Stability and change in young children’s linguistic experience during the COVID-19 pandemic: insight from a citizen-science sample in the United States

Abstract

Early in the COVID-19 pandemic, extensive lockdowns interrupted daily routines, including childcare. We asked whether these interruptions, and the inevitable changes in the people with whom children spent their waking hours, caused changes in the languages that children heard. We retrospectively queried parents of young children (0–4 years) in the US about childcare arrangements and exposure to English and non-English languages at four timepoints from February 2020 to September 2021. Despite discontinuity in childcare arrangements, we found that children’s exposure to English versus other languages remained relatively stable. We also identified demographic variables (child age at pandemic onset, parental proficiency in a non-English language) that consistently predicted exposure to non-English languages. Thus, multilingually-exposed children in this population did not appear to significantly gain or lose the opportunity to hear non-English languages overall. These results provide insight into the experiences of this unique cohort and inform our understanding of how language development can be shaped by complex environmental systems.

1. Introduction

The COVID-19 pandemic created disruptions for families around the world, leading to speculation about how pandemic-related changes in day-to-day life affected children. One notable change was interruptions in childcare due to widespread lockdowns in March 2020. In the US, approximately 40 % of families with children under 5 reported making adjustments to childcare (Patrick et al. 2020). These changes in childcare introduced drastic changes in the people with whom young children spent their waking hours. Adults in children’s everyday environments provide the input from which children learn language, and language skills depend on experiences both at home (Hart and Risley 1995; Hurtado et al. 2008) and in childcare settings (McCartney 1984). Thus, alterations in childcare arrangements could change children’s experience with language and subsequent language development. In the United States, 20 % of individuals 5 and older are reported to speak a language in addition to English (Federal Interagency Forum on Child and Family Statistics 2021). Therefore, if children spent more time at home, they might have had more exposure to languages other than English. Alternatively, if children were exposed to other languages in daycare settings, they could abruptly have lost exposure to these languages. We investigated children’s language experience over the course of the pandemic to determine whether these changes affected US children’s opportunity to learn non-English languages.

To consider how children’s language exposure might have been affected by the pandemic, we first review how children’s day-to-day experience changed upon the arrival of COVID-19. We then describe how children’s language skills are built on their experience, focusing on studies from multilingual environments. Informed by this research, our study tested whether US children received more or less exposure to languages other than English as a result of pandemic-related disruptions in childcare. We asked parents to retrospectively report on childcare arrangements and their child’s language exposure from February 2020 to September 2021 to explore how the pandemic may have affected children’s opportunity to become bilingual.

1.1. Changes in daily routines during the COVID-19 pandemic

At the onset of the pandemic, researchers rapidly set out to investigate how extensive lockdowns changed day-to-day experiences and activities for young children, as families adapted to spending more time at home. For example, many families both in the US and in a variety of Western and non-Western countries reported that children spent more time using screens (Bergmann et al. 2022; Hartshorne et al. 2021). Likewise, parents reported more screen-mediated reading, but not engaging in more shared reading overall (Read et al. 2021). Importantly, not all children experienced the same degree of change. Older children and children from families of lower socioeconomic status (SES) experienced more screen time (Hartshorne et al. 2021). Furthermore, increases in screen time were associated with losing access to childcare (Hartshorne et al. 2021; Read et al. 2021), suggesting that parents may have relied on media to fill children’s time. Encouragingly, increased screen usage does not appear to have interfered with vocabulary development for young children. Kartushina et al. (2022) found that for a large sample of 8- to 36-month-old children from 13 different countries (including Western and non-Western countries), gains in vocabulary size during the first set of lockdowns (∼March 2020–September 2020) were larger than would be expected from a pre-COVID normative sample. Thus, changes in their routines did not seem to impede children’s language learning. Nonetheless, children who spent more time passively viewing media showed smaller gains in vocabulary, suggesting that electronic devices were not directly supportive of language development.

One possibility is that although children spent more time using screens, they also engaged in other enriching activities (e.g. joint play, caregiver-child interactions) positively associated with language learning (e.g. Chen et al. 2021; Liebeskind et al. 2014). For example, parents of 3-to-6-year-olds in Brazil reported an increase in music-related activities, such as listening to music or dancing (Ribeiro et al. 2021). Thus, parents may have made up for the sudden lack of childcare by finding other ways to engage with children, and in some cases, these changes may have provided children with richer language experience.

