[PS-2.1] Cross-situational word learning: a decade of research

Dal Ben, R. 1 , Souza, D. 1 & Hay, J. 2

1 Universidade Federal de Sao Carlos
2 University of Tennessee

Word learning involves dealing with a great deal of ambiguity. Probabilistic word-referent co-occurrences may provide learners with statistical cues that may facilitate learning. Cross-situational word learning (CSWL) research, in which co-occurrences between words and referents across encounters provide relevant mapping information, has shown that adults and infants exploit this statistical cue to learn words. Since the pioneering work of Yu and Smith (2007), evidence for CSWL has been well documented. Here, we present an ongoing systematic review and meta-analysis on CSWL empirical research. As of 31st December 2017, a total of 52 articles, reporting 86 experiments (N = 9,455), spanning a wide range of topics, including the role of cognitive abilities (attention, memory), linguistic information (mutual exclusivity, naming frames, prosody, minimal pairs), social cues (looking direction), and levels of ambiguity (word-referent relation and contextual ambiguity), have been published. Overall the results support word-referent co-occurrence as an informative statistical cue for word learning. However, extensions to word learning in natural contexts and longevity of the learned words are uncertain. Furthermore, the learning strategies that underlie CSWL continue to be a matter debate. We are currently reviewing the articles published through 2018 and will present our up-to-date meta-analytic findings.