View in EDS HTML Full Text PDF Full Text

Language Acquisition in a Multilingual Society: English Vocabulary Norms and Predictors in Singaporean Children

Saved in:
Bibliographic Details
Title: Language Acquisition in a Multilingual Society: English Vocabulary Norms and Predictors in Singaporean Children
Language: English
Authors: Singh, Leher (ORCID 0000-0001-9423-4956), Cheng, QiQi (ORCID 0000-0002-2295-330X), Tan, Seok Hui, Tan, Agnes, Low, Yen Ling
Source: Child Development. Jan-Feb 2022 93(1):288-305.
Availability: Wiley. Available from: John Wiley & Sons, Inc. 111 River Street, Hoboken, NJ 07030. Tel: 800-835-6770; e-mail: cs-journals@wiley.com; Web site: https://www.wiley.com/en-us
Peer Reviewed: Y
Page Count: 18
Publication Date: 2022
Document Type: Journal Articles
Reports - Research
Descriptors: Infants, Child Development, Language Acquisition, Socioeconomic Status, Vocabulary Development, Predictor Variables, Multilingualism, Language Skills, Measures (Individuals), English, Mandarin Chinese, Indonesian Languages, Foreign Countries
Geographic Terms: Singapore
Assessment and Survey Identifiers: MacArthur Communicative Development Inventory
DOI: 10.1111/cdev.13676
ISSN: 0009-3920
Abstract: In this study, infant vocabulary development was tracked in a multilingual society (Singapore) within a socioeconomically diverse sample. The sample comprised 1316 infants from 17.4 to 27.7 months (669 females, 647 males; 88% Chinese race, 4% Malay, 4% Indian, and 0.004% mixed-race [4% declined to provide race information]). Children varied in English language exposure and socioeconomic status. Analyses focused on identifying demographic predictors of English vocabulary size in multilingually exposed infants. Adaptations of the Macarthur-Bates Communicative Development Inventory for English, Mandarin, and Malay are provided as well as English vocabulary norms that account for variation in English exposure. This manuscript reports the first set of English language norms--calibrated to English exposure--for multilingual infants in a non-Western setting.
Abstractor: As Provided
Entry Date: 2022
Accession Number: EJ1327059
Database: ERIC
Full text is not displayed to guests.
FullText Links:
  – Type: pdflink
    Url: https://content.ebscohost.com/cds/retrieve?content=AQICAHj0k_4E0hTGH8RJwT4gCJyBsGNe_WN95AvKlDbXJGqwxwGCtgSRo19jgfSWSqafmFlbAAAA4jCB3wYJKoZIhvcNAQcGoIHRMIHOAgEAMIHIBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDMWcAP2vk_O26PXcJwIBEICBmiK-qvrqcMmsiYgsnC8_J-4UltDH3O6Oxp_83eXZP5JIjMm8V92983nE26124MjoZKhiGT7cSYa35J262XeUuOM16BJL9DX5x4mehfEuHQejEbH3fMEyq11mlv9YNNU1KJ10QjPY4NlmFmXKnv51YeGc2L3FZER-ZfpNVbRxe7DluTaC1ypGn0dnLDb8nrbnCO77sUgIdl7JQNE=
Text:
  Availability: 1
  Value: <anid>AN0154716031;cdv01jan.22;2022Jan19.03:22;v2.2.500</anid> <title id="AN0154716031-1">Language acquisition in a multilingual society: English vocabulary norms and predictors in Singaporean children </title> <p>In this study, infant vocabulary development was tracked in a multilingual society (Singapore) within a socioeconomically diverse sample. The sample comprised 1316 infants from 17.4 to 27.7 months (669 females, 647 males; 88% Chinese race, 4% Malay, 4% Indian, and 0.004% mixed‐race [4% declined to provide race information]). Children varied in English language exposure and socioeconomic status. Analyses focused on identifying demographic predictors of English vocabulary size in multilingually exposed infants. Adaptations of the Macarthur‐Bates Communicative Development Inventory for English, Mandarin, and Malay are provided as well as English vocabulary norms that account for variation in English exposure. This manuscript reports the first set of English language norms—calibrated to English exposure—for multilingual infants in a non‐Western setting.</p> <p></p> <ulist> <item> Abbreviations</item> <p></p> <item> HSD honestly significant different</item> <p></p> <item> MCDI Macarthur‐Bates Communicative Development Inventory</item> <p></p> <item> SCDI Singapore Communicative Development Inventory</item> <p></p> <item> SES socioeconomic status</item> </ulist> <p>Most children around the world are raised with more than one native language. A large proportion of these children are raised in multilingual societies. Defining the language learning trajectories of such children is of major significance, for both basic and applied reasons. In terms of basic research, identifying how multilingual children develop language is crucial to our understanding of normative language development given the high proportion of children raised in multilingual societies. From an applied perspective, identifying clinical risk in multilingual populations depends crucially on accurate estimates of normative language acquisition within multilingual settings. In this study, we report the results of a large‐scale study with over 1300 participants living in Singapore, a quintessentially multilingual society. Vocabulary development was sampled from socioeconomically diverse families, whose children varied in their input languages, both in terms of the specific languages to which they were exposed and in the quantity of exposure to each language. Children's vocabulary size was determined using an adaptation of the Macarthur‐Bates Communicative Development Inventory (MCDI) for Singapore. The MCDI (Fenson et al., 2007) is a widely used measure of parent‐reported vocabulary size in young children below 30 months of age. It has proven to be instrumental in providing a standardized measure of children's vocabulary size prior to the point at which children can participate in direct testing (Law & Roy, 2008). The survey is generally easy for parents to complete and vocabulary size estimates derived from the MCDI have good concurrent and predictive validity (see Feldman et al., 2005). Using an adaptation of the MCDI for multilingual children in Singapore, we pursued two major goals in this study. First, we sought to identify effects of relevant environmental predictors of vocabulary acquisition in a multilingual setting in a large‐scale study of socioeconomically diverse families. Second, we sought to establish norms and expectations for English vocabulary development for multilingually exposed children in Singapore in relation to age, gender, and English language exposure.</p> <hd id="AN0154716031-2">An overview of language use in Singapore</hd> <p>Singapore has four official languages: English, Mandarin Chinese, Malay, and Tamil. The medium of education is English with a mandatory second language as of first grade. The vast majority of children learn Mandarin, Malay, or Tamil as second languages. Cross‐language contact is rampant in Singapore, which can be traced in part to structural factors. For example, the social composition of public housing estates, where approximately 80% of the population lives, is regulated by the government to ensure that different racial and ethnic groups and by extension, different language groups are represented within each housing estate. As a result, children typically have regular contact with languages other than their home languages. As such, while children learn two languages formally, many children may receive additional exposure to more than two languages in their home or school environments. Moreover, Singapore's bilingual policy was implemented relatively recently in 1966. Prior to this, many citizens were not schooled in English and spoke a heritage language other than Mandarin and English (e.g., Hokkien, Teochew, or Cantonese) as a first language. Although no longer widely spoken, these languages remain relevant to child language acquisition as grandparents, often proficient only in these languages, frequently serve as joint caregivers for young children. Therefore, it is not uncommon for young children to receive input in English and one official language (Malay, Mandarin, or Tamil) as well as an additional heritage language (e.g., Hokkien, Teochew, or Cantonese). Therefore, Singaporean children typically encounter two or more languages from a very early age.</p> <hd id="AN0154716031-3">Challenges and complications in establishing multilingual normative trajectories</hd> <p>Characterizing typical language development in multilingual populations is often challenging. Below, we articulate three significant challenges in developing multilingual norms. First, as noted in past research (see Hoff & Core, 2015), there is enormous heterogeneity in multilingual environments. Children within multilingual societies receive input that is structured in very different ways. Some children receive bilingual input from several people, while others receive first or second language input from one person; these differences that can be significant for language uptake (De Houwer, 2007; Place & Hoff, 2011). Others learn multiple languages on account of immigrating to a new country, while others still are raised immersed in multiple languages within the same country (Winsler et al., 2014). Bilingual children also differ in the specific language combinations that they are learning, with some children learning similar languages and others learning more dissimilar languages (Floccia et al., 2018). Finally, even seemingly trivial factors, such as whether bilingual vocabulary size estimates are reported by the child's mother or father, are relevant to estimating vocabulary size (De Houwer, 2019). Each of these sources of heterogeneity in input conditions has been shown to influence parental reports of vocabulary size.</p> <p>This relates to a second challenge, which is that establishing the typical range of bilingual vocabulary remains a "work‐in‐progress" for many multilingual societies. Early vocabulary inventories and their corresponding norms were originally generated for monolingual English‐speaking children belonging to high socioeconomic strata (Fenson et al., 2007). Subsequent adaptations and language norms reveal significant variability in vocabulary size estimates across different societies, both within monolingual and bilingual samples. For example, vocabulary size estimates are higher for American English monolingual children than for British English monolingual children (Hamilton et al., 2000). To account for this, Hamilton et al. (2000) appeal to possible cultural differences in parental expectations of children's language use, differences in childcare situations across the American and British samples which could affect parent reports, and differences in referent and/or word frequency across the two populations. Likewise, vocabulary size estimates also vary considerably across different bilingual populations within the same society (e.g., Floccia et al., 2018).</p> <p>A third factor is the definition of multilingualism. In experimental studies, the field has often subscribed to a ratio mis‐attributed to Pearson et al. (1993) of minimally 25% exposure to a second language and maximally 75% exposure to a first language to qualify as bilingual. Similarly, a commonly adopted definition of monolingualism is a minimum criterion of 90% exposure to a first language. There is little empirical support for a meaningful threshold at 25% exposure for bilingual children. For example, Thordardottir (2011) reported that bilingual children with at least 35% or 40% exposure to a language scored within the monolingual range for receptive vocabulary, although higher thresholds (closer to 80%) were reported for expressive vocabulary. Cattani et al. (2014) demonstrated that bilingual children with at least 60% exposure to a first language did not differ from monolingual peers in first language proficiency. Finally, DeAnda et al. (2016) compared monolingual l6‐month‐old infants with 100% exposure to English to bilingual learners with 80% exposure to English. The two groups differed in vocabulary size estimates, a gap which later narrowed at 22 months (Friend et al., 2017). Therefore, it remains unclear whether language exposure thresholds are useful in defining bilingualism. In recognition of this, many studies have adopted a more liberal criterion for bilingual exposure (e.g., Hoff et al., 2012; Hurtado et al., 2014; Marchman et al., 2004, 2010, 2017; Place & Hoff, 2011) and/or have defined bilingual exposure along a continuum and dispensed with categories.</p> <hd id="AN0154716031-4">Predictors of vocabulary development in bilingual infants</hd> <p>Past research investigating predictors of vocabulary size in children with more than one native language has revealed several factors of influence. First, language input in each language positively predicts vocabulary size in that language (e.g., De Houwer, 2009; Gathercole & Thomas, 2009; Hoff et al., 2012; Legacy et al., 2018; Oller & Eilers, 2002; Pearson et al., 1997; Scheele et al., 2010) and negatively predicts vocabulary size in the other language (Hoff et al., 2012). This finding is not unique to bilingual children: "dose–response" relations between input and uptake are similarly evident in monolingual children (e.g., Hoff & Naigles, 2002; Huttenlocher et al., 1991), a correspondence that appears to be mediated by increased processing efficiency that results from heightened language exposure (Weisleder & Fernald, 2013).</p> <p>Vocabulary size in bilingual children is also strongly influenced by socioeconomic status (SES). This too is not unique to bilingual children. There is widespread evidence that monolingual vocabulary development is influenced by SES (e.g., Fernald et al., 2013; Hoff, 2003), which has been shown to be mediated specifically by maternal input factors (Hoff, 2003). Variation in SES also predicts vocabulary development within bilingual children (DeAnda et al., 2016; Winsler et al., 2014, also see Hoff, 2018).</p> <p>Finally, bilingual vocabulary size estimates differ based on the measure used. Common measures adopted in infant research include single‐language vocabulary size estimates as well as combined estimates that provide a single measure of vocabulary in both languages. Two types of commonly used combined measures are total conceptual vocabulary and total summed vocabulary. Total conceptual vocabulary is the sum of children's individual vocabulary size estimates from which translations are subtracted. In contrast, total summed vocabulary is the sum of children's individual vocabulary size estimates, whereby translated forms are counted as separate items. Summed vocabulary size estimates facilitate comparison with monolingual children in identifying a similar proportion of children at the lowest and highest percentile bands (Core et al., 2013). In contrast, when comparing vocabulary size estimates with monolingual learners, total conceptual vocabulary may over‐identify children at the lowest percentile bands, In older bilingual children, this measure may under‐identify children at the highest percentile bands. While the precise reasons for this are not clear, it is possible that conceptual vocabulary underestimates the number of words by counting concepts rather than words. A measure such as summed vocabulary, that counts words, may be more comparable to single‐language measures used for monolingual children (which also count words).