DIVORCE EFFECTS'AND CAUSALITY IN THE SOCIAL SCIENCES 51 and inferential methods,social scientists should Why are we Insufficiently Sceptical in the expect causal claims to be very rare indeed.We need accordingly to develop and maintain a culture of Social Sciences? restraint in relation to causal inference.Competition for research resources between and within disci- There appear to be many factors contributing to the lack of scepticism in the social sciences. plines militates against such a culture.A culture of restraint needs therefore to become part of the ethi- cal code of conduct of each of the social sciences,to Complexity be maintained internally by the same social process that regulates standards in any discipline.Excessive As social scientists we frequently defend ourselves claims should be the subject of adverse peer review against the real or imagined charge of slow progress for scientific publication.Premature publicity for in our disciplines by claiming greater complexity for findings that are not in the public domain-that matters social,economic,and demographic than is is,such that a full account of the research is available true of the physical,chemical,and biological in a public form to other social scientists-should domain.This is almost certainly true.Yet social preclude the research in question from formal pub- scientists,and I include myself in this comment, lication,as happens with certain medical journals in will typically have little appreciation of how Britain.Beyond professional convention,the best complex natural phenomena actually are and the way to promote scientific caution is to establish pro- lengths to which natural scientists go to establish cedures that ensure that causal claims are likely to be causality in their 'simpler'systems.Lacking day- examined carefully by other scientists.This could be to-day familiarity with the complexity of the done by fostering more attempts to replicate studies, natural,can we possibly have a real appreciation of a policy of accessibility of data-sets,encouraging the true complexity of social,economic,and demo- and allowing attempts at repeating published ana- graphic processes?There is probably an under- lyses,24 and a willingness by social scientists in appreciation of how demanding evidential investi- general both to give and receive evidence-based gation actually is in the natural sciences and how criticism of their work. much more so,because of greater complexity,it needs to be in the social sciences. We need to be mindful of the level of detail that natural scientists have to produce in order to estab- lish a causal link to the satisfaction of their Not Proven vs.No Effect colleagues.For example,apart from compiling extensive observational evidence,Darwin did some The argument of the earlier sections of the present painstaking experiments to check assumptions in paper is that the proposition that parental divorce his theory of evolution.In the case of buckminster- has long-term adverse effects on children is not fullerene,a previously unknown form of carbon,it proven.It may be true,but the evidence is in- was five years,and a great deal ofexperimental work, sufficient at present to support this conclusion and from its original discovery,to the experiments that current approaches to evidence are unlikely to confirmed its existence to the satisfaction of the substantiate the case.'Absence of evidence'is not at scientific community (Baggott,1996).This is the all the same as 'evidence of absence.To say (1)that normal state of affairs in the natural sciences. there is 'currently no evidence'that parental divorce Fastidiousness with evidence and inference is a hall- has an adverse effect in the long term on their mark of good science. children -the only defensible inference from avail- able evidence,in my view-is quite different from Experiments saying (2)that there is 'evidence of no adverse effect'The difference is between uncertainty We also sometimes defend ourselves by highlighting whether or not there is an effect and certainty that the impossibility of experiments in the social there is none. sciences.While on the whole true,this is no                    ?                      )                      +                                                      /           8               >                            >                       /   7  7                             "            5                                        9C                      " @  (   @ 6 5                         3       !           G                "         2+  4        2  4 5  #%*    2    4        3        >            >    3   #9*    2    3 4 5 3              3             ! "    # 5            "       
  +                                                 5       ,      !                                            24   J"                                 @ 5                                            ?                               "          F         
     "    "          !     "       "               "                          #7  %&&E* 5        3      F          "    6   ?                    ?