The Lexical Restructuring Hypothesis: Two Claims of PA Evaluated Michelle E. Erskine, Tristan Mahr, Benjamin Munsona, and Jan R. Edwards University of Wisconsin-Madison, aUniversity of Minnesota

Introduction Methods & Analysis Results Results Nonword Repetition • Phonological awareness (PA) is an important skill for learning to Mediation Model 1 Mediation Model 2 Stimuli read (McBride-Chang, 1990; Cunningham & Stanovich, 1997). • 22 pairs of nonsense words adapted from Edwards et al., 2004 • A variety of factors including phonological working memory Blending Results: • Pairs included a 2-phoneme sequence that contrasted in (PWM), speech perception, and vocabulary skills have been shown • Consistent with complete mediation. phonotactic probability Vocabulary to be related to phonological awareness. • A boot-strapping method to evaluate mediation yields a • Presented stimuli matched child’s native dialect significant indirect effect of speech perception on PA, p < .001, • However, the relationships among these factors are not well and the direct effect of speech perception on PA is no longer understood. Procedure significant. • The lexical restructuring hypothesis proposes that the association • Nonwords were paired with a picture of an unfamiliar object in a • 94% of the effect of speech perception on PA is mediated by among PWM, speech perception and PA are secondary to vocabulary picture-prompted auditory-word-repetition task Phonological Phonological expressive vocabulary size. development in children. Working Memory Awareness • The direct effect of speech perception on PA is no longer • Thus, PA primarily emerges as a result of the gradual significant when the model accounts for effects of reorganization of the lexicon and, to a lesser degree, the vocabulary size. encoding (e.g., speech perception) and storage (e.g., memory “tweket” vs. “pwagub” Elision Results:

• There is no evidence for mediation. capacity) of sub-lexical units. High PP Low PP • Using a boot-strapping method to evaluate mediation, there is a Purpose Summary & Discussion significant indirect effect of PWM on PA, p < .001, but the direct This study uses mediation analysis to evaluate: Analysis of the Nonword Repetition Task effect of PWM on PA remains significant at p = .04. Summary • Is the relationship between speech perception and PA explained by • The 2-phoneme target sequences were transcribed and scored as in • 54% of the effect of PWM on PA is mediated by expressive • Not surprisingly, words that were higher in phonotactic probability children’s vocabulary development? Edwards et al., 2004 by counting the number of target features vocabulary size. were produced more accurately than words that were lower in • Is the relationship between phonological working memory and PA produced correctly • PWM and vocabulary size independently predict phonotactic probability. explained by children’s vocabulary development? • We fit a generalized linear mixed-effects model to predict accuracy phonological awareness. • The results partially support the lexical restructuring hypothesis. from phonotactic probability: • For Blending, a substantial proportion of the effects of both speech • Fixed effects: Overall accuracy (intercept) and the sequence perception and PWM on PA was explained by a child’s lexical Participant Summary & Descriptive Statistics frequency effect (slope) Blending Results: knowledge. • Ages: 3;0 (+/- 2 months) at time 1 and 4;0 (+/- 1 month) at time 2. • Random effects: By-subject random intercept and random slope, • Consistent with partial mediation. • For Elision, the results were similar for the effect of speech • Monolingual English speakers with typical speech and language and the by-item random intercept • Using a boot-strapping method to evaluate mediation, there is a perception on PA. However, Mediation Model 1 did not support the development. • Outcome measure for mediation analysis: Estimated accuracy significant indirect effect of PWM on PA, p < .001, and the direct claim that the effect of PWM on PA was mediated by vocabulary score for each participant controlling for frequency (a measure effect of PWM on PA is no longer significant. knowledge. This finding suggests that both vocabulary size and Analysis 1: Accuracy & Frequency Effect of PWM) • 60% of the effect of PWM on PA is mediated by expressive PWM independently influence PA. N = 73 vocabulary size.

Mediation Model 1: PWM à Vocabulary à PA Mediation Analyses • The direct effect of PWM on PA is no longer significant N = 66 • Mediated Logistic Regression Model used for Elision scores when the model accounts for effects of vocabulary size. Discussion Mediation Model 2: Speech Perception à Vocabulary à PA • 1/3 of participants scored a raw score of 0 • A primary goal of speech and language services is to provide a N = 66 • Mediated Linear Regression Model used for Blending scores Mediation Model 2 platform and foundation for academic success. The results of this

Number Mean EVT-2 Mean Elision Mean study suggest that a comprehensive approach to treatment that • Independent variables: NWR child-level intercept, EVT-2 GSV emphasizes vocabulary may lead to natural improvements in the of males/ standard score scaled score at Blending score, and % correct on Minimal Pairs females at age 3 (SD) age 4 (SD) scaled score at Vocabulary child’s phonological awareness skills. • Dependent variables: CTOPP-2 Elision Scaled Score and age 4 (SD) • It is necessary to consider the differences among tasks used to CTOPP-2 Blending Scaled Score measure PA 35/38 116 (14) 11 (3) 10 (3) • Elision is particularly challenging and requires children to make

Methods: Tasks Analysis 1: Accuracy & the Frequency Effect use of both their lexical knowledge and phonological working • As in previous research, phonotactic probability is a significant Phonological memory. Minimal Pairs Speech Perception predictor of NWR accuracy Awareness • By contrast, children can succeed on blending simply by Stimuli relying on their lexical knowledge. • 15 minimal pairs of familiar words. • A one-unit increase in

• Children were presented with the stimuli in their native dialect phonotactic probability Elision Results: ACKNOWLEDGEMENTS above the average • Consistent with partial mediation corresponded to an We want to acknowledge the L2T research teams at UW-Madison, • A boot-strapping method to evaluate mediation yields a significant “peas” vs. “keys” increase in accuracy of OSU, and UMN who helped collect the data; Mary E. Beckman, who indirect effect of speech perception on PA, p < .001, and a 1.6% (b = 0.1, SE = 0.03, helped with the collection, aggregation, and analysis of the nonword marginally significant direct effect of speech perception on PA, z = 3.09, p = 0.002). repetition; Emily Wagner, who helped transcribe the nonwords; and Procedure p = .06. the children who participated in the study and the parents who gave • 60% of the effect of speech perception on PA is mediated by • Each pair of images presented and named by the computer before their permission. This research was supported by NIDCD Grant vocabulary size. both images were presented at once while the target word was played R01-02932 to Jan Edwards, Mary E. Beckman, and Benjamin Munson • The direct effect of speech perception on PA is marginally and by NICHD Grant P30-HD03352 to the Waisman Center. The Primary measure significant when the model accounts for effects of authors have no financial or non-financial conflicts of interest. vocabulary size. • Percent correct for each child was computed automatically