</p> <p>Although our focus is on environmental predictors of vocabulary size, we note that child characteristics also influence bilingual vocabulary. In particular, factors such as gender (Winsler et al., 2014) birth order (Lauro et al., 2020), phonological memory (Lauro et al., 2020; Parra et al., 2011), non‐linguistic working memory (Barbosa et al., 2017), non‐verbal IQ (Lauro et al., 2020), and cognitive control (Nicoladis & Jiang, 2018) predict single‐ and/or dual‐language proficiency in bilingual children. Individual differences in bilingual development are therefore not determined by environmental factors alone, but additionally, by child‐internal factors.</p> <hd id="AN0154716031-5">Previous norms of bilingual language development</hd> <p>A large‐scale adaptation of the MCDI for multilingual populations in the United Kingdom was conducted by Floccia and colleagues to develop norms for children raised with English and one additional language (Floccia et al., 2018). Sampling 374 toddlers from 23 to 25 months, the authors measured children's vocabularies in English using the Oxford Communicative Development Inventory (OCDI) as well as in their heritage language, using heritage language adaptations of each CDI. The children sampled were from high socioeconomic strata, attributed in part to the relatively high SES of bilingual families in the U.K. (Floccia et al., 2018). Their primary analyses consisted of children's acquisition of 30 words that were common across all CDIs administered. The authors reported several significant predictors of infants' vocabulary knowledge. First, the authors reported effects of language exposure: in the community language (English), the amount of direct input (child‐directed speech) and the amount of overheard speech predicted English vocabulary size. Exposure to English was negatively related to heritage language vocabulary scores. Other factors that predicted English proficiency included family income, gender, and linguistic distance. Floccia and reported higher heritage language vocabulary size estimates for closer language pairings than for more distant pairings. Finally, Floccia et al. present norms for 2‐year‐old children learning English and one additional language in the United Kingdom.</p> <hd id="AN0154716031-6">Overview of this study</hd> <p>In this study, we present the SCDI (Singapore Communicative Development Inventories), an adaptation of the MCDI, with accompanying norms for multilingual learners. Our study sampled families across a range of different socioeconomic strata. We sampled infants from two different age groups. We investigated environmental predictors of English vocabulary size. In addition, we present lexical norms for English vocabulary size in relation to age, gender, and proportion of English exposure.</p> <p>Although we collected data from both English and heritage language scales, we chose to focus on English vocabulary size for three reasons. First, English represents the primary community language and <emph>lingua franca</emph> of Singapore. In recent years, the prominence of English has heightened in Singapore. English is also the language in which all public schooling is conducted. A child with a low English vocabulary is likely to struggle with the educational system more so than one who has a relatively high English vocabulary and low heritage language proficiency. For this reason, a tool that measures English proficiency is likely to be more useful to families than one that measures single‐language proficiency in a heritage language. Second, as English is standardly used in educational settings beginning from daycare, the specific lexical items present on an English language CDI are likely to surface with greater consistency across children. Single‐language measures of heritage languages are precariously dependent on a faithful estimate of referents that surface consistently across individual homes. Based on pilot investigations, we believe that there may be greater referential constancy in English than across different heritage languages. Finally, on purely practical grounds, as also noted by Floccia et al. (2018), a community language CDI is potentially available to all multilingual learners of community languages. Heritage language CDIs are only available for a subset of heritage language communities, which may not include newer immigrant communities who do not speak native heritage languages. It can therefore be beneficial to have a single‐language CDI that provides benchmarks for community language proficiency, calibrated to different exposure levels to the community language.</p> <hd id="AN0154716031-7">METHOD</hd> <p></p> <hd id="AN0154716031-8">Participants</hd> <p>Potential participants were identified from a database maintained by the Abbott Research and Development Laboratories, Singapore. The database accounts for approximately 30% of all children born in Singapore within the age range of interest and surveys a range of socioeconomic strata in proportion to the SES distribution of Singapore. A total of 5215 parents or guardians of healthy children between either 18–20 months (±2 weeks) of age or 25–27 months (±2 weeks) of age were invited to participate. The overall response rate was 28.9% (<emph>N</emph> = 1509). Inclusion criteria were that infants had no known developmental or medical concerns, were full‐term births, and had some exposure to English. From the initial set of respondents, data from 72 infants were excluded due to prematurity. Data from an additional 115 infants were excluded as they reportedly had no exposure to English. Data from an additional six participants were excluded due to medical or health concerns reported by parents. Data were collected from March 2011 to May 2011.</p> <p>A total of 1316 CDI responses were collected from parents of children from two age groups. These responses were divided into a "Younger" group comprising children between the ages of 17 and 21 months old (<emph>n</emph> = 746, <emph>M</emph> = 19.1, <emph>SD</emph> = 0.86; 369 boys) and an "Older" group comprising children aged between 24 and 28 months old (<emph>n</emph> = 570, <emph>M</emph> = 26.0, <emph>SD</emph> = 0.88; 278 boys). For all participants, it was requested that the person who spent the most time with the child respond to the questionnaire, but input could be sought from other caregivers. For all participants, the mother was the primary respondent and also spent the most time with the child.</p> <hd id="AN0154716031-9">Adapting the MCDI to Singapore</hd> <p>The MCDI (Fenson et al., 2007), CDIs developed for other varieties of English (Hamilton et al., 2000), for Malay (Thow, 2002), and for Mandarin (Hao et al., 2008; Tardif et al., 2009) as well as other cross‐language adaptations (e.g., Jackson‐Maldonado et al., 1993) were consulted prior to adapting the CDI to Singapore. In addition, the cross‐language adaptations were discussed with developers of previous adaptations in Mandarin (Tardif et al., 2009) and Malay (Thow, 2002). Adaptations were registered with the MCDI Board.</p> <hd id="AN0154716031-10">Adaptation of the MCDI to the Singapore English CDI (SCDI—English)</hd> <p>The process of adapting the MCDI to the Singaporean context was conducted by two authors (LS and TSH). Individual items on the MCDI were discussed with native speakers of Singapore English, with heritage language speakers of Malay and Mandarin, and with parents of bilingual infants in Singapore. Each word on the MCDI was discussed and critically reviewed for inclusion in the SCDI. Initial adaptations of the CDI were piloted with a sample of Singaporean parents (see Tan, 2010) which led to the forms being adapted for use in Singapore. The final adaptation of the SCDI (English) contains a total of 653 words, distributed across 22 semantic categories. The 680 words in the MCDI were distributed across the same 22 semantic categories, all of which were deemed suitable for Singapore (Fenson et al., 2007). Although the semantic categories from the MCDI were retained, individual items were changed within those categories. First, words were deleted if their referents were unlikely to surface in Singapore (e.g., "snow", "snowman", "boots", and "mittens"). Second, items were interchanged within semantic categories if original items from the MCDI were uncommon in Singapore English (e.g., "cheese", "applesauce"). Food items that are more common in Singapore (e.g., "porridge" and "rice") were added. Some items were exchanged for local forms of the same word (e.g., "jelly" was replaced with "jam"). Third, items that were not standard forms in the Singapore English lexicon, but that had familiar lexical equivalents were replaced with familiar forms, such as "pavement" for "sidewalk" and "petrol station" for "gas station." Fourth, items judged to be familiar to Singaporean children, but not present in the MCDI, such as "umbrella" and "slippers" were added to the SCDI (English) questionnaire.</p> <hd id="AN0154716031-11">Adaptation of heritage language CDIs</hd> <p>Two adaptations of the SCDI were developed for the most common heritage languages of Singapore: Mandarin and Malay. The SCDI (adapted for Mandarin) has a total of 603 words across 22 semantic categories, while the SCDI (adapted for Malay) has a total of 558 words across the same 22 semantic categories. The heritage language versions were not translations, but adaptations. Here, we outline the steps taken to develop Mandarin and Malay adaptations. As a first step, a group of four adult bilingual native speakers of each of the language pairings (English‐Mandarin and English‐Malay) reviewed each item on the English SCDI, evaluating its suitability for inclusion in Mandarin and Malay adaptations. In reviewing each item on the English SCDI, for those that were deemed suitable for heritage language forms, these items were translated if three conditions were met. First, the item had to be translatable across languages. Second, the translations had to faithfully capture common meaning across languages, connotatively and denotatively. Third, the translated form had to be judged to be a word that infants would both regularly encounter and produce during their early years in the heritage language. Examples of excluded items from the Mandarin adaptation were words such as "paint" (verb) or "shower" (noun), both of which can be translated into Mandarin, but English‐Mandarin bilingual toddlers in Singapore tend to default to the English words even within a Mandarin context. These words—which typically trigger a language switch back to the English word—were identified through parental feedback in piloting the CDI with Singaporean parents.</p> <p>In addition to a base set of words from the English SCDI, further adaptations were made to the heritage language forms. First, some words could be translated, but there was no clear one‐to‐one correspondence between the English and Mandarin forms. For example, in some cases, two contrastive English forms corresponded to one heritage language form (e.g., "cooker" and "stove" in English translates to one word '火炉' in Mandarin). In this case, a single item was included in the heritage language CDI. Second, some words, if literally translated across languages, have different connotative meanings across languages. For example, the Mandarin translation of "dinner" in English is unlikely to be used in Mandarin by a young child. For such items, common alternative forms (in this case, "吃饭" literally meaning "eat rice"), were included, as this form would be more likely to be used in child‐directed speech as well as productively by a young child than the more literal translation of "dinner." Third, some words could be translated from English to the heritage language, but maintain dual semantic representations in Mandarin and Malay. For example, the word "cut" is one item in English, but in Mandarin, this concept is expressed via a different word depending on whether one cuts with scissors or a knife. However, both forms are early acquired words. Likewise, the word "break" is one word in the English CDI, but in Malay, this verb is differentiated into "break into two pieces" or "break into several pieces." In Malay, both forms of "break" are early acquired forms in Malay. As a result, in these instances, both words were listed on the heritage language adaptation. Finally, there were early acquired words that had specific cultural and linguistic relevance to one language and that were likely to be known in a heritage language, but not in English. These words were added to the relevant heritage language version, but did not appear on the English SCDI (e.g., "sarong" or "baju kurong" for traditional Malay dress, judged by developers of a Malay CDI to be early acquired words for young children, Thow, 2002).</p> <p>For the subset of items that were translated across versions, two balanced bilingual speakers of English and Mandarin and two balanced bilingual speakers of English and Malay independently translated the items from the SCDI (English) questionnaire into lexical equivalents in Mandarin and Malay, respectively (forward translation). These translations were then verified by another native balanced bilingual third speaker of each language via backward translation, where all Mandarin and Malay items were translated back into English to ensure translation faithfulness.</p> <hd id="AN0154716031-12">Procedure</hd> <p>Parents were asked to complete two surveys: a Child Background Questionnaire assessing their child's health history and language exposure and the SCDI in both English and their heritage language (Mandarin or Malay). Both questionnaires were administered via parental survey, detailed below. The Child Background Questionnaire collected data on the child's basic demographic details (e.g., date of birth, medical history, language (s) spoken at home, gender of child, gestation, birth details, and weight details), as well as language background information. This background questionnaire was used to determine the health status of participants. All included participants were reported by their parents to be in good health with no major medical conditions. Information about SES was also collected for each participant, which included years of maternal education, years of paternal education, and total household income.</p> <p>Part of the Child Background Questionnaire was devoted to surveying the language background of participants. This questionnaire was based on the Language Exposure Questionnaire developed by Bosch and Sebastián‐Gallés (1997). Parents were surveyed on the number of caregivers and other regular conversational partners in the child's environment, the amount of time spent with each caregiver/regular conversational partner, and language use by each person, as well as a final estimate of the proportion of language exposure that each child receives. From these questions, we extracted the proportion of exposure to the community language (English) and heritage languages (Mandarin and Malay), across participants. A table of children's language backgrounds is provided in Appendix S1.