This study investigated whether the semantic similarity and grammatical class of distracter words affects the naming of pictured actions (verbs) in Japanese. Three experiments used the picture-word interference paradigm with participants naming picturable actions while ignoring distracters. In all three experiments, we manipulated the semantic similarity between distracters and targets (similar vs. dissimilar verbs) and the grammatical class of semantically dissimilar distracters (verbs, verbal nouns, and also nouns in Experiment 3) in addition to task demands (single word naming vs. phrase/sentence generation). While Experiment 1 used visually presented distracters, Experiment 2 and 3 used auditory distracter words to rule out possible confounding factors of orthography (kanji vs. hiragana). We found the same results for all three experiments: robust semantic interference in the absence of any effects of grammatical class. We discuss the lack of grammatical class effects in terms of structural characteristics of the Japanese language.