Fig 1.
Sample stimulus pairs of pre- and post-evolution Pokémon characters used in Experiment 1.
These images are reproduced with the permission of the artist.
Fig 2.
Distribution of features to pre- and post-evolution categories in the Japanese training subset.
Features appear in order of importance from left to right. Asterisks denote significant features.
Table 1.
Confusion matrix for the Japanese RF.
Table 2.
Feature importance (Importance) and p values for features that achieved a feature importance greater than 0.1% in the Japanese RF.
Fig 3.
Distribution of features to pre- and post-evolution categories in the Chinese training subset.
Features appear in order of importance from left to right. Asterisks denote significant features.
Table 3.
Feature importance (Importance) and p values for features that achieved a feature importance greater than 0.1% in the Chinese RF.
Table 4.
Feature importance (Importance) and p values for features that achieved a feature importance greater than 0.1% in the Chinese RF.
Fig 4.
Distribution of features to pre- and post-evolution categories in the Korean training subset.
Features appear in order of importance from left to right. Asterisks denote significant features.
Table 5.
Confusion matrix for the Korean RF.
Table 6.
Feature importance (Imp.) and p values (p) for features that achieved a feature importance greater than 0.1% in the Korean RF.
Fig 5.
Distribution of features to pre- and post-evolution categories in the elicited training subset.
Features appear in order of importance from left to right. Asterisks denote significant features.
Table 7.
Confusion matrix for the Elicited RF.
Table 8.
Feature importance (Imp.) and p values (p) for features that achieved a feature importance greater than 0.1% in the Elicited RF.
Table 9.
Confusion matrix for the RF trained using the official Japanese Pokémon names and tested using the elicited samples.
Table 10.
Confusion matrix for the for the RF trained using the elicited samples and tested using the official Japanese Pokémon names.
Fig 6.
Boxplot of length for pre- and post-evolution Pokémon across the three languages.
Table 11.
Feature importance (Imp.) and p values (p) for features that achieved a feature importance greater than 0.1% in the Korean RF.
Fig 7.
Official Japanese Pokémon name length (Official) and elicited name length (Elicited) presented side by side in a boxplot.
Table 12.
OOB error rates for the RFs constructed in Experiment 1 (-L OOB), the OOB error for RFs constructed using length (+L OOB), and the feature importance of length in those RFs (L Imp).
Table 13.
Results of multiple random forests.
The mean OOB error for RFs constructed without length (-L OOBM) and their standard deviation (-L OOBSD), the mean OOB error for the RFS constructed with length (+L OOBM) and their standard deviation (+L OOBSD), and the mean feature importance of length (L ImpM) in the +L MRFs.
Table 14.
Confusion matrix for the Japanese MRF trained and tested on multiple subsets from the official Pokémon names that include length as a feature.
Table 15.
Feature importance of sounds for Japanese RF trained on official Pokémon names (RF Imp), mean feature importance of sounds for Japanese MRF trained on official names (MRF ImpM), and mean standard deviation for the Japanese MRF (MRF ImpSD).
Asterisks reflect a mean p value of less than 0.05.
Table 16.
Confusion matrix for the MRF trained on all official Japanese Pokémon names and tested on all elicited samples.
Table 17.
Confusion matrix for the MRF trained on all elicited samples and tested on all official Japanese Pokémon names.
Table 18.
Results of testing the Japanese MRF on the elicited data from Experiment 1 and the Elicited MRF on the Japanese data.
This includes both MRFs that do not contain length as a feature (-L OOBM) and those that do (+L OOBM) as well as their standard deviation (-L OOBSD, +L OOBSD).