Chapter 35. Quantifying Language Experience in HL Development

Chapter 35. Quantifying language experience in HL development

Sharon Unsworth

35.1Introduction

Bilingual children’s heritage language (HL) experience may vary in many different ways. For example, some children hear the HL from both parents whereas others from one parent only, some children grow up in a wider HL-speaking community whereas other families live in relative isolation, some children mostly hear input from attrited HL speakers whereas others do not, and some children have access to schooling in the HL, whereas others do not. To understand the impact of theseand other different experiences on children’s HL outcomes, researchers typically collate and quantity specific aspects of children’s language input, transforming or reducing them into other, more general variables, such as language richness as a measure of input quality and amount of language exposure as a measure of input quantity. As Montrul (2008: 270) notes, “a problem with using input as a key variable to explain much of language acquisition is that it is difficult to operationalize”.

This chapter presents an overview of the most frequently used method of operationalizing language input in bilingual language acquisition research, namely the parental questionnaire. It is organised as follows: after providing a brief summary of the main experiential factors known to affect bilingual/HL development (§35.2), I outline some conceptual and practical issues surrounding parental questionnaires as a means of quantifying bilingual language experience (§35.3), before reviewing a number of questionnaires used in recent studies in more detail (§35.4). The focus is on bilingual/HL development in childhood and as such, questionnaires designed for use with adult HL speakers (see e.g., work by Naomi Nagy) are not included.

35.2The effect of specific properties on bilingual experience on (heritage) language development

In recent years, a wealth of studies have demonstrated the impact of various experiential variables on bilingual children’s (rate of) language development. Most studies on this topic, however, focus on bilingual children’s development in the societal (i.e., majority) rather than (or at least in addition to) the HL (see Unsworth 2016a for a detailed review). In this section, I highlight the main findings relevant for HL development in order to contextualise the questionnaires discussed below and highlight potential avenues for further research.

Various recent studies have shown that HL children’s rate of acquisition is related to the amount of HL input to which they are exposed. This has been observed across different linguistic properties (e.g., verbal and nominal morphology, vocabulary) and for different populations (e.g., Spanish-English, Welsh-English) (e.g., Hoff et al. 2012; Gathercole & Thomas 2009). The positive contribution of parents’ HL at home to children’s development in that language has also been observed; continued parental HL input seems especially important once children start schooling in the majority language (e.g., Dixon, Zhao, Quiroz, & Shin 2012; Duursma et al. 2007), although it may not necessarily guarantee children’s continuing HL development (Sheng, Lu, & Kan 2011).

The amount of societal language input at home has been found to have a greater, negative impact on the HL than it has a positive impact on the acquisition of the societal language (De Houwer 2007; Hammer et al. 2009; Hoff et al. 2014). Furthermore, the effect of parental use of the societal language at home may be modulated by the proficiency of the input-provider (Chondrogianni & Marinis 2011; Goldberg et al. 2008).

Whilst the differential effect of variation in input quantity on the (rate of) acquisition in many domains of language is undisputed, it is also clear that this relationship is non-linear in nature. In other words, threshold effects exist (e.g., Cattani et al.2014; Gutiérrez-Clellen & Kreiter 2003; Thordardottir 2014) and not all linguistic properties are affected equally (Chondrogianni & Marinis 2011; Oller et al. 2007; Unsworth 2014, 2016b; cf. Hoff et al. 2012l; Paradis 2011).

Research on the impact of variation in input quality has for the most part been restricted to studies on societal language development, with the notable exception of the studies by Place and Hoff (2011, 2015) and LaMorgia (2011). For example, it has been shown that input from native speakers is associated with more advanced vocabulary in the societal language, even after controlling for amount of input in that language (Place & Hoff 2011, 2015; see also Driessen et al., 2002). Furthermore, linguistically rich input (e.g., from different sources, including TV and reading) predicts rate of acquisition (e.g., Paradis 2011), and the number of speakers providing input is also a significant predictor of children’s (HL) development (Place & Hoff 2015).

