How words can be learned by observation depends on what is meant by “learned”
Word learners experience naming events differing widely in their referential quality. Whereas referents of some naming events are transparent from their extralinguistic contexts, referents of many naming events are ambiguous. Word learning theories are divided in whether learners mainly learn from a few transparent events or whether learners also aggregate across ambiguous ones. Data consistent with the former view are evident in the Human Simulation Paradigm (HSP) in which naïve observers must identify parents’ words from muted vignettes of parent-toddler interactions. The HSP reveals that even adults struggle to identify the identity of parents’ words across ambiguous vignettes. Our work revisits the HSP by examining how alterations to its dependent variable affects the conclusions about the naming events that shape learning. This work underscores how one’s definition of learning has implications for both accounts of the mechanisms of learning as well as accounts of the relevant input into those mechanisms.