Bilingual parent–child dyads' language use and code-switching during shared reading: exploring variations across labeling and word explanations

1 Introduction

Parent–child shared reading invites dyads to co-construct narratives and provides a rich context for supporting children's language development (Dowdall et al., 2020; Grøver et al., 2020). During shared reading, parents can use various forms of talk to scaffold their children's language learning by exposing them to novel words that are not typically used in everyday conversations (Noble et al., 2018; Rowe and Snow, 2020). The dialogic nature of shared reading offers children the opportunity to expand beyond their current vocabulary knowledge as they acquire unfamiliar words through interactions with more knowledgeable individuals (Ninio and Bruner, 1978; Vygotsky, 1978).

Recent studies have suggested that parents of dual language learners (DLLs) use and mix their first language (L1) and second language (L2) for specific communicative purposes, such as supporting their child's understanding (e.g., Byers-Heinlein, 2013; Kremin et al., 2022; Read et al., 2021). However, little is known about (a) the specific nature of shared reading among bilingual parent–child dyads and the ways in which they turn attention to words across their two languages, and (b) during what kinds of interactions they code-switch (i.e., the alternating use of languages) most frequently when engaging with words.

Grounded in a social-interactionist framework, which posits that children's language learning emerges through social interactions (Bruner, 1983; Grøver et al., 2019; Vygotsky, 1978), the present study draws on observational data to examine how bilingual parent–child dyads naturally use their languages during shared reading of a wordless picture book at home. We focus specifically on two different strategies used to scaffold children's word understanding, namely, labeling and word explanations. Labeling, which involves simply naming objects, characters, or actions in the book (e.g., “this is a forest”), has been shown to support early vocabulary acquisition through simple and concrete talk (Rowe and Snow, 2020). Word explanations involve providing definitions, descriptions, or elaborations related to elements in the book (e.g., “when a place is filled with many trees, it's called a forest”) and are particularly effective in supporting children's word comprehension as their language abilities grow (Evans et al., 2011; Korat et al., 2018; Read et al., 2023). We explore how these two ways of turning attention to words differ between dyads who primarily read in their L1 compared to those who primarily read in their L2, and when code-switching is most likely to occur.

1.1 Bilingual parent–child shared reading

Bilingual parents from language-minority groups vary in their relative use of L1 and L2 when communicating with their children (Mancilla-Martinez and Lesaux, 2011; Quiroz et al., 2010; Rydland and Grøver, 2021; Scheele et al., 2010). These parents “often must balance the competing goals of supporting a minority heritage language at home while also preparing their children for the language of formal literacy instruction when they begin school” (Read et al., 2021, p. 362). Preschool entry often marks the beginning of systematic exposure to the majority language for many bilingual children (Halpin and Melzi, 2021), which can place the minority language at risk. In a Norwegian study, parents reported that they mostly used their L1 when addressing their children, whereas their children used it less when addressing them (Rydland and Grøver, 2021). Parents who wish for their children to maintain proficiency in their L1 play an important role in teaching it to them. Indeed, bilingual parents are often aware of this role and engage in various practices to support their children's L1 development (De Houwer, 2017). Variation in parents' language use has also been documented in literacy activities (Quiroz et al., 2010; Scheele et al., 2010) and these variations has been linked to parents' general language preference, language use, and adaptations to their children's school language, which is typically the L2 for DLLs (Gonzalez-Barrero et al., 2021; Quiroz et al., 2010; Read et al., 2021).

Research on bilingual parent-child dyads' shared reading practices has taken several approaches. One approach has been to compare monolingual and bilingual dyads from different language-minority backgrounds to examine differences in their reading interactions (Scheele et al., 2010). Others have examined within-dyad differences by having the same dyads read in one language and then the other in controlled experimental settings (Muysken et al., 1996; Quirk et al., 2024), while some have used surveys in which parents reported on their reading practices across both languages (Avelar et al., 2025; Gonzalez-Barrero et al., 2021). Finally, parent reports have also been used to examine between-dyad differences in reading practices by comparing those primarily using their L1, their L2, or both languages more equally (Read et al., 2021).