1.2. Language input and development

In the current investigation, our focus is children’s experience with language before and during the height of the pandemic. Early reports suggest that children’s language development has not been globally delayed (Kartushina et al. 2022), but whether there were changes in the particular languages that children learned is less clear. Even in places like the English-dominant US, there is growing recognition of both the prevalence and utility of bi- or multilingualism (Ryan 2013; see Fibla et al. 2022). However, not all children from bilingual homes or communities ultimately become proficient in multiple languages (e.g. Duursma et al. 2007; Wright et al. 2000). Thus, the mere presence of another language does not guarantee bilingual proficiency, but successful learning is related to factors such as chronological age and early bilingual experience (e.g. Ahn et al. 2017).

Contemporary research reliably demonstrates that children’s ability to learn a language is shaped by their experience with language (Hart and Risley 1995; Hoff 2006; Hurtado et al. 2008), and bilingual children’s knowledge depends on their experience with each of their languages (Marchman et al. 2017; Place and Hoff 2015). For example, Thordardottir (2014) found that children who experienced unequal amounts of input in different languages also had unequal levels of vocabulary and grammar performance in each language. While it is difficult to estimate exactly how much exposure children need to learn a language, research has found that below some threshold (e.g. less than ∼20 % of their input) may lead to more resistance on behalf of children to speak the language (Pearson et al. 1997). Collectively, research exploring the input that children receive across different languages and contexts suggests that differences in ultimate attainment may be at least in part related to differences in exposure to each language (Weisleder 2017).

Nonetheless, children do not necessarily learn languages in direct proportion to their exposure, and patterns of attainment may differ for majority and minority languages (see Hoff 2021). Factors such as motivation, community attitudes, and institutional support can also contribute to children’s eventual proficiency (e.g., Douglas Fir Group 2016). Children may need more support to learn languages that are not the dominant language in their community (Gathercole and Thomas 2009; Hoff et al. 2021), and improvement in children’s knowledge of a minority language does not impede the development of their skills in the majority language (Giguere and Hoff 2020). Importantly, there are factors beyond the quantity of exposure that influence children’s learning of different languages. For example, in the US, minority languages are more likely to be learned inside the home, relative to the majority language (English), which is widely learned both in and outside of the home (see Hoff et al. 2021). Relatedly, US children are more likely to hear English from more people, potentially minimizing the relative importance of non-English languages (Place and Hoff 2011). Additionally, minority languages can hold a less prestigious status, which could change how children prioritize speaking one language over the other (e.g. Giles and Watson 2013). It is possible that changes in children’s social environment during the pandemic would also affect their interest in majority versus minority languages.

1.3. Current study

Here, we explored whether and how pandemic-related changes in childcare arrangements coincided with changes in children’s exposure to different languages. If minority languages are more likely to be used inside the home, and children spent more time at home with caregivers using non-English languages, there may have been more opportunities to learn the minority language. In fact, recent research has offered some evidence that the pandemic changed home language use. Serratrice (2020) conducted a study of 4-to 11-year-olds in the UK and Ireland in March 2020 and reported that children had more opportunities to hear minority languages during lockdown. Similarly, a study comparing a pre-pandemic and a pandemic group of 4- to 8-year-old Mandarin-English bilinguals in the US found that the pandemic group had greater exposure to Mandarin from online and electronic sources and used more Mandarin at home (Sheng et al. 2021). Thus, there is evidence that the COVID-19 pandemic influenced some children’s experience with language. In the current study, we asked whether very young children in the US had different experiences with minority (non-English) languages because of the pandemic. We measured children’s exposure to non-English languages across four timepoints spanning February 2020 to September 2021. Using a retrospective survey administered during Fall 2021, participants responded to questions about childcare arrangements and language exposure at each timepoint. We selected timepoints designed to capture children’s experience at different phases of the pandemic: prior to the pandemic (February 2020), during the initial period of widespread lockdowns and daycare closures (April 2020), 6 months into the pandemic when parents may or may not have made alternative arrangements (October 2020), and after vaccines were available (September 2021). Thus, we were able to see how families adjusted both their childcare strategies and language use over an 18-month span.

To test whether pandemic-related influences increased young children’s opportunities to hear non-English languages, we focused on the language experience of children exposed to any language other than English at any timepoint. We refer to children who received any exposure to both English and a non-English language as “Multilingually-exposed”, which we contrast with “Monolingually-exposed” children, who were only exposed to English for the entire study period. Our goal was to target a representative sample of US children under 5 at the onset of the pandemic, and therefore not yet involved in formal schooling (which could change bilingual language development, see MacLeod et al. 2019). Broadly, we sought to document changes in children’s exposure to non-English languages and to examine factors that might predict those changes, including changes in childcare arrangements, demographic variables, and parents’ proficiency in the target language. More specifically, we asked:

1. What changes to childcare arrangements did Monolingually- and Multilingually-exposed children experience from February 2020 to September 2021?

