In my last two posts in this series, I suggested that science is a field of possible moments, with no set trajectory over what I called the “dance floor of science”. Some commentators have objected to this, arguing that there is no real difference between classification and theory building. I disagree, but first I had better make it clear that this is a continuum, and that any actual event in the history or development of science or a scientific career is almost never going to be exclusively theory or classification building.
One reason why the “theory-dependence” hypothesis was so successful is that it is almost trivial to find elements of theory in any set of observations, and likewise the inverse is true: it is almost trivial to find elements of classification activity in a theory. On my account, this is inevitable, not least because nobody ever starts a scientific process tabula rasa. We have prior conceptual commitments at any point; some being mediated by evolution itself (our Umwelten), and some being mediated by the general cultural context in which we begin. For example, when biological classification began in the 15th century, it was built on the traditions of herbalism, and bestiaries, that were the medieval traditions, but the recognition of kinds of plants and animals was not itself theoretical.
Even within science, in fact especially so, when we begin upon a new subject of investigation, we bring to that activity all the prior theories, methods, techniques, industries and educational substrates that exist at the start. The theory-dependence view has it that we can only begin to learn about the world because of our theories. On my account, we learn because we are learning machines;* and prior knowledge always biases how we do that. But we are not constrained in what we can observe merely by conceptual commitments, nor are we unable to “see” phenomena because theory doesn’t permit it. Either of these would inhibit our ability to learn.
But learn we do, and one way we do that is by identifying phenomena that are regular in that domain. If there is no theory, or the theory is inadequate, these phenomena trigger the development of explanations. So how is classification, which is one kind of conceptual construction, differ from theory building, which is another? My answer is to ask what the metrics on the axes are, and to answer them thusly: each asymptote is either active or passive. Experiment is intervention, but field observation is merely passive (and of course one can do field experiments). Theory construction, which is explanatory, is active conceptualisation, but classification is passive, it is “handed to you” by your cognitive dispositions and the data that you observe. Without going into the whole observation sentences mode of logical positivism, it remains true that we do have a distinction between observation and explanation, even if nothing is ever purely the one or the other.
To classify is to order the data, to find regularities even if you have no direct idea why they cluster so. It is to find identity classes empirically, setting up some problem or thing in need of explanation. Darwin used the regularities of largely atheoretical classification, which he referred to in the Origin as “group under group”, as the explanandum that his “theory” of common descent by branching evolution explained. The empirical regularities of the periodic table were explained by, first, valency theory, and then quantum mechanics. There is a distinction between observation and sorting on the one hand, and model building and explanation on the other.
A second comment made by Isaac at Think Deviant, is that I haven’t specified what the “context” is into which I place this scientific dynamic system. He’s right. I haven’t done this because it is an empirical matter in each case what that context is. There are no generalisations that I think are unexceptional about this. Sometimes a science will run more or less independently of its culture, and at other times a science will be beholden to its cultural context independently of the internal issues of the science. Consider, for example, the difference between late nineteenth century physics and psychoanalysis. The former was almost acultural, accessible to western, Indian, and Asian workers, while the latter was tied to the cultural context, first of Viennese society, and later New York. The recent paper on WEIRD (” Western, Educated, Industrialized, Rich, and Democratic”) psychological studies indicates that this remains an issue in psychology.
To think there is a general, universal and consistent cultural context for science is, I believe, a holdover of Comtean positivist thinking. You want to know what the relevant context was for the Hubble telescope, or for the discovery of aspirin? Go look. My schematic here merely indicates the general relations of external and internal movements in the science.
I hope this clears some of the concerns up.
* By “learning machine” I mean that we are what artificial intelligence researchers call “classifier systems” that take unstructured data and organise them into reliable recognition schemata. This is a result of our being neural networks, which can iteratively refine classifiers until they satisfy some function of accuracy. The best classifier system known occurs between the ears of systematists and other classifiers in science.