One other important aspect of HL children’s language experience is how much they use the language themselves. Indeed, children’s own HL use has been found to predict language outcomes in a number of studies (Bohman et al. 2010; Montrul 2008; Unsworth 2011, in prep). Furthermore, whilst output and input are clearly related to each other, they may nevertheless relate to outcomes differently (Bohman et al. 2010; Unsworth2015).

To summarise, recent research has shown that variation in input quantity impacts on rate of acquisition in vocabulary and morphosyntax, though not for all age groups, or all contexts (see Unsworth 2016a for an overview). Furthermore, input beyond a certain level does not facilitate further development and more qualitative properties of language input, such as whether it is from native speakers or from different speakers, have also been found to predict variation in children’s outcomes. On the whole, however, most of the research on input effects in child bilinguals is concerned with development in the societal language, with only a handful of studies focussing on children’s HL development.[1] In other words, whilst there are a number of relatively robust findings concerning the impact of experiential factors on bilingual children’s language development, there is certainly scope for further research, especially with respect to language use and the more quality-oriented factors, and with respect to development in the HL. The remainder of this chapter is concerned with the tools one might use to conduct such research.

35.3How to quantify bilingual / HL experience

There are a number of conceptual and practical issues which the researcher needs to consider when deciding how best to quantify bilingual / HL experience.

35.3.1Conceptual issues

From a conceptual point of view, the first step is to decide where experiential variables – whichever of the specific variables discussed above these turn out to be – fit into the research design. Are they independent variables (i.e., predictors) or control variables? For example, if the locus of interest in a given study is whether input quantity or input quality has a greater impact on children’s HL development, experiential variables such as amount of weekly exposure and contact with native speakers could be included as independent variables. On the other hand, if the study’s goal is for example to examine crosslinguistic influence across two groups of bilingual children with different language combinations (e.g., Spanish-Swedish with German-Swedish), experiential variables such as amount and richness of input should be included as control variables. In this latter example, then, bilingual experience is used to ensure that the two bilingual groups are comparable such that any differences between the two on the study’s dependent variable may plausibly be attributed to the independent variable of interest (i.e., the role of the ‘other’ language).

In pinpointing the experiential variables of interest, it is important to bear in mind that many such variables often strongly correlated with each other or with other variables of interest. Perhaps the most clear-cut case of this problem of multicollinearity is the interdependence between age of testing, age of onset and length of exposure. This is a well-known issue in L2 acquisition research (e.g., Muñoz 2008),but it is also relevant to HL acquisition. For example, in HL children, age of onset to the L2 will likely be correlated with the extent of early HL exposure, and hence an effect of (extensive) early HL experience may just as easily be interpreted as an effect of later L2 onset. Similarly, and just as in monolinguals, experiential variables such as (absolute) amount of input may correlate with other external factors, such as SES (Hoff, 2006).

Researchers in bilingual child language research have addressed the problem of multicollinearity in different ways.[2] For example, in a study on the impact of various internal and external factors on the child L2 acquisition of English, Paradis (2011) observed a strong correlation between age of onset and months of exposure (r = -.758), and consequently decided to include only the latter variable in her regression analyses. In the results, months of exposure turned out to be one of the significant predictors for children’s vocabulary and verb morphology scores; Paradis (2011: 230) notes that this is in line with earlier research finding an effect of length of exposure. However, given the multicollinearity with age of onset, we cannot rule out that these results may in fact reflect the effect of this latter variable instead. Blom and Bosma (2016) adopt a different approach: in their study on the acquisition of vocabulary and verb morphology in Frisian-Dutch bilingual children, they regressed length of exposure on age of onset and used the residualised (or decorrelated) values in their subsequent analyses (cf. Wurm Fisicaro 2014 for the problems and pitfalls in using residualisation as a means of addressing multicollinearity).

