Perceptual statistical learning over one week in child speech production

Abstract

What cognitive mechanisms account for the trajectory of speech sound development, in particular, gradually increasing accuracy during childhood? An intriguing potential contributor is statistical learning, a type of learning that has been studied frequently in infant perception but less often in child speech production. To assess the relevance of statistical learning to developing speech accuracy, we carried out a statistical learning experiment with four- and five-year-olds in which statistical learning was examined over one week. Children were familiarized with and tested on word-medial consonant sequences in novel words. There was only modest evidence for statistical learning, primarily in the first few productions of the first session. This initial learning effect nevertheless aligns with previous statistical learning research. Furthermore, the overall learning effect was similar to an estimate of weekly accuracy growth based on normative studies. The results implicate other important factors in speech sound development, particularly learning via production.

A Learning Interaction Bbetween Statistical Learning Experiments

“…Accuracy was marginally lower in the low experimental frequency condition for the phonotactic learning first data (β = -.08, SE = .05, t = -1.66, p = .097); accuracy was significantly higher in the low experimental frequency condition for the phonotactic learning second data (β = .29, SE = .04, t = 6.51, p < .001). The results for the phonotactic learning second condition are surprising because high experimental frequency has typically been reported to increase children's production accuracy relative to low experimental frequency (for example, Plante et al, 2011;Richtsmeier et al, 2009;Richtsmeier & Goffman, 2017;Richtsmeier & Good, 2018).…”

“…The four target consonant sequences were chosen because consonant sequences are relatively difficult, making it likely that children would sometimes produce them in error, and learning could be measured. We note that data from the phonotactics experiment-when that experiment was completed firstare reported in Richtsmeier and Goffman (2017). All other data have not been reported elsewhere.…”

“…Mean accuracy across all conditions and participants was 5.53. Visual analysis of the figure suggests that participants were slightly more accurate in the high experimental frequency condition when the phonotactics experiment came first (see also Richtsmeier & Goffman, 2017), but when the phonotactics experiment came second, they were more accurate in the low experimental frequency condition. The baseline mixed effects model included experiment order, experimental frequency, and session as main effects.…”

<strong>Order in the Statistical Learning of Phonotactics</strong>

“…The targets appeared medially in nonwords (referred to hereafter to as items) with a CVCCVC shape and with stress on the initial CVC syllable. Although there is some evidence that item-medial consonant sequences are more difficult to learn than word-initial and word-final consonants (Endress & Mehler, 2010), other studies have confirmed that participants are sensitive to the experimental frequencies of these sequences (Richtsmeier, 2011(Richtsmeier, , 2016Richtsmeier & Goffman, 2017). Recordings of the items for this study were culled from those latter studies.…”

“…Participants heard items from 7-10 talkers, with the number of talkers depending on the list, and with 13 different talkers being used across all lists. In statistical learning research, talker variability is often beneficial (Bulgarelli & Weiss, 2018;Plante, Bahl, Vance, & Gerken, 2011;Richtsmeier, Gerken, Goffman, & Hogan, 2009;Richtsmeier & Goffman, 2017). However, the benefits are small, and talkers beyond the first two or three do not appear to provide an additional benefit (Richtsmeier & Moore, In preparation).…”