Free word association tasks are used to evaluate different hypotheses proposed by interactive and cascade models of speech processing. The interactive model predicts a small semantic and phonological distance between the target and the response words, whereas the cascade model predicts that the responses are semantically close to the targets but are phonologically far from the targets. One hundred forty-five stimuli tested with 22 participants resulted in 2289 tokens available for testing. The phonological and semantic distances were automatically measured using Levenshtein distance and word embeddings; additional metadata over 10M drawn from the Academia Sinica Corpus in Taiwan was computed. The results show that the stimuli and the responses are closer than random semantically and phonologically, supporting the predictions from the interactive models. However, we also observe that the semantic distance is shorter than the phonological distance. A concomitant increase in chronometry is found with longer semantic distance.