</p> <p>Following the Child Background Questionnaire, the SCDI was administered in both English and the respondents' heritage language of either Mandarin or Malay if children received exposure to these languages. For reasons articulated in the Introduction, we decided to focus on English vocabulary size in relation to the amount of English experience to provide an instrument that could be broadly applied across the population. In the event that the adaptations are helpful for further study, the full SCDIs for English, Malay, and Mandarin can be found in Supporting Information.</p> <hd id="AN0154716031-13">RESULTS</hd> <p></p> <hd id="AN0154716031-14">Descriptive statistics</hd> <p>Parents' education level and household income were both scored on an ordinal scale of 1–5. For all analyses, education levels were scored as follows: <emph>1</emph> = Elementary and below, <emph>2</emph> = Middle school; <emph>3</emph> = High School (GCE "A" Levels [equivalent to 10th grade]); <emph>4</emph> = ITE (vocational training) or other professional diplomas, and <emph>5</emph> = University and above. Household income was scored as follows: <emph>1</emph> = SGD$2000; <emph>2</emph> = SGD$2000–SGD$3999; <emph>3</emph> = SGD$4000–SGD$5999; <emph>4</emph> = SGD$6000–$9999; <emph>5</emph> = SGD$10,000 and above. Descriptive statistics for each measure are reported in Table 1. The distribution of income levels in the sample reflects the mean and overall distribution of household income levels in Singapore as provided by the Singapore Department of Statistics during the year of data collection.</p> <p>1 TABLEDescriptive statistics on SES variables, English exposure, and number of words reported on CDI by age group</p> <p> <ephtml> <table><thead><tr><th align="left" /><th align="left"><p>Younger group</p><p>(<italic>N</italic> = 746)</p></th><th align="left"><p>Older group</p><p>(<italic>N</italic> = 570)</p></th></tr><tr><th align="left"><italic>M</italic> (<italic>SD</italic>)</th><th align="left">Range</th><th align="left"><italic>M</italic> (<italic>SD</italic>)</th><th align="left">Range</th></tr></thead><tbody><tr><td align="left">Maternal education (range: 1–5)</td><td align="char" char=".">4.11 (1.15)</td><td align="char" char="–">1–5</td><td align="char" char=".">4.10 (1.15)</td><td align="char" char="–">1–5</td></tr><tr><td align="left">Paternal education (range: 1–5)</td><td align="char" char=".">4.12 (1.16)</td><td align="char" char="–">1–5</td><td align="char" char=".">4.11 (1.19)</td><td align="char" char="–">1–5</td></tr><tr><td align="left">Household income (range: 1–5)</td><td align="char" char=".">3.55 (1.08)</td><td align="char" char="–">1–5</td><td align="char" char=".">3.62 (1.07)</td><td align="char" char="–">1–5</td></tr><tr><td align="left">Composite SES score (range: 1–5)</td><td align="char" char=".">11.78 (2.76)</td><td align="char" char="–">3–15</td><td align="char" char=".">11.84 (2.80)</td><td align="char" char="–">3–15</td></tr><tr><td align="left">Amount of English exposure (%)</td><td align="char" char=".">54.71 (28.25)</td><td align="char" char="–">1–100</td><td align="char" char=".">56.59 (29.22)</td><td align="char" char="–">1–100</td></tr><tr><td align="left">English vocabulary size (SCDI)</td><td align="char" char=".">109.90 (98.48)</td><td align="char" char="–">0–590</td><td align="char" char=".">253.27 (148.73)</td><td align="char" char="–">1–644</td></tr></tbody></table> </ephtml> </p> <p>1 Note</p> <ulist> <item>4 <emph>1</emph> = Elementary and below, <emph>2</emph> = Middle school; <emph>3</emph> = High school (GCE "A" Levels [equivalent to 10th grade]); <emph>4</emph> = ITE (vocational training) or other professional diplomas, and <emph>5</emph> = University and above. Household income was scored as follows: <emph>1</emph> = SGD$2000; <emph>2</emph> = SGD$2000–SGD$3999; <emph>3</emph> = SGD$4000–SGD$5999; <emph>4</emph> = SGD$6000–$9999; <emph>5</emph> = SGD$10,000 and above. SES composite score was a sum of maternal education, paternal education, and combined household income.</item> <item>5 Abbreviations: CDI, Communicative Development Inventory; SCDI, Singapore Communicative Development Inventory; SES, socioeconomic status.</item> </ulist> <p>All analyses presented were confirmatory unless specified as exploratory. An initial set of analyses designed to determine an optimal measure of SES revealed a significant correlation between SES factors: maternal education, paternal education, and combined household income (Spearman's rank correlation coefficient, maternal education, and paternal education: <emph>r</emph>(1237) = .53, <emph>p</emph> < .001; maternal education and household income: <emph>r</emph>(1213) = .47, <emph>p</emph> < .001; paternal education and household income: <emph>r</emph>(1143) = .47, <emph>p</emph> < .001). As such, a composite score for SES was derived by computing a summed score of these three variables. One hundred and seventy‐one (12.99%) participants had missing SES scores as they opted not to report information on paternal education (<emph>n</emph> = 77), household income (<emph>n</emph> = 101), or both (<emph>n</emph> = 7). For the missing values, multiple imputation was performed to fill the non‐responses for parental education and household income using the Hmisc package (4.5‐0) for R (Harrell, 2021). The SES composite score ranged from a minimum of 3 to a maximum of 15.</p> <hd id="AN0154716031-15">Bilingual vocabulary size estimates: Combined measures</hd> <p>Our primary analyses focused on estimates of English vocabulary size in relation to language exposure and age. However, a subset of children (<emph>N</emph> = 460) was exposed to only two languages (English and Mandarin/English and Malay) so CDI adaptations were available and administered for both. For these children, vocabulary size in each language was assessed for participants whose parents completed CDIs in both English and the heritage language. Single‐language (English) and language‐combined measures (conceptual vocabulary and summed vocabulary) were calculated. Summed vocabulary and conceptual vocabulary are reported for these participants in Table 2 by age group.</p> <p>2 TABLESingle language and language‐combined vocabulary size estimates for bilingually exposed learners</p> <p> <ephtml> <table><thead><tr><th align="left" /><th align="left"><p>English L1/Mandarin or Malay L2</p><p><italic>N</italic> = 249</p></th><th align="left"><p>Mandarin‐Malay L1/English L2</p><p><italic>N</italic> = 211</p></th></tr><tr><th align="left">Younger (<italic>SD</italic>)</th><th align="left">Older (<italic>SD</italic>)</th><th align="left">Younger (<italic>SD</italic>)</th><th align="left">Older (<italic>SD</italic>)</th></tr><tr><th align="left"><italic>N</italic> = 132</th><th align="left"><italic>N</italic> = 117</th><th align="left"><italic>N</italic> = 125</th><th align="left"><italic>N</italic> = 86</th></tr></thead><tbody><tr><td align="left">English vocabulary</td><td align="char" char="(">142.52 (110.81)</td><td align="char" char="(">293.28 (138.01)</td><td align="char" char="(">104.58 (90.88)</td><td align="char" char="(">203.64 (135.00)</td></tr><tr><td align="left">Conceptual vocabulary</td><td align="char" char="(">151.20 (114.27)</td><td align="char" char="(">304.46 (139.35)</td><td align="char" char="(">125.22 (103.37)</td><td align="char" char="(">268.55 (141.06)</td></tr><tr><td align="left">Summed vocabulary</td><td align="char" char="(">173.02 (139.04)</td><td align="char" char="(">354.35 (178.47)</td><td align="char" char="(">158.56 (148.39)</td><td align="char" char="(">348.01 (201.03)</td></tr></tbody></table> </ephtml> </p> <hd id="AN0154716031-16">Predictors of English vocabulary size</hd> <p>A hierarchical regression analyses were conducted to determine if effects of age, gender, SES, and the amount of exposure to English significantly predicted English vocabulary size. Preliminary analyses revealed heteroscedasticity in the residuals of a multivariate linear regression model with raw English vocabulary size as a dependent variable. Therefore, the raw test scores were Box–Cox‐transformed with <emph>λ</emph> = .4 (scaled score = (raw score<sups>0.4</sups> − 1)/0.4). Age in months (e.g., age = calendar age − 22) and English exposure (e.g., English exposure = calculated English exposure − 55) were centered before computing the quadratic terms and interactions to avoid multicollinearity (Marquardt, 1980). Gender was dummy‐coded as male = 0 and female = 1. SES scores were Box–Cox‐transformed with <emph>λ</emph> = 2 (scaled SES score = (raw SES composite score<sups>2</sups> − 1)/2) to contain effects of skew in the dataset. English exposure was operationalized as a continuous variable. In the following analyses, demographic variables (gender, age, and SES) were entered into the first step. In the second step, amount of exposure to English (%) was added to determine if this significantly predicted vocabulary size over and above the extent of variance accounted for by background variables. In the third step, interactions between age and other three variables were included. Finally, we investigated the effects of other potential terms, age<sups>2</sups> (to enable modeling of quadratic age associations) and English exposure<sups>2</sups> (to enable modeling of the non‐linear effect of English exposure).</p> <p>The hierarchical regression model with background variables as predictors—age, gender, and SES—in the first step was significant compared with a null model in predicting English vocabulary size, Δ<emph>F</emph>(<reflink idref="bib3" id="ref1">3</reflink>, 1315) = 189.35, <emph>p</emph> < .001, Δ<emph>R</emph><sups>2</sups> = .288. All three background variables significantly predicted English vocabulary size estimates (Table 3). The second model with amount of English exposure added was also significant against the first model, Δ<emph>F</emph>(<reflink idref="bib1" id="ref2">1</reflink>, 1312) = 67.31, <emph>p</emph> < .001, Δ<emph>R</emph><sups>2</sups> = .034, suggesting that English exposure significantly predicted English vocabulary size. In this step, SES was no longer a significant predictor. Adding the interaction terms of age with each of the other three variables (SES, gender, English exposure) in the third model further improved the model, Δ<emph>F</emph>(<reflink idref="bib3" id="ref3">3</reflink>, 1311) = 6.15, <emph>p</emph> < .001, Δ<emph>R</emph><sups>2</sups> = .009. In this step, age × English exposure was significant (<emph>b</emph> = .006, <emph>t</emph> = 4.07, <emph>p</emph> < .001), but the other two interaction terms (age × gender; age × SES) were not (<emph>p</emph>s > .29). The improvement of the model with age × English exposure added was attributable to the relations between English exposure and vocabulary size being stronger in the older group than the younger group. The last step with the higher‐order term of age and English exposure improved the model significantly, Δ<emph>F</emph>(<reflink idref="bib2" id="ref4">2</reflink>, 1308) = 5.34, <emph>p</emph> < .005, Δ<emph>R</emph><sups>2</sups> = .005. In this step, the squared term of English exposure was significant (<emph>b</emph> = −.0006, <emph>t </emph>= −3.11, <emph>p</emph> = .002), but the squared term of age was not (<emph>b</emph> = −.03, <emph>t</emph> = −0.99, <emph>p</emph> = .32), suggesting that there was a non‐linear relation between vocabulary size estimates and English exposure. To summarize, age, gender, SES, the age × English exposure interaction, and the squared term of English exposure were significant predictors of English vocabulary. The final model is therefore expressed as b<subs>0</subs> + b<subs>1</subs> × Gender + b<subs>2</subs> × Age + b<subs>3</subs> × SES + b<subs>4</subs> × English Exposure + b<subs>5</subs> × Age × English Exposure + b<subs>6</subs> × English Exposure<sups>2</sups> (13.91 + 2.00 × Gender + 1.01 × Age + 0.0095 × SES + 0.043 × English Exposure + 0.0064 × Age × English Exposure − 0.00064 × English Exposure<sups>2</sups>).</p> <p>3 TABLEHierarchical regression of predictors of scaled English words produced as reported on the CDI</p> <p> <ephtml> <table><thead><tr><th align="left">Predictor</th><th align="left"><italic>b</italic></th><th align="left"><p><italic>b</italic></p><p>95% CI</p><p>[LL, UL]</p></th><th align="left">Fit</th><th align="left">Difference</th></tr></thead><tbody><tr><td align="left">(Constant)</td><td align="char" char=".">13.37<xref ref-type="fn" rid="tfn9" /></td><td align="left">[12.48, 14.25]</td><td align="left" /><td align="left" /></tr><tr><td align="left">Gender</td><td align="char" char=".">1.80<xref ref-type="fn" rid="tfn9" /></td><td align="left">[1.17, 2.44]</td><td align="left" /><td align="left" /></tr><tr><td align="left">SES</td><td align="char" char=".">0.012<xref ref-type="fn" rid="tfn8" /></td><td align="left">[0.001, 0.022]</td><td align="left" /><td align="left" /></tr><tr><td align="left">Age</td><td align="char" char=".">1.02<xref ref-type="fn" rid="tfn9" /></td><td align="left">[0.93, 1.11]</td><td align="left" /><td align="left" /></tr><tr><td align="left" /><td align="char" char="." /><td align="left" /><td align="left">R<sup>2</sup> = .29<xref ref-type="fn" rid="tfn9" /></td><td align="left" /></tr><tr><td align="left" /><td align="char" char="." /><td align="left" /><td align="left">95% CI [0.25, 0.32]</td><td align="left" /></tr><tr><td align="left">(Constant)</td><td align="char" char=".">13.54<xref ref-type="fn" rid="tfn9" /></td><td align="left">[12.67, 14.40]</td><td align="left" /><td align="left" /></tr><tr><td align="left">Gender</td><td align="char" char=".">1.91<xref ref-type="fn" rid="tfn9" /></td><td align="left">[1.29, 2.53]</td><td align="left" /><td align="left" /></tr><tr><td align="left">SES</td><td align="char" char=".">0.0082</td><td align="left">[−0.0020, 0.0185]</td><td align="left" /><td align="left" /></tr><tr><td align="left">Age</td><td align="char" char=".">1.01<xref ref-type="fn" rid="tfn9" /></td><td align="left">[0.92, 1.09]</td><td align="left" /><td align="left" /></tr><tr><td align="left">English exposure</td><td align="char" char=".">0.045<xref ref-type="fn" rid="tfn9" /></td><td align="left">[0.034, 0.056]</td><td align="left" /><td align="left" /></tr><tr><td align="left" /><td align="char" char="." /><td align="left" /><td align="left">R<sup>2</sup> = .32<xref ref-type="fn" rid="tfn9" /></td><td align="left">ΔR<sup>2</sup> = .034<xref ref-type="fn" rid="tfn9" /></td></tr><tr><td align="left" /><td align="char" char="." /><td align="left" /><td align="left">95% CI [0.28, 0.36]</td><td align="left">95% CI [0.02, 0.05]</td></tr><tr><td