Phonological working memory and non-verbal complex working memory as predictors of future English outcomes in young ELLs

Abstract

Aim of the study:

The aim of this longitudinal study was twofold: first, to determine whether the relationship between working memory measures and language performance in young English language learners (ELL) remains constant over the year. The second aim was to determine if performance on working memory tasks predicts future performance on language measures.

Methodology:

The participants were 27 ELLs between the ages of 5 and 6 years who were in their first year of formal schooling and attended the same mid-low socioeconomic status school in South Africa. The participants were tested three times throughout the year on tasks of working memory and an English assessment battery.

Data analysis and results:

Mixed effects models and multiple linear regression were used to address the aims of the study. The first aim of the study showed that there are significant correlations between all working memory measures and all language measures in varying strengths across the year. The second research aim further elaborated on this by showing that both phonological working memory and non-verbal complex working memory are implicated in the acquisition of syntax, semantics and pragmatics at different points throughout the year.

Implications and originality:

Language acquisition in ELLs is not a stand-alone process and working memory measures may be able to predict future language outcomes. This could indicate that working memory measures may be used as an indicator for who may need language intervention, at a time when the ELL only has limited English proficiency and limited English exposure. This research is the first of its kind to originate from Africa, with a sample from low socioeconomic, culturally and linguistically diverse circumstances who are exposed to English consistently for the first time and are tested with working memory tasks with less strong language components.

Keywords

English language learners bilingualism working memory English proficiency

Introduction

Children develop language skills at different rates and do not necessarily follow one standard trajectory (Paradis, 2015). A great challenge is how to ascertain whether or not a slow rate of language acquisition is an indication of an underlying language disorder. This is particularly difficult in the context of children who are still in the process of acquiring another language, such as English language learners (ELL). It is becoming increasingly common to have children in English-medium schools worldwide who are tasked with learning English relatively quickly, most commonly because it is the school’s sole medium of instruction, while having no prior knowledge of English. To accurately identify language learners with a language disorder and provide appropriate intervention, measures that take into account not only language abilities but also cognitive abilities have been used to disentangle bilingual development from language disorders. Complicating the matter further is that not all ELLs reach native English proficiency, which impacts negatively on their performance in English-medium language-related tasks, such as reading comprehension (Kieffer, 2011; Mancilla-Martinez & Lesaux, 2010). To this end, measures that take into account not only language abilities but also more cognitive abilities have been used in an attempt to disentangle bilingualism from language disorders. Such measures include the use of nonword repetition, sentence repetition and other working memory measures (e.g. Henry & Botting, 2016; Kohnert, 2010).

Previous research indicates that measures of non-linguistic processing may provide important information about language development in multilingual contexts (Paradis, 2010; Sandgren & Holmström, 2015), especially in the preschool years (Chiat & Roy, 2008). It is therefore important to determine which cognitive measures underpin language processes and which of these measures can predict future language outcomes. Cognitive measures such as working memory capacity have been found to be related to language acquisition, also in studies with ELLs (e.g. Gorman, 2012; Swanson, 2014; Swanson et al., 2015). ‘Working memory’ is defined as a limited-capacity resource which is involved with the short-term storage and use of information (Baddeley & Hitch, 1974; Engle et al., 1999). According to Wen and Skehan (2011), working memory is important for second language development and has a central role in second language acquisition, to the extent that it should be considered as a component of foreign language aptitude.

To explore the link between working memory and language skills further, the current study assesses mid-low socioeconomic status (SES) ELLs between 5 and 6 years old in their first year of formal English-medium schooling. The children are presented with tasks of working memory and language in order to determine (a) whether there is a consistent relationship between working memory measures and language performance throughout the year and (b) whether performance on working memory tasks can predict future performance on language measures. Standardised language assessments are not available for every language spoken in South Africa, therefore, to use working memory along with language tests to determine the child’s development would be greatly beneficial for children who are still in the early stages of second language (L2) learning. The first year of formal schooling is a crucial point, as it is the first time children receive five hours a day of structured and consistent English-medium input. It is during this time that children might already start falling behind, and decisions need to be made whether or not to refer the child for intervention. This study follows children’s progress throughout this first year of language immersion longitudinally and is the first of its kind to track the relationship between working memory and language in a South African mid-low SES school, using working memory tests that do not have a strong English component that could negatively influence performance.

Working memory and language in childhood

There are several models that have been proposed to describe the structure of working memory (see Miyake & Shah, 1999 for an overview). However, the multi-component model, which was developed by Baddeley and Hitch (1974) and elaborated on by Baddeley (2000), is the most suitable for describing working memory development during childhood (Alloway et al., 2006; Bayliss et al., 2005; Henry, 2011). The multi-component model consists of a central executive, a phonological loop and a visuospatial sketchpad. The updated version of this model includes an episodic buffer that serves to integrate information from the above-mentioned subcomponents (the phonological loop and the visuospatial sketchpad) and store this information temporarily (Baddeley, 2000). This constitutes a domain-general approach to working memory which states that working memory tasks place the processing load on the central executive, while the storage aspect of the task is supported by the relevant specific component – that is, the phonological loop or the visuospatial sketchpad (Baddeley & Hitch, 1974).

The central executive component is responsible for the processing of information and serves to coordinate and control the three subsystems, namely the phonological loop, the visuospatial sketchpad and the episodic buffer. The phonological loop consists of a short-term store and a verbal rehearsal process, while the visuospatial sketchpad integrates spatial and visual information that can be used and stored. Previous studies that have investigated the development of working memory have found that these components are in place from as young as 4 years of age (e.g. Alloway et al., 2006).

The aforementioned phonological loop is tapped into with tasks of phonological working memory. Phonological working memory and its association with vocabulary learning and language acquisition make up a large body of research (e.g. Baddeley et al., 1998; Farnia & Geva, 2011; Gathercole, 2006). Phonological working memory facilitates the long-term learning of the phonological structure of a given language (Baddeley et al., 1998) and is therefore integral to vocabulary learning in both L1 and L2 acquisition. The most commonly used task to assess phonological working memory is nonword repetition, which involves the participant hearing and repeating a novel word. Close associations between nonword repetition and vocabulary measures have been found in L1 acquisition (e.g. Engel de Abreu et al., 2011; Gathercole, 2006; Gathercole & Baddeley, 1989) as well as in L2 acquisition (e.g. Masoura & Gathercole, 2005; Service & Kohonen, 1995; Szewczyk et al., 2018). Specifically, the greater the child’s phonological memory span, the better the child’s vocabulary scores. Outcomes from nonword repetition tasks are highly related with vocabulary measures in young children, but this relationship becomes weaker as children get older (Gathercole, 2006, p. 514). A longitudinal study of children between the ages of 4 and 8 years showed that vocabulary and nonword repetition scores were highly correlated at the ages of 4, 5 and 6 years (r = .52–.56) but that by the age of 8 this correlation was weaker (r = .28) (Gathercole & Baddeley, 1989; Gathercole et al., 1992). The same weakening over time is found in studies of L2 acquisition. For instance, Cheung (1996) found that the relationship between scores on English nonword repetition tasks and English vocabulary tasks was significant for Cantonese-speaking children learning English, yet only for those who had low English vocabulary scores. For the children who had high English vocabulary scores, these scores did not exhibit the same significant relationship with their nonword repetition scores. From these and other studies, it is concluded that phonological working memory is implicit mainly in new word learning when vocabulary size is small, which explains the weakening of the relationship between it and vocabulary tasks over time (e.g. Cheung, 1996; Gathercole et al., 1992).

