View article

In early language learning, children have to learn what object a word refers to. Given that there are seemingly infinite number of possible word-object mappings when a word is heard for the first time, the question is how young learners can solve this puzzle. This is the problem of referential uncertainty famously framed by Quine (1960).

Many recent studies claim that when children learn to find what object a new word refers to, they may use very different learning strategies, such as the wholeobject assumption (Markman, 1991) and the mutual-exclusivity assumption (Mather & Plunkett, 2012). Children may also benefit from a number of different sources of information, including social and pragmatic knowledge (Baldwin, 1991; Bloom, 2000; Tomasello, 2003) and knowledge of syntactic structure (Fisher, Klinger, & Song, 2006; Yuan & Fisher, 2009). In addition to these useful strategies that give children a better idea of what to guess, recent evidence argues that both adult and infant learners can correctly identify a word’s referent by using cross-situational information (eg, Yu & Smith, 2007; Smith & Yu, 2008). For example, Yu and Smith (2007) presented adults a series of trials in which they saw four novel objects and heard four novel words on every training trial. Although each novel object has a fixed label associated with it, participants have no information indicating which object mapped on to which label on any given trial. They found that even though participants could not find the correct wordobject mappings initially, they were able to gradually learn which novel label consistently co-occurs with which object across multiple training trials. This …