align="left">(Constant)</td><td align="char" char=".">13.48<xref ref-type="fn" rid="tfn9" /></td><td align="left">[12.62, 14.35]</td><td align="left" /><td align="left" /></tr><tr><td align="left">Gender</td><td align="char" char=".">1.98<xref ref-type="fn" rid="tfn9" /></td><td align="left">[1.36, 2.60]</td><td align="left" /><td align="left" /></tr><tr><td align="left">SES</td><td align="char" char=".">0.088</td><td align="left">[−0.002, 0.019]</td><td align="left" /><td align="left" /></tr><tr><td align="left">Age</td><td align="char" char=".">0.90<xref ref-type="fn" rid="tfn9" /></td><td align="left">[0.66, 1.14]</td><td align="left" /><td align="left" /></tr><tr><td align="left">English exposure</td><td align="char" char=".">0.044<xref ref-type="fn" rid="tfn9" /></td><td align="left">[0.033, 0.055]</td><td align="left" /><td align="left" /></tr><tr><td align="left">Age × English exposure</td><td align="char" char=".">0.0064<xref ref-type="fn" rid="tfn9" /></td><td align="left">[0.0033, 0.0094]</td><td align="left" /><td align="left" /></tr><tr><td align="left">Age × SES</td><td align="char" char=".">0.0014</td><td align="left">[−0.0016, 0.0043]</td><td align="left" /><td align="left" /></tr><tr><td align="left">Age × gender</td><td align="char" char=".">0.0076</td><td align="left">[−0.1673, 0.1826]</td><td align="left" /><td align="left" /></tr><tr><td align="left" /><td align="char" char="." /><td align="left" /><td align="left">R<sup>2</sup> = .33<xref ref-type="fn" rid="tfn9" /></td><td align="left">ΔR<sup>2</sup> = .009<xref ref-type="fn" rid="tfn9" /></td></tr><tr><td align="left" /><td align="char" char="." /><td align="left" /><td align="left">95% CI [0.29, 0.37]</td><td align="left">95% CI [0.00, 0.02]</td></tr><tr><td align="left">(Constant)</td><td align="char" char=".">14.26<xref ref-type="fn" rid="tfn9" /></td><td align="left">[13.16, 15.37]</td><td align="left" /><td align="left" /></tr><tr><td align="left">Gender</td><td align="char" char=".">1.98<xref ref-type="fn" rid="tfn9" /></td><td align="left">[1.37, 2.60]</td><td align="left" /><td align="left" /></tr><tr><td align="left">SES</td><td align="char" char=".">0.0092</td><td align="left">[−0.0011, 0.0194]</td><td align="left" /><td align="left" /></tr><tr><td align="left">Age</td><td align="char" char=".">0.93<xref ref-type="fn" rid="tfn9" /></td><td align="left">[0.68, 1.18]</td><td align="left" /><td align="left" /></tr><tr><td align="left">English exposure</td><td align="char" char=".">0.043<xref ref-type="fn" rid="tfn9" /></td><td align="left">[0.032, 0.054]</td><td align="left" /><td align="left" /></tr><tr><td align="left">Age<sup>2</sup></td><td align="char" char=".">−0.026</td><td align="left">[−0.076, 0.025]</td><td align="left" /><td align="left" /></tr><tr><td align="left">English exposure<sup>2</sup></td><td align="char" char=".">−0.00065<xref ref-type="fn" rid="tfn9" /></td><td align="left">[−0.00106, −0.00024]</td><td align="left" /><td align="left" /></tr><tr><td align="left">Age × English exposure</td><td align="char" char=".">0.0063<xref ref-type="fn" rid="tfn9" /></td><td align="left">[0.0033, 0.0093]</td><td align="left" /><td align="left" /></tr><tr><td align="left">Age × SES</td><td align="char" char=".">0.0014</td><td align="left">[−0.0015, 0.0044]</td><td align="left" /><td align="left" /></tr><tr><td align="left">Age × gender</td><td align="char" char=".">0.0092</td><td align="left">[−0.1652, 0.1836]</td><td align="left" /><td align="left" /></tr><tr><td align="left" /><td align="char" char="." /><td align="left" /><td align="left">R<sup>2</sup> = .337<xref ref-type="fn" rid="tfn9" /></td><td align="left">ΔR<sup>2</sup> = .005<xref ref-type="fn" rid="tfn9" /></td></tr><tr><td align="left" /><td align="char" char="." /><td align="left" /><td align="left">95% CI [0.29, 0.37]</td><td align="left">95% CI [−0.00, 0.01]</td></tr></tbody></table> </ephtml> </p> <ulist> <item>2 Note</item> <item>6 A significant <emph>b</emph>‐weight indicates the semi‐partial correlation is also significant. <emph>b</emph> represents unstandardized regression weights. LL and UL indicate the lower and upper limits of a confidence interval, respectively.</item> <item>7 Abbreviations: SES, socioeconomic status.</item> <item>8 * <emph>p</emph> < .05.</item> <item>9 ** <emph>p</emph> < .01.</item> </ulist> <p>As can be viewed in Table 3, there was a quadratic effect of English exposure on English vocabulary size. This effect was examined via new models within each of the two age groups (older and younger) to identify predictors of vocabulary size with quadratic terms included via exploratory analysis. The goal of this was to determine whether thresholds exist for language exposure in relation to English vocabulary size. For the younger group, the model was 8.59 + 1.32 × Gender + 0.93 × (Age − 19) + 0.014 × (Exposure − 55) − 0.00037 × (Exposure − 55)<sups>2</sups>. In this model, there was an increase in predicted English vocabulary as English exposure increased (see Figure 1a) until English exposure reached an inflection point at 73.9%. After this inflection point, English vocabulary began to plateau as English exposure increased. In the older group, a different pattern of results was observed. Using the following model, 37.12 + 5.23 × Gender + 0.052 × SES + 3.22 × (Age − 26) + 0.21 × (Exposure − 55) − 0.0021 × (Exposure − 55)<sups>2</sups>, there was no clear threshold effect as the mathematical inflection point exceeded the range of exposure values (<reflink idref="bib105" id="ref5">105</reflink>). There was, however, a decrease in the marginal effect of increasing English exposure on English vocabulary size. These findings suggest that the significance of English exposure for vocabulary uptake is age‐dependent, with a more pronounced threshold effect evident between 17 and 21 months.</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/CDV/01jan22/cdev13676-fig-0001.jpg?ephost1=dGJyMNXb4kSepq84yOvqOLCmsE6epq5Srqa4SK6WxWXS" alt="cdev13676-fig-0001.jpg" title="1 (a) Relationship between English vocabulary size and English exposure (younger group). (b) Relationship between English vocabulary size and English exposure (older group). For both charts, dashed line indicates inflection point." /> </p> <p></p> <hd id="AN0154716031-18">Disentangling SES and English exposure: A mediation analysis</hd> <p>We sought to examine the significant effect of SES on English vocabulary as the amount of exposure to English was significantly correlated with SES scores (<emph>r</emph>(1213) = .09, <emph>p</emph> = .001). In contrast, English exposure was not correlated with other background variables (age and gender, <emph>p</emph>s > .18). Furthermore, the effect of SES in the second step in the hierarchical regression was no longer significant after including English exposure as predictors. Mediation analyses based on the final model confirmed that effects of SES on vocabulary size were indeed mediated by English exposure (conducted using the <emph>mediation</emph> 4.5.0 package for <emph>R</emph> with a bootstrap method; simulations: 15,000). In particular, higher‐SES families were associated with higher levels of English exposure, which returned higher English vocabulary sizes. The indirect effect, 0.0036 (<emph>p</emph> = .004) and the total effect, 0.013 (<emph>p</emph> = .01) of SES on productive vocabulary were statistically significant, and the direct effect, 0.0095 (<emph>p</emph> = .06) was marginally significant. Within the total effect, the indirect effect accounted for 27.4% (<emph>p</emph> = .02) of the full effect of SES on vocabulary size.</p> <p>To delve more deeply into the effects of SES, we conducted two additional regressions based on the model expressed above (b<subs>0</subs> + b<subs>1</subs> × Gender + b<subs>2</subs> × Age + b<subs>3</subs> × SES + b<subs>4</subs> × English Exposure + b<subs>5</subs> × Age × English Exposure + b<subs>6</subs> × English Exposure<sups>2</sups>). In the first regression, we included SES, but excluded all terms relating to English exposure in the model above. In the second regression, we excluded SES from the model, but included all terms related to English exposure in the model above. We conducted an ANOVA to compare the two models to determine the unique effects of SES on English vocabulary. The results demonstrated that the variance accounted for by adding English exposure (together with its interaction term and quadratic term) was 4.8% (Δ<emph>R</emph><sups>2</sups> = .048, <emph>p</emph> < .001). The variance accounted for by adding SES in the second step was marginally significant at 0.2% (Δ<emph>R</emph><sups>2</sups> = .002, <emph>p</emph> = .07). This suggests that the contribution of SES to vocabulary size estimates was largely due to the increased English exposure present in higher SES families, indicating a limited unique effect of SES.</p> <hd id="AN0154716031-19">Summary of regression analyses</hd> <p>To summarize, a consistent set of predictors were identified for English vocabulary size. First, girls had higher vocabulary sizes than boys. Second, older children had higher vocabulary sizes than younger children, which accounted for the largest part of the variance explained by the model. Third, English exposure positively predicted English vocabulary size and its squared term was negatively significant, which suggests a non‐linear relation between English vocabulary size and English exposure. The effect of English exposure was moderated by age, reflected by the significant effect of the age × English exposure interaction term. This was accounted for by the fact that the relation between English exposure and vocabulary size was stronger in the older group than in the younger group. Over and above effects of gender, age, and English exposure, the interaction terms (gender × English exposure and SES × English exposure) did not improve the fit of the model for English. The mediating role of English exposure on the relations between English vocabulary and SES was confirmed via mediation analysis, suggesting that the benefits of SES for vocabulary size were largely attributable to the larger amount of English exposure in higher‐SES families. In contrast, the unique effect of SES alone was comparatively small.</p> <hd id="AN0154716031-20">Monolingual and bilingual comparisons of English vocabulary size: A categorical approach</hd> <p>To determine how multilingual learners compared with monolingual learners using a categorical approach, we compared English vocabulary size across three groups of learners: monolingual infants (>90% exposure to English), multilingual infants with dominant English exposure (<=90% exposure to English with English as the first language), and multilingual infants with dominant heritage language exposure (<=90% exposure to English with English as the second language). A one‐way ANOVA was performed to compare English vocabulary across each exposure group within each age group. In the younger age, there was a main effect of exposure group, <emph>F</emph>(<reflink idref="bib2" id="ref6">2</reflink>, 743) = 5.31, <emph>p</emph> = .005. English vocabulary size was highest for the English monolingual group, followed by multilingual infants (English‐dominant exposure), and then by multilingual infants (heritage language‐dominant exposure), Tukey's honestly significant different (HSD) post hoc comparisons revealed no significant difference in English vocabulary between English monolingual and English‐dominant multilingual infants (<emph>p</emph> = .77). English vocabulary size estimates for English monolingual infants and English‐dominant multilingual infants were significantly greater than for English non‐dominant multilingual infants (<emph>p</emph> = .03 and <emph>p</emph> = .02, respectively). A similar pattern of results emerged for the older group. There was a main effect of exposure group, <emph>F</emph>(<reflink idref="bib2" id="ref7">2</reflink>, 567) = 26.41, <emph>p</emph> < .0001. Tukey's HSD post hoc comparisons revealed no significant difference in English vocabulary between English monolingual and English‐dominant multilingual infants (<emph>p</emph> = .34). However, vocabulary size for English monolingual infants and English‐dominant multilingual infants was significantly greater than for English non‐dominant multilingual infants (both <emph>p</emph>‐values < .0001). Vocabulary size estimates are plotted for each group in Figure 2.</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/CDV/01jan22/cdev13676-fig-0002.jpg?ephost1=dGJyMNXb4kSepq84yOvqOLCmsE6epq5Srqa4SK6WxWXS" alt="cdev13676-fig-0002.jpg" title="2 Number of words produced by age and exposure group. Error bars reflect SEM" /> </p> <p></p> <hd id="AN0154716031-22">Language norms for multilingual children</hd> <p>In this section, we report our Singaporean children's English vocabulary norms based on a regression‐based norming approach (see Floccia et al., 2018). The regression‐based norms approach offers two methodological advantages over simply carving the sample into different tiers of vocabulary size and assigning each tier a percentile value. First, this approach builds on the assumption that scaled score distributions change smoothly with English exposure and age, revealing change in predicted vocabulary in relation to English exposure and age, both of which were shown to predict English vocabulary size. Second, this approach allows for all observations within the normative sample be used to estimate the distribution without sub‐dividing the sample into smaller groups based on age or other factors (Oosterhuis et al., 2017). We report both age‐dependent and age‐aggregated norms.</p> <p>In developing norms, we had four goals. First, we sought to identify a means to derive population‐relative scores (percentiles) that correspond to vocabulary scores from the English SCDI. Second, we sought to determine whether age‐dependent or age‐aggregated norms provided similar results. Third, we investigated whether the norms could be used to identify cases of potential clinical concern. Finally, we tested our norms fitted to 80% of our sample to a randomly generated 20% of the sample to determine whether these norms could be broadly used to predict percentiles within the larger population of Singaporean children in the age bands sampled.