Evolution, Biology & Society
Total Page:16
File Type:pdf, Size:1020Kb
__ EEvvoolluuttiioonn,, BBiioollooggyy & Society & Society Fall 2010 Newsletter of the ASA Section on Evolution, Biology & Society Volume 7, No. 1 Chair In this issue: 2010-2011 Essay:” Replications and Genetic Jeremy Freese, Northwestern University Studies of Behavior” by Michael Shanahan, Shawn Bauldry& Ross Past-Chair Macmillan 2009-2010 Essay: “Are Evolutionary Theory and Stephen Sanderson, UC-Riverside Rational Choice Theory Compatible?” by Richard Hutchinson Chair-Elect Essay: Why Rational Choice Theory 2011-2012 and Sociobiology* Are Natural Allies Jonathan Turner, UC-Riverside by Stephen K. Sanderson New publications of Section Secretary-Treasurer Members 2010-2012 Announcement: Special Issue of F. Scott Lewis, Penn-State Harrisburg Journal of Mathematical Sociology: “Micro-Macro Links and Micro- Council Members Foundations” Alan Booth, Penn State University Section Award Winners 2010 (2009-2012) David Franks, Virginia Commonwealth University (2010-2013) Replications and Genetic Christine Horne, Washington State Studies of Behavior University (2008-2011) Richard Machalek, University of Michael Shanahan & Shawn Bauldry, UNC- Wyoming Chapel Hill (2009-2012) Ross Macmillan, University of Minnesota Patrick Nolan, University of South Carolina (2008-2011) Michael Shanahan, University of North The behavioral geneticist Lydon Eaves is Carolina at Chapel Hill (2010-2013) fond of noting that Saint Augustine was among the first to make genetically-informed Newsletter editor and Webperson observations when, in arguing against the Rosemary L. Hopcroft, UNC-Charlotte validity of horoscopes, he pointed out how very www2.asanet.org/sectionevol/ different twins could be. Fast- forward 1,600 years from the Bishop of Hippo to the publication of two papers in Science by Avshalom Caspi, Terri Moffitt, and their Evolution, Biology and Society Vol. 7, No. 1 Spring 2010 - 2 – colleagues in 2002 and 2003. Both papers when self-reports are used (Uher & McGuffin, were empirical studies that focused on genetic 2010; see also Caspi et al., 2010, for a recent polymorphisms and their interactions with discussion of supporting evidence with non- social experiences in the prediction of human primates). psychiatric outcomes. And both papers have put the issue of replication in the scientific So things stand at this writing. And, quite spotlight. understandably, a major concern in this area of research is now replication, the reproducibility The first such study (the “MAOA study”) of a study’s methods and findings. Yet some reported that a genetic variant associated with basic lessons about replications have been low MAOA activity coupled with child forgotten and, with these lessons in mind, maltreatment was associated with antisocial replication can proceed with a more behavior in young adulthood. A 2006 meta- constructive sense for building scientific analysis by Kim-Cohen and her colleagues knowledge bases and theories. concluded that, across five extant studies, the interactive effect held. The original MAOA At the outset, it is noteworthy that study (cited over 1,800 times to date) and the replication is not a hot topic beyond gene meta-analysis have spawned a cottage candidate research. Turn to any issue of AJS industry of discussion and attempted or ASR in sociology and ask yourself how replications, with some scholars presently many studies constitute replications or will concluding that results are “mixed” (an generate replications or will require replications undesirable descriptor to which we return to be validated. Although there are certainly below). areas of research that have been heavily replicated in the behavioral sciences, explicit The Caspi team’s second Science paper concerns for replication in sociology are (the “5HTT paper”) reported an interaction minimal. This unevenness may be thought between the serotonin transporter linked “unfair” given the extensive attention to polymorphic region (5-HTTLPR) and stressful replication in gene-environment research, but it life events in the prediction of major depression actually works to the disadvantage of subfields in young adulthood. If the MAOA paper requiring little replication. In the long run, such spawned a cottage industry, the 5HTT paper fields cannot generate a cumulative body of gave rise to a veritable rust belt of heavy knowledge. industry, with the original paper now cited almost 3,000 times. A highly influential meta- Historians of science hold that replications analysis appearing in Journal of the American were a major breakthrough for the scientific Medical Association concluded that, across 14 method because they required that scientists reviewed studies, no such interaction was who were largely isolated from one another