A possible confounding factor regarding the relationship between nonword repetition and vocabulary is the finding that the familiarity of the phonological structures of the nonword items determines performance on the task (e.g. Gathercole, 1995). This has been referred to as ‘the wordlikeness problem’, which entails the finding that nonwords which have a high phonotactic probability are easier to remember than nonwords which are more irregular in their phonotactics (e.g. Edwards et al., 2004). Children might therefore be using previous linguistic knowledge to aid in recall and repetition. Gathercole (1995) reported that monolingual children’s performance on wordlike nonwords was correlated with their performance on vocabulary measures but that their performance on less wordlike nonwords was correlated with digit span. The conclusion from this study was that vocabulary knowledge had a causal relationship with scores on the wordlike nonwords but that this causal relationship did not exist with the low wordlike nonwords. Baddeley (2003) posits that this may be due to the two systems that make up the phonological loop, namely the storage component, which is responsible for immediately storing phonological input, and the articulatory component, which is in charge of rehearsing the input. The storage component is said to not be influenced by previous linguistic knowledge, whereas the articulatory component is indeed dependent on previous knowledge of a given language, such as morphological rules. This follows that ELLs, who have less English exposure and a lower English proficiency, would be expected to perform poorly on the nonword repetition tasks that are based on English phonotactics.

Studies have also found a link between phonological working memory and L2 grammar. Authors such as Ellis and Sinclair (1996) assert that this link is found because children with more developed memory spans are more apt at creating long-term linguistic representations. Data from 11-year-old French-speaking children, who were learning English, was gathered from two nonword repetition tasks. One task was based on English phonotactics and the other was based on Arabic phonotactics, which the authors assumed was far detached from the children’s prior linguistic knowledge and thus would not be affected by it (French & O’Brien, 2008). The results showed that both the Arabic and English nonword repetition tasks were significant predictors of the outcomes for the L2 grammar tasks. Similar results were found in a study by Verhagen et al. (2015) of bilingual 4-year-old children. The authors found that the nonword repetition tasks had moderate but significant correlations with L2 grammar scores.

As opposed to the phonological loop, the central executive component is thought to play a more general role in the early acquisition of language, especially in the acquisition of language comprehension. It stands to reason that there is a relationship between this working memory component and language learning, as this component is responsible for the complex cognitive action of actively processing information while storing additional information (Baddeley, 2000). Complex working memory tasks are used to measure the central executive. The most commonly used tasks are the backward digit span and listening recall, which are classed as complex verbal working memory tasks. These are said to furnish a truthful indication of higher-level cognition, making them more accurate than simple memory tasks that only require the storage of information (Daneman & Carpenter, 1980).

Processing and storage are both crucial cognitive abilities for reading and comprehending a text, where words have to be read individually and held in mind while continuing to read subsequent words. Performance on reading and language comprehension tasks is linked to scores obtained on complex memory span tasks that tap into the central executive component of working memory (e.g. Cain et al., 2004; Swanson, 2014; Swanson et al., 2015; Swanson & Beebe-Frankenberger, 2004). The syntactic and semantic interpretation of sentences is also affected by working memory capacity; individuals with lower working memory capacity have lower scores on tasks involving the comprehension of unfamiliar or complex syntactic structures (see Kidd, 2013 for a critical review). Complex verbal working memory has also been found to be implicated in monolingual children’s receptive syntax (e.g. Ellis Weismer et al., 1999), sentence comprehension (e.g. Montgomery, 1995) and accuracy in grammaticality judgement (e.g. McDonald, 2008).

Complex verbal working memory has not only been considered in monolingual populations but also in bilingual populations. For example, Verhagen and Leseman (2016) investigated the relationship between complex verbal working memory, grammar and vocabulary in 5-year-old Turkish-Dutch bilinguals as well as Dutch monolinguals. They found that complex verbal working memory was a significant predictor for both L1 and L2 morphology and syntax knowledge. Similar results have also been found in previous studies with children (e.g. Engel de Abreu & Gathercole, 2012; Masoura & Gathercole, 2005). Children between the ages of 7 and 8 years who were trilingual in Luxembourgish, German and French were tested on backward digit span and counting recall tasks by Engel de Abreu and Gathercole (2012). Their results yielded that the complex verbal working memory tasks were a predictor of syntax, reading comprehension and spelling across all three of the children’s languages. Andersson (2010) investigated the role that the phonological loop and the central executive play in children’s foreign language comprehension. The children’s working memory abilities were tested between the ages of 9 and 10 years, and foreign language comprehension was tested one to two years later. Results showed that working memory was associated with foreign language proficiency but it was found that the phonological loop and the central executive were independent predictors for future foreign language comprehension. These results are in line with what previous studies have found in both children and adults (Geva & Ryan, 1993; Miyake & Friedman, 1998; Service et al., 2002).

All the aforementioned studies make use of verbal working memory tasks. As their names state, these tasks have a specific language component; the child is expected to have enough prior knowledge of the language of testing to be able to repeat what the examiner is presenting. In the case of early bilinguals, who have limited knowledge of their L2, these tasks are less suitable due to the strong language component. As stated previously, the articulatory component of the phonological loop is dependent on previous knowledge of a given language (Baddeley, 2003). One cannot be sure that what is being tested is working memory abilities and that low scores on these complex verbal working memory tasks are not an effect of a lack of knowledge of the language of testing. Children with less English knowledge are therefore expected to perform poorly on the nonword repetition tasks that are closely based on English phonotactics, the poor performance being an artefact of low proficiency and not low working memory. A similar uncertainty about complex verbal working memory tasks, such as digit span or reading span, has been put forward recently by a handful of authors (Gangopadhyay et al., 2015; MacDonald et al., 2001; MacDonald & Christiansen, 2002). A further problem is that the majority of complex verbal tasks are considered to be too complex for 5-year-old children, as floor effects have been found in previous studies, which may be due to the instructions having been too difficult to grasp (Petruccelli et al., 2012; Pickering & Gathercole, 2001).

Therefore, previous studies found a relationship between complex verbal working memory tasks and language aptitude, but it is unclear whether these tasks measured what they set out to measure, as the participant’s poor knowledge of English could have negatively affected the results. Against this background, the current study sets out to use a complex non-verbal visuospatial working memory task, which necessitates both storage and retrieval of information, to tap the central executive (following the findings of e.g. Ellis Weismer et al., 1999; Montgomery, 1995). This will render a more accurate estimate of an ELL’s central executive capacity than verbal working memory tasks do, due to the former not being dependent on language knowledge. Also, in the current study, two nonword repetition tasks are employed, one task that has a high wordlikeness with English and another task that strives to be linguistically independent. The use of two different nonword repetition tasks is to address whether the strong language specific component of verbal tests affect performance in ELLs with limited knowledge of English. The current research focuses on two core research questions:

Is there a consistent relationship between working memory measures and language performance at each time point throughout the year?

Does performance on working memory tasks predict future performance on language measures?

