Negative results are disappearing from most disciplines and countries

Analyses over 4,600 papers published in all disciplines between 1990 and 2007, measuring how often papers that declared they tested a hypothesis reported positive support for it. The proportion reporting a positive result was 70.2% in 1990-1991, peaked at 88.6% in 2005 and was 85.9% in 2007; the overall frequency of positive supports grew by over 22%, more strongly in the social and some biomedical disciplines, with significant differences between countries. The author argues methodological artefacts cannot explain the pattern.