Factors Affecting the Speed of Word Retrieval in Children Learning English as a Foreign Language

Ting Zhao and Victoria A. Murphy

Introduction

When a child comes to know a word, he or she must store its semantic, phonological and orthographical representations, as well as associations between these representations. The child's stored knowledge needs to be accessed in the service of language comprehension and production. The speed with which a word is retrieved hinges on and reflects the child's ability to make use of his or her semantic-lexical knowledge. Thus, identifying the possible factors underlying children's lexical accessibility would likely help understand the sources of individual differences observed in language learners' performance.

In the second language (L2) literature, a large body of research has examined the determinants of L2 speaking ability or proficiency (e.g. De Jong et al., 2013; Derwing et al., 2009; Riazantseva, 2001). Most of these studies have been conducted with adolescents and adults, whereas much less attention has been paid to primary-aged children, an L2 population of increasing importance around the world. Against this background, it is necessary to investigate L2, or more precisely English as a foreign language (EFL) children's emerging accessibility to already known words by assessing the relative contribution of several demographic and language proficiency variables (e.g. English vocabulary size, first language [L1] proficiency level and residential areas) to their retrieval speed of foreign language (FL) spoken words.

Two types of oral word production are common to FL children: naming pictures that represent objects and entities, and translating individual words from one language into another. Pictures are thought of as symbols that approximate imagistic representations in the mind, and in this sense how children name pictures could provide insight into how they perform in communication. Bilingual language use also involves an ability to translate. Every bilingual speaker has once engaged in translation, not to mention professional translators and interpreters. Additionally, translation is a pedagogical activity that often occurs in FL classes.

To sum up, the present study identified significant predictors of English lexical accessibility in picture naming and translation among EFL children and advanced theoretical and practical implications for current conceptualisations of EFL learning and teaching within input-limited contexts.

Literature Review

Describing and defining EFL children

In the course of globalisation, English has become an important world language. Many countries in the Asia-Pacific region have made English education compulsory at the primary level over the past decades. For example, in mainland China, English has been nationally recognised as a compulsory subject from the third grade onwards since September 2001 (Ministry of Education of China, 2001). Exact statistics on the number of Chinese EFL children are hard to come by, but the following figures present a rough picture. It was recently estimated that there were 379,253 primary-level Chinese teachers of English and 1,794,614 classes of third to sixth graders across the country (Ministry of Education of China, 2014). Conceivably, there must be a large population of children learning EFL outside English-speaking countries; nevertheless, this FL population has hitherto been woefully underrepresented in the L2 literature.

Bilingual learners can be defined using different criteria (see Li [2000] for a summary), such as language dominance (i.e. balanced vs. unbalanced), the sequence of acquisition (i.e. simultaneous vs. sequential), the onset age of L2 acquisition (i.e. early vs. late) and the preferential domain of language use (receptive vs. productive). According to these criteria, EFL children are a group of sequential bilingual beginners who use the L1 dominantly in almost any setting and learn EFL through formal instruction. This population, when compared to their immersion counterparts, receives much less exposure to the target language, especially outside the classroom.

Process of spoken word production

Language speakers, regardless of age, constantly access their mental lexicons. Drawing on Levelt's (1989) blueprint for the speaker, word production proceeds through conceptualisation, grammatical encoding, articulation and self-monitoring. Research on spoken production has followed two main approaches: the study of speech errors (e.g. Poulisse & Bongaerts, 1994) and the measurement of production latencies (for a review, see Jiang [2012]).

The linguistic function of naming objects and entities develops early in young children (Bates et al., 1979). As a psycholinguistic tool, picture naming has been extensively used to define different stages of speech production and to uncover the mechanisms underlying this process (e.g. Glaser, 1992; Levelt et al., 1991; Potter et al., 1984). Picture naming is typically conducted in such a way that participants are presented with a series of pictures one by one, and as a picture appears on the computer screen, they name it as accurately and rapidly as possible. There are three main stages involved in picture naming: (1) object recognition and concept activation; (2) lexical selection and activation; and (3) response execution and production (Johnson et al., 1996). Additionally, picture naming is a common classroom activity, especially for young children who learn and recycle words with images and other visual aids.