2. Did Multilingually-exposed children experience changes in exposure to non-English languages from February 2020 to September 2021?

3. Were changes in exposure to non-English languages related to changes in childcare arrangements?

2. Methods

2.1. Participants

Participants were 96 caregivers of US children born between 2016 and 2019. An additional 41 participants began the survey, but did not fill out enough information to be included in the analyses or their language exposure data could not be interpreted (i.e. the proportions did not add up to 100 % of the time). Children were on average 24 months old at the onset of the pandemic in March 2020 (SD = 13, range = 3–50). 78 of the children were White or Caucasian, 13 were biracial, 2 were Black or African American, 1 was Asian, and 2 identified as other. 87 reported their Ethnicity as not Hispanic or Latino, 7 reported as Hispanic or Latino, and 2 opted to not report this information.

Participants could be located anywhere in the United States, see Figure 1. 50 % were located in North Carolina, while the remainder were spread out across 26 additional states.

Figure 1:

Location (by zip code) of participants.

To measure parental education (a proxy for family SES), we used the education reported by the participating caregiver. Overall, participants were highly educated, 71 % had a graduate degree (MA, PhD, JD or MBA), 26 % had a BA, and 3 % had some college or post-high school education or a high school diploma/GED. According to US Census Data, 14.4 % of the population 25 years and older has an advanced or professional degree, and 23.5 % have a bachelor’s degree (Census.gov 2022). Since our sample was considerably more highly educated than the average US sample, we created a binary variable with “more highly educated” referring to those with a graduate degree, and “less highly educated” referring to all others.

Fifty-five of these participants reported their children heard at least one language other than English during at least one timepoint; we refer to this group as Multilingually-exposed. Multilingually-exposed children were exposed to English, and one or more of the following languages: Arabic, Azerbaijani, French, German, Hebrew, Indonesian, Italian, Japanese, Korean, Mandarin, Marathi, Norwegian, Russian, Spanish, Turkish. The remaining 41 participants, who were not exposed to a language other than English at any of our time points will be referred to as Monolingually-exposed. Parental education did not differ for families with Monolingually- or Multilingually exposed children; for both groups, 71 % of caregivers were considered “more highly educated.”

2.2. Materials

Our questionnaire asked parents about household members and the languages spoken by those household members, as well as childcare arrangements and children’s language exposure at four timepoints spanning different phases of the pandemic (February 2020, April 2020, October 2020, September 2021). The full questionnaire, data, and analysis scripts are available on OSF: https://osf.io/etqra/. On average, these households included 2.06 adults and 1.65 children.

For each timepoint, participants were asked to select all applicable situations from a list of possible childcare arrangements. Options included “care outside the home”, (e.g., daycare centers, in-home daycares, or family member’s houses), “care inside the home – non-household member” (e.g., babysitters, nannies, or family members who did not typically live with the child), or “care inside the home – household member” which referred to parents or other people living in the child’s home. For options involving non-household members, parents indicated whether children experienced that arrangement for more or less than 20 hours per week. Caregivers were also asked to evaluate whether the amount of time their child spent with household members had changed since the beginning of the pandemic, on a 5-point scale ranging from “A lot less time with household members” to “A lot more time with household members”, with the middle point referring to “No change”.

To assess children’s language exposure, caregivers were asked to think about all the words their child heard directed to them at a given timepoint. For each language (including English), caregivers were asked to report the percentage of the words that the child heard in that language, and whether they heard the language inside the home, outside the home, or both. Caregivers were also asked to report whether they thought there had been changes in the amount of English their child heard, using a 5-point scale ranging from “A lot less English” to “A lot more English”, with the middle referring to “No change”. Caregivers also indicated whether they considered their child ‘bilingual’, and in what language the child knew the most words.

2.3. Procedure

We recruited participants through a citizen science approach, sharing our survey in local (to the authors) parenting groups, as well as on listservs and social media. Responses were collected with a single survey administered between September 2021 and December 2021, which required parents to retrospectively report about their families’ experiences from the prior 18 months.

2.4. Data analysis

We used R (Version 4.0.2; R Core Team 2020) and the R-packages papaja (Version 0.1.0.9999; Aust and Barth 2022), and tinylabels (Version 0.2.3; Barth 2022) for all our analyses. R packages used can be found in the references and on OSF.

