During the initial stage of language development, nouns corresponding to “object categories” are known to be acquired earlier than other parts of speech, including verbs. However, it is unclear whether the semantic content of words is the same for toddlers who have just begun to learn them and adult speakers. This preliminary study experimentally investigated the theoretical hypothesis of Werner and Kaplan, which posited that the initial meanings of noun-like words do not differentiate into specific categories of objects themselves but rather refer to holistic event categories. Toddlers aged 19–35 months (N = 36) were prompted to select one of two juxtaposed video stimuli (e.g., “putting shoes on” vs. “rubbing two baskets” in the match condition, “putting baskets on” vs. “rubbing shoes” in the mismatch condition) using questions about noun-like words (e.g., “Which is shoes?”). Statistical modeling demonstrated that the meanings of noun-like words for toddlers under 21 months of age or with a vocabulary size of less than 140 words were deeply influenced by conventional actions related to the objects (e.g., “putting shoes on”), whereas they subsequently differentiated into specific object categories (e.g., “shoes” alone), becoming “true nouns.” These findings support Werner and Kaplan’s hypothesis and provide the first experimental evidence for the theoretical and observational assumptions that early words are not easily classified into parts of speech only by their vocables (e.g., nouns, verbs). We discuss the flexibility of vocable–meaning connections during early language development by proposing the “semantic pluripotency hypothesis.”