Recent research has begun examining whether the language bilinguals use for reading affects how they read (e.g., Avelar et al., 2025; Gonzalez-Barrero et al., 2021; Quirk et al., 2024; Read et al., 2021). For instance, Gonzalez-Barrero et al. (2021) found that English–French parents in Quebec reported reading more frequently and for longer durations in their dominant language. Of particular relevance to the present study are the studies by Avelar et al. (2025) and Read et al. (2021), which investigated differences in parents' self-reported talk during shared reading among Hispanic families in the U.S. Both studies categorized participants into three groups (only or mostly Spanish, only or mostly English, and both English and Spanish more equally) based on parents' self-reported primary language use at home (Read et al., 2021) or their preferred language with their child (Avelar et al., 2025). Avelar et al. (2025) found an interaction between parents' preferred language and the book language. Parents engaged in interactive reading strategies (e.g., asking questions and defining words) most frequently when reading in their preferred language, particularly among those who preferred and read in English. Overall, parents read more frequently in English. Read et al. (2021) found that parents' use of dialogic reading strategies (e.g., asking questions and commenting on word meanings) did not differ between the three groups. Overall, parents read more frequently in the language they primarily used at home.

Given that parental input during literacy activities can vary from concrete to more elaborated talk, an important question is whether the nature of parent–child interactions differs depending on their dominant reading language. While few studies have examined bilinguals' language use and reading practices through surveys (e.g., Avelar et al., 2025; Read et al., 2021) and experimental designs (e.g., Quirk et al., 2024), less is known about how distinct forms of talk emerge in naturalistic reading interactions. Because bilingual parents tend to match their reading language with the text when using textbooks (Muysken et al., 1996), wordless picture books provide a particularly useful format which allow bilinguals to use their languages freely.

1.2 Labeling and word explanations during shared reading

Researchers have distinguished between different ways in which parents draw children's attention to words. Two common types that differ in their degree of elaboration include labeling, which is concrete and direct (Muhinyi and Rowe, 2019; Quiroz et al., 2010), and word explanations, which involve providing definitions and elaborations of words (Evans et al., 2011; Korat et al., 2018).

While labeling and word explanations represent only two of the many forms of talk (Read et al., 2023), they illustrate distinct ways in which parents can scaffold children's word understanding. As children grow and become more linguistically competent, parents may shift from labeling to word explanations (Martin, 1997; Weizman and Snow, 2001). Such shifts reflect parents' adaptation to their children's early communicative behaviors and developmental stage (Ninio and Bruner, 1978). In other words, this sensitivity is likely part of a broader adaptation in which parents align their support strategies with the child's zone of proximal development, by tailoring their talk to match their child's abilities and learning needs (Vygotsky, 1978). Generally, parents appear to use word explanations infrequently during shared reading of textbooks, as reported by Evans et al. (2011) and Korat et al. (2018). These studies also examined the relationship between mothers' word explanations and educational level, with mixed findings.

An important question is whether parent–child dyads differ in their use of labeling and word explanations depending on their dominant reading language, particularly when reading wordless books. In a previous qualitative study of seven bilingual parent–child dyads' reading interactions (Nomat et al., 2024), we observed that parents who relied more on their L1 (using their L2 less) attempted to maintain the conversation in L1. Specifically, when a child used a word in their L2, the parent would redirect the dialogue by asking the child to say the word in their L1. Overall, research is limited on how the nature of parent–child talk varies depending on their dominant reading language. Understanding this variation can provide insight into how DLLs are exposed to language during their early literacy experiences.

1.3 Code-switching as a dialogic resource in shared reading

A growing line of research highlights the frequent use of code-switching during parent–child reading interactions (e.g., Gonzalez-Barrero et al., 2021; Kremin et al., 2022; Muysken et al., 1996; Quirk et al., 2024; Read et al., 2021). In parent-child dyads from language-minority backgrounds, children often initiate most of the switches to the L2 when reading primarily occurs in their L1, whereas their parents tend to mostly switch from the L2 back to the L1 (Nomat et al., 2024; Pan, 1995). Children's code-switching patterns may reflect their language proficiency, as young DLLs are more likely to incorporate their stronger language when using their weaker language (Halpin and Melzi, 2021). Parents' L2-to-L1 switching behavior has also been observed in contexts where two majority languages have equal status (Quebec), where parents used their dominant language during discussions, even when reading in their non-dominant language (Quirk et al., 2024).