3. Results

All data reflect parents’ retrospective reports of the child’s experience at that time point.

3.1. Changes in childcare arrangements

We first report on childcare arrangements for both Monolingually- and Multilingually-exposed children before and during the pandemic. We classified childcare arrangements by whether children received no external care from non-household members, part-time childcare from non-household members, or full-time childcare from non-household members (collapsing across whether childcare took place in or outside the home). Collapsing across groups (Monolingually- and Multilingually-raised), the proportion of children receiving no childcare from non-household members increased from February 2020 (25 %) to April 2020 (81 %), and then returned to its prepandemic average by October 2020 (31 %), before dropping lower in September 2021 (11 %). The proportion of children receiving part-time care remained relatively stable across our four time points, with a small reduction in April 2020 (February 2020 = 18 %, April 2020 = 6 %, October 2020 = 19 %, September 2021 = 19 %). Patterns were similar for Monolingually- and Multilingually-exposed children, see Figure 2.

Figure 2:

Proportion of children receiving full time, part time, or no childcare from non-household members at different time points. Left includes participants who did not report any exposure to a non-English language; right includes participants who did report exposure to a non-English language.

We quantified changes in childcare by comparing each individual set of consecutive timepoints (e.g. was the amount of childcare that child received the same or different in February 2020 and April 2020, in April 2020 and October 2020, and in October 2020 and September 2021). On average, children experienced 1.45 changes (range: 0–3) in childcare. Linear effects models tested whether the number of changes children experienced was related to the child’s age at the pandemic’s onset, or by parental education. The number of changes in childcare was predicted by the child’s age in March 2020 (t = 4.65, p < 0.001); younger children experienced fewer changes. This could be because, on average, younger children were more likely to only receive childcare at home before the pandemic (children not receiving external care in February 2020 were 12.12 months on average vs. 27.32 months for children receiving external care). These changes were not significantly related to parental education (t = 1.93, p = 0.057). Whether children were Monolingually- or Multilingually-raised was also not significantly related to the number of changes in childcare children experienced (t(94) = 1.91, p = 0.059).

We found that most children experienced at least one change in childcare arrangements during the pandemic, and 61 % of caregivers reported that their children spent more time with household members, relative to before the pandemic (see Supplementary Table on OSF). However, these changes did not differ systematically for Monolingually versus Multilingually-exposed children.

3.2. Changes in language exposure

Our next set of analyses were designed to address our primary research question: did children’s exposure to non-English languages change over our target time period? For these analyses, we focused only on children who were reported to receive exposure to a non-English language at least 1 % of the time during at least one of our time points. In February 2020, 46 children were reported to receive exposure to at least one language besides English. Of these, 33 % were reported to receive exposure to other languages outside the home, 39 % inside the home, and 26 % both in and outside the home (1 participant did not report this information). In April 2020, 34 children were reported to receive exposure to at least one language besides English. While the overall number decreased, the proportion of participants who received exposure to languages other than English inside the home (as opposed to outside the home or both) increased to 85 % in April 2020. The number of participants exposed to a language other than English increased in October 2020 (n = 39), and in September 2021 (n = 45), and the proportion of children receiving exposure inside, outside the home, or both, returned to their pre-pandemic levels by October 2020, see Figure 3.

Figure 3:

Distribution of where (inside or outside the home or both) Multilingually-exposed children heard non-English languages. One participant did not specify the location of exposure.

We next tested whether children’s exposure to non-English languages changed across timepoints. For children exposed to multiple non-English languages, we aggregated all non-English languages to calculate their total percentage of exposure to languages other than English. We converted timepoints into a numeric variable representing time since pandemic onset in March 2020, such that February 2020 was −1 (1 month prior to the pandemic), April 2020 was 1, October 2020 was 5 and September 2021 was 18 months. This approach allowed us to analyze time as a repeated measure, while also accounting for the fact that the amount of time between time points varied. Models also included by-participant random intercepts.

On average, children exposed to non-English languages heard those languages 17.67 % of the time in February 2020, a percentage which remained highly consistent across our four timepoints (April 2020: 13.91 %; October 2020: 15.67 %; September 2021: 15.58 %). The effect of time was not significant, t(164) = −0.26, p = 0.795, confirming that non-English exposure was stable across time. Despite little group change, roughly half of parents reported changes in the amount of English exposure their children received (see Supplementary Table on OSF).