</p> <p>To determine which predictors would be used in the regression models used to derive norms, our analyses were aligned with norming procedures pursued by Floccia et al. (2018). First, the full models, including all significant predictors identified in the predictive model described above, were reduced via stepwise elimination of the least significant predictor with a two‐tailed <emph>p</emph> value above.05. Second, a predictor was required to reach an effect size of at least.005 in the analyses of covariance to prevent over‐fitting (see Floccia et al., 2018). The assumptions of the final regression analysis were tested for each model. Once the final model was determined, we determined percentiles for English vocabulary size for any given infant using a four‐step method. First, the final regression model (predicted scale score = <emph>B</emph><subs>0</subs> + <emph>B</emph><subs>1</subs><emph>X</emph><subs>1</subs> + ... + <emph>B<subs>i</subs>X<subs>i</subs></emph>, with <emph>B</emph><subs>0</subs> = the constant intercept, <emph>B<subs>i</subs></emph> = coefficients of the predictors, and <emph>X<subs>i</subs></emph> = the value of predictors) was used to calculate the infant's predicted scaled score. Second, the residuals were calculated (<emph>e<subs>i</subs></emph> = observed scaled score − predicted scaled score). Third, the residuals were standardized (<emph>Z<subs>i</subs></emph> = <emph>e<subs>i</subs></emph>/<emph>SD</emph>(residual), with <emph>SD</emph>(residual) = the standard deviation of the residuals). Fourth, the standardized residuals were converted into percentiles via the standard normal cumulative distribution function (if the model's assumption of normality of the residuals was met in the normative sample) or via the empirical cumulative distribution function of the standardized residuals (if the standardized residuals were not normally distributed in the normative sample).</p> <p>Norms were constructed separately for older and younger age groups as well as combined across age groups. After developing both types of norms, these norms were evaluated by computing the percentage of agreement via kappa coefficients (Cohen, 1960) using the 5th and 95th percentile cut‐offs for age‐aggregated and age‐dependent norms. Classification agreement is considered poor when <emph>κ</emph> value is below.40, fair when the values lie between.40 and.60, good when values lie between.60 and.75, and values above.75 indicate excellent agreement (Watkins & Pacheco, 2000). In addition, agreement was also evaluated by computing the correlations between the <emph>z</emph>‐score measures generated from the age‐aggregated and age‐dependent regression models (i.e. <emph>r</emph>[<emph>Z</emph>(combined), <emph>Z</emph>(separated)]. All analyses were conducted with R 4.1. An alpha level of.05 was used as a significance criterion for in all analyses in the regression models.</p> <hd id="AN0154716031-23">Regression‐based norms: Final models</hd> <p>The final regression models of scaled scores are presented in Table 4. At the outset, we checked for outliers, heteroscedasticity, multicollinearity, and normality. There was no significant influence of outliers (maximum Cook's distance in all three models <0.02) or multicollinearity (maximum VIF in all three models <1.02) observed for the final models. The Breusch Pagan test suggested that there was no heteroscedasticity for any of the three models with fitted values (with all <emph>p</emph>s > .56). The standardized residuals were normally distributed (Kolmogorov–Smirnov test: all <emph>p</emph>s > .15). The final model demonstrated that infant's gender, age, English exposure, the quadratic term of English exposure, and the interaction between age and English exposure were significant predictors of infant's English vocabulary scores. As before, infant age explained most of the variance of the model. English exposure also significantly positively predicted English vocabulary size. The significant interaction term age × English exposure suggests that the relation between English exposure and vocabulary size was stronger in the older group than the younger group.</p> <p>4 TABLEFinal multiple linear regression models for English vocabulary scaled scores norms</p> <p> <ephtml> <table><thead><tr><th align="left">Model</th><th align="left">Predictor</th><th align="left">Coef.</th><th align="left"><italic>SE</italic></th><th align="left"><italic>T</italic></th><th align="left"><italic>R</italic><sup>2</sup></th><th align="left"><italic>SD</italic></th><th align="left"><italic>λ</italic></th></tr></thead><tbody><tr><td align="left">Combined</td><td align="left" /><td align="char" char="." /><td align="char" char="." /><td align="char" char="." /><td align="char" char=".">.33</td><td align="char" char=".">5.69</td><td align="char" char=".">.4</td></tr><tr><td align="left">(Constant)</td><td align="char" char=".">14.07</td><td align="char" char=".">0.22</td><td align="char" char=".">62.68<xref ref-type="fn" rid="tfn16" /></td><td align="char" char="." /><td align="char" char=".">5.99 for boys</td><td align="char" char="." /></tr><tr><td align="left">Gender<xref ref-type="fn" rid="tfn11" /></td><td align="char" char=".">2.00</td><td align="char" char=".">0.31</td><td align="char" char=".">6.36<xref ref-type="fn" rid="tfn16" /></td><td align="char" char="." /><td align="char" char=".">5.37 for girls</td><td align="char" char="." /></tr><tr><td align="left">Age (months)<xref ref-type="fn" rid="tfn12" /></td><td align="char" char=".">1.00</td><td align="char" char=".">0.04</td><td align="char" char=".">22.57<xref ref-type="fn" rid="tfn16" /></td><td align="char" char="." /><td align="char" char="." /><td align="char" char="." /></tr><tr><td align="left">English exposure<xref ref-type="fn" rid="tfn13" /></td><td align="char" char=".">0.044</td><td align="char" char=".">0.0055</td><td align="char" char=".">8.09<xref ref-type="fn" rid="tfn16" /></td><td align="char" char="." /><td align="char" char="." /><td align="char" char="." /></tr><tr><td align="left">Age × English exposure</td><td align="char" char=".">0.0064</td><td align="char" char=".">0.0015</td><td align="char" char=".">4.16<xref ref-type="fn" rid="tfn16" /></td><td align="char" char="." /><td align="char" char="." /><td align="char" char="." /></tr><tr><td align="left">Older</td><td align="left" /><td align="char" char="." /><td align="char" char="." /><td align="char" char="." /><td align="char" char=".">.18</td><td align="char" char=".">15.67</td><td align="char" char=".">.6</td></tr><tr><td align="left">(Constant)</td><td align="char" char=".">37.12</td><td align="char" char=".">1.91</td><td align="char" char=".">19.44<xref ref-type="fn" rid="tfn16" /></td><td align="char" char="." /><td align="char" char=".">16.97 for boys</td><td align="char" char="." /></tr><tr><td align="left">Gender</td><td align="char" char=".">5.23</td><td align="char" char=".">1.33</td><td align="char" char=".">3.93<xref ref-type="fn" rid="tfn16" /></td><td align="char" char="." /><td align="char" char=".">14.35 for girls</td><td align="char" char="." /></tr><tr><td align="left">Age (months)</td><td align="char" char=".">3.22</td><td align="char" char=".">0.75</td><td align="char" char=".">4.29<xref ref-type="fn" rid="tfn16" /></td><td align="char" char="." /><td align="char" char="." /><td align="char" char="." /></tr><tr><td align="left">English exposure</td><td align="char" char=".">0.21</td><td align="char" char=".">0.02</td><td align="char" char=".">9.01<xref ref-type="fn" rid="tfn16" /></td><td align="char" char="." /><td align="char" char="." /><td align="char" char="." /></tr><tr><td align="left">English exposure<sup>2</sup></td><td align="char" char=".">−0.0021</td><td align="char" char=".">0.0009</td><td align="char" char=".">−2.42<xref ref-type="fn" rid="tfn14" /></td><td align="char" char="." /><td align="char" char="." /><td align="char" char="." /></tr><tr><td align="left">SES</td><td align="char" char=".">0.052</td><td align="char" char=".">0.022</td><td align="char" char=".">2.36<xref ref-type="fn" rid="tfn14" /></td><td align="char" char="." /><td align="char" char="." /><td align="char" char="." /></tr><tr><td align="left">Younger</td><td align="left" /><td align="char" char="." /><td align="char" char="." /><td align="char" char="." /><td align="char" char=".">.083</td><td align="char" char=".">3.75</td><td align="char" char=".">.3</td></tr><tr><td align="left">(Constant)</td><td align="char" char=".">8.30</td><td align="char" char=".">0.20</td><td align="char" char=".">42.30<xref ref-type="fn" rid="tfn16" /></td><td align="char" char="." /><td align="char" char=".">3.87 for boys</td><td align="char" char="." /></tr><tr><td align="left">Gender</td><td align="char" char=".">1.32</td><td align="char" char=".">0.28</td><td align="char" char=".">4.79<xref ref-type="fn" rid="tfn16" /></td><td align="char" char="." /><td align="char" char=".">3.63 for girls</td><td align="char" char="." /></tr><tr><td align="left">Age (months)</td><td align="char" char=".">0.93</td><td align="char" char=".">0.16</td><td align="char" char=".">5.84<xref ref-type="fn" rid="tfn16" /></td><td align="char" char="." /><td align="char" char="." /><td align="char" char="." /></tr><tr><td align="left">English exposure</td><td align="char" char=".">0.014</td><td align="char" char=".">0.005</td><td align="char" char=".">2.92<xref ref-type="fn" rid="tfn15" /></td><td align="char" char="." /><td align="char" char="." /><td align="char" char="." /></tr></tbody></table> </ephtml> </p> <ulist> <item>3 Note</item> <item>10 The lambda (<emph>λ</emph>) for Box–Cox transformation differs in different models, so the coefficients are not comparable between models.</item> <item>11 a Male = 0 and female = 1.</item> <item>12 b Age was centered (age = age − 26 in older infants model; age = age − 19 in younger infants model; age = age − 22 in combined model).</item> <item>13 c English exposure was centered (English exposure = English exposure − 55).</item> <item>14 * <emph>p</emph> < .05.</item> <item>15 ** <emph>p</emph> < .01.</item> <item>16 *** <emph>p</emph> < .001.</item> </ulist> <hd id="AN0154716031-24">Use of raw vocabulary size estimates to determine percentiles</hd> <p>Norms for scaled scores transformed from the raw scores of English vocabulary size were established by the four‐step procedure described above. For example, if an 18‐month‐old female infant with 50% English exposure obtained a raw score of 200 from the English CDI for Singaporean infants, by using a regression‐based norm for both age‐groups combined, the percentile can be calculated by using the first model in Table 4 via the following steps. First, the raw score is transformed through a Box–Cox function to a scaled score, 18.31 (=(200<sups>0.4</sups> − 1)/0.4). The regression‐based norm can be used to calculate the expected scaled score for this infant, i.e., 11.98 (=14.07 + (2.00 × 1) + (1.00 × (18 − 22)) + (0.044 × (50 − 55)) + (0.0064 × (18 − 22)) × (50 − 55)). Second, the residual is calculated, that is, 6.33 (=18.31 − 11.98). Third, the residual is standardized, that is, 1.18 (=6.33/5.37). Finally, the standardized residual is converted into a percentile value via the standard normal cumulative distribution function. A standardized residual of the scaled score, which equals 1.18, corresponds to a percentile value of 88.10.</p> <p>If we were to calculate the percentile for this infant based on the regression exclusively for the younger infant group rather than the age‐aggregated measures, the third regression in Table 4 would be used. The scaled score for the above‐mentioned infant is 13.00 (=(200<sups>0.3</sups> − 1)/0.3). The predicted score would be 8.62 (=8.30 + (1.32 × 1) + (0.93 × (18 − 19)) + (0.014 × (50 − 55))). The residual is 4.38 (=13.00 − 8.62), and the standardized residual is 1.21 (=4.38/3.63). A standardized residual of the scaled score (1.21) corresponds to a percentile value of 88.69. As age is a predictor in the norm‐based regression model, the calculated percentiles for the same infant from the two norms (age‐aggregated and age‐dependent) vary only slightly (88.1% vs. 88.7%) and remain within the same 5‐percentile bracket (85–90).</p> <p>The calculated agreement between age‐aggregated and age‐dependent norms show that both norms provide a suitable population‐relative metric of vocabulary size. We then explored whether both sets of norms would be equally useful in identifying cases at the tails of the distribution (below the 5th percentile and above the 95th percentile). Comparisons between the age‐aggregated and age‐dependent regression‐based norms when the 5th percentile was used provided agreement in 98.8% of cases (Cohen's <emph>κ</emph> = .81). At the other extreme, when the 95th percentile was used as a cut‐off, there was agreement in 98.9% of the cases for the scores (Cohen's <emph>κ</emph> = .86). These results suggest that the differences between age‐aggregated and age‐dependent regressions are very small, with <emph>κ</emph> values of.81 at the 5th percentile cut‐off and.86 at the 95th percentile cut‐off, indicating excellent agreement (Watkins & Pacheco, 2000). Therefore, the results point to the use of either set of norms for identifying percentiles at the tails of the distribution.</p> <hd id="AN0154716031-25">User‐friendly normative tables</hd> <p>We have provided normative tables for infants aged from 17 to 28 months based on the combined regression model in Supporting Information for general use. Our norming sample ranged from 17 to 21 months and 24 to 28 months. The benefit of this table is that it allows parents to evaluate their child's English raw scores from the CDI without having to go through the previously mentioned four‐step calculation. The cost of using this normative table is the loss of some degree of differentiation: percentiles are provided in intervals of 5 percentile points, rather than an individualized percentile. However, intervals of 5 percentile points are standardly used when computing vocabulary norms (Fenson et al., 2000).</p> <hd id="AN0154716031-26">The robustness of the approach and validity of the equation in predicting percentile values f...</hd> <p>We evaluated the robustness of our approach in constructing the norms and validity of the regression equations presented in Table 4 by investigating how accurately they can be used to predict new data, as well as clinically relevant cases. Here, we asked two questions. First, we investigated whether vocabulary percentiles could be predicted in novel participants by using obtained data from a subset of the sample to predict data for the remainder of the sample, to examine the generalizability of our obtained norms (see Floccia et al., 2018). Second, we examined whether the norms obtained could be used to identify clinically relevant subsamples of participants. We note that this is not an external validation of our norms and does not tell us how CDI scores relate to other measures of language development. Instead, it tells us whether the norms that we have developed generalize internally to new CDI data within our sample.