How exactly are the history and philosophy of science supposed to come together? In the case of the scientific realism debate there is a relatively straightforward answer to this. In short, scientific realists are keen to make some sort of success-to-truth inference. Typically they state that when scientific success is sufficiently impressive, we ought to infer that the hypotheses (or parts or aspects of the theory) that generated this success are at least approximately true. This allows for the possibility that, in the history of science, one might find just that sort of success, born of a theory/set of hypotheses that are definitely not approximately true (whatever your take on ‘approximate truth’). Even allowing for one or two exceptions, the possibility arises that there might be many such cases in the history of science. This makes many scientific realist positions falsifiable—loosely speaking, at least. But as things stand, nobody knows which contemporary realist positions (if any) are indeed falsified because we just don’t have at hand the relevant historical ‘facts’ (again, speaking loosely). What we need to make progress is careful and detailed history of science, dealing with relevant historical episodes.
There may still be doubters(!) that this approach can actually work in practice. To such doubters I can do no better than to give a specific example that currently has me quite excited. Inspired by the work of Kyle Stanford, I started looking into the case of J. F. Meckel’s 1811 prediction of gill slits in the development of the human embryo. Certainly we have significant success here. In fact, the prediction was temporally novel: the phenomenon was completely unknown at the time Meckel made his prediction. In 1825, Heinrich Rathke discovered gill slits in pig and chick embryos, and eventually gill slits in human embryos were observed in 1827 by Rathke, von Baer, and others.
Having established that the success was significant in a sense the realist would accept (namely, novel predictive success), the next question is whether the hypotheses (or parts or aspects of the theory) that generated this success are at least approximately true. Our answer starts with the fact that Meckel was part of the Naturphilosophie movement within Germany at the turn of the nineteenth century. This movement was based on some basic assumptions concerning biological development:
- The animal series assumption: Animals can be put in a series from ‘lower’ to ‘higher’, with human beings at the top of the pile.
- Teleology: ‘Lower’ animals are striving to be human beings, and fail to become human beings because their development halts (Oken described lower animals as ‘human abortions’).
- A ‘single biological force’ assumption: A single force or ‘power’ underlies all biological development. Thus similar organisms must develop in the same general biological ‘direction’ (remaining similar after a period of development).
- A developmental theory of the animal series: The series of animals from ‘lower’ to ‘higher’ is caused by each animal developing from the same starting place, but to a different extent.
From here it is a very short step to the conclusion that the series of adult organisms from ‘lower’ to ‘higher’ must parallel the stages of human development from a ‘primal zero’ or ‘initial chaos’ (Oken) to a final adult human being. Thus there is a period where the human embryo is a fish, and it follows that there is a period where the human embryo has gill slits.
Can the realist claim that the hypotheses doing work here are (approximately) true? Prima facie it seems highly unlikely, simply because there is so little within Meckel’s assumptions that is today considered (approximately) true. It goes radically against evolutionary biology to try to place animals in a single series from ‘lower’ to ‘higher’, and teleology and the ‘single biological force’ assumption have no place in modern developmental biology. Even Meckel’s parallelism is misguided: it is not the case that the development of the human embryo parallels (even roughly) the (alleged) animal series. Thus one might describe Meckel’s predictive success as ‘lucky’. And, if the realist thinks it’s a miracle to get lucky in this way, then there are miracles in science!
How can the realist respond? As before in this debate, the realist might adjust her position, perhaps insisting that for realist commitment we need quantitative novel predictive success. The problem here is that it starts to look like the realist position is too flexible to ever really be challenged, thus making nonsense of the whole debate.
But a better realist response might just be possible. It turns out that Meckel’s 1811 prediction is phrased as follows: ‘Perhaps [Vielleicht] there is even a much earlier period when the embryos of the higher animals are also furnished with inner gills’. With the use of the word ‘vielleicht’, it is clear that Meckel is speculating here, as opposed to making the sort of deductive prediction we are familiar with in the realism debate. To put it another way, the prediction isn’t ‘risky’ in the way the prediction of the Poisson white spot was for Fresnel’s theory of light.
This reduces the significance of the prediction somewhat, but it is still telling that Meckel was confident enough to state this (tentative) prediction in print. However, the realist can add to this consideration the suggestion that, if we more fully contextualise Meckel’s prediction, we can come to understand why it wasn’t so surprising that he reached this idea. The realist strategy here is to argue that given the purely empirical/observational knowledge of the day, one could—with a bit of imagination—come to the gill slit prediction without any substantial theoretical ideas. For example, Meckel knew of a stage of development of the human heart where it very closely resembles a fish heart. And he also knew that frogs—which breathe with lungs—have gills at an early stage of their development (when they are tadpoles). Thus perhaps the realist can argue that Meckel really reached his conclusion not via his (false) theoretical ideas, but rather via his empirical knowledge.
It is illuminating at this point to make use of Bayes’s formula (which I am starting to think doesn’t feature often enough in the realism debate). Consider the formula:
In a case where we have a remarkable novel predictive success, the antecedent probability of the evidence (p(E)) is very small, and thus the probability of the theory (or relevant theory parts) after the prediction is verified and found to be correct (p(T, E)) is very high. Or, to put it another way, one’s degree of belief should increase dramatically when a (precise, surprising) prediction is found to be correct. However, in the case of the gill slit prediction, the antecedent probability of finding gill slits in the human embryo was not so small, given the empirical knowledge at hand. In addition, given the use of the word ‘perhaps’, Meckel’s prediction is only tentative, meaning that p(E, T) is less than 1. These two facts serve to decrease the extent to which one’s degree of belief should increase in this case. Some possible numbers one could use in the formula are:
p(T) = 0.4 (we are initially quite sceptical of the theory);
p(E, T) = 0.8 (the theory does not entail that we will find gill slits in the human embryo);
p(E, ¬T) = 0.5 (we have reasons independent of Meckel’s theory to expect to find gill slits in the human embryo).
Crunching the numbers, we find that our degree of belief should only increase from 0.4 to 0.52. That is, we remain sceptical of the theory. And if the reader doesn’t like these numbers, he/she should feel free to try some others. The major lesson remains the same as the numbers vary.
These are early stages, and much more work—especially historical work—remains to be done. What other predictions did Meckel (and others) make? If there were unsuccessful predictions (even tentative predictions), then these should also play a role in determining our final degree of belief. Just as importantly, was there further empirical knowledge suggesting that gill slits would be found, quite independent of Meckel’s theory (is 0.5 a fair number for p(E, ¬T)?)? It would be interesting (at least) to get a better sense of what number(s) might be reasonable for p(T), considering the community at large as well as key individuals within that community. There is also a philosophical task here: to consider how the divide et impera (selective realist) strategy can be incorporated in this story. Is it feasible to simply swap ‘T’ for ‘Twp’, where the latter stands for the ‘working posits’ of T?
Whatever the case, this fascinating episode in the history of science deserves its place within the realism debate. And it is undoubtedly crucial to integrate the history and philosophy of science if we want our conclusions to be meaningful and without bias.