Word and child characteristics in vocabulary learning of native English speakers and bilingual learners

Abstract

This study investigated how word and child characteristics affect children's ability to learn the meanings of novel words. Participants were fourth- and fifth-graders representing native English speakers (NE) and bilingual learners with fluent English proficiency (FEP) and designated English Learners (EL). Students were taught the meanings of a series of novel words that were either morphologically related or unrelated. Results showed that compared to the EL group, the NE and FEP groups were better able to learn the meanings of morphologically related words than morphologically unrelated words. The effect of morphological relatedness on semantic learning was stronger for the target words with familiar suffixes than with novel suffixes. Students learned the meanings of derivatives with familiar suffixes faster than the derivatives with novel suffixes. Verbal working memory and word reading fluency significantly predicted word learning performance. Implications for vocabulary instruction for learners with varying English proficiency are discussed.

Appendices

Appendix 1

Novel word learning task stimuli

Item no.	Subtest with familiar suffixes	Meaning Version A	Meaning Version B	Morphological relation
1	Blick	Cry	Like	Related
2	Blickable	Sad	Pleasant	Related
3	Flur	Play	High	Related
4	Flurness	Game	Top	Related
5	Ralp	Pain	Trip	Related
6	Ralpify	Hurt	Travel	Related
7	Torb	Like	Cry	Unrelated
8	Meckable	pleasant	Sad	Unrelated
9	Yord	High	Play	Unrelated
10	Slomness	Top	Game	Unrelated
11	Sark	Trip	Pain	Unrelated
12	Clorify	Travel	Hurt	Unrelated

Item no.	Subtest with novel suffixes	Meaning Version A	Meaning Version B	Morphological relation
1	Bilen	Car	Food	Related
2	Bilentod	Drive	Eat	Related
3	Dod	Fire	Lamp	Related
4	Dodlin	Hot	Bright	Related
5	Prap	Chair	Car	Related
6	Praptod	Sit	Drive	Related
7	Sod	Joy	Fire	Unrelated
8	Tiglin	Glad	Hot	Unrelated
9	Stug	Lamp	Joy	Unrelated
10	Chuklin	Bright	Glad	Unrelated
11	Tot	Food	Chair	Unrelated
12	Kegtod	Eat	Sit	Unrelated

Appendix 2

• 1. Model equation:

  Final model for measured variable ${\eta }_{tij}$ of individual i for item j at test trial t, may be written as:

  Level 1: ${\eta }_{tij}= {\pi }_{0ij}+ {\pi }_{1i}*\mathrm{Test trials}$

  Level 2:

  $${{\pi }_{0ij}= \theta }_{000}+{ \theta }_{001}*\mathrm{Female}+{ \theta }_{002}*\mathrm{NE}+ { \theta }_{003}*\mathrm{FEP}+{ \theta }_{004}* \mathrm{Grade}+{ +{ \theta }_{005}*\mathrm{Silent} \mathrm{word reading fluency}+ \theta }_{006}*\mathrm{Vocabulary}+ { \theta }_{007}*\mathrm{Reading comprehension}+ { \theta }_{008}*\mathrm{Extract base}+{ \theta }_{009}*\mathrm{Suffix Choice}+{ \theta }_{010}*\mathrm{Sentence span}+{ \theta }_{011}*\mathrm{Backward digit span}+{ \theta }_{012}*\mathrm{Test versions}{ + \theta }_{013}*\mathrm{Word type}+{ \theta }_{014}*\mathrm{Morphological relatedness}+{ \theta }_{015}*\mathrm{Morphological complexity}+{ \theta }_{016}*\mathrm{Native }\times \mathrm{Morphological relatedness}+{ \theta }_{017}*\mathrm{FEP }\times \mathrm{ Morphological relatedness}+ { \theta }_{018}*\mathrm{Word type }\times \mathrm{ Morphological relatedness}+ {b}_{00i}+ {c}_{00j}$$

  , with ${b}_{00i}\sim N (0, {\uptau }_{b00})$ and ${c}_{00j}\sim N (0, {\uptau }_{c00})$;

  $\pi_{1i} = \theta_{100} + \theta_{101} *{\text{NE}} + \theta_{102} *{\text{FEP}} + \theta_{103} *{\text{Word}}\,{\text{type}} + { }b_{10i}$, with $b_{10i} \sim N \left( {0, {\uptau }_{b10} } \right)$.

  The joint distribution of the random student effects follows a bivariate normal distribution with a variance–covariance matrix: $\left( {\begin{array}{*{20}c} {b_{00i} } \\ {b_{10i} } \\ \end{array} } \right)$ ~ N $\left[ {\left( {\begin{array}{*{20}c} 0 \\ 0 \\ \end{array} } \right)\left( {\begin{array}{*{20}c} {{\uptau }_{b00} } & 0 \\ 0 & {{\uptau }_{b10} } \\ \end{array} } \right)} \right]$. The covariance of the random student effects is fixed to zero for convergence purpose. The 95% credible interval of the estimate of their correlation coefficient includes zero.

• 2. Priors:

  The following priors were used: a Student's t-distribution ($v = 1$,$\mu =$ 0,$\sigma =$ 1) for the standard deviations of random effects, while we specified a normal distribution ($\mu =$ 0, $\sigma =$ 0.005) for the priors for the regression coefficients of silent word reading fluency, vocabulary, reading comprehension, extract the base, suffix choice, sentence span, and backward digit span, and a normal distribution ($\mu =$ 0, $\sigma =$ 0.25) for the priors for the regression coefficients of students' test trials, gender, language status, grade, test version, word type, morphological relatedness, morphological complexity, and interaction effects. These priors were selected for two reasons. First, we would like to be more conservative about the estimates. Second, we encountered convergence issues when using the default priors of brms for some regression coefficients. In order to improve the model convergence, more informative priors were selected.

• 3. Seeds and number of iterations requested:

  A seed value of 123,456 was used for all the analyses. We requested four sampling chains for 5000 iterations with a warm-up period of 2500 iterations for each imputed data set. Thus, overall 10,000 samples were generated for each parameter for each imputed data set.