Another way of examining spoken word production is through wordby-word translation. Any individual who has receptive and productive vocabulary knowledge of two languages is able to perform a kind of translation. As an experimental tool, translation is typically conducted in such a way that participants are presented with a series of stimulus words one by one, and as a word appears on the computer screen, they translate it into the target language as accurately and rapidly as possible. There are two directions of translation: forward translation (L1-to-L2) and backward translation (L2-to-L1). Given that the present study specifically addressed L2 lexical accessibility in production, only forward translation was examined. The stages involved in L1-to-L2 translation are similar to those in picture naming, except that translation starts with visual word recognition rather than object recognition (Snodgrass, 1993). Additionally, translation as a pedagogical tool has been used in FL classrooms since the 19th century (Howatt, 1984; Richards & Rodgers, 2001), though it has had its ups and downs. Investigating translation illustrates how such a classroom activity proceeds at the individual level.

Factors affecting spoken word production

A variety of factors are believed to affect the retrieval speed of spoken word production, as reviewed below.

L2 vocabulary knowledge and L2 proficiency

Vocabulary knowledge is a multidimensional construct, which incorporates size (or breath, i.e. the number of words a learner knows), depth (i.e. how well a learner knows individual words) (Nation, 2001; Schmitt & Meara, 1997) and the automaticity with which words are accessed or processed during language use (Meara, 1996). The literature abounds with studies examining the correlation between the first two dimensions. For instance, Vermeer (2001) observed that two measures of size (i.e. receptive vocabulary and description tasks) and a depth measure (i.e. an association task) were strongly correlated among Dutch bilingual kindergarteners. Vermeer interpreted this result from a network perspective, that is, lexical elements in the mind are connected to each other, and conceivably the size and depth dimensions hinge on the same or a similar underlying construct. Moreover, the dimension of depth was found to explain additional variance in Dutch children's reading comprehension ability beyond what was explained by the dimension of size (Schoonen & Verhallen, 1998). The study by Cremer and Schoonen (2013) showed that the speed of access explained more variance in reading comprehension beyond decoding and the availability of semantic knowledge.

Despite this research, there is a paucity of research examining the relationship between instructed L2 children's vocabulary knowledge and their speaking ability, which is understandable given that beginning learners have limited knowledge of productive words and limited practice of productive skills (but see Koizumi [2005] for studies on L2 adolescents). This is also one of the reasons why we investigated the child participants' production of single words instead of running speech. The adult literature, however, has shown that vocabulary knowledge is crucial to spoken fluency (e.g. De Jong et al., 2013; Hilton, 2008). For example, De Jong et al. (2013) examined the relationship between L2 linguistic knowledge and spoken fluency among intermediate or advanced speakers of Dutch, demonstrating a strong correlation between the adults' vocabulary knowledge and their speaking rate (mean syllable times) (r=-0.58).

Vocabulary knowledge has been shown to be a robust indicator of language proficiency (Cummins, 2000). In effect, language proficiency likely affects the ease of access to individual words in spoken production. An obvious example is that L2 speakers have slower speech and articulation rates, longer pause times and shorter runs than L1 speakers (Wiese, 1984). As shown in Riazantseva (2001), university-level Russian speakers with high English proficiency made shorter pauses than those with intermediate English proficiency. Note that child participants are clearly limited in producing spoken or written sentences, in which case the variable of L2 proficiency was defined by the children's receptive knowledge of English.

Despite the common assumption that vocabulary knowledge and language proficiency are predictive of lexical accessibility in spoken production, few studies have empirically addressed this relationship, particularly with instructed FL children.

L1 proficiency and L1 lexical accessibility

The issue of cross-language influence has received much attention in the literature. According to the developmental interdependence hypothesis (DIH; Cummins & Swain, 1986), the development of L1 literacy goes hand in hand with that of L2 literacy. As argued by Cummins and Swain (1986: 87), 'To the extent that instruction in Lx is effective in promoting proficiency in Lx, transfer of this proficiency to Ly will occur provided there is adequate exposure to Ly (either in school or environment) and adequate motivation to learn Ly'. In explaining such a close relationship between L1 and L2 literacy, Geva and Ryan (1993) emphasised the important roles of common cognitive components and processes (e.g. working memory, executive control function and self-regulation) in sustaining the development of both languages.