Bilingual parents often use code-switching deliberately for specific purposes, such as scaffolding their child's understanding (Byers-Heinlein, 2013; Kremin et al., 2022; Muysken et al., 1996; Quirk et al., 2024). For instance, parents code-switch to extend and confirm the child's contributions to the dialogue during shared reading (Nomat et al., 2024). In a scale development and validation study, Byers-Heinlein (2013) unexpectedly found that parents reported in their open-ended responses using code-switching to teach new words. The author subsequently revised the scale to include this reason as an item, which became the most commonly reported motivation for code-switching. Similarly, parents' open-ended responses in Read et al. (2021) showed them reporting using both languages to translate unfamiliar words and when the complexity of the extra-textual dialogue increased. Evidence also suggests that parents' code-switching may be sensitive to children's developmental stage. Kremin et al. (2022) reported descriptive findings suggesting that parents code-switch to teach new words, particularly when their infant was older.

However, teaching or emphasizing words can be accomplished in different ways. Thus, an interesting question is whether parents code-switch more frequently during word explanations than during labeling. Bail et al. (2015) proposed a very similar idea by suggesting that parents might vary in their use of code-switching when teaching new words compared to simply labeling familiar objects. Such differences in code-switching may be more common among parents than preschool-aged children, as young children tend to code-switch less frequently and in ways that are less deliberate than older children and adults (Halpin and Melzi, 2021).

Taken together, research on code-switching during shared reading has largely used parental self-reports (e.g., Gonzalez-Barrero et al., 2021; Read et al., 2021) and experimental designs (e.g., Muysken et al., 1996; Quirk et al., 2024), while observational studies in naturalistic settings are few and typically involve small samples (e.g., Nomat et al., 2024; Pan, 1995). Notably, scholars have highlighted the need for research on bilingual parent–child dyads during shared reading to better understand when code-switching occurs (e.g., Gonzalez-Barrero et al., 2021; Read et al., 2021).

1.4 The present study

A distinct feature of bilingual parent–child interactions is that they can choose and mix their two languages. While research indicates that bilingual parents use their languages to varying degrees during shared reading (e.g., Read et al., 2021; Quiroz et al., 2010; Scheele et al., 2010), more research is needed to better understand the potential implications for their reading behaviors. Based on our previous qualitative analysis of seven bilingual dyads' reading interactions (Nomat et al., 2024), in which parents who adhered strongly to their L1 appeared to maintain the dialogue in that language by asking their children to repeat words in the L1, we expect that parents who primarily read in their L1 will be more likely to engage in talk that generally focuses on words. Moreover, given that code-switching functions as a dialogic resource that parents use to enhance children's understanding (Byers-Heinlein, 2013) and tends to occur when the extra-textual dialogue becomes more complex (Read et al., 2021), code-switching might be more likely to emerge during elaborative talk. Overall, we aim to better understand how bilingual parent–child dyads naturally attend to words and code-switch during shared reading, with a focus on labeling and word explanations. We categorize dyads into groups based on the language mainly used during shared reading: those who primarily read in their L1 are referred to as the “Mostly First Language Group” and those who primarily read in their L2 are referred to as the “Mostly Second Language Group.” We aim to address the following two research questions:

• RQ1: do dyads in the Mostly First Language Group differ from those in the Mostly Second Language Group in their use of labeling and word explanations?
• RQ2: do parents' and children's code-switching differ across labeling and word explanations?

2 Method

2.1 Sample

The present sample comprised 151 bilingual parent–child dyads living in the larger Oslo area in Norway. The children (81 girls, 70 boys) were 3 to 5 years old (M = 54.18 months, SD = 9.49 months), with a mean preschool entry age of 25 months (SD = 9.71). All participants were part of a larger language intervention study by Grøver et al. (2020), which aimed at supporting bilingual development through shared reading in preschool and at home. As part of the intervention, all parents received for home reading some of the books that were shared in the preschool classrooms. In the present study, we focused on parent–child reading interactions involving the first wordless picture book sent home. Bilingual research assistants fluent in the families' L1 conducted telephone interviews in the parents' preferred language with one parent to gather information about family demographics. The dyads' first languages included Urdu (19%), Polish (17%), Somali (15%), Arabic (13%), Tamil (8%), Kurdish (7%), Albanian (6%), Turkish (6%), Bosnian/Serbian (4%), Vietnamese (4%), and Russian (1%). All dyads shared Norwegian as their L2.