Method

Participants

Typically developing children in their first year of formal schooling, aged between 5 and 6 years, were eligible for participation. Three public schools were approached to take part in the study. The first grade of the only public school that responded was in a multicultural area, consisting of three classes with a total of 60 children. Of these children, 45 were ELLs. Eighteen of these ELLs did not meet the inclusion criteria and/or parental consent was not obtained for their participation. Therefore, 27 children who are ELLs consented to participate in the study (see Table 1 for participant details). These 27 children made up approximately one half of the total size of the three classes (60 children) in their grade. All ELLs’ age of first exposure to English occurred between the age of 2;6 and 3;1. All children were attending the same multicultural, English-medium, mid-low SES school in South Africa. English-medium schools are those which use English as the language of teaching for the entirety of the school day, even though most of the children do not speak English as their L1. The child participants were from three classes which often mixed for activities throughout the school day and the children all played together at recess.

Table 1.

Number, sex and mean age across the year.

	N	Sex		Age
		(M, F)	Time	Mean	Range	SD
ELL	27	12, 15	T1	5;6	5;1–6;0	3.14
			T2	5;10	5;5–6;4	3.06
			T3	6;3	5;9–6;8	3

All the participants were from mid-low SES households, which was determined by the primary caregiver’s highest level of education, measured on a scale from 1 to 8: early childhood education (1), primary education (2), lower secondary education (3), completed secondary education (4), post-secondary short study (5), tertiary education diploma (6), tertiary education degree (7), and Master’s level (8). The primary caregiver’s education was chosen as the main measure of SES, as previous research has shown that it is one of the strongest predictors of income and occupation (e.g. Hoff et al., 2012; Sirin, 2005), compared to other SES metrics it has the highest predictive power for cognitive performance (e.g. Noble et al., 2006), and it is also the most widely used indicator of SES in child language research (e.g. Ensminger & Fotherhill, 2003). Moreover, because the school itself is classified as mid-low SES (based on geographical area, cost of school fees and the provision of free lunches), the primary caregiver’s education serves as an additional more precise metric of SES. Table 2 gives the participants’ SES and age of first exposure to English.

Table 2.

Mean, range and standard deviation of age of first exposure and SES across the year.

	Mean	Range	SD
SES	3	2–4	0.77
Age of first exposure	34	30–37	2.52

For the overall aim of the study to be reached, it was necessary that all children were to be ELLs – in other words, children who have an L1 other than English and who are still in the process of learning English. Children were considered to be ELLs if they are sequential bilinguals and if both of their parents have any language(s) other than English as their L1. It was not the focus of this study to examine the effects of any particular language on English language learning; therefore, the decision was made to include all combinations of mother tongue languages in the sample. Table 3 shows the different primary languages that are spoken by the participants. Some languages represented in this sample are primarily spoken outside of Southern Africa, which indicates the heterogeneity of language backgrounds of the participants. All children were nonetheless born and raised in South Africa.

Table 3.

Primary languages and the number of children speaking them.

Primary languages of ELLs	Number of speakers
isiXhosa	11
Shona	6
French	3
Swahili	2
isiZulu	1
Sesotho	1
Oshiwambo	1
Igbo	1
Cameroonian Pidgin English	1

The age of first exposure to English had to be no earlier than in the third year of life; in other words, children who attended a bilingual or English-only day-care in the year that they turned three qualified for the study. Children who had consistent exposure to English at home for at least one year prior to the study were excluded. Consistent exposure was defined as hearing English daily in the home. Hearing English from the radio or television did not qualify as consistent exposure in this study because it has been found that these sources of input are not supportive of language development at younger ages, such as those in the current sample (Patterson, 2002). Answers about parents’ and children’s language repertoires were obtained from a questionnaire completed by the parents, namely the Bilingual Language Exposure Calculator (BiLEC) (Unsworth, 2013), which informed the author’s ultimate decision about participant suitability. All parents completed the questionnaire in interview format with the researcher. Considering the self-reported and author-observed poor level of parental English (which pertained to the quality of English input) and their self-reported propensity for code-switching (pertaining to the quantity of input), it is unlikely that the English language input measured by the BiLEC was beneficial for the children’s English development (e.g. Golberg et al., 2008; Paradis, 2011; Paradis & Kirova, 2014) and was therefore unable to be included in the analysis. Therefore, the only information that was used in the study from the BiLEC and parent interview was the child’s age of first exposure to English, languages spoken at home and primary caregiver’s education level.

Procedure and materials

The same tasks assessing English proficiency and working memory aptitude were presented to the child at each of his/her three testing sessions, in randomised order so as to prevent any unforeseen priming effects. These tasks included a vocabulary task, a language assessment battery, two phonological working memory tasks (nonword repetition) and a complex non-verbal visuospatial working memory task (odd-one-out). Children were tested three times during this longitudinal study: at the beginning of the school year (T1), in the middle of the year (T2) and at the end of the year (T3), with four-month gaps between testing sessions. These four months comprised approximately 250 hours of structured English-medium input. Testing was conducted individually in a quiet room at their school, and it took an average of 60 minutes to administer all the tests to the child. Children were given breaks when they showed signs of fatigue and/or if they requested a break, sessions could be broken up. However, all tests were completed within one school day.

The longitudinal nature of the study necessitated the use of language and working memory tests which were able to be used three times in one year without eliciting practice effects. This consideration was addressed by using the PPVT-4 (Dunn & Dunn, 2007) and the odd-one-out (Vugs et al., 2014), which all increase with difficulty as the participant responds correctly. The participant will therefore see new items when s/he completes the test at the next testing session. This was not the case for the DELV-CR (Seymour et al., 2003) or the CL-NWR and E-NWR (Chiat et al., 2012). However, only modest practice effects were found in previous studies with an interval of 19 days between test administrations (Seymour et al., 2005). Although the CL-NWR and E-NWR do not have such data yet, another nonword repetition task showed some practice effects for a language-impaired group but not for a typically developing group, with only an interval of one day between testing sessions (Gray, 2003). Therefore, the time between testing sessions for the current study was maintained at a set duration, with four months between each testing session. This was a sufficient length of time to minimise practice effects for the aforementioned tests.

Language measures

The Peabody Picture Vocabulary Test (PPVT-4) (Dunn & Dunn, 2007) was used to measure receptive vocabulary, and the Developmental Evaluation of Language Variation – Criterion Referenced Edition (DELV-CR) (Seymour et al., 2003) was used as the language assessment tool. In the DELV-CR, comprehension and production skills are assessed across the linguistic domains of syntax, semantics and pragmatics. The syntax domain has a maximum score of 43 and consists of three target areas: Wh-questions, passives and articles. The semantics domain has a maximum score of 45 and consists of verb contrast, preposition contrast, quantifiers, and fast mapping of verbs, while the pragmatics domain consists of communicative role-taking, short narratives and question-asking and has a maximum score of 25. Children are shown pictures which they need to point at, to answer questions about or to describe, depending on the task. The greatest advantage of the DELV is that it targets the linguistic structures that are most commonly found across the different varieties of English, which allows the tool to be dialect neutral (Seymour et al., 2003).

The PPVT-4 consists of colour pictures that are presented to the participant four pictures at a time. The examiner says a word aloud and the participant then points to the picture that matches the word. There are three distractor pictures and one target picture. The test consists of 228 items, divided into 19 sets of 12 items, where sets increase with difficulty. Items include nouns, verbs and adjectives. The participant continues with the test until eight or more mistakes are made in one set, at which time the testing is discontinued.

Neither the PPVT-4 nor the DELV-CR has been standardised for use and normed with English L2 South African children. For this reason, raw scores instead of standard scores are considered, in order for the child to be compared to himself/herself across three data collection points and not to the norming sample of the PPVT-4 or the DELV-CR.