We next explored whether some children were more likely to experience changes in language exposure. First, we examined general participant characteristics: child age at the onset of the pandemic and parental education (as a binary variable indexing more vs. less highly educated). We then explored variables specific to bilingual experience: parent proficiency in non-English languages and child bilingual status. In each model, we sought to predict the amount of exposure to non-English languages using main effects of time and each of these additional variables (in separate models), and their interactions.

We first tested whether child age at the onset of the pandemic was related to their exposure to languages other than English, reasoning that parents might speak differently to younger children versus older children. This model included a significant main effect of child age (t(59.84) = −3.00, p = 0.004), such that children who were younger at the onset of the pandemic received more exposure to non-English languages than children who were older at the onset of the pandemic; r = −0.35, 95 % CI [−0.46, −0.23], t(218) = −5.49, p < 0.001. The effect of time (t(163) = −0.07, p = 0.942) and the interaction (t(163) = −0.06, p = 0.953) were not significant, suggesting that exposure to languages other than English remained consistent across time, with children who were younger at the onset of the pandemic continuing to hear more non-English input even as they grew older. In a follow-up analysis, we tested the possibility that younger children were always more likely to hear non-English languages. For the follow-up model, we included age at each timepoint, instead of age at pandemic onset. Child age at each time point was a significant predictor of the percentage of non-English exposure (t(64.24) = −3.00, p = 0.004), suggesting that younger children were more likely to have more exposure to non-English languages, independent of the time when exposure was measured. The effect of time (t(213.10) = 1.49, p = 0.137) and the interaction (t(164.03) = −0.02, p = 0.984) were not significant. Thus, regardless of time point, children who were younger had more exposure to non-English languages, and children who were younger at pandemic onset were more likely to maintain that exposure across the 18-month span.

Our next model tested the influence of parental education. The main effect of parental education was not significant (t(58.50) = −0.73, p = 0.465); children with more versus less highly educated parents had similar exposure to non-English languages. However, the main effect of time was significant (t(163) = −2.48, p = 0.014), as was the interaction between parental education and time (t(163) = 2.77, p = 0.006). To better understand this interaction, we conducted correlations to test for patterns of language exposure across time points for children with more versus less highly educated parents. Children of more highly educated parents heard a relatively consistent proportion of languages other than English (r = 0.06, 95 % CI [−0.10, 0.21], t(154) = 0.69, p = 0.494), while children of less highly educated parents received less exposure to languages other than English at our later timepoints (r = −0.15, 95 % CI [−0.38, 0.10], t(62) = −1.21, p = 0.232). However, neither correlation reached significance on its own (see Figure 4 for raw data).

Figure 4:

Median percentage of non-English exposure across time points for children of less and more highly educated parents. Each dot reflects an individual data point, thick line represents the median, the lower and upper hinges correspond to the first and third quartiles.

We then examined variables that are more likely to be specific to bilingual experience, beginning with caregiver proficiency in a language other than English. For this analysis, we used the highest proficiency rating for any non-English language for any caregiver (e.g. if Parent 1 was rated as a 3, but Parent 2 was rated as a 4 for non-English language, we used 4). 64 % of Multilingually-exposed children had caregivers who reported at least minimal proficiency in a second language. The average highest proficiency (excluding participants whose parents did not report being proficient in a non-English language) was 3.54/5. This model revealed a significant main effect of parent proficiency in a non-English language (t(59.52) = 3.55, p = 0.001), such that children whose parents had higher proficiency in non-English language had a greater proportion of exposure to other languages (r = 0.34, 95 % CI [0.22, 0.45], t(218) = 5.35, p < 0.001). The effect of time (t(163) = 1.62, p = 0.106) was not significant, but the interaction between caregiver proficiency and time was significant (t(163) = −2.55, p = 0.012). Visual inspection of the data (see Figure 5) suggests that children with parents who were highly proficient in a non-English language had more exposure and more consistent exposure to non-English languages, while children whose parents were less proficient in a non-English language had more variable exposure across time.

Figure 5:

Median non-English exposure at each time point, organized by caregivers’ self-reported proficiency in a non-English language. Each dot reflects an individual data point, thick line indicates the median, the lower and upper hinges correspond to the first and third quartiles.

We next asked whether children who were considered to be bilingual (n = 12) differed in exposure to other languages. This model revealed a significant main effect of child bilingual status, (t(65.02) = 7.65, p < 0.001). Bilingual children heard a higher percentage of other languages (mean = 42.90 %) than Multilingually-exposed children reported to not be bilingual (mean = 8.12 %). The main effect of time (t(163) = 0.46, p = 0.649) and the interaction with time (t(163) = −1.47, p = 0.143) were not significant, suggesting that differences did not change over the course of the pandemic.

