Samara, A. 1, 2 , Wonnacott, E. 2 & Ambridge, B. 1
1 School of Psychology, University of Liverpool
2 Division of Psychology and Language Sciences, University College London
How do speakers tell apart novel utterances which are acceptable from those which are not, e.g.: (i) *he carried me the ball? Under statistical preemption (Goldberg, 1995, 2019), this learnability ?paradox? (Baker, 1979) is resolved via probabilistic inference, such that the acceptability of (i) is negatively correlated with the frequency of ?carry? in an acceptable alternative (?he carried the ball to me?) (Ambridge et al., 2014). Preemption effects have been shown in artificial language learning and are predicted by models where cues ?compete? in determining the use of linguistic forms (e.g., Ramscar et al., 2010). Here we ask how broadly competition occurs: Are only near-synonymous uses relevant or can any usage (e.g., hearing he carried me home) help to block sentences like (i) (aka entrenchment).
In two sessions, 5-6-year-olds (n = 20) learned a novel language where nouns could be singular/plural, with some plurals unattested, followed by a grammaticality judgment task. Frequently encountering a noun as singular led to lower ratings for its unattested plural (compared with both attested-plurals and unattested-plural with novel nouns), despite the different semantics of singulars/plurals. This is direct evidence for entrenchment in children. An (ongoing) experiment tests whether this also holds for adults.