Repeatability and transparency in ecological research

Abstract

A fundamental tenet of science is that results must be reproducible by other scientists before they are accepted as factual. However, because ecological phenomena are context-dependent, and because that context changes through time and space, it is virtually impossible to reproduce precisely or quantitatively any single experimental or observational field study in ecology. Yet many ecological studies can be repeated. In particular, ecological synthesis—the assembly of derived data sets and their subsequent analysis, reanalysis, and metaanalysis—should be easy to repeat and reproduce. Such syntheses also demonstrate qualitative and quantitative consistency among many ecological studies (Gurevitch et al. 1992, Warwick and Clarke 1993, Jonsen et al. 2003, Walker et al. 2006, Cardinale et al. 2006, Marczak et al. 2007, Vander Zanden and Fetzer 2007) and provide strong support for general ecological theories. It should come as no surprise that meta-analysis by Mittelbach et al. (2001) of the effect of productivity on species richness has led to the development of a cottage industry focused on empirical testing of this relationship (post-2001 examples abound in Appendix A of Whittaker 2010). But it is much more surprising that continual reanalyses of the same data sets (Whittaker and Heegaard 2003, Gillman and Wright 2006, Partel et al. 2007) have yielded such disparate results that Whittaker (2010) has suggested abandoning the effort to obtain consistent results from the available data. He goes even further, suggesting that ecology may not yet be ready for meta-analysis and data synthesis. For two reasons, I respectfully suggest that Whittaker’s critique is misplaced. First, of all the studies critiqued by Whittaker (2010), only Mittelbach et al. (2001) actually conducted a formal meta-analysis. The others, as pointed out by Whittaker (2010), undertook extensive primary analyses, but the authors did not conduct formal meta-analyses (Gurevitch and Hedges 1999). Second, and more importantly, if ecological synthesis is transparent—data, models, and analytical tools are available freely to the research community—then it should yield consistent, repeatable results. We may then disagree on the interpretation of the resulting synthesis, but at least we will be able to agree on the reproducibility of the results themselves.