Working memory measures

The two tasks used to assess phonological working memory were the language specific (English) nonword repetition (E-NWR) (Chiat, 2015) and the cross-linguistic nonword repetition (CL-NWR) (Chiat, 2015) tasks. The E-NWR uses nonwords which are based on English phonotactics (e.g. flɑnəmuzə) whereas the CL-NWR consists of items which are based on the most commonly observed phonotactics across languages (e.g. lumigɑ). Both tests of nonword repetition include two practice items, which are not scored, followed by 16 items for the CL-NWR and 24 items for the E-NWR. Both tasks were presented on a laptop through the Bead Game (Polišenská & Kapalkova, 2014), which is a game where the child hears a target nonword being played, after which s/he is expected to repeat the nonword. The child is rewarded with a colourful bead that appears on the screen after every repetition. Each nonword corresponds to a bead, and by the end of each task, all the beads together formed a necklace. Scoring was based on whole item correctness where a score of one was awarded for a correct repetition and zero was given when the repetition was inaccurate. Self-corrections were allowed and a correct self-correction was awarded a score of one.

The odd-one-out task (based on Henry, 2001) is a non-verbal visuospatial working memory task that was used to tap into the central executive by the storage and processing of information via the visuospatial sketchpad. This task relies very little on language, which makes it ideal for children with low proficiency. In this task, the participant is asked to look at a grid containing three shapes, which are simple black and white line drawings, such as stars or arrows (see Figure 1 for an example). The participant is required, first, to point to which of the three presented shapes is different, and second, to remember the position of the different shape. If the child correctly remembers the location of the odd shape, one point is awarded. The task starts with Level 1, which only requires the child to remember one odd-one-out item’s position. The test gradually increases in difficulty until Level 6, where six positions need to be recalled. There are three trials for every level and if the child fails to get two out of the three trials correct, the task is discontinued.

Figure 1.

Example of an item at level 1 of the odd-one-out task (Henry, 2001).

Results

The descriptive statistics for all measures at all three testing sessions (T1, T2 and T3) are reported in Table 4, in the form of raw scores. A cursory look at the scores shows a marked improvement across the three testing sessions. Moreover, there were no floor or ceiling effects recorded for any of the measurements, which illustrates the suitability of these tests to these participants.

Table 4.

Means and standard deviations for the tasks across T1, T2 and T3 as well as the maximum obtainable score of each test.

Testing session	Working memory measures	Mean	SD	Maximum test score
T1	Odd-one-out	4.81	1.08	24
T1	CL-NWR	10.26	2.75	16
T1	E-NWR	13.7	3.91	24
T2	Odd-one-out	5.41	1.72	24
T2	CL-NWR	11.67	2.27	16
T2	E-NWR	16.11	4.61	24
T3	Odd-one-out	7.89	1.28	24
T3	CL-NWR	14.19	1.67	16
T3	E-NWR	19.63	3.80	24
	Language measures
T1	Syntax	14.63	5.53	43
T1	Semantics	17.96	5.37	45
T1	Pragmatics	8.04	4.82	25
T1	PPVT	50	15.92	228
T2	Syntax	21.04	7.07	43
T2	Semantics	21.44	5.41	45
T2	Pragmatics	12.3	4.66	25
T2	PPVT	63.26	16.81	228
T3	Syntax	26.30	5.75	43
T3	Semantics	27.19	4.49	45
T3	Pragmatics	17	4.07	25
T3	PPVT	78.33	14.98	228

Bivariate correlations were calculated in order to address the research question concerning the longitudinal relationships between working memory measures and language measures. All correlations are reported in Table 5. As can be seen from this table, the E-NWR is significantly correlated with syntax, pragmatics and the PPVT vocabulary score across all three testing sessions. As for the odd-one-out task, the only correlation that is constantly significant across the testing year is that with syntax. All correlations are positive, indicating a positive growth relationship.

Table 5.

Correlations between all working memory and language measures across T1, T2 and T3.

Tasks		Odd-one-out	CL-NWR	E-NWR	Syntax	Pragmatics	PPVT	Semantics
Visuospatial working memory	Odd-one-out (T1)	–	.290	.380	.486*	.350	.274	.459*
	Odd-one-out (T2)	–	.431*	.388*	.522**	.403*	.311	.340
	Odd-one-out (T3)	–	−.134	.323	.469*	.605**	.662**	.325
Phonological working memory	CL-NWR (T1)	.290	–	.622**	.474*	.307	.155	.162
	CL-NWR (T2)	.431*	–	.739**	.580**	.556**	.457*	.432*
	CL-NWR (T3)	−.134	–	.655**	.147	.369	.065	.412*
	E-NWR (T1)	.380	.622**	–	.671**	.505**	.414*	.258
	E-NWR (T2)	.388*	.739**	–	.481*	.432*	.555**	.347
	E-NWR (T3)	.323	.655**	–	.538**	.773**	.420*	.520**
Language tasks	Syntax (T1)	.486*	.474*	.671**	–	.584**	.609**	.500**
	Syntax (T2)	.522**	.580**	.481*	–	.723**	.712**	.752**
	Syntax (T3)	.469*	.147	.538**	–	.610**	.477*	.492**
	Pragmatics (T1)	.350	.307	.505**	.584**	–	.639**	.637**
	Pragmatics (T2)	.403*	.556**	.432*	.723**	–	.698**	.751**
	Pragmatics (T3)	.605**	.369	.773**	.610**	–	.522**	.440*
	PPVT (T1)	.274	.155	.414*	.609**	.639**	–	.565**
	PPVT (T2)	.311	.457*	.555**	.712**	.698**	–	.732**
	PPVT (T3)	.662**	.065	.420*	.477*	.522**	–	.521**
	Semantics (T1)	.459*	.162	.258	.500**	.637**	.565**	–
	Semantics (T2)	.340	.432*	.347	.752**	.751**	.732**	–
	Semantics (T3)	.325	.412*	.520**	.492**	.440*	.521**	–

Correlation is significant at the 0.05 level (2-tailed).

Correlation is significant at the 0.01 level (2-tailed).

On account of the correlations between the working memory tasks, particularly between the E-NWR and the CL-NWR, which was as high as (r = 0.739, p < .001), multicollinearity between predictors was checked. It is generally accepted that if the variance inflation factor (VIF) is between 2 and 3, there may be cause for concern (Field et al., 2012, p. 276). There was evidence of collinearity between the E-NWR and the CL-NWR predictors, but not for the odd-one-out (VIF = 2.597 for E-NWR; VIF = 2.469 for CL-NWR; VIF = 1.554 for odd-one-out). Due to the collinearity between E-NWR and CL-NWR, it was decided to only use one of the nonword repetition tasks in the analysis. The decision was made to retain E-NWR. The evidence of collinearity indicates that the wordlikeness problem is not evident in this dataset and therefore retaining a nonword repetition measure that is alike to English phonotactics is more in line with previous literature (e.g. Gathercole, 2006).

Analysis was then conducted to address the first research question, which is concerned with the relationship between the working memory measures and language outcomes longitudinally. A mixed effects model was used, and the analysis was done using the R-software version 3.5.1 (R Core Team, 2018), by means of the nlme package (Pinheiro et al., 2018: version 3.1). A model was fit for each language measure with random effects for subject, and fixed effects for age of first exposure to English, Sex and SES. Odd-one-out, Time and E-NWR were entered as main effects, with two-way interaction terms for Time and Odd-one-out, and Time and E-NWR. Reduced models were run along with the full model in order to compare the Akaike’s information criterion (AIC)¹ values, which returned the lowest value for the full model (refer to Appendix A for the likelihood ratio comparisons of the full and reduced models). The above described full model was therefore kept for the analysis.

