS TO GENETICS
One problem with applying the traditional approach of scientific metaphysics to genetics is that the results are unstable. It leads to results that might seem plausible 
FIG. i The molecular gene concept enables biologists to partition DNA in multiple ways. With respect to this figure, if biologists are interested in the synthesis of the primary RNA transcript (see above), then the DNA segment covered by the bracket directly under ‘gene for primary RNA segment’ is the relevant gene (the continuous region of five segments in the DNA molecule starting with the solid white segment and ending with the solid black segment). If biologists are interested in the synthesis of polypeptide B occurring in some tissue at a particular stage of development, then the relevant gene is the discontinuous region marked by the brackets directly under ‘Gene for polypeptide B’ (the white and grey segments in DNA and not the segments between).
at one time, but turn out to be implausible afterward. For example, the fundamental interpretation of classical genetics set out in section 4 turns out to be simply false. Subsequent developments in genetics, developmental biology, and evolutionary biology have shown that genes are notfundamental units of heredity, development, and evolution. Genes are not the fundamental units of heredity and development because it is plainly false that knowledge about the functioning of each and every gene would provide the fundamental basis for explaining everything about heredity or development (Griffiths and Stotz 2013). Genes are not the fundamental units of evolution. Even if one could identify each and every gene that has occurred in evolutionary history, each and every difference in each gene that has occurred in evolutionary history, the first appearance of each of these genes and gene differences, and even if you could trace and explain the changes in frequencies of these gene differences, you would still not have a comprehensive basis for understanding all of evolution.
The attempt to interpret the best theories of the 1930s to identify the fundamentals of heredity, development, and evolution would have led to false conclusions (and the conclusions would not have been trivially false; they would have been false in significant ways). I wish to emphasize that the problem is with trying to interpret the transmission theory in a way that will reveal the fundamentals. The problem is not with the transmission theory as it was used in practice. The theoretical explanations, as sketched in section 3 (see Waters 2007 for more details), have withstood the test of time. Subsequent findings at the molecular level indicate that these explanations were correct. The transmissions of differences in functional units arranged linearly in chromosomes were indeed causally responsible for the patterns of phenotypic transmission produced in the laboratory. History does not undermine a realist interpretation of local explanations of classical genetics. But history does undermine an attempt to interpret the basic theory in a way that reveals the fundamental structure of the world (or the fundamental structure of inheritance, development, and evolution).
But perhaps the theories of genetics in the 1930s were not sufficiently developed to provide a basis for drawing stable metaphysical conclusions. Traditional scientific metaphysicians might respond that we should look to today’s science, not yesterday’s. And does not my analysis of the molecular gene concept support the idea that genes are fundamental units ? After all, the analysis identifies a kind of internal structure and a kind of immediate functional role (in RNA synthesis) allegedly shared by all molecular genes. So this class of genes could be considered to be a natural (or fundamental) class. But this is a natural class (or kind) in a weak sense, in the sense of sharing an internal makeup and structure and having the same kind of proximate causal impact.
In a stronger sense, natural classes (or kinds) are supposed to provide the single correct parsing that reveals the joints of nature. Such classes would be categories of being that key into the fundamental structure of the world. If we looked at the gene concept out of the context of investigative and manipulative practices, it might appear to designate a natural class in this strong sense. But when we see how the concept is employed in practice, and why it is a useful concept, we see it designates a natural class only in the weak sense.
The molecular gene concept is useful not only because of its precision, but also because of its flexibility. The concept is relational. It is a gene for concept (Waters 2000). It can be applied differentially, to pick out different causal chains passing through DNA. These chains of causation extend from bewildering causal complexities of cellular functioning through different portions of DNA and then continue out into equally bewildering causal complexities of the cell. Multiple causal chains are entangled with one another and proceed through overlapping segments of DNA. The molecular concept enables biologists to identify and manipulate individual chains in precise ways while other chains are stable (often systems are held stable via experimental manipulation). This enables biologists to slip and slide through causal entanglements that they do not understand in their natural complexity. In multicellular eukaryotes, differential splicing leads to different proteins in different functional contexts in different tissues at different stages of development. In some contexts, one set of exons plays a key role determining linear sequences; in other contexts other segments play key roles in determining linear sequences. The molecular gene concept enables biologists to specify the exons relevant to the synthesis of different proteins in particular cellular contexts without knowing much about the overall complexity of the natural situation.
The fact that this kind of conceptual practice works so well in genetics supports the idea that molecular genes are not a natural class in the stronger sense. That is, they do not provide a single correct parsing of DNA that reveals fundamental joints of nature. If molecular genes were a natural class in this strong sense, they would provide the unique division of DNA that would line up with the uniquely correct and comprehensive partition of development processes. Parsing DNA at its joints would line up with a parsing developmental at its joints. But geneticists have not identified such a canonical parsing. So, the metaphor of dividing nature at its joints does not seem to fit. There are simply way too many “joints” in DNA and way too many useful ways to parse the molecular and developmental processes. I have already explained that some parsings of DNA key into particular processes in particular tissues at particular developmental stages, other parsings key into different developmental processes in other tissues and/or other developmental stages. But the same is true from the other direction. Biologists have not found a parsing of genetic control processes that keys into a canonical set of joints in DNA. In sum, what they have found is that just as DNA complexly impinges on a causal mess, a causal mess also complexly impinges on DNA. That is why conceptual practice takes the form it does. Biologists have designed the molecular gene concept so they can navigate through a mess.
My concern is not that individuating genes in DNA depends on employing a relational concept. Rather, my point is that the way this relational concept is employed in practice indicates that the overall situation is far too messy to apply the metaphor of cutting nature at its joints. The remarkable feat of biologists is that they have succeeded in developing conceptual and technical tools that enable them to maneuver within bewildering messy complexities without having an overall theory or understanding of these complexities. My metaphysical claim is that scientific practices in genetics and allied sciences take this form because they are adapted to a reality that has no overall structure. The reality has lots of structure, but no overall structure.
-  The relevant sense of ‘determine’ is analyzed in Waters 2007.
-  Griffiths and Stotz (2013) supports this account of the causal situation, though it does not argue for the kind ofmetaphysical view advanced in this chapter. In fact, its argumentation seems to assume a much different epistemology and metaphysics than the one being developed in this chapter and the corpus of work this chapterdraws upon (Waters 1994 through Waters 2014).