While meta-analysis is typically used to identify value estimates for benefit transfer, applications also provide insights into the potential influence of design, study and methodological factors on results of non-market valuation experiments. In this paper, a metaanalysis of sixteen separate choice modelling studies in Australia with 130 individual value estimates relating to river health are reported. The studies involved different measures and scales of river health, so consistency was generated by transforming implicit prices from each study into a common standard of WTP per kilometer of river in good health. Tobit models have been used to identify the relationships between the dependent variable (WTP/km) and a number of variables. The results demonstrate that values are sensitive to marginal effects, with lower WTP/km for larger catchments, and higher WTP/km when river health is in decline. Values are also lower when river health has been defined by a subset of benefit types, such as recreation uses, vegetation health, fish health or bird populations. While there is evidence that the framing of the choice sets and descriptions of attributes have systematic impacts on values, there is very little evidence that choice dimensions, collection methods, sample sizes, response rates, statistical methods or publication status have influenced value estimates. Tests of apparent author effects show that these become insignificant when other explanatory variables are included in the models.