All results from the four models can be found in Appendix B, as only the significant findings are reported in the text. The results of the models showed that E-NWR was significantly related to scores on Syntax (t(46) = 3.02, p = 0.0041) as well as scores on Pragmatics (t(46) = 3.51, p = 0.0010). The model containing the Semantics scores revealed a significant positive relation between T3 and Semantics (t(46) = 2.49, p = 0.0164) as well as a significant interaction between Odd-one-out and T3 (t(46) = −2.44, p = 0.0187). While finally, no factor was significantly related to Vocabulary scores.

In order to address research question two, which focuses on determining whether working memory measures at T1 and T2 can predict performance on language measures at T3, a multiple linear regression model was fit using the lm function in the R-software, version 3.5.1 (R Core Team, 2018). The two working memory measures (E-NWR and Odd-one-out) were used as dependent variables in separate models. The predictors used in the models were the age of first exposure to English and the language measures: Syntax, Semantics, Pragmatics and Vocabulary. Participant was entered as a random effect to account for the inevitable non-independence of observations, due to the longitudinal nature of the study. SES and Sex were also entered into the models but were found to not be significant, and also contributed to a higher Akaike’s information criterion (AIC). Therefore, these two predictors were left out of further analyses.

The model was fit, first, with the T1 working memory scores and the T3 language measures, and, secondly, with the T2 working memory scores and the T3 language measures. The Odd-one-out scores failed to predict any language outcomes at either T1 or T2; all results of these analyses are to be found in Table 6. The E-NWR scores at T1 did however predict outcomes on Semantics at T3 (F(5, 21) = 3.18, p = 0.04).

Table 6.

Results of regression analysis for predicting future language outcomes.

	B	SE	t	P	R ²
T1 predict T3
Odd-one-out _T1					0.15
Syntax _T3	0.02	0.05	0.37	0.72
Semantics _T3	0.05	0.06	0.88	0.39
Pragmatics _T3	0.03	0.07	0.42	0.68
Vocabulary _T3	0.00	0.02	0.13	0.90
AFE	0.00	0.09	0.02	0.98
E-NWR _T1					0.43
Syntax _T3	0.00	0.15	0.03	0.97
Semantics _T3	0.4	0.18	2.18	0.04 *
Pragmatics _T3	0.22	0.21	1.03	0.32
Vocabulary _T3	0.02	0.05	0.3	0.77
AFE	0.48	0.26	1.86	0.08
T2 predict T3
Odd-one-out _T2					0.28
Syntax _T3	0.12	0.07	1.56	0.13
Semantics _T3	0.08	0.09	0.96	0.35
Pragmatics _T3	0.1	0.10	0.93	0.36
Vocabulary _T3	0.01	0.03	0.4	0.7
AFE	0.01	0.13	0.09	0.93
E-NWR _T2					0.54
Syntax _T3	0.01	0.16	0.06	0.96
Semantics _T3	0.26	0.19	1.37	0.19
Pragmatics _T3	0.65	0.23	2.89	0.01**
Vocabulary _T3	0.00	0.06	0.07	0.95
AFE	0.45	0.28	1.64	0.12

Significance at the 0.05 level.

Significance at the 0.01 level.

Subsequent analysis was conducted in order to determine if working memory measures at T2 predict language outcomes at T3. It was found that E-NWR scores at T2 were a significant predictor for pragmatics outcomes at T3 (F(5, 21) = 4.84, p = 0.01).

Discussion

The aim of this study was twofold: first, to uncover the relationship between working memory measures and language measures in young ELLs, and, secondly, to determine whether the working memory measures can predict future outcomes on language measures. The overall findings show clear correlations between the working memory measures and the language outcomes, which is in line with previous research with ELLs (e.g. Swanson, 2014; Swanson et al., 2015). The mixed effects analysis indicates an interaction between some working memory measures over time, while the regression analysis found that some working memory measures could predict later language outcomes.

In order to answer the first research question, a correlation analysis was performed on all measures across the three testing sessions. All correlations are positive, which entails that all measures are increasing across the year in a positive relationship. This is to be expected as the English language immersion that the first year of school offers will promote language learning. Notably, the odd-one-out is significantly correlated with syntax at every time point throughout the year. The odd-one-out, which tests the central executive, has been found to be correlated with the central executive in previous studies (e.g. Ellis Weismer et al., 1999; Montgomery, 1995). The same consistency of correlations across the testing points is seen between the E-NWR and syntax, and the E-NWR and pragmatics. Syntax has been found to be correlated with nonword repetition in previous studies such as Verhagen et al. (2015) who showed that bilingual 4-year-old children’s L2 grammar was correlated with nonword repetition. The pragmatics section assessed productive language skills such as communicative role-taking, question asking and producing short narratives, which would also have tapped the phonological loop, thus tapping the same process as the nonword repetition task. However, there are apparent significant correlations between all working memory measures and all language measures in varying strengths, and in order to gain a better understanding of these observed relationships, a mixed effects analysis was run.

The mixed effects analysis showed that E-NWR, measuring phonological working memory, is implicated in syntax and pragmatics, while the odd-one-out, measuring the central executive, had an interaction effect with time (T3) on semantics. The involvement of the central executive in the semantics tasks is in line with what has been found previously, namely that complex working memory is implicit in the comprehension and production of complex syntactic structures, which is what the DELV-CR’s semantics section mainly consists of – for example, fast mapping and preposition and verb contrasts (Kidd, 2013). This could indicate that the central executive is relied upon throughout the acquisition of semantics but that it might become more involved after a year of formal schooling and English language exposure. Moreover, E-NWR was found to be significantly involved with syntax. In foreign language learning, nonword repetition, and therefore phonological working memory, has been found to play a role in grammar learning (Verhagen et al., 2015), which is in line with what has been found in this study. The E-NWR task was also found to be significantly involved in the children’s pragmatics. The mixed effects analysis therefore confirms the significant correlations found and the importance that especially phonological working memory plays in language outcomes, also shown in previous studies (e.g. Baddeley et al., 1998; Farnia & Geva, 2011; Gathercole, 2006).

The correlation between vocabulary and E-NWR was stable and significant throughout the year (T1: r = 0.41; T2: r = 0.56; T3: r = 0.42). Past studies have ascertained that children rely on phonological working memory from the ages of 4 to 6 years, but that this diminishes by the age of 8 as their vocabulary grows (Gathercole, 2006). The participants of this study fall in this age range as they are between 5 and 6 years old. Although strong correlations between E-NWR and vocabulary were found throughout the year, the lack of significant results from the mixed effects analysis is an interesting finding, especially considering that vocabulary is one of the most commonly observed outcomes to be associated with phonological working memory (e.g. Baddeley et al., 1998). However, mixed effects analysis on longitudinal data specifies the predictor’s influence on the developmental trajectory (Liu, 2016), which entails that phonological working memory, while correlated with vocabulary measures, does not influence the developmental trajectory. This causality was directly explored by Gathercole et al. (1992), who found that the causal relationship between vocabulary and nonword repetition was weak after 5 years of age, however a correlation remained. This is in line with our findings.