Most parents were born outside of Norway, with only seven mothers and 11 fathers born in Norway. Parents' years of residency in Norway ranged from one to 38 years (mothers' years of residency in Norway: M = 12.35, SD = 7.52; fathers' years of residency in Norway: M = 14.34, SD = 8.70). Education level was measured on a scale from one to six (1 = 0–4 years of schooling, 2 = 5–7 years of schooling, 3 = junior high school, 4 = high school, 5 = bachelor's degree, 6 = master's degree or higher). Most parents reported high school or less as their highest level of education (mothers' mean level of education: M = 4.07, SD = 1.25; fathers' mean level of education: M = 4.10, SD = 1.22). Mothers' and fathers' education levels were correlated [r (121) = 0.41, p < 0.001]. Most parents were employed or studying either full-time or part-time outside the home (mothers: 65%; fathers: 80%).

2.2 Audiotaped parent–child shared reading interactions

The first wordless picture book sent home for shared reading was Frog, Where Are You? (Mayer, 1969), which we expected to be unfamiliar to children in Norway. Parents were asked to read the book together with their child at home as they normally would, in their preferred language, and to audio-record one reading. In these audio-recorded reading interactions, 113 (75%) children read with their mothers and the remaining read with their fathers. The parents were also provided with information, in both their languages, about four target words that the preschools focused on when reading the book. The four target words were jar, swarm, resemble and persistent. Parents were not specifically instructed to teach these words when sharing the book, but rather, the primary goal was to ensure that the dyads enjoyed the book together.

2.3 Transcription and coding

Audio recordings were transcribed using the standardized CHAT format of the Child Language Data Exchange System (MacWhinney, 2000), with utterances serving as the unit of analysis. Utterances were delimited based on syntactic structure, intonation patterns, pause markers, and turn-taking. Bilingual research assistants fluent in the families' L1 transcribed the recordings, and all transcripts were translated into Norwegian prior to coding. Several steps were taken to ensure the validity of the transcription and translation. First, a bilingual research assistant translated and transcribed the recordings into Norwegian, marking all words spoken in the dyads' L1. Second, another bilingual research assistant validated the translations and transcriptions by comparing them against the audio recordings. Finally, the first author further verified all translated transcripts for language use by comparing them to the audio recordings. This last step ensured accurate annotation of code-switching by verifying that all words spoken in the dyads' L1 were appropriately marked. For dyads primarily reading in their L1, words spoken in their L2 were annotated at the end of each word, and conversely, for dyads primarily reading in their L2, words spoken in their L1 were similarly marked at the end of each word.

2.4 Dominant reading language

To explore differences in dyads' reading practices, we grouped them based on the language parents mainly read in during the audio-recorded reading. Dyads who primarily read in their L1 were classified as the “Mostly First Language Group” and those who primarily read in the L2 were classified as the “Mostly Second Language Group.” The dominant reading language was determined by listening to the recordings and evaluating which language was primarily used. Classifying the dominant reading language was straightforward for 144 (95%) dyads, as they mainly read in one language, but it was less straightforward for seven (5%) dyads who used both languages more frequently compared to the others. To address this, a closer review of their transcripts was conducted by counting utterances in each language. This closer review revealed that despite frequent use of both languages, one language could still be identified as the dominant reading language. Overall, the Mostly First Language Group comprised 96 (64%) dyads, and the Mostly Second Language Group comprised 55 (36%) dyads, which suggests that the majority primarily read in their L1. Notably, 111 (74%) dyads used both of their languages to varying degrees, and our categorization of the dyads into two groups is therefore a broad description of the dominant reading language.

2.5 Code-switching

We used the dyads' dominant reading language to operationalize code-switching by identifying utterances that contained words from the non-dominant reading language. For dyads who primarily read in their L1, utterances containing words in their L2 were coded as code-switching utterances. Similarly, for dyads who primarily read in their L2, utterances containing words in their L1 were coded as code-switching utterances. We followed the procedure as previous studies to obtain an overall measure of code-switching (e.g., Kaushanskaya and Crespo, 2019; Read et al., 2021). Specifically, all occurrences of code-switching—both inter-sentential code-switches (e.g., Example 1, line 2, where the child switched entirely to Norwegian in an utterance) and intra-sentential code-switches (e.g., Example 4, line 4, where the mother used both Norwegian and Kurdish words within a single utterance)—were summed to create an overall code-switching score that indicated the absolute number of code-switching utterances for each dyad. Examples of code-switching (in italics) are provided in Example 1 (lines 3–4), Example 3 (line 1), and Example 4 (line 4). All examples are translated into English.

