This study investigates and analyzes native Japanese speakers' evaluations of the literal translation and semantic similarity of English metaphors. Using an existing model for calculating the meanings of metaphors, we will determine whether the model can be applied to translated metaphors as well. The hypothesis is that there is a correlation between the evaluations calculated by the model and those made by humans. The results, however, showed a lack of correlation in three sets of items and only a weak positive correlation in one set. This result may be attributed to a particular method used in the model.