When a child first begins to acquire a lexicon, the sources of word-meanings must be available from the situational context. However, it has been argued that the situational availability of the meanings of relational terms, such as verbs, is lower than that of whole-object labels, such as nouns. In this paper, we present a corpus of child-directed language, paired with situational descriptions, that enables us to explore the situational availability of word-meanings using a computational learner.