Brown, Smith, Samara, & Wonnacott (pre-print). Semantic cues in language learning: An artificial language study with adult and child learners

Many languages contain phonological and/or semantic cues to word class membership. Previous research suggests that when these two cues are both present in the input, learners often favour phonological cues. In the current study we ask whether child and adult learners can use semantic cues to form novel word classes, and whether they can generalize these classes to novel words from the same semantic class. In addition, we examined whether the number of nouns exemplifying each semantic class (type-frequency) determined the extent of the generalization. To do this, we taught children (6 year olds) and adults a semi-artificial language in which English nouns were followed by novel particles. Particle use was either fully or partially determined by the semantics (animal vs. vehicle) of the noun (with an additional control condition in which particle use was unconditioned by semantic cues). Crucially there were no phonological cues to word class membership. Production and forced choice tests showed that both child and adult learners were able to form word categories, and generalize these to novel nouns, when the semantic cues in the input were fully consistent. Adults (but not children), also generalized semantic cues to novel nouns based on partially consistent semantic cues. However, in almost all cases, successful generalization of the semantic categories to novel items was dependent upon explicit awareness of the semantic cues in the learning input. There was no evidence for a role of type-frequency in generalization.