Empirically, cross-language associations have been demonstrated in examining adolescent and adult EFL learners' reading and writing skills (e.g. Lee & Schallert, 1997; Sasaki & Hirose, 1996; Yamashita, 2002). Most relevant to the present study, Derwing et al. (2009) compared L1 and L2 spoken fluency over two years among Russian- and Ukrainian- and Mandarin-speaking adult immigrants to Canada. The result showed that L2 fluency was significantly and positively correlated with L1 fluency in the early stages of L2 exposure, although the correlation coefficient was higher in the Slavic learners (r=0.62) than in the Mandarin learners (r = 0.53). Spoken fluency partially hinges on the speed of accessing individual words. It is therefore conceivable that L2 lexical accessibility is intertwined with L1 lexical accessibility. Note that the present study included both L1 proficiency and L1 lexical accessibility as predictors for the purpose of cross-validating L1 influence on the processing of L2 words.

Non-verbal ability

As a form of a human's mental abilities, intelligence (in particular, reasoning ability) is fundamental to the development of language proficiency (Hulstijn, 2015). A classic view, represented by Piaget (1954), holds that children's language development is contingent on the knowledge and understanding accumulated through cognitive development. Most of the empirical studies have demonstrated strong correlations between cognitive ability and different linguistic aspects in monolingual and bilingual learners (Andringa et al., 2012; but also see Oller et al., 2001; Saklofske et al., 2000; Sasaki & Hirose, 1996; Verhoeven & Vermeer, 2009; Wechsler, 1999), indicating that performing reasonably well in most situations of language use is intricately associated with a relatively high degree of cognitive ability.

In the literature on L2 children, inconsistent results have been reached with regard to the relationship between cognitive ability and L2 proficiency. Verhoeven and Vermeer (2009) investigated the extent to which L2 overall proficiency was predicted by a wide range of cognitive and social-cultural predictors (e.g. rule discovery, parental education and home language use) among Turkish children (4 years old) of Dutch in the Netherlands. The result showed that cognitive ability was the strongest predictor of L2 proficiency above and beyond other predictors. This result, however, has not been observed in Oller et al.'s (2001) study, in which Oller and colleagues examined 50 primary-aged Spanish-speaking children learning English in a bilingual school in Mexico City. The participants were tested on nonverbal IQ, English vocabulary, articulation and morphology. The children's non-verbal ability did not turn out to be significantly correlated with any measures of L2 proficiency. It should be acknowledged that these two studies are not directly comparable due to different research contexts and target populations. At the moment, it would be difficult to interpret these inconsistent results until more empirical studies are conducted to address the relationship between children's cognitive ability and L2 proficiency.

To sum up, the present study assessed the relative contribution of a range of learner variables to L2 lexical accessibility in spoken production among EFL children, a population of increasing importance but underrepresented in the L2 literature.

The Present Study

Research questions

The present study addressed the following questions:

(1) What is the relationship between a range of learner variables and the retrieval speed of FL words in spoken production?

(2) What variable(s) best predict the retrieval speed of FL words in spoken production?

Population and sample

The target population was primary-aged children learning EFL in public schools in mainland China. With the release of Guidelines for Vigorously Promoting the Teaching of English in Primary Schools (Ministry of Education, 2001), English education was made compulsory from the third grade onwards on a national scale.

The present sample consisted of 39 fifth-grade students (M =10.31 years) who had learned English for two and a half years (on average, 120 minutes per week). The participants were selected using convenience sampling due to practical considerations. They were from two classes at a public primary school located on the urban-rural periphery of a middle-income city. All participants were taught by the same EFL teacher who had about seven years of teaching experience. The sampling procedure is presented in Figure 1.1.

In advance of participating in the reaction time-based experiments, a measure of non-verbal IQ (the Wechsler Abbreviated Scale of Intelligence test [WASI], matrix reasoning) and the Word Associates test (WAT) were administered to candidate participants for screening out those children with limited cognitive and linguistic abilities. Twenty-two children whose WASI or WAT scores were lower than three standard deviations below the mean were excluded from our study. Subsequently, picture naming and L1-to-L2 translation were found to be beyond 12 participants' ability to complete, reducing the sample size down to 39 (19 boys and 20 girls).