The second research question was posed to determine whether children’s working memory scores at T1 and T2 could have predictive power for language scores at T3. The odd-one-out score failed to predict any of the language outcomes. The results did, however, show that the T1 E-NWR scores could predict the outcomes of the semantics task at T3. The domain of pure semantics and its relationship to working memory is relatively under-researched, especially in terms of its development in ELLs. Considering that the semantics subtest of the DELV-CR that was used for the current study investigates lexical retrieval, fast mapping of verbs and the comprehension of quantifiers, it follows that phonological working memory would be implicit in these specific tasks. The child would have to recognise and remember phonotactic patterns and use them productively, especially in the fast mapping task in which the items contain both inflected existing words and inflected novel words. This seemingly taps the same process that is needed for nonword repetition.

Moreover, E-NWR scores at T2 were a highly significant predictor of pragmatics scores at T3. As mentioned previously, the DELV-CR pragmatics subtest especially focuses on the ability to ask questions and to build narratives. Both these tasks demand a significant amount of working memory as the child hears the relevant information and is then expected to productively use this information. The phonological working memory in this scenario would be highly taxed, especially in the case of ELLs, who have to hold foreign phonological representations in mind while formulating a response in their L2 (e.g. Miyake & Friedman, 1998). These findings are therefore reasonable based on our knowledge of the tasks and the workings of phonological working memory.

Finally, two nonword repetition tasks were included in this study; one which adhered to English phonotactics, and one which did not adhere to only a single language’s phonotactics but which rather strived to be quasi-universal. The inclusion of both tests was in order to determine whether there was a wordlikeness effect on nonword repetition performance. However, the two nonword repetition tasks were very highly correlated in the current study. These results were unlike those found by Gathercole (1995), who found that performance on wordlike nonwords was correlated with vocabulary measures and low wordlike nonwords were correlated with a digit span task. This is commonly attributed to the amount of familiarity the participant has with the phonological structures of the nonword items (e.g. Gathercole, 1995). Children included in this study were all born in South Africa, where the lingua franca is English in most environments. Therefore, the children would have been exposed to spoken English in public spaces and, through this, may have become familiarised with English phonotactics.

One can conclude from the analyses that were conducted, as well as the correlations that were calculated, that both phonological working memory and the central executive are implicated, to some extent, in the acquisition of syntax, semantics, pragmatics and vocabulary at different points throughout the first year of formal schooling. It is therefore suggested by the current longitudinal study that working memory and English language acquisition are related processes which interact throughout the language acquisition process. Due to the participating children having a limited knowledge of English at the time of the study, parsing complex syntactic structures, productively formulating syntactic structures and comprehending the language tasks are taxing for not only the central executive system, which would be responsible for comprehending the input, but also the phonological working memory system, which would be responsible for storing the decoded input (e.g. Andersson, 2010; Miyake & Friedman, 1998).

The current study’s main limitation is that the sample size is relatively small. Although it is not uncommon for studies focusing on samples other than monolingual mid-SES and typically developing children to have slightly smaller sample sizes (see for example Archibald & Gathercole, 2006; Chiat & Polišenská, 2016), bigger sample sizes are essential for drawing conclusions which can be generalised to the wider population. The current study attempted to keep the sample as homogenous as possible to reduce confounding factors: participants were from the same neighbourhood, had comparable SES and attended the same school (hence diminishing the variability of different school environments, teachers, teaching methods/strategies, peer contact, etc.). The inclusion of additional children from other schools would have created a more heterogenous sample which may have contributed to unexplained variability in English language learning and working memory development. The participants in this study represented one half of their school grade; therefore, the results are representative of that school environment and can be treated as a starting point which can be explored by future research using larger sample sizes.

Despite the small sample size, the current study demonstrates that language acquisition in ELLs is not a stand-alone process and that working memory measures could be a predictor of language outcomes, which can provide important information about multilingual language development (Chiat & Roy, 2008; Paradis, 2010; Sandgren & Holmström, 2015). This can be particularly useful in furnishing educators and clinicians with important information about whether precautionary language intervention is necessary for an ELL, as a low score on working memory tests combined with low scores on language measures could be an indicator that a child needs extra support so that, by implication, their future academic success is supported. Importantly, results from working memory measures can aid in the decision regarding the need for intervention, at a time when the ELL has had very little exposure to English and only has low English proficiency levels.

Footnotes

Appendix

Appendix B

	Coefficient	SE	t	p
Syntax
Intercept	10.75	12.47	0.86	0.3932
T2	6.96	4.81	1.45	0.1552
T3	10.85	6.59	1.65	0.1066
Odd-one-out	0.70	0.85	0.83	0.4124
E-NWR	0.79	0.26	3.02	0.0041**
AFE	−0.23	0.34	−0.69	0.5000
SES	−0.53	1.07	−0.49	0.6254
SexM	−1.43	1.77	−0.81	0.4283
T2:Odd-one-out	0.42	0.91	0.46	0.6488
T3:Odd-one-out	−0.31	1.02	−0.31	0.7593
T2:NWRLS	−0.32	0.27	−1.19	0.2392
T3:NWRLS	−0.18	0.29	−0.62	0.5362
Semantics
Intercept	17.63	13.32	1.32	0.1921
T2	3.82	3.92	0.98	0.3346
T3	13.51	5.42	2.49	0.0164**
Odd-one-out	1.26	0.70	1.78	0.0810
E-NWR	−0.06	0.23	−0.27	0.7841
AFE	−0.19	0.37	−0.53	0.6015
SexM	−0.42	1.90	−0.22	0.8279
SES	0.59	1.15	0.52	0.6103
T2:Odd-one-out	−1.03	0.75	−1.38	0.1751
T3:Odd-one-out	−2.05	0.84	−2.44	0.0187**
T2:NWRLS	0.29	0.22	1.33	0.1913
T3:NWRLS	0.43	0.23	1.83	0.0743
Pragmatics
Intercept	8.51	8.68	0.98	0.3319
T2	6.09	3.45	1.77	0.0839
T3	1.90	4.72	0.40	0.6887
Odd-one-out	0.58	0.60	0.96	0.3401
E-NWR	0.65	0.18	3.51	0.0010**
AFE	−0.28	0.23	−1.18	0.2516
SexM	−1.79	1.23	−1.46	0.1591
SES	−0.59	0.74	−0.80	0.4294
T2:Odd-one-out	−0.07	0.65	−0.11	0.9158
T3:Odd-one-out	−0.22	0.73	−0.30	0.7656
T2:NWRLS	−0.21	0.19	−1.09	0.2796
T3:NWRLS	0.16	0.21	0.78	0.4387
Vocabulary
Intercept	48.92	41.37	1.18	0.2431
T2	8.16	8.61	0.95	0.3483
T3	20.41	12.04	1.69	0.0969
Odd-one-out	2.08	1.57	1.32	0.1933
E-NWR	0.41	0.56	0.73	0.4701
AFE	−0.34	1.15	−0.30	0.7685
SexM	1.29	5.90	0.22	0.8283
SES	−1.03	3.61	−0.28	0.7783
T2:Odd-one-out	−1.31	1.64	−0.79	0.4307
T3:Odd-one-out	0.29	1.86	0.16	0.8757
T2:NWRLS	0.62	0.48	1.30	0.2004
T3:NWRLS	−0.16	0.51	−0.32	0.7521

Significance at the 0.05 level.

Significance at the 0.01 level.

Acknowledgements

We acknowledge the financial assistance of the National Research Foundation (NRF) towards this research. The opinions expressed and conclusions arrived at are those of the authors, and are not necessarily to be attributed to the NRF.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Michelle J White

Notes

Author biography

Michelle J White is a postdoctoral fellow at the University of Cape Town.