</p> <p>For the first question of whether the norms obtained could be used to predict percentiles in novel participants, we randomly divided our data into two subsamples by a ratio of 80:20. This led to the creation of a training dataset (<emph>n</emph> = 1052) and a testing dataset (<emph>n</emph> = 264). We then used the training dataset to re‐fit the regression equation (the combined model in Table 4) to obtain new regression coefficients. Following this, we applied the new model to the testing database to obtain the predicted scaled score and associated percentiles. Similar to Floccia et al. (2018), we calculated the correlation between the predicted and observed scaled scores of infants in the testing dataset and compared the differences between the predicted and actual values using <emph>t</emph>‐tests and associated <emph>p</emph>‐values. There was a strong correlation between the observed and predicted scaled scores (0.57, 95% CI [0.51, 0.63]), which was significant in all 1000 simulations. The distribution of <emph>p</emph>‐values for the <emph>t</emph>‐test of the mean of the observed and predicted scores over 1000 simulations demonstrated no significant difference (<emph>p</emph> < .05) between the predicted and observed scaled scores in 98.9% of the 1000 simulations based on having applied the re‐fitted model from the training dataset to the testing dataset. This indicates no systematic underprediction or overprediction of scaled vocabulary scores in the testing dataset. Among the 1000 randomly selected training datasets for which we obtained model estimates, the mean value of the coefficient for the constant term was 14.07, 95% CI [13.88, 14.27], the average coefficient for gender was 2.01, 95% CI [1.74, 2.28], the average value of the coefficient for age was 1.00, 95% CI [0.97, 1.04], the coefficient mean for English exposure was 0.045, 95% CI [0.04, 0.05], the coefficient mean for English exposure × age was 0.0064, 95% CI [0.0051, 0.0078]. These values were very close to the coefficients of the model we fit to the data in Table 4.</p> <p>For the second question, to determine risk for a potential language delay, we obtained an estimate of the proportion of children judged to be language‐delayed in the training dataset, forming a distribution of these participants. This distribution comprised infants at the 15th percentile and below. This threshold was guided by past evidence that the 15th percentile serves as a useful cut‐off point to identify clinical concern in young children's vocabulary sizes (Thal et al., 2004). We then sought to identify the proportion of children in the testing dataset who would be identified as being of clinical concern from the training data. We repeated the steps identified in the first analysis in this section 1000 times to obtain correlations, the proportion of cases of clinical concern identified, and the distribution of the model coefficients, calculating means and confidence intervals. The mean value of the resulting proportion of children identified was 15.7%, which is very close to the criterion we set for those scoring below the 15th percentile (15.7%, 95% CI [11.7, 19.7]).</p> <p>These results indicated that the sensitivity of our regression‐based norms is very high. Using the vocabulary development norms obtained from our data, we were able to generate percentiles for different subsamples of data. We were also able to identify infants who may present with risk for language delay with high reliability, providing a key test of the predictive validity and generalizability of the obtained norms within the age groups sampled.</p> <hd id="AN0154716031-27">DISCUSSION</hd> <p>In this study, we have provided predictors of vocabulary size as well as normative data drawn from a large and socioeconomically diverse sample of monolingual and multilingual infants from Singapore. We present quantitative norms for English vocabulary in relation to age, gender, and English exposure. These norms were developed via regression‐based analysis, which can be used to derive a population‐relative measure of English vocabulary in typically developing children as well as to identify cases that may be useful in identifying risk for language delay. In addition to providing norms for early vocabulary development in Singaporean children, this study expands our understanding of predictors of community language vocabulary size in multilingual settings. The specific age ranges for which norms were developed reflect developmentally relevant milestones. Within the younger group of 17–21 months, having a low score on the CDI at 18 months predicts slower vocabulary growth over the following year (Fernald & Marchman, 2012). Within the older group of 24–28 months, children who have low CDI scores at the 2‐year mark go onto to have lower linguistic proficiency in middle childhood (Marchman & Fernald, 2008). The 2‐year‐mark has been proposed as an appropriate time to evaluate children for potential clinical risk using CDI estimates (Can et al., 2013), lending significance to these age intervals as informative evaluation points.</p> <p>In addition to the development of norms, we sought to identify predictors of English vocabulary size in a multilingual environment. In this vein, our study yielded three primary findings. First, higher vocabulary size estimates were reported in girls versus boys, in older infants versus younger infants, and in those with higher amounts of exposure to English. Second, the relation between English exposure and vocabulary size was stronger for older infants than for younger infants, with a more pronounced relationship observed around the 2‐year mark versus at 1.5 years. Third, SES also predicted vocabulary size estimates; however, SES exerted both direct and indirect effects on vocabulary size. A test of indirect effects of SES demonstrated that SES effects were mediated by the amount of English exposure. Even though our sample was socioeconomically diverse, the unique effect of SES on vocabulary size was relatively small.</p> <p>Situated within the broader context of past research on first and second language acquisition, both familiar and novel patterns emerged. We review both types of patterns. First, the presence of "dose–response" relations, where the amount of exposure infants receive in their native language contributes significantly to variance in vocabulary size, is highly consistent with prior findings in monolingual infants (e.g., Weisleder & Fernald, 2013). Similar relations between first and second language input and first and language uptake in each language have been attested in bilingual children (Floccia et al., 2018; Place & Hoff, 2011). However, our study demonstrates a tighter coupling of language exposure and language uptake in older children than in younger children. This suggests that when evaluating bilingual children on their community language acquisition, community language exposure should be more carefully considered in the interpretation of vocabulary size estimates as children approach the two‐year mark.</p> <p>A distinguishing feature of our study was that it sampled a socioeconomically diverse set of families. In contrast, past norming studies both in monolingual and multilingual populations have typically sampled higher‐SES populations, reporting no or limited effects of SES. For example, within the MCDI norming sample, consisting predominantly of high‐SES families, SES accounted for <1% of the variance in vocabulary size (Fenson et al., 2007). Similarly, other adaptations of the CDI have typically under‐represented lower SES families (Dale & Goodman, 2005; Floccia et al., 2018; Hamilton et al., 2000), finding no significant effect of SES. The under‐representation of low‐SES families in past studies on vocabulary development is of consequence in light of the fact that more pronounced effects of SES on vocabulary have been reported in lower‐SES groups than in higher‐SES groups (Arriaga et al., 1998; Fernald et al., 2013). Although our study contained significant socioeconomic diversity, SES alone accounted for <1% of the variance, consistent with past findings drawing from narrower and higher socioeconomic samples. In our study, indirect effects of SES were more clearly evident. In particular, effects of SES were mediated by exposure to English, suggesting that the risk introduced by SES alone, independent of other factors, is relatively small. We note that while it is not uncommon for SES to be intertwined with community language exposure, the nature of the association can vary across societies. In the United Kingdom, high second language exposure relative to English exposure is reportedly associated with higher SES (Floccia et al., 2018). By contrast, in parts of the United States, low community language exposure aggregates within lower SES strata, a pattern consistent with the findings within our sample (Hoff, 2013; also see Hernandez, 2004).</p> <p>Identifying indirect effects of SES on bilingual vocabulary acquisition, specifically, effects of English language exposure, relates to the highly important issue of how to narrow SES‐linked differences in language acquisition. Here, we appeal to an important distinction raised by Hoff (2006), which is whether SES confers advantages in terms of language learning ability or language experience. There is considerable evidence to suggest that SES modifies language learning experience in young children in multiple ways (e.g., Hoff, 2003). In particular, children from higher SES families enjoy a richer conversational world in terms of the amount of speech they hear (Hart & Risley, 1995). They also profit from a home literacy environment that is conducive to language learning (Korat et al., 2013), receive a higher quality of language input (Huttenlocher et al., 2007), as well as increased exposure to conversational routines and opportunities for learning sociopragmatic skills (Hirsh‐Pasek et al., 2015). This study corroborates effects of SES on language experience, specifically, via community language exposure, which in turn appears to increase community language uptake.</p> <p>A limitation of our study is that our measure of community language input was operationalized by the proportion of time spent with a caregiver who spoke English to the child. This is a composite measure of different facets of exposure that include measures of quantity and quality. Future research could explore how these different components of English exposure compare in their effects on children's vocabulary size and which specific components prove to be causal antecedents of vocabulary size in multilingual settings. In addition, our study measured SES via household income and parental education levels. While these are widely‐used proxies for SES, there are likely additional factors that contribute to increased English exposure beyond these variables. For example, parents may place value on English due to its social and cultural "capital" both at a domestic and international level, which may contribute to the link between SES, English exposure, and English vocabulary. Further studies could explore the effects of family attitudes toward the community language, independent of SES, on English exposure and English vocabulary.</p> <p>Our findings bear on the question of how to assess bilingual vocabulary and furthermore, how to interpret vocabulary size estimates in multilingual societies. In particular, it remains a matter of some debate whether single‐language instruments can be meaningfully used to assess bilingual children. Some researchers have argued for the importance of multilingual assessment in cases of clinical concern (e.g., Paradis, 2010). Others have emphasized the practical value of having a community language estimate available to bilingual children with varying home languages within a society (e.g., Floccia et al., 2018). While both approaches hold merit, we propose four reasons that may favor the use of culturally adapted single language assessment in the community language in settings such as Singapore, assuming that these assessments are calibrated to variation in community language exposure.</p> <p>First, as others have acknowledged, in many bilingual communities, the "gravitational pull" of the community language can be strong and the heritage language often becomes less dominant as children mature (Gathercole & Thomas, 2009; Hoff, 2018; Pearson, 2007). The process of heritage language loss in bilingual children has tended to accelerate over time in many multilingual societies and thus, community language proficiency has become increasingly relevant for bilingual families. Singapore is no exception to this trend even as a multilingual nation, reflected in recent government campaigns to encourage heritage language use, such as the recent "Speak Mandarin" campaign. Therefore, in multilingual societies where one community language is clearly dominant, estimates of community language acquisition may reveal meaningful variation in children's language trajectories. Additionally, in such societies, such as Singapore, it remains an inescapable reality that delays in English vocabulary development can have real consequences for school readiness and educational achievement. For this reason, estimates of vocabulary size for the community language may be of direct benefit to the child in evaluating preparedness of the education system that awaits them.</p> <p>Second, as mentioned in the Introduction, concepts represented in the community language may overlap across different groups of bilingual learners to a greater extent than those represented across heritage languages. For example, in Singapore, different groups of heritage language learners may overlap in the words to which they are exposed in English given the widespread communal use of English in Singapore. In contrast, concepts introduced at home to heritage language users of Malay may differ from those of heritage language speakers of Tamil or Mandarin, with particular words in the heritage language perhaps having greater cultural relevance. This was observed in pilot studies where there was greater uniformity across participants in words acquired in English than in words acquired in heritage languages across language communities. We acknowledge that overlap 'differentials' may vary across semantic categories. For example, there may be less cross‐language overlap in food items and kinship terms than in words for animals or body parts. In general, however, the development of a community language tool may capitalize on greater referential constancy across different bilingual learners in English more so than heritage language inventories, which may contain more culturally specific items.</p> <p>Third, prior norms established for bilingual infants have demonstrated that the number of heritage‐language words acquired varies depending on the similarity between the heritage language and the community language (Floccia et al., 2018). However, in Floccia's study, community language (English) vocabulary size estimates did not differ on account of cross‐linguistic similarity, suggesting that community language vocabulary may be more robust against variation in the specific heritage languages being learned. As further articulated by Floccia et al. (2018), heritage language adaptations for the CDI are frequently not available for each language and often, only a community language version exists. It is therefore often for practical reasons that community language CDIs are used. Given the findings of our study, regression equations that incorporate the degree of English exposure, both independently and in relation to age, when predicting vocabulary percentiles allow for calibrated norms in multilingual children with some community language exposure.