Across a variety of demographic variables, our analyses demonstrated that despite slight fluctuations in the number of children reported to hear non-English languages, the overall proportion of children’s exposure to non-English languages remained remarkably consistent. Still, we found several reliable predictors of children’s language exposure. Children who were younger, whose parents were more proficient in non-English languages, or who were considered to be bilingual themselves heard higher proportions of non-English languages at all timepoints.

3.3. Links between childcare changes and language exposure

While our models found very few effects of time on exposure to non-English languages, 47 % of participating parents reported that their children’s experience with language changed as a result of the pandemic. Thus, we tested whether exposure to non-English languages was related to changes in childcare. If a child spent more or less time in external childcare in sequential timepoints, they were considered to have experienced a change between those two timepoints (e.g., they received full-time childcare outside the home at time 1 and part-time care outside the home at time 2). To measure changes in language exposure, we calculated positive or negative changes in the percentage of non-English language exposure between any two consecutive timepoints (e.g., if a child heard 30 % non-English in February 2020 and 50 % non-English in April 2020, their percentage change from February 2020 to April 2020 would be +20 %). We predicted percentage change across timepoints based on time, changes in childcare (yes/no), age at pandemic onset, and their interactions.

$$\mathrm{N}\mathrm{o}\mathrm{n}\mathrm{E}\mathrm{n}\mathrm{g}\mathrm{l}\mathrm{i}\mathrm{s}\mathrm{h}\mathrm{P}\mathrm{e}\mathrm{r}\mathrm{c}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{C}\mathrm{h}\mathrm{a}\mathrm{n}\mathrm{g}\mathrm{e}\sim \mathrm{C}\mathrm{h}\mathrm{i}\mathrm{l}\mathrm{d}\mathrm{c}\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{A}\mathrm{m}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t}\mathrm{C}\mathrm{h}\mathrm{a}\mathrm{n}\mathrm{g}\mathrm{e}\times \mathrm{T}\mathrm{i}\mathrm{m}\mathrm{e}\mathrm{P}\mathrm{o}\mathrm{i}\mathrm{n}\mathrm{t}\times \mathrm{A}\mathrm{g}\mathrm{e}\mathrm{P}\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{m}\mathrm{i}\mathrm{c}\mathrm{O}\mathrm{n}\mathrm{s}\mathrm{e}\mathrm{t}$$

This model included a significant interaction between childcare changes and time (F(2,153) = 5.66, p = 0.004), see Figure 6. There was a significant difference in how language exposure changed from February 2020 to April 2020 for children who did versus did not experience a childcare change (t-test: ΔM = 7.25, 95 % CI [1.56, 12.93], t(47.32) = 2.57, p = 0.014). There was a 6 % decrease in exposure to non-English languages for children whose childcare changed, while those who did not experience childcare changes maintained similar non-English exposure (+1 %). There was also a significant difference from April 2020 to October 2020 for children who did or did not experience a childcare change (t-test: ΔM = −8.55, 95 % CI [−14.88, −2.21], t(50.52) = −2.71, p = 0.009); children with childcare changes had a 5 % increase in non-English exposure, while those without changes experienced a small reduction in non-English exposure (−3%). Lastly, there were no such associations from October 2020 to September 2021; changes in exposure to non-English were not more likely for children who experienced changes in childcare (t-test: ΔM = 6.88, 95 % CI [−5.71, 19.48], t(20.98) = 1.14, p = 0.269). All other effects were not significant (all ps > 0.11).

Figure 6:

Predicted values of change in percentage of non-English exposure. Each facet represents change between each set of consecutive time points, for example the “FebApril” facet shows the change in percentage of exposure to non-English languages between February 2020 and April 2020, for children who did or did not experience changes in childcare.