References

Akaike

(1981). This week’s citation classic. Current Contents Engineering, Technology, and Applied Sciences, 12, 42.

Alloway

T. P.

Gathercole

S. E.

Pickering

S. J.

(2006). Verbal and visuospatial short term and working memory in children: Are they separable? Child Development, 77, 1698–1716.

Andersson

(2010). The contribution of working memory capacity to foreign language comprehension in children. Memory, 18, 458–472. https://doi.org/10.1080/09658211003762084

Archibald

L. M. D.

Gathercole

S. E.

(2006). Short-term and working memory in specific language impairment. International Journal of Language & Communication Disorders, 41, 675–693.

Baddeley

(2000). The episodic buffer: A new component of working memory? Trends in Cognitive Science, 4, 417–423.

Baddeley

(2003). Working memory: Looking back and looking forward. Nature Reviews Neuroscience, 4, 829–839.

Baddeley

A. D.

Gathercole

S. E.

Papagno

(1998). The phonological loop as a language learning device. Psychological Review, 105, 158–173.

Baddeley

A. D.

Hitch

(1974). Working memory. Psychology Learning & Motivation, 8, 47–89.

Bayliss

D. M.

Jarrold

Baddeley

A. D.

Gunn

D. M.

Leigh

(2005). Mapping the developmental constraints on working memory span performance. Developmental Psychology, 41, 579.

10.

Cain

K. E.

Bryant

P. E.

Oakhill

(2004). Children’s reading comprehension ability: Concurrent prediction by working memory, verbal ability, and component skills. Journal of Educational Psychology, 96, 31–42. https://doi.org/10.1037/0022-0663.96.1.31

11.

Cheung

(1996). Nonword span as a unique predictor of second-language vocabulary learning. Developmental Psychology, 32, 867–873.

12.

Chiat

(2015). Nonword repetition. In Armon-Lotem

de Jong

Meir

(Eds.), Methods for assessing multilingual children: Disentangling bilingualism from Language Impairment. Multilingual Matters.

13.

Chiat

Polišenská

(2016). A framework for crosslinguistic nonword repetition tests: Effects of bilingualism and socioeconomic status on children’s performance. Journal of Speech, Language, and Hearing Research, 59, 1179–1189.

14.

Chiat

Polišenská

Szewczyk

(2012). Crosslinguistic Nonword Repetition Tasks: British English version. A part of LITMUS COST IS0804 Battery. Unpublished material.

15.

Chiat

Roy

(2008). Early phonological and sociocognitive skills as predictors of later language and social communication outcomes. Journal of Child Psychology and Psychiatry, 49, 635–645.

16.

Daneman

Carpenter

P. A.

(1980). Individual differences in working memory and reading. Journal of Verbal Learning and Verbal Behavior, 19, 450–466.

17.

Dunn

L. M.

Dunn

D. M.

(2007). Peabody picture vocabulary test (4th ed.). Pearson.

18.

Edwards

Beckman

M. E.

Munson

(2004). The interaction between vocabulary size and phonotactics probability effects on children’s production accuracy and fluency in nonword repetition. Journal of Speech, Language, and Hearing Research, 47, 421–436.

19.

Ellis

N. C.

Sinclair

(1996). Working memory in the acquisition of vocabulary and syntax. Quarterly Journal of Experimental Psychology, 49, 234–250.

20.

Ellis Weismer

Evans

Hesketh

. (1999). An examination of verbal working memory capacity in children with specific language impairment. Journal of Speech, Language, and Hearing Research, 42, 1249–1260.

21.

Engel de Abreu

P. M. J.

Gathercole

S. E.

(2012). Executive and phonological processes in second-language acquisition. Journal of Educational Psychology, 104, 974–986.

22.

Engel de Abreu

P. M. J.

Gathercole

S. E.

Martin

(2011). Disentangling the relationship between working memory and language: The roles of short-term storage and cognitive control. Learning and Individual Difference, 21, 569–574. https://doi.org/10.1016/j.lindif.2011.06.002

23.

Engle

R. W.

Tuholski

S. W.

Laughlin

J. E.

Conway

A. R.

1999 Working memory, short-term memory, and general fluid intelligence: A latent variable approach. Journal of Experimental Psychology: General, 128, 309–331

24.

Ensminger

M. E.

Fothergill

(2003). A decade of measuring SES: What it tells us and where to go from here. In Bornstein

M. H.

Bradley

R. H.

(Eds.), Socioeconomic status, parenting and child development (pp. 13–28). Erlbaum.

25.

Farnia

Geva

(2011). Cognitive correlates of vocabulary growth in English language learners. Applied Psycholinguistics, 32, 711–738.

26.

Field

A. P.

Miles

Field

(2012). Discovering statistics using R. SAGE.

27.

French

L. M.

O’Brien

(2008). Phonological memory and children’s second language grammar learning. Applied Psycholinguistics, 29, 463–487.

28.

Gangopadhyay

Davidson

Ellis

W. S.

Kaushanskaya

(2015). The role of nonverbal working memory in morphosyntactic processing by school-aged monolingual and bilingual children. Journal of experimental child psychology, 142, 171–194. https://doi.org/10.1016/j.jecp.2015.09.025

29.

Gathercole

S. E.

(1995). Is nonword repetition a test of phonological memory or long-term knowledge? It all depends on the nonwords. Memory & Cognition, 23, 83–94.

30.

Gathercole

S. E.

(2006). Nonword repetition and word learning: The nature of the relationship. Applied Psycholinguistics, 27, 513–543.

31.

Gathercole

S. E

Baddeley

A. D.

(1989). Evaluation of the role of phonological STM in the development of vocabulary in children: A longitudinal study. Journal of Memory and Language, 28, 200–213.

32.

Gathercole

S. E.

Willis

Emslie

Baddeley

A. D.

(1992). Phonological memory and vocabulary development during the early school years: A longitudinal study. Developmental Psychology, 28, 887–898.

33.

Geva

Ryan

E. B.

(1993). Linguistic and cognitive correlates of academic skills in first and second language. Language Learning, 43, 5–42.

34.

Golberg

Paradis

Crago

(2008). Lexical acquisition over time in minority L1 children learning English as a L2. Applied Psycholinguistics, 29, 41–65.

35.

Gorman

B. K.

(2012). Relationships between vocabulary size, working memory, and phonological awareness in Spanish speaking English language learners. American Journal of Speech-Language Pathology, 21, 109–123.

36.

Gray

(2003). Diagnostic accuracy and test-retest reliability of nonword repetition and digit span tasks administered to preschool children with specific language impairment. Journal of Communication Disorders, 36, 129–151.

37.

Henry

(2011). The development of working memory in children. SAGE.

38.

Henry

L. A.

(2001). How does the severity of a learning disability affect working memory performance? Memory, 9, 233–247.

39.

Henry

L. A.

Botting

(2016). Working memory and developmental language impairments. Child Language Teaching and Therapy, 33, 19–32. https://doi.org/10.1177/0265659016655378

40.

Hoff

Laursen

Bridges

(2012). Measurement and model building in studying the influence of socioeconomic status on child development. In Lewis

Mayes

(Eds.), The Cambridge handbook of environment in human development (pp. 590–606). Cambridge University Press.

41.

Kidd

(2013). The role of verbal working memory in children’s sentence comprehension: A critical review. Topics in Language Disorders, 33, 208–223.

42.