2.6 Labeling and word explanation

Sequences of labeling and word explanations were coded in the transcripts, with the two categories being mutually exclusive. Labeling was examined by identifying utterances that involved requesting labeling (e.g., “What do you call this?”) (Muhinyi and Rowe, 2019; Quiroz et al., 2010; Read et al., 2021), providing labeling (e.g., “This is a dog”) (Muhinyi and Rowe, 2019; Quirk et al., 2024), and confirming labeling (e.g., “That is correct, a dog”) (Nomat et al., 2024). Labeling occurred in various ways, some of which are illustrated below.

2.6.1 Example 1: labeling “the moon” in a Somali dyad (Mostly First Language Group)

• ADU: and what is this then?
• CHI: it is the moon.
• ADU: moon.

In Example 1, the mother initiated a labeling request in Somali (line 1). The child responded in Norwegian, identifying it as “the moon” (line 2). The mother confirmed the child's response by repeating “moon” in Norwegian (line 3).

2.6.2 Example 2: labeling “horn” in an Albanian dyad (Mostly Second Language Group)

• ADU: where does the boy hold on to?
• CHI: In +...
• ADU: he is holding on to?
• ADU: do you remember what they are?
• CHI: Ehh +...
• ADU: horn.
• CHI: horn.

In Example 2, the mother repeatedly asked what the boy was holding (lines 1, 3, and 4). Eventually, she provided the label “horn” (line 6), which the child confirmed by repeating it (line 7). This example illustrates how labeling could involve multiple labeling requests.

We examined word explanations by focusing on the specific word targeted in the explanation. Following (Martin 1997) framework, we identified utterances that involved requesting a word explanation (e.g., asking the child about the meaning of a word), providing a word explanation (e.g., providing the meaning of a word through synonyms, explanations, or elaborations), and confirming a word explanation (Nomat et al., 2024). Word explanations occurred in various ways, some of which are illustrated below.

2.6.3 Example 3: explaining “bottle” in a Somali dyad (Mostly First Language Group)

• ADU: bottle in Somali is bottle.
• CHI: bottle.
• ADU: a container you drink water from.

In Example 3, the mother introduced the word “bottle” in both Somali and Norwegian (line 1), which the child repeated in Somali (line 2). She clarified the meaning by providing additional information (line 3). This example illustrates the mother's effort to support the child's understanding of “bottle.”

2.6.4 Example 4: explaining “swarm” in a Kurdish dyad (Mostly First Language Group)

• ADU: how are the bees?
• ADU: swarm, so swarm means that it is a large number.
• ADU: that means one, two, three, four, five, plenty, plenty.
• ADU: plenty, plenty (.) plenty, plenty (.) plenty, plenty.
• ADU: it becomes a flock.

In Example 4, the mother attempted to explain the concept of “swarm.” While observing it depicted in the book, she elaborated by describing it in Kurdish. To emphasize the meaning, she used multiple utterances (lines 2–5) and emphasized the quantity of bees in both Kurdish and Norwegian (line 4). This example illustrates the mother's elaborated description of “swarm” and her use of dual languages.

2.7 Interrater reliability

A trained researcher and the first author independently coded 20 transcripts to identify: (1) occurrences of labeling and word explanations, (2) the utterance that initiated the start of a labeling and word explanation sequence, and (3) the utterance that marked the end of each of these sequences. Interrater agreements (Cohen's kappa) were 0.82, 0.95, and 0.93, respectively, indicating good reliability. Disagreements were resolved through discussion.

2.8 Analytical approach

All language transcript analyses were conducted using CLAN (MacWhinney, 2000). Descriptive statistics and correlations (Table 1), the linear regression models (Table 2), and the linear mixed models (Tables 3, 4) were performed in JAMOVI (The Jamovi Project, 2026). To investigate how labeling and word explanations differed between dyads based on their dominant reading language (RQ1), we calculated the proportion of labeling and word explanation utterances relative to the total number of utterances. Thus, “labeling” refers to the proportion of labeling utterances, and “word explanation” refers to the proportion of word explanation utterances. We used standard linear regression modeling to compare these proportions between dyads (Table 2).

Table 1. Means, SDs, Ns, and correlations across key variables.

