The structure of science

Ian Stewart, in his book Does God Play Dice?, describes how science is structured. The explanations and theories provided by science are hierarchic; they start from the theories of fundamental particles and atoms, follow through theories of fluid dynamics, ecosystem, etc and finish with theories of sociology and art. Each explanation is constructed on top of the theories that are at a lowest level but in the same time, it does not care of the detail of these lower-level theories: the equations of fluid dynamics are constructed for a small water parcel, typically several moles of water, but it does not care about the individual atoms, nor about the fact that gravity has yet to be explained by the physics of particles. This important view of science is also shared by Edward O. Wilson in his book Consilience. Here is Ian Stewart's quote:

"Current science possesses no truly fundamental theories - not in the sense that they describe what nature actually does. They are all approximations, valid within some reasonably well-defined domain. Quantum mechanics work well at the submicroscopic level. General relativity is great for describing entire universes [...]. Science is a patchwork of models, each of which has been enormously refined within its own domain. The models habitually disagree when those patches overlap. Some disagreements are relatively harmless: atomic theory and continuum fluid mechanics disagree on the fine structure of water, holding it to be respectively to be discrete and infinitely divisible, but on macroscopic scales continuity and discreteness effectively approximate each other. Others are fatal: for example, as I write, the best current theory of astrophysics and the best current theory of cosmology compel us to accept stars older than the universe that contains them. Today's science is a pluralist patchwork of locally valid models, not a global monolith. Indeed it succeeds because it is a pluralist pacthwork of locally valid models.
Our concept of explanation is also a patchwork. A philosophical model that fits it well is what Richard Dawkins calls 'hierarchical reductionism', which sees scientific theories as a hierarchical structure, with some on different levels from others, corresponding to different levels of description of phenomena. (The hierarchy is not rigid and the levels need not be like layers of bricks in a wall.) For example, the complexities of ecosystems are explained by referring them back to those of organisms; organisms are explained by the growth of spatially organized proteins and other macromolecules; the complex organization of organisms is referred back to the linear complexity of their DNA code; the complexity of DNA is referred back to combinations of simpler atoms - and so on, right back to the Theory of Everything.
As Dawkins rightly remarks, it is not necessary to trace every phenomenon right back down this chain of reductions in order to understand it. Chemistry can be considered as 'given' for the purposes of understanding DNA; DNA can be taken as 'given' for the purpose of understanding protein manufacture in organisms, and so on.
[...]
What we tend to forget, when told a story with this structure, is that it could have had many different beginnings. Anything that lets us start from the molecular level would have done just as well. A totally different subatomic theory would be an equally valid starting-point for the story, provided it led to the same general feature of a replicable molecule. [...] It has to be or else we would never be able to keep a goat [within a wooden fence] without first doing a Ph.D. in subatomic physics."

Ian Stewart, Does God Play Dice?, Farewell, Deep Thought.