Kieffer

M. J.

(2011). Converging trajectories: Reading growth in language minority learners and their classmates, kindergarten to grade 8. American Educational Research Journal, 48, 1187–1225. https://doi.org/10.3102/0002831211419490

43.

Kohnert

(2010). Bilingual children with primary language impairment: Issues, evidence and implications for clinical actions. Journal of Communication Disorders, 43, 456–473.

44.

Liu

(2016). Latent growth, latent growth mixture, and group-based models. In Methods and applications of longitudinal data analysis. Higher Education Press, Elsevier Inc.

45.

MacDonald

M. C.

Almor

Henderson

V. W.

Kempler

Andersen

E. S.

(2001). Assessing working memory and language comprehension in Alzheimer’s disease. Brain and Language, 78, 17–42.

46.

MacDonald

M. C.

Christiansen

M. H.

(2002). Reassessing working memory: Comment on Just and Carpenter (1992) and Waters and Caplan (1996). Psychological Review, 109, 35–54.

47.

Mancilla-Martinez

Lesaux

N. K.

(2010). Predictors of reading comprehension for struggling readers: The case of Spanish-speaking language minority learners. Journal of Educational Psychology, 102, 701–711. https://doi.org/10.1037/a0019135

48.

Masoura

E. V.

Gathercole

S. E.

(2005). Contrasting contributions of phonological short-term memory and long-term knowledge to vocabulary learning in a foreign language. Memory, 13, 422–429.

49.

McDonald

J. L.

(2008). Grammaticality judgments in children: The role of age, working memory, and phonological ability. Journal of Child Language, 35, 247–268.

50.

Miyake

Friedman

N. P.

(1998). Individual differences in second language proficiency: Working memory as language aptitude. In Healy

A. F.

Bourne

L. E. J.

(Eds.), Foreign language learning: Psycholinguistic studies in training and retention (pp. 339–364). Psychology Press.

51.

Miyake

Shah

(Eds.). (1999). Models of working memory: Mechanisms of active maintenance and executive control. Cambridge University Press.

52.

Montgomery

(1995). Sentence comprehension in children with specific language impairment: The role of phonological working memory. Journal of Speech and Hearing Research, 38, 177–189.

53.

Noble

K. G.

Wolmetz

M. E.

Ochs

L. G.

Farah

M. J.

McCandliss

B. D.

(2006). Socioeconomic status modulates brain-behaviour relationships in reading. Developmental Science, 9, 642–654.

54.

Paradis

(2010). The interface between bilingual development and specific language impairment. Applied Psycholinguistics, 31, 227–252.

55.

Paradis

(2011). Individual differences in child English second language acquisition: Comparing child-internal and child external factors. Linguistic Approaches to Bilingualism, 1, 213–237.

56.

Paradis

(2015). The development of English as a second language with and without specific language impairment: Clinical implications. Journal of Speech, Language, and Hearing Research, 59, 171–182.

57.

Paradis

Kirova

(2014). English second language learners in preschool: Profile effects in their English abilities and the role of home language environment. International Journal of Behavioral Development, 38, 342–349.

58.

Patterson

J. L.

(2002). Relationships of expressive vocabulary to frequency of reading and television experience among bilingual toddlers. Applied Psycholinguistics, 23, 493–508.

59.

Petruccelli

Bavin

E. L.

Bretherton

(2012). Children with specific language impairment and resolved late talkers: Working memory profiles at 5 years. Journal of Speech, Language, and Hearing Research, 55, 1690–1703. https://doi.org/10.1044/1092-4388(2012/11-0288)

60.

Pickering

Gathercole

(2001). Working memory test battery for children (WMTB-C) manual. The Psychological Corporation.

61.

Pinheiro

Bates

DebRoy

Sarkar

D., &

R Core Team (2018). nlme: Linear and nonlinear mixed effects models. R package version 3.1-137. https://CRAN.R-project.org/package=nlme

62.

Polišenská

Kapalkova

(2014). Improving child compliance on a computer administered nonword repetition task. Journal of Speech, Language, and Hearing Research, 57, 1060–1068. https://doi.org/10.1044/1092-4388(2013/13-0014)

63.

R Core Team. (2018). R: A language and environment for statistical computing. R Foundation for Statistical Computing. https://www.R-project.org/

64.

Sandgren

Holmström

(2015). Executive functions in mono- and bilingual children with language impairment: Issues for speech-language pathology. Frontiers in Psychology, 6, 1664–1078.

65.

Service

Kohonen

(1995). Is the relation between phonological memory and foreign language learning accounted for by vocabulary acquisition? Applied Psycholinguistics, 16, 155–172.

66.

Service

Simola

Metsänheimo

Maury

(2002). Bilingual working memory span is affected by language skill. European Journal of Cognitive Psychology, 14, 383–408.

67.

Seymour

H. N.

Roeper

T. W.

de Villiers

(2003). Diagnostic evaluation of language variation, criterion referenced: Examiners manual. Psychological Corporation, A Harcourt Assessment Company.

68.

Seymour

H. N.

Roeper

T. W.

de Villiers

J. G.

de Villiers

P. A.

Pearson

B. Z.

Ciolli

. . .Cooper

(2005). Development and validation of a language test for children speaking non-standard English: A Study of Children Who Speak African American English. (Final Report: 3/1/98 – 8/29/05). Harcourt Assessments, Inc.

69.

Sirin

S. R.

(2005). Socioeconomic status and academic achievement: A meta-analytic review of research. Review of Educational Research, 75, 417–453.

70.

Swanson

H. L.

(2014). Growth in working memory and inhibition predicts literacy in English language learners: A cross-sectional and longitudinal study. Memory, 23, 748–773. https://doi.org/10.1080/09658211.2014.927504

71.

Swanson

H. L.

Beebe-Frankenberger

(2004). The relationship between working memory and mathematical problem solving in children at risk and not a risk for serious math difficulties. Journal of Educational Psychology, 96, 471–491. https://doi.org/10.1037/0022-0663.96.3.471

72.

Swanson

H. L.

Orosco

M. J.

Lussier

C. M.

(2015). Growth in literacy, cognition, and working memory in English language learners. Journal of Experimental Child Psychology, 132, 155–188. https://doi.org/10.1016/j.jecp.2015.01.001

73.

Szewczyk

J. M.

Marecka

Chiat

Wodniecka

(2018). Nonword repetition depends on the frequency of sublexical representations at different grain sizes: Evidence from a multi-factorial analysis. Cognition, 179, 23–36.

74.

Unsworth

(2013). Assessing age of onset effects in (early) child L2 acquisition. Language Acquisition, 20, 74–92. http://doi.org/10.1080/10489223.2013.766739

75.

Verhagen

Leseman

(2016). How do verbal short-term memory and working memory relate to the acquisition of vocabulary and grammar? A comparison between first and second language learners. Journal of Experimental Child Psychology, 141, 65–82.

76.

Verhagen

Messer

M. H.

Leseman

P. P. M.

(2015). Phonological memory and the acquisition of grammar in child L2 learners. Language Learning, 65, 417–448.

77.

Vugs

Hendriks

Cuperus

Verhoeven

(2014). Working memory performance and executive function behaviors in young children with SLI. Research In Developmental Disabilities, 35, 62–74.

78.

Wen

Skehan

(2011). A new perspective on foreign language aptitude research: Building and supporting a case for ‘working memory as language aptitude’. Ilha do Desterro, 60. http://doi.org/10.5007/2175-8026.2011n60p015