A modest proposal for research.

Complexity is conducive to dependence on specialists on the one hand, and mistrust of them on the other. If you read in the Bulletin, “A dendrogram of sequence-relatedness was constructed by the method of the VPI maximumlikelihood bootstrap consensus tree (PUZZLE, version 4.0)” (1), and this is not in your area of specialization, you might respect it for its apparent professionalism, resent it for its obscurity, or just tune it out. The usefulness of such information must be judged in relation to what the researchers were trying to do. In this case they were trying to calculate the implications of three cases of paralysis caused by wild poliovirus which had occurred in Roma children in the eastern harbour city of Bourgas, in Bulgaria, 10 years after the last case had been reported. In the context of the polio eradication campaign, the objective of the study was valid and it came to a plausible conclusion, so the Bulletin’s editorial committee was inclined to trust the authors, together with the peer reviewers whose verdict was that the methods were OK. Trust is indispensable but so is wariness, as Uwe Reinhardt & Tsung-mei Cheng pointed out in their review of The world health report 2000 (2). They objected that “to see what was actually done, one must plough through the cryptic commentary that accompanies the tables in the Annex or dig up and read sundry sources cited in the references”. Having done that, they suspected that the measures were flawed and the findings invalid. More fundamentally they questioned the value of the attempt itself, which was to rank the national health systems of 191 countries in order