Writing about the philosophy of science sometimes has a tendency to become a little abstract in nature. It also has some historical connotations: although there is heavy discussion whether or not philosophical arguments can be supported by examples, it is customary to do so (at least to help the reader understand the arguments). Most of the time, these examples are rather old and famous, the most recent often being the transition from Newtonian physics to the improvements made by Einstein.
But today I realized that it all happened, right in front of my eyes. Recently, I’ve been busy writing on the problems of theory-ladenness of observations when researchers try to test hypotheses against empirical observations. Basically, I’m applying the Lakatos-Feyerabend debate to present-day examples of social-science methodology in the hope to find a satisfactory answer on how to get the most out of the methodological pluralism that present-day social science disciplines are characterized by.
So, I’ve gathered some nice examples to illustrate my arguments on the implications of theory-ladenness of observations. One of those arguments comes down to the statement that when researchers apply a method to reduce uncertainty about a specific research problem, they do not fundamentally solve that uncertainty, but transform it to uncertainty about the validity their method.
What does this mean? One aspect of the design researchers apply is the sampling strategy that led to the observations on which the conclusions are based. Due to monetary and other practical limitations, several sampling strategies are used in such a way, that many disturbing factors are held constant (and thereby cannot form an alternative explanation of findings). An excellent example of this is formed by research on monozygotic twins which aims to rule out any influence from genetic differences. This is an often used strategy in social science to be able to make statements on, for instance, the effect of parental social characteristics on social characteristics of their children, without the effect of the genetic traits children receive from their parents.
It has long been thought that monozygotic twins are indeed genetically identical. So, if we sample monozygotic twins, we can safely make the assumption that any differences found between those twins cannot be due to genetic influences, right? Although this assumption has often been made, it appears to be invalid in general.
Yesterday I read a very short headline in a quite sensationalistic newspaper: “Identical twins not identical”, it stated. Today I realized the consequences: the often made assumption made in research based on monozygotic twins is not valid. A report published yesterday (( Bruder et al., Phenotypically Concordant and Discordant Monozygotic Twins Display Different DNA Copy-Number-Variation Profiles, The American Journal of Human Genetics (2008), doi:10.1016/j.ajhg.2007.12.011 )) shows that there are indeed genetic differences between monozygotic twins.
This is what makes the philosophy of science interesting: what we see now is that an assumption that for long has been thought of as acceptable, is not valid anymore. Research based on this assumption will need re-evaluation. This is exactly what I mean when I wrote that when researchers apply a method to solve a problem, the uncertainty about that problem becomes the uncertainty of the method applied. No matter how valid an assumption may seem or how often used an assumption is, it is always falsifiable. And no matter how hard we may try, performing research without making any assumptions seems to be impossible to me.
It is nice to see how what you are trying to describe from a philosophical perspective appears to happen during the process of writing. At least it helps to gain some confidence in your own arguments and it allows for some nice illustration. Additionally, without scientific development, there would probably be no need for an actively developed philosophy of science at all.
Obviously, all implications of these findings for sampling strategies in social sciences need to be investigated more closely, before all previous findings that are based on it are thrown out of the window. Perhaps, when more is known, we can accurately (to our own opinion) show in what cases the existing sampling strategies can be maintained, and under what conditions the assumption it is based on indeed is not valid.
But, at least until then, I would say that studies that employ sibling-paired sampling strategies should be interpreted with a little more care and new studies should argue whether or not this new study (and other that will surely follow) violates the assumption of genetic similarity in their specific situation.
BRUDER, C., PIOTROWSKI, A., GIJSBERS, A., ANDERSSON, R., ERICKSON, S., DIAZDESTAHL, T., MENZEL, U., SANDGREN, J., VONTELL, D., POPLAWSKI, A. (2008). Phenotypically Concordant and Discordant Monozygotic Twins Display Different DNA Copy-Number-Variation Profiles. The American Journal of Human Genetics, 82(3), 763-771. DOI: 10.1016/j.ajhg.2007.12.011
“what we see now is that an assumption that for long has been thought of as acceptable, is not valid anymore. Research based on this assumption will need re-evaluation.”
You really make this way more dramatic than it is. Based on phenotype it is easy to deduce that their must be some genetic differences between identical twins (e.g. Frank and Ronald de Boer look different). The paper just confirms that a long suspected cause for this difference is indeed there. Still, the difference is small and in most cases probably insignificant compared to environmental factors.
Dear Martijn,
first of all, thank you for the reaction to my post and: welcome to my blog!
Obviously, I bring my argument quite strongly, to emphasize the importance of this finding. But, I chose my words carefully: I explicitly wrote that I do not want to throw all findings based on this assumption “out of the window’, but I do argue that re-evaluation is called for.
The argument you give about the difference between genotype and phenotype actually is not that strong. Rather, I think it implicitly emphasized the importance of my interpretation. You write that for long we now that phenotype of monozygotic twins can differ. This is true, but until now, the differences in phenotype have not been attributed to genetic differences. Thus, they were attributed to environmental influences (i.e. personal choices made, social environment, etc.), but this cannot be done any longer from now on. At least not as an automatic assumption, for we now know that monozygotic twins can differ both genetically and phenotypically.