replicated (Risch et al. 2008; see also Munafo clearly communicate their procedures and et al., 2009 for a similar meta-analysis and findings. If Robert Boyle could confirm the conclusion). Numerous deficiencies in the inverse relationship between an ideal gas’s meta-analyses were noted (see Letters in pressure and volume—and his work was then JAMA, November 4, 2009, for some examples) replicated across Western Europe--then and a qualitative assessment of empirical something indeed had been accomplished studies published at about the same time as (Daston, 1994). But herein lies the first problem Risch concluded that replication results were with replications: the concept of the replication patterned by the measure of life-events, with was originally developed in the physical semi-structured interviews being associated sciences where essentially exact replications with rejection of the null (Uher & McGuffin, were necessary. Indeed, most “facts” in the 2008). More recently, a review concludes that physical sciences rest on hundreds of nearly the interaction does indeed replicate across identical replications, sometimes performed studies using objective measures of adversity purposively, and sometimes performed as the among females, but the effect size attenuates Evolution, Biology and Society Vol. 7, No. 1 Spring 2010 - 3 – first step in an elaboration of the original measure of childhood maltreatment that shares research. no common items with the original MAOA study? And so on. The behavioral sciences simply have not proceeded in this manner (for reasons The second stage in the replication process discussed by Lindsay and Ehrenberg [1993]); is where life gets yet more interesting because at best, replications are typically “incidental” in differentiated replications help to establish the sense that data collected for one purpose scope conditions to the original finding; further prove reasonably suitable to replicate a finding validate the initial finding by establishing from a different dataset. As Lindsay and convergent validity; and strengthen confidence Ehrenberg note, this incidental quality is both in the original finding by eliminating potential good and bad. The good part: the ultimate biases due to unobserved heterogeneity goals of replications are (1) the validation of a (Rosenbaum, 2001). Through the second finding and (2) a determination of the range of stage, perhaps some replications will reject the conditions in which a finding holds true. null and some will not and, ideally, patterns Incidental replications typically bear on the among the replications will reveal the second purpose, which they view as more conditions under which a gene-environment important than the first purpose. They are effect is valid. The second stage is not without “differentiated” in their methodologies from the its challenges, however. Foremost, because original study and the resulting pattern of most (virtually all) differentiated replications are findings establishes scope conditions to the incidental, they vary many methodological original finding (hopefully). However, they also features of the original study simultaneously. If note that replications should ideally begin with the null hypothesis is not rejected (i.e., a the first purpose in mind: to validate the finding genetic finding is not replicated), it may be with “close” replications, ones that are very unclear which features of the incidental similar, methodologically (i.e., sampling, replication account for this failure to reject. Or measures, research design) to the original perhaps the original finding is a false positive. study. The Caspi 5HTT-Risch-Uher sequence can In other words, in a perfect world, a finding be read as a jumbled variant of this perfect of interest is followed by close replications and world scenario. The 5HTT study was followed then, if validated, by increasingly differentiated by many replications that, not ideally, fell replications. The first stage of close across the continuum from close to replications is an exercise in cost- differentiated replications but were decidedly effectiveness—why bother with differentiated toward the latter end of the continuum. The replications if the basic finding is very likely Risch meta-analysis revealed that across all of untrue? However, what is a close replication, these studies, the interaction did not hold. particularly in gene-environment studies? Does However Uher showed that the interaction is a close replication necessarily refer to the replicable among females when objective same population but a different sample? Must measures of adversity are used. Our own it use the same measures? The issue of review also suggests that the interaction holds measurement is especially vexing because the among clinical samples with high levels