Since about one-third of participants reported only hearing non-English languages outside the home in February 2020, a follow-up analysis asked whether changes in language exposure between timepoints were related to whether children heard non-English languages exclusively outside the home, or inside the home (combining inside-only and both) in February 2020.

$$\mathrm{N}\mathrm{o}\mathrm{n}\mathrm{E}\mathrm{n}\mathrm{g}\mathrm{l}\mathrm{i}\mathrm{s}\mathrm{h}\mathrm{P}\mathrm{e}\mathrm{r}\mathrm{c}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{C}\mathrm{h}\mathrm{a}\mathrm{n}\mathrm{g}\mathrm{e}\sim \mathrm{N}\mathrm{o}\mathrm{n}\mathrm{E}\mathrm{n}\mathrm{g}\mathrm{l}\mathrm{i}\mathrm{s}\mathrm{h}\mathrm{L}\mathrm{o}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{F}\mathrm{e}\mathrm{b}2020\times \mathrm{T}\mathrm{i}\mathrm{m}\mathrm{e}$$

This model revealed a significant interaction between time and other-language exposure location prior to the pandemic (F(2,129) = 6.33, p = 0.002). Follow-up t-tests revealed that while children who heard non-English languages only outside the home in February 2020 experienced a significant reduction (−13 %) in non-English language exposure from February 2020 to April 2020 (95 % CI [−21.67, −3.93], t(14) = −3.10, p = 0.008), non-English language exposure remained consistent between all other pairs of time points regardless of whether participants initially heard non-English languages in or outside the home (all ps > 0.05).

These analyses suggest that children’s language exposure was influenced by changes in childcare arrangements, likely as they spent more or less time with caregivers who used different languages. Children appeared to receive less exposure to non-English languages when the majority of their care happened at home, particularly among children who initially were receiving all of their non-English exposure outside the home. For children who always experienced at least some non-English input inside the home, there were fewer changes overall.

4. Discussion

We investigated whether and how the pandemic changed young US children’s opportunities to hear non-English languages. We hypothesized that changes in childcare arrangements might offer adults proficient in non-English languages more opportunities to speak those languages with their children. To test this hypothesis, we asked parents of young children to retrospectively provide information about childcare arrangements and children’s language exposure over an 18-month period. As expected, the parents in our sample reported changes in their children’s daily care, particularly early in the pandemic, and that on average, children spent more time with household members. However, contrary to our predictions, these children’s language exposure remained surprisingly consistent during a period of widespread upheaval.

First, we aimed to describe the changes in childcare arrangements that children experienced from February 2020 to September 2021. We expected that many families would have removed their children from daycare settings or stopped inviting outside caregivers into their homes when lockdowns began, but the duration of these arrangements was unknown. Consistent with other studies (e.g. Patrick et al. 2020), we found that the majority of the children in our study experienced changes in childcare arrangements before and after April 2020. However, these changes may have been relatively short-lived; six months later, the number of children receiving childcare from non-household members had returned to pre-pandemic rates and continued to increase thereafter. In addition, we found no evidence that multilingual or less highly educated families in our sample experienced disproportionate changes to childcare arrangements.

Our second goal was to determine whether children’s exposure to non-English languages changed over our sampled timepoints. Contrary to our expectations, we did not see evidence for increases in non-English exposure at the onset of the pandemic; if anything, parents reported that children might have heard more English. However, we found some suggestive evidence that some parental and demographic factors were related to the consistency of children’s language experience. Children with more highly educated parents tended to maintain the same amount of non-English exposure across our sampled timepoints, while children of less highly educated parents received less exposure to non-English languages at later times. Since parental education was not a predictor of changes in childcare arrangements, the reasons for these different tendencies are unclear. It could be that more highly educated parents, who on average are likely to have higher incomes (in the US, U.S. Bureau of Labor Statistics 2022), had more or different options for spending time with their children. For example, workers in upper-income brackets may be more able to work from home, relative to those in middle-income or lower-income brackets (Parker et al. 2022). The patterns we observed could also reflect differences in how languages are prioritized. Research suggests that in US communities, working-class parents (compared to middle-class parents, as labeled by Lambert and Taylor 1996) may express more concern about their children’s skills in the community-dominant language (Lambert and Taylor 1996), and may have therefore been more likely to increase their use of English as children got older. In the literature to date, income and education are often conflated, and we look forward to future research exploring whether parents with different educational backgrounds or different kinds of professions differently emphasize the use of particular languages.

While non-English exposure was relatively stable for our sample as a whole, some children heard more input in a non-English language. One factor that influences chidlren’s experience with non-English languages was parent’s proficiency in a non-English language. Parent proficiency has been linked to children’s language skills, and the presence of bilingual adults increases the likelihood that children will hear multiple languages (Schott et al. 2021; Unsworth et al. 2019). Children with more proficient caregivers heard more non-English input, which remained more consistent over time, while children with less proficient caregivers heard less non-English input overall, and the amount of that input varied more over time. This pattern likely reflects that the return to childcare outside the home lowered exposure to non-English languages for some participants, while increasing exposure for others, as children came in contact with more and less proficient users of different languages. However, given the size of our sample and the limited variability in parental education in particular, we cannot determine whether these patterns of input would be likely to be present in other populations.