</p> <p>The fourth point derives from our experience in piloting adaptations of the CDI for Singapore and from administering the CDI over a multi‐year period in Singapore. The validity of parental reports is crucially dependent on parental knowledge of the child's language knowledge. As noted in the Introduction, tracking knowledge of heritage languages can be more challenging for several reasons. First, parents in Singapore are generally very vigilant with respect to their child's English proficiency, because of the widespread use of English as a common community language. Second, it is not uncommon for multilingual children to demonstrate a "productive preference" for the community language over heritage languages. Often, parents' questions or conversational bids in the heritage language are responded to in the community language, both in Singapore and in other multilingual settings (see Hoff, 2018). For this reason, parents may have more tangible positive evidence for their child's language knowledge in English than in the heritage language, because of the fact that more concepts are expressed in English than in the heritage language by children. In pilot studies, several parents did comment that they did not know what their child knew in Mandarin or Malay because their child only spoke to them in English, even though they reportedy understood the heritage language. Parents generally expressed greater confidence in completing the English CDI than heritage language CDIs, even if they themselves were providing heritage language input. Future studies could compare community language and heritage language CDI scores with direct measures of infant language knowledge, such as the Computerized Comprehension Task developed by Friend and Keplinger (2003) to determine whether cross‐instrument agreement is higher in community languages versus heritage languages. Finally, we qualify these arguments by stating that we do not believe that community language estimates are unequivocally a better choice than heritage language estimates across all multilingual societies. The utility of each approach depends on the desired language knowledge that investigators seek to estimate, the context within which vocabulary data are collected, and the accuracy with which data in each language can be collected by caregivers.</p> <p>Finally, we address limitations and constraints of our study. Our study examined predictors and norms for English vocabulary within Singapore, one example of a multilingual society. We do not presume that these norms or predictors apply to all multilingual societies. Indeed, the determinants of vocabulary acquisition have been shown to differ within multilingual and monolingual societies (e.g., Floccia et al., 2018; Hamilton et al., 2000). While there have been very few large‐scale norms developed for the CDI, some of our findings have been reliably observed in prior studies in both monolingual and multilingual societies. For example, both SES and the amount of exposure to English have been shown to predict CDI scores in prior studies (Floccia et al., 2018; Hoff, 2006; Place & Hoff, 2011). However, it remains to be seen whether the precise interactions between these factors observed here generalize to other multilingual societies. As noted earlier, in some societies (e.g., the United Kingdom), multilingualism aggregates within higher SES levels (Floccia et al., 2018), whereas in some communities within the United States, it traditionally aggregates within lower SES levels (Hoff, 2013). The relations between SES and community language exposure are therefore likely variable across societies and the nature of this relation may determine the extent to which the present findings generalize to other multilingual societies. In addition, Singapore is a multilingual society that places a high premium on English, which is associated with greater social prestige than heritage languages. It is possible that variation in how community and heritage languages are perceived in multilingual societies modifies patterns of language uptake in young learners.</p> <hd id="AN0154716031-28">SUMMARY AND CONCLUSIONS</hd> <p>The goal of this study was to develop vocabulary norms and identify predictors for multilingual children in a multilingual society. Results point to important influences of SES, gender, age, English language exposure as well as interactions between English exposure with age and SES on English vocabulary size. We provide norms for English exposure that are calibrated to effects of predictors identified in regression models. Our findings add to a growing set of studies attempting to capture and define underlying heterogeneity in language experience endemic to multilingual populations.</p> <hd id="AN0154716031-29">ACKNOWLEDGMENTS</hd> <p>This study was funded by the Abbott Nutrition Research & Development Asia‐Pacific Center. We thank Twila Tardif, Philip Dale, and Thow Mei Kuen for helpful discussions on adaptations of the MCDI for Singapore.</p> <p>GRAPH: Supplementary Material</p> <p>GRAPH: Supplementary Material</p> <ref id="AN0154716031-30"> <title> Footnotes </title> <blist> <bibl id="bib1" idref="ref2" type="bt">1</bibl> <bibtext> Leher Singh and QiQi Cheng contributed equally to this work.</bibtext> </blist> </ref> <ref id="AN0154716031-31"> <title> REFERENCES </title> <blist> <bibtext> Arriaga, R. I., Fenson, L., Cronan, T., & Pethick, S. J. (1998). Scores on the MacArthur Communicative Development Inventory of children from low and middle‐income families. Applied Psycholinguistics, 19, 209 – 223. https://doi.org/10.1017/s0142716400010043</bibtext> </blist> <blist> <bibl id="bib2" idref="ref4" type="bt">2</bibl> <bibtext> Barbosa, P. G., Jiang, Z., & Nicoladis, E. (2017). The role of working and short‐term memory in predicting receptive vocabulary in monolingual and sequential bilingual children. International Journal of Bilingual Education and Bilingualism, 22, 801 – 817. https://doi.org/10.1080/13670050.2017.1314445</bibtext> </blist> <blist> <bibl id="bib3" idref="ref1" type="bt">3</bibl> <bibtext> Bosch, L., & Sebastián‐Gallés, N. (1997). Native‐language recognition abilities in 4‐month‐old infants from monolingual and bilingual environments. Cognition, 65, 33 – 69. https://doi.org/10.1016/S0010‐0277(97)00040‐1</bibtext> </blist> <blist> <bibl id="bib4" type="bt">4</bibl> <bibtext> Can, D. D., Ginsburg‐Block, M., Golinkoff, R. M., & Hirsh‐Pasek, K. (2013). A long‐term predictive validity study: Can the CDI Short Form be used to predict language and early literacy skills four years later? Journal of Child Language, 40, 821 – 835. https://doi.org/10.1017/S030500091200030X</bibtext> </blist> <blist> <bibl id="bib5" type="bt">5</bibl> <bibtext> Cattani, A., Abbot‐Smith, K., Farag, R., Krott, A., Arreckx, F., Dennis, I., & Floccia, C. (2014). How much exposure to English is necessary for a bilingual toddler to perform like a monolingual peer in language tests? International Journal of Language and Communication Disorders, 49, 649 – 671. https://doi.org/10.1111/1460‐6984.12082</bibtext> </blist> <blist> <bibl id="bib6" type="bt">6</bibl> <bibtext> Cohen, J. (1960). A coefficient of agreement for nominal scales. Educational and Psychological Measurement, 20, 37 – 46. https://doi.org/10.1177/001316446002000104</bibtext> </blist> <blist> <bibl id="bib7" type="bt">7</bibl> <bibtext> Core, C., Hoff, E., Rumiche, R., & Señor, M. (2013). Total and conceptual vocabulary in Spanish‐English bilinguals from 22 to 30 months: Implications for assessment. Journal of Speech, Language, and Hearing Research, 56, 1637 – 1649. https://doi.org/10.1044/1092‐4388(2013/11‐0044)</bibtext> </blist> <blist> <bibl id="bib8" type="bt">8</bibl> <bibtext> Dale, P. S., & Goodman, J. C. (2005). Commonality and individual differences in vocabulary growth. In M. Tomasello & D. I. Slobin (Eds.), Beyond nature‐nurture: Essays in honor of Elizabeth Bates (pp. 41 – 78). Lawrence Erlbaum Associates Publishers.</bibtext> </blist> <blist> <bibl id="bib9" type="bt">9</bibl> <bibtext> De Houwer, A. (2007). Parental language input patterns and children's bilingual use. Applied Psycholinguistics, 28, 411 – 424. https://doi.org/10.1017/S0142716407070221</bibtext> </blist> <blist> <bibtext> De Houwer, A. (2009). Bilingual first language acquisition. Multilingual Matters. https://doi.org/10.21832/9781847691507</bibtext> </blist> <blist> <bibtext> De Houwer, A. (2019). Equitable evaluation of bilingual children's language knowledge using the CDI: It really matters who you ask. Journal of Monolingual and Bilingual Speech, 1, 32 – 54. https://doi.org/10.1558/jmbs.11184</bibtext> </blist> <blist> <bibtext> DeAnda, S., Arias‐Trejo, N., Poulin‐Dubois, D., Zesiger, P., & Friend, M. (2016). Minimal second language exposure, SES, and early word comprehension: New evidence from a direct assessment. Bilingualism: Language and Cognition, 19, 162 – 180. https://doi.org/10.1017/S1366728914000820</bibtext> </blist> <blist> <bibtext> Feldman, H. M., Dale, P. S., Campbell, T. F., Colborn, D. K., Kurs‐Lasky, M., Rockette, H. E., & Paradise, J. L. (2005). Concurrent and predictive validity of parent reports of child language at ages 2 and 3 years. Child Development, 76, 856 – 868. https://doi.org/10.1111/j.1467‐8624.2005.00882.x</bibtext> </blist> <blist> <bibtext> Fenson, L., Marchman, V. A., Thal, D. J., Dale, P. S., Reznick, J. S., & Bates, E. (2007). MacArthur‐Bates Communicative Development Inventories: User's guide and technical manual (2nd ed.). Brookes.</bibtext> </blist> <blist> <bibtext> Fenson, L., Pethick, S., Renda, C., Cox, J. L., Dale, P. S., & Reznick, J. S. (2000). Short‐form versions of the MacArthur Communicative Development Inventories. Applied Psycholinguistics, 21, 95 – 116. https://doi.org/10.1017/S0142716400001053</bibtext> </blist> <blist> <bibtext> Fernald, A., & Marchman, V. A. (2012). Individual differences in lexical processing at 18 months predict vocabulary growth in typically developing and late‐talking toddlers. Child Development, 83, 203 – 222. https://doi.org/10.1111/j.1467‐8624.2011.01692.x</bibtext> </blist> <blist> <bibtext> Fernald, A., Marchman, V. A., & Weisleder, A. (2013). SES differences in language processing skill and vocabulary are evident at 18 months. Developmental Science, 16, 234 – 248. https://doi.org/10.1111/desc.12019</bibtext> </blist> <blist> <bibtext> Floccia, C., Sambrook, T., Delle Luche, C., Kwok, R., Goslin, J., White, L., Cattani, A., Sullivan, E., Abbot‐Smith, K., Krott, A., Mills, D., Rowland, C., Gervain, J., & Plunkett, K. (2018). Vocabulary of 2‐year‐olds learning English and an additional language: Norms and effects of linguistic distance. Monographs for the Society for Research in Child Development, 83 (1), 1 – 135. https://doi.org/10.1111/mono.v83.1</bibtext> </blist> <blist> <bibtext> Friend, M., DeAnda, S., Arias‐Trejo, N., Poulin‐Dubois, D., & Zesiger, P. (2017). Developmental changes in maternal education and minimal exposure effects on vocabulary in English‐ and Spanish‐learning toddlers. Journal of Experimental Child Psychology, 164, 250 – 259. https://doi.org/10.1016/j.jecp.2017.07.003</bibtext> </blist> <blist> <bibtext> Friend, M., & Keplinger, M. (2003). An infant‐based assessment of early lexicon acquisition. Behavior Research Methods, Instruments, & Computers, 35, 302 – 309. https://doi.org/10.3758/BF03202556</bibtext> </blist> <blist> <bibtext> Gathercole, V. C. M., & Thomas, E. M. (2009). Bilingual first‐language development: Dominant language takeover, threatened minority language take‐up. Bilingualism: Language and Cognition, 12, 213 – 237. https://doi.org/10.1017/S1366728909004015</bibtext> </blist> <blist> <bibtext> Hamilton, A., Plunkett, K., & Schafer, G. (2000). Infant vocabulary development assessed with a British Communicative Development Inventory. Journal of Child Language, 27, 689 – 705. https://doi.org/10.1017/S0305000900004414</bibtext> </blist> <blist> <bibtext> Hao, M., Shu, H., Xing, A., & Li, P. (2008). Early vocabulary inventory for Mandarin Chinese. Behavior Research Methods, 40, 728 – 733. https://doi.org/10.3758/BRM.40.3.728</bibtext> </blist> <blist> <bibtext> Harrell, F. E. (2021). Hmisc: Harrell Miscellaneous. R package version. https://CRAN.R‐project.org/package=Hmisc</bibtext> </blist> <blist> <bibtext> Hart, B., & Risley, T. R. (1995). Meaningful differences in the everyday experience of young American children. Paul H Brookes Publishing.