One final conceptual issue to consider when deciding how to quantify bilingual / HL experience using parental questionnaires is whether relative or absolute measures of input are required. Most of the studies reviewed in section 35.2 correlated relative measures of language experience (e.g., 70% English input, 30% Spanish input) with absolute measures of language development (e.g., standardised language scores). Grüter, Hurtado, Marchman and Fernald (2014) illustrate how this relationship may change considerably when absolute measures of language experience (e.g., number of words heard each day) are used instead. For example, two children may have the same relative exposure to their HL, but the number of words they hear may differ, depending on – amongst other things – the talkativeness of their input-providers, and this in turn will likely impact on their rate of acquisition for certain properties of that language. As Carroll (2016: 4) notes, deriving estimates of language input from how many hours children spend with various interlocutors and the languages these interlocutors speak to children is problematic becausesuch “[t]emporal units are crude measures of exposure and they tell us nothing about input”. Whilst the tone of her conjecture is probably too categorical, Carroll raises an important point, pertinent to the present discussion, namely that individual variation in terms of how much input is heard in a given hour cannot be captured using methods such as parental questionnaires. This is something which should be borne in mind when evaluating the results of studies using parental questionnaire data and the claims being made (see also De Houwer 2014 for discussion of absolute and relative measures).

35.3.2Practical issues

One of the most practical considerations in selecting a parental questionnaire is the resources available, not only with respect to money but also in terms of time, and not only for data collection but also for analysis.

Like anything else, the potential costs of collecting detailed data on children’s HL experience should be weighed up against the benefits in terms of the relative contribution said data will make to the overall research design. In order to make such a cost-benefit analysis, it is important to consider whether more general or specific measures of HL experience are needed, and what the minimum level of detail is that is required to answer the research question. For example, in studies where amount of HL experience is a control variable, a general measure will largely suffice, whereas in studies focussing on the impact of variation in HL input more detailed measures will clearly be required.

Other factors which may contribute to data collection costs are whether the questionnaire can be completed by parents only, or whether a research assistant is needed. In this latter case, it may be necessary for the research assistant to be a member of the relevant HL community in order for data collection to be a success. Whether the questionnaire is completed on paper, online or via an interview (either face to face, via telephone or the internet) will not only contribute to data collection costs, it will also impact on the time and money needed for data analysis. For example, many online survey services deliver data in an analysis-ready, spreadsheet format. Finally, it is of course important to consider whether a questionnaire is available in the target language(s) or whether it will need translating, something which will again impact on overall (time/money) costs.

Another important practical consideration when selecting a parental questionnaire is the characteristics of the target population. This may include factors such as parents’ level of education, literacy and numeracy skills and access to internet, their proximity to you, and whether they will be present during the child’s test session, in which case parent and child data collection can be combined.

35.4 Survey of available parental questionnaires

This section provides an overview of a number of available parental questionnaires.[3]All are freely accessible (either in appendices, online or directly from the authors) and all have been used in published studies. They have each been designed to collect information about children’s bilingual language experience in general, that is, their experience with the L2 as well as the HL. There are two broad types: i) questionnaires which involve (automatic) calculations and usually result in composite scores, and ii) questionnaires which capture raw data only.Table 1 provides information about how each questionnaire is administered, its output, the population for which it was designed, and the language(s) in which it is available.

Table 1. Overview of parental questionnaires measuring bilingual language experience in children

Questionnaire / Administration / (Main) output / Population / Language
ALEQ (Alberta Language Exposure Questionnaire; Paradis 2011) /

Interview with research assistant
Paper

Current language use, incorporating child’s input and output
Language richness scores in both languages
Calculated by hand

Primary school children, but could be used with toddlers and preschoolers
Sequential bilinguals but could be used with simultaneous bilinguals

English

Cattani et al. (2014) /

Parent
Paper
Optional Excel file, available from the authors

% exposure in typical week in last year of child’s life
Calculated by hand or using Excel file

