Competition between phonological and semantic cues in noun class learning

Abstract

Learning noun classification systems, like gender, involves inferring a language-particular set of (often probabilistic) cues to class membership. Previous work has shown that learners rely disproportionately on phonological cues (e.g., Karmiloff-Smith, 1981; Gagliardi & Lidz, 2014). Surprisingly, this occurs even when competing semantic cues are more reliable predictors of class. We investigate two possible explanations for this: first, that phonological cues are more salient to learners than semantic cues, and second that phonological cues are generally available earlier than semantic cues. We show that adult learners' treatment of conflicting cues to noun class in a miniature artificial language depends on both cue saliency and early availability. Importantly, learners prioritize earlier-available cues even when they are less salient than competitor cues. Our findings suggest a possible mechanism for children's over-reliance on phonology: children start building their classifications systems very early, when phonological information is available, but word meanings are not.

Children’s sensitivity to phonological and semantic cues during noun class learning: evidence for a phonological bias

“…Nevertheless, staging cue availability can amplify children's use of both types of cues, just as it does for adults. This suggests, contrary to the argument developed in Culbertson et al (2017) (and against our own expectations), that children are in fact biased to attend to phonological cues when acquiring noun classes. An a priori bias for phonology and earlier availability of phonological cues may therefore in combination lead to the robust finding of children's over-reliance on phonology at early stages of acquisition of noun classification systems across natural languages.…”

“…Finnish does not have a noun class system. 18 This is in contrast with the adult experiments reported in Culbertson et al (2017) where participants received twice the training in staged experiments. We did this in order to keep the experimental sessions relative short (around 20-25 minutes) for our child participants.…”

“…Since these exact cues have not been previously tested in adults, and because the method we are using here differs to some degree from Culbertson, Gagliardi & Smith (2017) (see below), we also tested adults using the same paradigm that we used for children.…”

“…The set of experiments reported above aimed to uncover whether child learners have a preference for phonological cues when learning a novel noun class system, and whether they are sensitive to early availability of cues in the same way as adults in Culbertson et al (2017). In particular, Culbertson et al (2017) found that adults were sensitive to cue saliencesome types of semantic cues (e.g., animacy) were learned more easily than others (e.g., flexibility), and some types of phonological cues were easier than others (e.g., a suffix ending on the noun vs. both a prefix and suffix together). However, there was no evidence of any bias in adults against phonology; when two cues were both equally predictive in training, but conflicted at test, cue salience rather than cue type appeared to determine which cue adults relied on.…”

“…Interestingly, no parallel studies have been conducted with phonological cues; we know of no studies which would allow us to compare children's acquisition of deterministic and non-deterministic phonological cues to class (for adults see Frigo & McDonald 1998). 5 This cue is similar in terms of phonological substance to high salience phonological cues used in Culbertson et al (2017) and Frigo & MacDonald (1989) which featured e.g., both a prefix and a suffix added to nouns. Here we used reduplication for two reasons: first, we wanted to keep word length to a minimum; second, previous studies have suggested that children have trouble learning classification systems in the lab, therefore we wanted to create cues that were as high salience as possible, and there is reason to believe that reduplication might be such a cue for children (Ota & Skarabela 2016, 2017.…”