A structural-sociological approach to suicide research holds that an aggregate-level cause of suicide should correlate with the suicide rates in a population. In 1980, Sainsbury, Jenkins, and Levey published the article "The Social Correlates of Suicide in Europe" which related the suicide rates in 1961-63 and the changes in them in the following 11 years to 15 social variables in 18 European countries. Its main findings were that the changes in suicide rates could be attributed to specific changes in the social environment. Complementary discriminant analyses showed that it was possible accurately to divide the countries into low- and high-change suicide rate groups on the basis of a combination of the social variables. Although criticized for its method, the study has been widely quoted and sometimes presented as the most definitive current study on the subject. In order to see whether its results held for similar data 16 years later it was replicated for 1977-79 and the ensuing 11 years, with data and method as similar as possible to the original. The results agreed with those of the original study on only one point: the correlations between the levels of the social variables and those of the suicide rates were similar in both periods. However, changes in the suicide rates were unrelated to either the levels of the social variables or the changes in them: correlations found in the original study tended to change profoundly or disappear. Moreover, the results of the original discriminant analyses were a property of the method employed and thus independent of the data. Statistical artefacts or social processes such as changing expectations are unlikely to explain the suddenly changing or vanishing correlations. The original correlations seem to have been largely spurious and dependent on the fact that the more modern countries in Europe experienced a "suicide boom" in the 1960s. As the boom waned in these, it was beginning in the less modern countries: the correlations between the processes indicated by the social variables and the suicide rates were reversed or disappeared. The results call the existence of clear relations between these "suicidogenic" social circumstances and the suicide rates into question. Since many of the variables used are traditional "Durkheimian" indicators of the integration of society, a critique of this still-dominant view of the relationship between society and suicide mortality, or its common operationalization, is implied.