Assessor consistency when scoring different impact types. Results from the 12.7% of the 1,000 randomisations, i.e. models including only the single effect of the detailed impact types as explanatory variable, when using the dataset including all protocol questions on impact (GQuest-Taxon). The unit of the x-axis is residuals; note that these estimates are from a model using the residuals of a previous model as dependent variable. The dot depicts the mean and the brackets the confidence level at 95%. See consensus Tukey adhoc-test in Suppl. material 1: Table S11.