Here we briefly describe some of the analysis history in the paper, because we feel they may be useful for other authors. In our first write up of the paper, our permutation variable importance varied slightly. We include the plot of the order of the variable importance here. In the first version of the paper, we reported four graphs (available [here][1] under "Version 1"). The top predictors are comparable. However, the first and last author conducted some of the analyses in parallel, and discovered that their orders differed. The last author used an unfiltered version of the dataset, while the first author used a version of the dataset, filtered by temperature in the lab. Both authors used mtry = 6 and trees = 1,000, and two seeds. We then found out where the differences where, and instead used the unfiltered data, and the data without MTurk participants for the final analyses, changed the number of trees to 10,000, and - for the full dataset - we ran 7 different versions to be sure of the stability of the model (4 seeds instead of 2 to be certain of the stability of the data. We also in- and excluded variable "site" to examine whether lab differed in Stroop performance, and used mtry = 8 in addition to mtry = 6). The graphs and partial dependency plots are also available [here][1] under "Version 2". In our revision of the paper, we also discovered we had used the wrong value of mtry. The final version of mtry was 4 (but the variable order remained largely comparable). We then also added extra thresholds based on random noise that we modelled. The graphs for the revisited forests with mtry = 4, the random noise, and the revisited forests with extra thresholds can also be found [here][1] under version 3. [1]: https://osf.io/q9tbr/files/