ChildAge, children's age in months; MotherEdu, maternal education level; LAB, proportion of labeling utterances; WE, proportion of word explanation utterances; LAB_CS_P, proportion of parents' code-switching utterances during labeling; LAB_CS_C, proportion of children's code-switching utterances during labeling; WE_CS_P, proportion of parents' code-switching utterances during word explanation; WE_CS_C, proportion of children's code-switching utterances during word explanation; L1 vs. L2, Mostly First Language Group coded as “1” and Mostly Second Language Group coded as “2.” M and SD represent mean and standard deviation, respectively. ^***p < 0.001; ^**p < 0.01; ^*p < 0.05.^† < 0.10.

Table 2. Linear regression models predicting the proportions of labeling utterances and word explanation utterances out of total number of utterances (N = 151).

L1 vs. L2, mostly first language group coded as “1” and mostly second language group coded as “2”; MotherEdu, maternal education level; ChildAge, children's age in months.

Table 3. Linear mixed models predicting parent and child proportion of code-switching utterances out of total number of utterances (N = 151).

LAB vs. OTH, proportion of code-switching utterances during labeling vs. other utterances, labeling coded as the reference; MotherEdu, maternal education level; ChildAge, children's age in months; L1 vs. L2, Mostly First Language Group coded as “1,” Mostly Second Language Group coded as “2.”

Table 4. Linear mixed models in the subset sample predicting parent and child proportion of code-switching utterances out of total number of utterances (N = 41).

WE vs. LAB, proportion of code-switching utterances during word explanation vs. labeling, word explanation coded as the reference; WE vs. OTH, proportion of code-switching utterances during word explanation vs. other utterances, word explanation coded as the reference; MotherEdu, maternal education level; ChildAge, children's age in months. L1 vs. L2, Mostly First Language Group coded as “1” and Mostly Second Language Group coded as “2.”

To examine dyads' code-switching during labeling and word explanations (RQ2), we calculated the proportion of code-switching utterances relative to the total number of utterances within each category of labeling and word explanation to account for differences in the amount of talk, separately for parents and children. Thus, “code-switching” refers to the proportion of code-switching utterances within each category. Additionally, we calculated the proportion of code-switching utterances relative to the total number of utterances in the remaining talk that we did not code. These remaining utterances do not include labeling or word explanations, which we hereinafter refer to as “other utterances.” This allowed us to compare whether code-switching differed between labeling and word explanations, and also whether either category differed from other utterances. Thus, for each parent and child, three code-switching variables were created: one that indicated the proportion of code-switching during labeling, one during word explanations, and one during other utterances. Note that our main focus is the comparison between labeling and word explanations, but we include other utterances to provide additional context to assess whether differences in code-switching are specific to the comparison between labeling and word explanations.

To compare code-switching across different types of talk, we restructured the dataset to long format, so that each dyad had a single code-switching variable repeated across three rows (one indicating code-switching during labeling, one during word explanations, and one during other utterances). We created two such code-switching variables, one for parents and one for children. A dummy variable was used to index each observation, indicating whether code-switching occurred during labeling, word explanations, or other utterances. We used linear mixed modeling to compare these proportions within dyads (Tables 3, 4), as it accounts for the dependence among the multiple observations from the same dyad.

Since only a subset of dyads used word explanations and most used labeling, we first compared code-switching during labeling to other utterances using the full sample (Table 3). We then analyzed the subset who engaged in word explanations by comparing their code-switching during word explanations to both labeling and other utterances (Table 4). This approach allowed us to use the available data before restricting analyses to the subset. This subsetting was done for RQ2 because our goal was to examine within-dyad differences in code-switching, unlike in RQ1, where we focused on between-dyad differences in labeling and word explanations. All models are presented both without and with control variables.

2.9 Control variables

For RQ1, we controlled for maternal education level, as it has been associated with word explanations (Evans et al., 2011) and child-directed talk. Because 113 (75%) children read with their mothers and since maternal education is a strong indicator of family SES, we applied maternal education as a control across all dyads, including those who read with their fathers. We also controlled for children's age, given its association with parental talk (e.g., Martin, 1997; Quiroz et al., 2010) and code-switching (Kremin et al., 2022). For RQ2, we additionally controlled for dominant reading language, as previous research has shown that parental code-switching varies depending on their dominant reading language (Gonzalez-Barrero et al., 2021).