We also found that characteristics of the child influenced their language experience. Younger children and children who were younger at the onset of the pandemic heard more non-English input across all timepoints. These findings are consistent with other research suggesting that exposure to heritage languages decreases as children age (Hoff et al. 2014) and could also suggest that lockdowns early in a child’s life affected these families’ language use. In addition, children described as bilingual, like those whose parents reported higher proficiency in a non-English language, heard more non-English input. Indeed, differences in child and parent proficiency (as discussed above) could explain why our results contrast with prior studies suggesting that in some communities already identified as bilingual, minority language use increased during the pandemic (Serratrice 2020; Sheng et al. 2021). In our study, it may be that when older children appeared to be more competent in English than in the minority language, parents tended to use more English, whereas with very young children or children who were equally comfortable in either language, they used the language(s) in which they themselves were most proficient, leading children to have different experiences. Together, these findings suggest that child factors (such as age and bilingual status), and some caregiver factors (such as parental education and language proficiency) shaped the early patterns of language use.

Finally, we explored associations between changes in childcare arrangements and changes in exposure to non-English languages. Children who experienced more changes in childcare showed larger fluctuations in language exposure, likely reflecting that children spent different amounts of time with people more or less able to speak different languages. At the beginning of the pandemic, changes in childcare were associated with decreased exposure to non-English languages. Unsurprisingly, the largest change was that children exposed to non-English languages exclusively outside the home showed a significant decrease in hearing those languages in April 2020. However, in October 2020, these same children tended to have increased exposure to non-English languages if they experienced changes in childcare. In contrast, children who received at least some exposure to non-English languages inside the home were more likely to maintain that exposure. These patterns of change underscore the complexity of multilingual language environments. They also highlight possible differences between the experiences of heritage language learners, whose primary source of non-English input is likely at home, relative to those learning an additional language outside the home.

While these results provide new insight into the language experience of a potentially unusual cohort of children, it is important to acknowledge some limitations of our study. To try to include as representative a group of US families as possible, we employed a citizen science approach, recruiting participants through social media and sharing the survey openly. While we recruited participants from 29 different states, our final sample was relatively small, highly educated and primarily White, limiting the conclusions and generalizations that can be made. Notably, we might expect children whose primary caregivers were not proficient in English (and therefore unable to respond to our questionnaire) and/or who heard English exclusively outside the home prior to the pandemic, to have reduced exposure to English and increased exposure to minority languages. Another consideration is that we relied on retrospective reports. Research suggests that parental report is an accurate measurement of language exposure (Orena et al. 2020), but these reports may be less accurate over a longer delay, particularly when families experienced many changes. Lastly, we cannot link children’s early experience with non-English languages to their later fluency or interest in those languages. Future research may be able to compare the experience and long-term proficiency of children raised before, during, and after pandemic-related disruptions. Including later measures of proficiency could be used to determine whether the modest changes in children’s exposure to different languages will (or will not) have a permanent effect on their language skills and can help us understand cascading consequences of changes in children’s early language experience.

The COVID-19 pandemic provided an unfortunate natural experiment highlighting the need to consider the wide range of influences on children’s development (Bronfenbrenner and Ceci 1994). As young children raised during the pandemic enter formal schooling, educators, clinicians, and parents need to understand how their language experience may have differed from prior cohorts (see Hoff et al. 2021) and to consider how the pandemic created or impeded opportunities for learning different languages. Our initial intuition that children’s increased time at home would facilitate exposure to non-English languages was not supported. In fact, despite frequent changes in childcare arrangements, we found striking continuity in children’s language exposure over the course of the pandemic. There was little evidence that families suddenly introduced or eliminated their use of a heritage language. Thus, while patterns of language use are undoubtedly complex, our study suggests that perhaps for very young children, local factors such as family environment and parent preferences exert a more direct influence on the input they receive, and early tendencies may be quite robust. More generally, it is possible that patterns of language use were not as sensitive to COVID-related disruptions as might have originally been thought. At the same time, it is clear that the input parents provide to their children can be shaped by factors such as socioeconomic burden and stress (e.g. Roby and Scott 2022; Ellwood-Lowe et al. 2022). Thus, research needs to continue to explore how early language environments are shaped by both immediate and systemic variables in order to explain why some children are more likely than others to become bilingual.