Blute Blog

I enjoyed the annual 4S meeting (Society for Social Studies of Science) in Cleveland this week. I gave a talk with the above title which flowed from a paper titled “The Reinvention of Grand Theories of the Scientific/Scholarly Process” published with a graduate student, Paul Armstrong, in the current issue of Perspectives on Science (a paper which MIT press seems to have unusually made available for free here). That paper dealt with the work of ten contemporary sociologists and sociologically-minded philosophers of science who have presented general theories of the scientific/scholarly process on eleven issues. Methodologically it was done by means of an analysis of texts as well as interviews with the majority and was designed to assess the compatibility or lack thereof of their theories with each other and whether a new general theory is emerging. It ultimately concluded that it is a powerful argument in its favour that a Darwinian-style sociocultural evolutionary theory (of the general kind pioneered by Stephen Toulmin and David Hull) “can incorporate both all of the common and all of the useful unique features of contemporary grand theories of the scientific/scholarly process”.

The talk however was about four empirical generalizations not discussed in the paper that an evolutionary theory of science/scholarship can explain – the first from David Hull’s book Science as a Process, the second from my book, the third blogged about here on June 14, 2010, and the last from an article by Jonah Lehrer in The New Yorker on December 13, 2010 here.

i) Fraud and plagiarism. It is well known that the scientific community treats fraud (such as falsifying data) more harshly than it does plagiarism (although journal editors have begun to crack down more on the latter recently – perhaps because the availability of electronic data bases makes it easier to do so). According to Hull the reason for the traditional attitude is that plagiarism hurts only the ‘ancestor’ upstream i.e. the author or authors of the paper plagiarized while fraud hurts all those ‘descendants’ downstream – those whose time and energy spent building on falsified results has been wasted. You may have noticed that this has been one of the major complaints about the recent scandal involving Andrew Wakefield’s claims about the relationship between the MMR vaccine and autism and bowel disease. It resulted not only in damage to some of the children whose parents avoided having them vaccinated, but also in effort wasted on subsequently testing the theory – effort which could have gone into other better approaches to the causes of and treatments for these diseases.

ii) Citations to long papers. I was once puzzled to learn that longer papers gain more citations than shorter ones because the evolutionary ecology of life histories would predict the reverse. Without getting technical, the reason for that prediction is that under low density (plentiful resources) selection favours ‘productivity’ – eating a lot and producing a lot of (hence necessarily small) offspring. Under high density (scarce resources), it favours ‘efficiency’ – deriving more breakdown products from each unit of resources acquired and deriving more grand-offspring from each offspring produced (hence necessarily few, large offspring). In the light of this expectation that short papers would garner more citations but long ones would garner more citations to the papers that cite them, I was subsequently relieved to find out that studies have shown that many citations are not taken from original papers at all but from the citations of others to them. In short, many of the apparent “offspring” of long papers are likely not offspring at all, but grand-offspring. However, the necessary definitive study, comparing the “copied” citation rates of long and short papers has not to my knowledge been done.

iii) Mentors, students and their students. A study of roughly a century and a half of the lineages of mathematicians published in Nature last year showed a similar life history phenomena with respect to students rather than papers. The students of those who produce few students go on to produce about a third more students themselves than expected – i.e. less prolific teachers produce higher quality students who yield more grand-students.

iv) The truth wears off. The last example came to my attention from an article by Jonah Lehrer published in the New Yorker titled “The truth wears off.” It has long been known that there is a bias in science in favour of the publication of positive results. In the 1950’s for example, a statistician found that ninety-seven per cent of psychology papers found positive results. Perhaps authors themselves, and certainly referees and editors are loathe to publish results that say, “Nope I didn’t find what I was looking for, sorry”. What seems to happen is that if a novel, interesting result is found, a rash of confirmatory studies follow. What has not been so well known (and certainly not to me) is that such a trend is often followed by a trend in the opposite direction! What seems to happen is that by that time, the opposite result is the novel and interesting one and is therefore published, after which a rash of confirmatory studies of that follow. A cultural fad in science in one direction is followed by one in the opposite direction. Other observers of science have noticed more or less the same phenomena. Both continuity and innovation are valued in science but according to Hull, there is a tension between them. Citations claiming one’s work follows from that of others gains support but detracts from its apparent innovativeness, while originality is admired but detracts from the support that showing continuity provides.

The point I would like to make about this is that evolutionary theory has long been familiar with the phenomenon and has basically explained the reason for it. Evolutionists call it “negative frequency-dependent selection” – it being commonly adaptive to do the opposite of what the majority are doing because that reduces competition. So for example if most members of a population of birds are eating small seeds, an innovation (mutant) that eats large seeds could be favoured and spread until it becomes most common, upon which small seed eaters would again be favoured and so on. Actually, I suspect that avoiding competition may not be precisely the actual proximate explanation. The reason why large seed eaters are favoured when small seed eaters are common is because not only are the two kinds of seed eaters dependent on the exogenous availability of the two kinds of seeds, but because eating small seeds alters the ecological environment (depleting small seeds but permitting the population of large seeds to recover) thus favouring large seed eaters and of course vice-versa. Effects like these have recently come to be called “niche construction” i.e. the ecological environment not only structures populations but populations also construct their ecological environment. In science, both continuity and originality matter but not always simultaneously apparently.

Lehrer briefly discusses a number of other reasons for the truth “wearing off” – plain poor science, regression towards the mean in subsequent studies, data mining of studies with a large number of variables and comes to a more radical conclusion that I think justified – that “when the experiments are done, we still have to choose what to believe”. It is worth noting however that the sociological phenomena of negative frequency-dependent selection – cultural fads or social movements in one direction being followed by those in the opposite direction – does not readily explain the truth wearing off in a sequence of studies done by the same individual unless negative frequency-dependent selection works psychologically as well as socioculturally – at least one example of which Lehrer highlights. Maybe the story there is that we just get bored!