3 Results

3.1 Descriptives and correlations

The total audio recordings yielded 21 h and 53 min of data, with individual recordings ranging from a minimum of 1.29 min to a maximum of 26.3 min (M = 8.70, SD = 4.74), indicating notable variation in how long dyads spent reading the book. The total absolute number of utterances for all dyads combined was 26,418 utterances. Labeling was observed in 143 (95%) dyads, word explanations in 41 (27%) dyads, and 40 (26%) dyads demonstrated both labeling and word explanations. Among the dyads who labeled objects and events in the book, code-switching during labeling occurred in 59 (41%) dyads, and among the dyads who provided word explanations of elements in the book, code-switching during word explanations occurred in 22 (54%) dyads. In total, labeling comprised 2,361 utterances, and word explanations comprised 458 utterances, indicating that labeling was overall more frequent than word explanations. Not surprisingly, dyads used more utterances to explain words than they did when labeling, with a labeling sequence consisting of on average four utterances, whereas a word explanation sequence consisted of eight. Parents produced 1,414 (60%) of the utterances during labeling and 298 (66%) utterances during word explanations, suggesting that both labeling and word explanations were largely parent-driven.

Dyads who primarily read in their L1 and those who primarily read in their L2 did not differ in reading duration [t (149) = 0.99, p = 0.325], maternal education [t (134) = −0.27, p = 0.786], nor whether the child read with their mother or father [X²(1) = 0.00, p = 0.951]. However, dyads who primarily read in their L1 produced more utterances [t (149) = 2.60, p = 0.010], included more boys [X²(1) = 8.30, p = 0.004], and consisted of younger children [t (149) = −2.46, p = 0.015]. Since dyads who primarily read in their L1 included more boys, we conducted all main analyses controlling for gender, which did not change any of the results.

As shown in Table 1, labeling and word explanations were positively correlated [r (149) = 0.17, p = 0.039], indicating that dyads who engaged in more labeling also provided more word explanations. Parents' code-switching during labeling was positively correlated with children's code-switching during labeling [r(135) = 0.75, p < 0.001]. Similarly, parents' code-switching during word explanations was positively correlated with children's code-switching during word explanations [r (37) = 0.57, p < 0.001]. These results suggest that parents' and children's code-switching were closely related in both labeling and word explanations.

3.2 Labeling and word explanation between dyads (RQ1)

Table 2 presents the linear regression analyses predicting labeling and word explanations. There was a significant association between dominant reading language and labeling, whereas no significant association was found for word explanations. This indicates that dyads who primarily read in their L1 engaged in more labeling than those who primarily read in their L2, while word explanations did not differ based on the dominant reading language. These results remained consistent when control variables were included. Furthermore, there was a significant negative association between children‘s age and labeling, indicating that labeling was more frequent in dyads with younger children. Maternal education was positively associated with word explanations, suggesting that word explanations were more frequent in dyads where the mother had a higher education level.

3.3 Code-switching during labeling and word explanation (RQ2)

Table 3 presents the results from the linear mixed models predicting parents' and children's code-switching in the full sample. In the full sample, there was no significant difference in either parents' or children's code-switching between labeling and other utterances (labeling and word explanation utterances excluded from the count of other utterances). These results indicate that neither parents' nor children's use of both languages differed across labeling and other utterances. Children's overall code-switching was significantly associated with the dominant reading language, indicating that children in dyads who primarily read in their L1 alternated languages more than those who primarily read in their L2. Children's overall code-switching was only marginally negatively associated with maternal education, suggesting that children with mothers who had a higher education level alternated less. These associations were in the same direction but not significant for parents' overall code-switching.

Table 4 presents the results from the linear mixed models predicting parents' and children's code-switching in the subset of dyads who engaged in word explanations. In this subset, there was a significant association between parents' code-switching and word explanation vs. labeling, as well as word explanation vs. other utterances. The negative estimate indicates that parents alternated languages more during word explanations, both compared to labeling and other utterances (as word explanation was coded as the reference). There was no significant association in children's code-switching between word explanations and labeling, but only between word explanation and other utterances. Children's overall code-switching in this subset was also significantly associated with the dominant reading language, indicating that children in dyads who primarily read in their L1 alternated languages more than those who primarily read in their L2. This association was in the same direction but not significant for parents' overall code-switching.