</bibtext> </blist> <blist> <bibtext> Hernandez, D. J. (2004). Demographic change and the life circumstances of immigrant families. The Future of Children, 14, 17 – 47. https://doi.org/10.2307/1602792</bibtext> </blist> <blist> <bibtext> Hirsh‐Pasek, K., Adamson, L. B., Bakeman, R., Owen, M. T., Golinkoff, R. M., Pace, A., Yust, P. K. S., & Suma, K. (2015). The contribution of early communication quality to low‐income children's language success. Psychological Science, 26, 1071 – 1083. https://doi.org/10.1177/0956797615581493</bibtext> </blist> <blist> <bibtext> Hoff, E. (2003). The specificity of environmental influence: Socioeconomic status affects early vocabulary development via maternal speech. Child Development, 74, 1368 – 1378. https://doi.org/10.1111/1467‐8624.00612</bibtext> </blist> <blist> <bibtext> Hoff, E. (2006). How social contexts support and shape language development. Developmental Review, 26, 55 – 88. https://doi.org/10.1016/j.dr.2005.11.002</bibtext> </blist> <blist> <bibtext> Hoff, E. (2013). Interpreting the early language trajectories of children from low‐SES and language minority homes: Implications for closing achievement gaps. Developmental Psychology, 49, 4 – 14. https://doi.org/10.1037/a0027238</bibtext> </blist> <blist> <bibtext> Hoff, E. (2018). Lessons from the study of input effects on bilingual development. International Journal of Bilingualism, 24, 82 – 88. https://doi.org/10.1177/1367006918768370</bibtext> </blist> <blist> <bibtext> Hoff, E., & Core, C. (2015). What clinicians need to know about bilingual development. Seminars in Speech and Language, 36, 89 – 99. https://doi.org/10.1055/s‐0035‐1549104</bibtext> </blist> <blist> <bibtext> Hoff, E., Core, C., Place, S., Rumiche, R., Señor, M., & Parra, M. (2012). Dual language exposure and early bilingual development. Journal of Child Language, 39, 1 – 27. https://doi.org/10.1017/S0305000910000759</bibtext> </blist> <blist> <bibtext> Hoff, E., & Naigles, L. (2002). How children use input to acquire a lexicon. Child Development, 73, 418 – 433. https://doi.org/10.1111/1467‐8624.00415</bibtext> </blist> <blist> <bibtext> Hurtado, N., Grüter, T., Marchman, V., & Fernald, A. (2014). Relative language exposure, processing efficiency and vocabulary in Spanish‐English bilingual toddlers. Bilingualism: Language and Cognition, 17, 189 – 202. https://doi.org/10.1017/S136672891300014X</bibtext> </blist> <blist> <bibtext> Huttenlocher, J., Haight, W., Bryk, A., Seltzer, M., & Lyons, T. (1991). Early vocabulary growth: Relation to language input and gender. Developmental Psychology, 27, 236 – 248. https://doi.org/10.1037/0012‐1649.27.2.236</bibtext> </blist> <blist> <bibtext> Huttenlocher, J., Vasilyeva, M., Waterfall, H. R., Vevea, J. L., & Hedges, L. V. (2007). The varieties of speech to young children. Developmental Psychology, 43, 1062 – 1083. https://doi.org/10.1037/0012‐1649.43.5.1062</bibtext> </blist> <blist> <bibtext> Jackson‐Maldonado, D., Thal, D., Marchman, V., Bates, E., & Gutierrez‐Clellen, V. (1993). Early lexical development in Spanish‐speaking infants and toddlers. Journal of Child Language, 20, 523 – 549. https://doi.org/10.1017/S030500090000846</bibtext> </blist> <blist> <bibtext> Korat, O., Arafat, S. H., Aram, D., & Klein, P. (2013). Book reading mediation, SES, home literacy environment, and children's literacy: Evidence from Arabic‐speaking families. First Language, 33, 132 – 154. https://doi.org/10.1177/0142723712455283</bibtext> </blist> <blist> <bibtext> Lauro, J., Core, C., & Hoff, E. (2020). Explaining individual differences in trajectories of simultaneous bilingual development: Contributions of child and environmental factors. Child Development, 91, 2063 – 2082. https://doi.org/10.1111/cdev.13409</bibtext> </blist> <blist> <bibtext> Law, J., & Roy, P. (2008). Parental report of infant language skills: A review of the development and application of the communicative development inventories. Child and Adolescent Mental Health, 13, 198 – 206. https://doi.org/10.1111/j.1475‐3588.2008.00503.x</bibtext> </blist> <blist> <bibtext> Legacy, J., Zesiger, P., Friend, M., & Poulin‐Dubois, D. (2018). Vocabulary size and speed of word recognition in very young French‐English bilinguals: A longitudinal study. Bilingualism: Language and Cognition, 21, 137 – 149. https://doi.org/10.1017/s1366728916000833</bibtext> </blist> <blist> <bibtext> Marchman, V. A., & Fernald, A. (2008). Speed of word recognition and vocabulary knowledge in infancy predict cognitive and language outcomes in later childhood. Developmental Science, 11, F9 – F16. https://doi.org/10.1111/j.1467‐7687.2008.00671.x</bibtext> </blist> <blist> <bibtext> Marchman, V. A., Fernald, A., & Hurtado, N. (2010). How vocabulary size in two languages relates to efficiency in spoken word recognition by young Spanish‐English bilinguals. Journal of Child Language, 37, 817 – 840. https://doi.org/10.1017/S0305000909990055</bibtext> </blist> <blist> <bibtext> Marchman, V. A., Martínez, L. Z., Hurtado, N., Grüter, T., & Fernald, A. (2017). Caregiver talk to young Spanish‐English bilinguals: Comparing direct observation and parent‐report measures of dual‐language exposure. Developmental Science, 20, e12425. https://doi.org/10.1111/desc.12425</bibtext> </blist> <blist> <bibtext> Marchman, V. A., Martínez‐Sussmann, C., & Dale, P. S. (2004). The language‐specific nature of grammatical development: Evidence from bilingual language learners. Developmental Science, 7, 212 – 224. https://doi.org/10.1111/j.1467‐7687.2004.00340.x</bibtext> </blist> <blist> <bibtext> Marquardt, D. W. (1980). Comment. Journal of the American Statistical Association, 75, 87 – 91. https://doi.org/10.1080/01621459.1980.10477430</bibtext> </blist> <blist> <bibtext> Nicoladis, E., & Jiang, Z. (2018). Language and cognitive predictors of lexical selection in storytelling for monolingual and sequential bilingual children. Journal of Cognition and Development, 19, 413 – 430. https://doi.org/10.1080/15248372.2018.1483370</bibtext> </blist> <blist> <bibtext> Oller, D. K., & Eilers, R. E. (2002). Language and literacy in bilingual children. Multilingual Matters. https://doi.org/10.21832/9781853595721</bibtext> </blist> <blist> <bibtext> Oosterhuis, H. E. M., van der Ark, L. A., & Sijtsma, K. (2017). Standard errors and confidence intervals of norm statistics for educational and psychological tests. Psychometrika, 82, 559 – 588. https://doi.org/10.1007/s11336‐016‐9535‐8</bibtext> </blist> <blist> <bibtext> Paradis, J. (2010). The interface between bilingual development and specific language impairment. Applied Psycholinguistics, 31, 227 – 252. https://doi.org/10.1017/S0142716409990373</bibtext> </blist> <blist> <bibtext> Parra, M., Hoff, E., & Core, C. (2011). Relations among language exposure, phonological memory, and language development in Spanish‐English bilingually developing 2‐year‐olds. Journal of Experimental Child Psychology, 108, 113 – 125. https://doi.org/10.1016/j.jecp.2010.07.011</bibtext> </blist> <blist> <bibtext> Pearson, B. Z. (2007). Social factors in childhood bilingualism in the United States. Applied Psycholinguistics, 28, 399 – 410. https://doi.org/10.1017/S014271640707021X</bibtext> </blist> <blist> <bibtext> Pearson, B. Z., Fernández, S. C., Lewedeg, V., & Oller, D. K. (1997). The relation of input factors to lexical learning by bilingual infants. Applied Psycholinguistics, 18, 41 – 58. https://doi.org/10.1017/s0142716400009863</bibtext> </blist> <blist> <bibtext> Pearson, B. Z., Fernández, S. C., & Oller, D. K. (1993). Lexical development in bilingual infants and toddlers: Comparison to monolingual norms. Language Learning, 43, 93 – 120. https://doi.org/10.1111/j.1467‐1770.1993.tb00174.x</bibtext> </blist> <blist> <bibtext> Place, S., & Hoff, E. (2011). Properties of dual language exposure that influence 2‐year‐olds' bilingual proficiency. Child Development, 82, 1834 – 1849. https://doi.org/10.1111/j.1467‐8624.2011.01660.x</bibtext> </blist> <blist> <bibtext> Scheele, A. F., Leseman, P. P. M., & Mayo, A. Y. (2010). The home language environment of monolingual and bilingual children and their language proficiency. Applied Psycholinguistics, 31, 117 – 140. https://doi.org/10.1017/S0142716409990191</bibtext> </blist> <blist> <bibtext> Tan, S. H. (2010). Multilingual infant vocabulary development. In M. Cruz‐Ferreira (Ed.), Multilingual norms (pp. 113 – 140). Peter Lang.</bibtext> </blist> <blist> <bibtext> Tardif, T., Fletcher, P., Liang, W., & Kaciroti, N. (2009). Early vocabulary development in Mandarin (Putonghua) and Cantonese. Journal of Child Language, 36, 1115 – 1144. https://doi.org/10.1017/S0305000908009185</bibtext> </blist> <blist> <bibtext> Thal, D. J., Reilly, J., Seibert, L., Jeffries, R., & Fenson, J. (2004). Language development in children at risk for language impairment: Cross‐population comparisons. Brain and Language, 88, 167 – 179. https://doi.org/10.1016/S0093‐934X(03)00096‐8</bibtext> </blist> <blist> <bibtext> Thordardottir, E. (2011). The relationship between bilingual exposure and vocabulary development. International Journal of Bilingualism, 15, 426 – 445. https://doi.org/10.1177/1367006911403202</bibtext> </blist> <blist> <bibtext> Thow, M. K. (2002). Norm for Malay Adapted Communicative Development Inventory—Word and gestures for 12 and 18 months old Malay infants (Bachelor of Speech Science thesis). Universiti Kebangsaan Malaysia. <ulink href="http://mash.org.my/wp‐content/uploads/Speech%202002%20Thow%20Mei%20Kuen.pdf">http://mash.org.my/wp‐content/uploads/Speech%202002%20Thow%20Mei%20Kuen.pdf</ulink></bibtext> </blist> <blist> <bibtext> Watkins, M. W., & Pacheco, M. (2000). Interobserver agreement in behavioral research: Importance and calculation. Journal of Behavioral Education, 10, 205 – 212. https://doi.org/10.1023/A:1012295615144</bibtext> </blist> <blist> <bibtext> Weisleder, A., & Fernald, A. (2013). Talking to children matters: Early language experience strengthens processing and builds vocabulary. Psychological Science, 24, 2143 – 2152. https://doi.org/10.1177/0956797613488145</bibtext> </blist> <blist> <bibtext> Winsler, A., Burchinal, M. R., Tien, H. C., Peisner‐Feinberg, E., Espinosa, L., Castro, D. C., LaForett, D. R., & Kim, Y. K. & De Feyter, J. (2014). Early development among dual language learners: The roles of language use at home, maternal immigration, country of origin, and socio‐demographic variables. Early Childhood Research Quarterly, 29, 750 – 764. https://doi.org/10.1016/j.ecresq.2014.02.0</bibtext> </blist> </ref> <aug> <p>By Leher Singh; QiQi Cheng; Seok Hui Tan; Agnes Tan and Yen Ling Low</p> <p>Reported by Author; Author; Author; Author; Author</p> </aug> <nolink nlid="nl1" bibid="bib105" firstref="ref5"></nolink>
Header DbId: eric
DbLabel: ERIC
An: EJ1327059
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Language Acquisition in a Multilingual Society: English Vocabulary Norms and Predictors in Singaporean Children
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Singh%2C+Leher%22">Singh, Leher</searchLink> (ORCID <externalLink term="https://orcid.org/0000-0001-9423-4956">0000-0001-9423-4956</externalLink>)<br /><searchLink fieldCode="AR" term="%22Cheng%2C+QiQi%22">Cheng, QiQi</searchLink> (ORCID <externalLink term="https://orcid.org/0000-0002-2295-330X">0000-0002-2295-330X</externalLink>)<br /><searchLink fieldCode="AR" term="%22Tan%2C+Seok+Hui%22">Tan, Seok Hui</searchLink><br /><searchLink fieldCode="AR" term="%22Tan%2C+Agnes%22">Tan, Agnes</searchLink><br /><searchLink fieldCode="AR" term="%22Low%2C+Yen+Ling%22">Low, Yen Ling</searchLink>
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="SO" term="%22Child+Development%22"><i>Child Development</i></searchLink>. Jan-Feb 2022 93(1):288-305.
– Name: Avail
  Label: Availability
  Group: Avail
  Data: Wiley. Available from: John Wiley & Sons, Inc. 111 River Street, Hoboken, NJ 07030. Tel: 800-835-6770; e-mail: cs-journals@wiley.com; Web site: https://www.wiley.com/en-us
– Name: PeerReviewed
  Label: Peer Reviewed
  Group: SrcInfo
  Data: Y
– Name: Pages
  Label: Page Count
  Group: Src
  Data: 18
– Name: DatePubCY
  Label: Publication Date
  Group: Date
  Data: 2022
– Name: TypeDocument
  Label: Document Type
  Group: TypDoc
  Data: Journal Articles<br />Reports - Research
– Name: Subject
  Label: Descriptors
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Infants%22">Infants</searchLink><br /><searchLink fieldCode="DE" term="%22Child+Development%22">Child Development</searchLink><br /><searchLink fieldCode="DE" term="%22Language+Acquisition%22">Language Acquisition</searchLink><br /><searchLink fieldCode="DE" term="%22Socioeconomic+Status%22">Socioeconomic Status</searchLink><br /><searchLink fieldCode="DE" term="%22Vocabulary+Development%22">Vocabulary Development</searchLink><br /><searchLink fieldCode="DE" term="%22Predictor+Variables%22">Predictor Variables</searchLink><br /><searchLink fieldCode="DE" term="%22Multilingualism%22">Multilingualism</searchLink><br /><searchLink fieldCode="DE" term="%22Language+Skills%22">Language Skills</searchLink><br /><searchLink fieldCode="DE" term="%22Measures+%28Individuals%29%22">Measures (Individuals)</searchLink><br /><searchLink fieldCode="DE" term="%22English%22">English</searchLink><br /><searchLink fieldCode="DE" term="%22Mandarin+Chinese%22">Mandarin Chinese</searchLink><br /><searchLink fieldCode="DE" term="%22Indonesian+Languages%22">Indonesian Languages</searchLink><br /><searchLink fieldCode="DE" term="%22Foreign+Countries%22">Foreign Countries</searchLink>
– Name: Subject
  Label: Geographic Terms
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Singapore%22">Singapore</searchLink>
– Name: SubjectThesaurus
  Label: Assessment and Survey Identifiers
  Group: Su
  Data: <searchLink fieldCode="SU" term="%22MacArthur+Communicative+Development+Inventory%22">MacArthur Communicative Development Inventory</searchLink>
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1111/cdev.13676
– Name: ISSN
  Label: ISSN
  Group: ISSN
  Data: 0009-3920
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: In this study, infant vocabulary development was tracked in a multilingual society (Singapore) within a socioeconomically diverse sample. The sample comprised 1316 infants from 17.4 to 27.7 months (669 females, 647 males; 88% Chinese race, 4% Malay, 4% Indian, and 0.004% mixed-race [4% declined to provide race information]). Children varied in English language exposure and socioeconomic status. Analyses focused on identifying demographic predictors of English vocabulary size in multilingually exposed infants. Adaptations of the Macarthur-Bates Communicative Development Inventory for English, Mandarin, and Malay are provided as well as English vocabulary norms that account for variation in English exposure. This manuscript reports the first set of English language norms--calibrated to English exposure--for multilingual infants in a non-Western setting.
– Name: AbstractInfo
  Label: Abstractor
  Group: Ab
  Data: As Provided
– Name: DateEntry
  Label: Entry Date
  Group: Date
  Data: 2022
– Name: AN
  Label: Accession Number
  Group: ID
  Data: EJ1327059
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=eric&AN=EJ1327059
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1111/cdev.13676
    Languages:
      – Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 18
        StartPage: 288
    Subjects:
      – SubjectFull: Infants
        Type: general
      – SubjectFull: Child Development
        Type: general
      – SubjectFull: Language Acquisition
        Type: general
      – SubjectFull: Socioeconomic Status
        Type: general
      – SubjectFull: Vocabulary Development
        Type: general
      – SubjectFull: Predictor Variables
        Type: general
      – SubjectFull: Multilingualism
        Type: general
      – SubjectFull: Language Skills
        Type: general
      – SubjectFull: Measures (Individuals)
        Type: general
      – SubjectFull: English
        Type: general
      – SubjectFull: Mandarin Chinese
        Type: general
      – SubjectFull: Indonesian Languages
        Type: general
      – SubjectFull: Foreign Countries
        Type: general
      – SubjectFull: Singapore
        Type: general
      – SubjectFull: MacArthur Communicative Development Inventory
        Type: general
    Titles:
      – TitleFull: Language Acquisition in a Multilingual Society: English Vocabulary Norms and Predictors in Singaporean Children
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Singh, Leher
      – PersonEntity:
          Name:
            NameFull: Cheng, QiQi
      – PersonEntity:
          Name:
            NameFull: Tan, Seok Hui
      – PersonEntity:
          Name:
            NameFull: Tan, Agnes
      – PersonEntity:
          Name:
            NameFull: Low, Yen Ling
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 01
              Type: published
              Y: 2022
          Identifiers:
            – Type: issn-print
              Value: 0009-3920
          Numbering:
            – Type: volume
              Value: 93
            – Type: issue
              Value: 1
          Titles:
            – TitleFull: Child Development
              Type: main
ResultId 1