Toddlers, but could be used with older children
Simultaneous and sequential bilinguals

English

Gathercole & Thomas (2007)a /

Interview with research assistant
Similar versions also available for parent to complete
Paper

Current language use by parents and child
Previous language use by parents and child
Parents’ current language skills
Frequency of contact with languages from other sources

Primary school children

English
Welsh
Spanish

Gutiérrez-Clellen & Kreiter (2003) /

Parent
Paper

No. of years’ exposure
Current % exposure at home
No. of hours of reading and other literacy activities in L2
Calculated by hand

Up to 8 years, but could be used with older children
Simultaneous and sequential bilinguals

English

Language Exposure Questionnaire (LEQ) (Hoff et al., 2012)b /

Interview with research assistant
Paper

Current % exposure at home
No. of hours per week of input in each language for range of activities and input-providers
No. of monolingual / bilingual / native input-providers
Calculated by hand

Toddlers / preschoolers

English
Spanish

Language Input Diary (LID)
(De Houwer 2011) /

Parent
Paper

Who is with child and language(s) spoken in 30 min slots for each day in the week

Toddlers
Simultaneous and early sequential bilinguals
Need to hear both languages at home

Mayo & Leseman (2006) /

Parent
Paper

Composite scores for various language/literacy activities (how often, in which language)
Calculated by hand

Toddlers / preschoolers
Simultaneous and sequential bilinguals

Dutch
Parts available in English

PABiQ (Questionnaire for Parents of Bilingual Children; Tuller2015) /

Parent or interview with research assistant
Paper

Current language use, incorporating child’s input and output
Language richness scores in both/all languages
Current language skills in both/all languages
Calculated by hand

Toddlers and preschoolers

English
Dutch
Others available via COST Action ISO804

BiLEC (Bilingual Language Experience Calculator; Unsworth 2013) /

Interview with research assistant
Excel file, available via IRISc

% current exposure
Cumulative length of exposure
% current output
Average quality of exposure (nativelikeness)
No. of different (native/non-native) speakers providing input at home
No. of single language conversational partners
Calculated automatically

Toddlers through 18 years
Simultaneous and sequential bilinguals

English
Dutch

a Various, similar but not identical versions of this questionnaire are available from Ginny Gathercole’s webpage.

bThe LEQ is available directly from Erika Hoff.

c Go to and search for BiLEC.BiLEC was originally dubbed the Utrecht Bilingual Language Exposure Calculator (UBiLEC). Its name has since been changed to better reflect its content (and my current affiliation); its contents are essentially the same, however.

In the remainder of this section, I discuss the relative advantages and disadvantages of the questionnaires given in Table 1. Whilst the questionnaires differ in emphasis, most are designed for use with young (toddler/preschooler) children and use a series of questions about the child’s daily routine to derive compositemeasures of current exposure to and/or use of their two (or more) languages. Most, if not all, have been used in studies exploring bilingual children’s exposure in their two languages and its relation to language development, although these studies rarely explicitly refer to children as HL speakers. The exception to this is Cattani et al.’s (2014) questionnaire, which focuses on early sequential bilinguals’ L2 acquisition; however, given that the % exposure score is relative, their questionnaire could quite easily be adapted to address HL development.

35.4.1Administration

With respect to how the questionnaire is administered, most can be distributed to participating families for completion at home, whereas others (e.g., ALEQ, BiLEC and LEQ at least in their current forms) involve the parent being interviewed by a research assistant, cultural broker or SLP. The Language Input Diary (De Houwer 2011) differs from the others in that it involves parents keeping a record of children’s patterns of language exposure one day a week over a 7-week period. Using a paper booklet, parents note down every 30 minutes who is with the child and which language they are using. By collecting information from Monday in week 1, Tuesday in week 2, etc., the researcher builds up a picture of an “average” week’s language exposure. Parents can be reminded via text messaging, email or Whatsapp when they need to fill in the diary.