Finally, we further explored whether the association between parents and children's code-switching and word explanation vs. labeling was moderated by dominant reading language. To do this, we extended the linear mixed models presented in Table 4 by including an interaction between code-switching and word explanation vs. labeling, for parents and children separately. The interaction was not significant for parents (model without control variables: B = 3.16, p = 0.727; model with control variables: B = 2.89, p = 0.750), which indicates that parents used more code-switching during word explanation compared to labeling irrespective of the language they primarily used for reading. The interaction was also not significant for children (model without control variables: B = −2.78, p = 0.769; model with control variables: B = −3.26, p = 0.730), which indicates that children's use of both languages did not differ during word explanations and labeling, regardless of the language they primarily used for reading.

4 Discussion

The present study is among the few observational studies that examine interactions between bilingual parent–child dyads during shared reading. Drawing on parent-child dyads' naturally occurring talk, we examined how two ways of turning attention to words differed across dyads based on the language they mainly used for reading, and we identified when dyads alternated between languages most frequently. The study revealed the following main findings: (1) dyads who primarily read in their L1 engaged in more labeling but did not explain words more than dyads who primarily read in their L2, and (2) parents code-switched more when explaining words than when labeling, irrespective of the language they primarily used for reading.

Regarding the first main finding, parents who primarily read in their L1 may have adapted their communicative behaviors by using labeling as a strategy to support their children's word knowledge. Because parents are typically attuned to their children's language abilities and needs, their use of labeling likely reflects an effort to scaffold language learning in their L1. This builds on our qualitative study of seven bilingual dyads' reading interactions (Nomat et al., 2024), where we noted that parents who used their L1 more appeared to maintain dialogue in that language by asking children to say words in their L1. Indeed, language-minority children have been shown to use their L2 more compared to their parents (Pan, 1995). Thus, parents who wish for their children to maintain their L1 play an important role in using communicative strategies that create opportunities for language learning in their L1. Parents of young DLLs, who are exposed to the majority language in preschool, as is the case for language-minority children in Norway, are usually aware of this role (De Houwer, 2017).

The lack of a difference in word explanations between dyads may be related to their infrequent occurrence. The infrequent use of word explanations is consistent with previous studies of monolinguals reading textbooks (e.g., Evans et al., 2011; Korat et al., 2018). These findings on how two ways of turning attention to words differed between dyads based on their dominant reading language builds on an experiment by Quirk et al. (2024) and parent-reported studies by Read et al. (2021) and Avelar et al. (2025). Specifically, Read et al. (2021) found that parents' self-reported dialogic reading strategies did not differ across languages, whereas Avelar et al. (2025) found that parents' self-reported interactive reading strategies were more common when parents read in their preferred language. Findings from the present study both align with and diverge from these prior studies, suggesting that some types of talk might differ between dyads with different dominant reading languages. By examining the audio-recorded interactions, this study contributes new insight by illustrating bilingual parent-child reading interactions in a natural context.

While maternal education was not related to the language dyads primarily read in, it was positively associated with dyads' use of word explanations. This suggests that maternal education may be more relevant for features of talk than which language dyads primarily read in. Previous studies have reported mixed findings regarding the relationship between maternal education and word explanations, with Evans et al. (2011) reporting a positive association and Korat et al. (2018) finding a negative relationship.

Regarding the second main finding, parents may have drawn on their bilingual resources as a communicative tool or strategy to scaffold their children's understanding. Parents are known to be attuned to their child's signals during their interactions and in turn they can adapt their response and input according to their child's needs (Luo and Tamis-LeMonda, 2017; Tamis-LeMonda et al., 2001). Although we cannot draw conclusions about parents' and children's intentionality from our observational data, it is conceivable that the more elaborative nature of word explanations compared to labeling may have encouraged parents to use their resource of using both of their languages to scaffold their child's comprehension. This finding provides empirical support for previous suggestions, such as Bail et al. (2015), who proposed that parents' code-switching might differ when teaching words compared to simply labeling. It also builds on previous studies where parents reported using both languages when the extra-textual dialogue became more complex (Read et al., 2021) and to teach new words (Byers-Heinlein, 2013). The present study extends their work by examining whether parents' and children's code-switching differed across two ways of engaging with words. Thus, it provides insight into when parents and children naturally code-switch during reading interactions, thereby contributing to ongoing research on the nature of code-switching in parent–child shared reading (e.g., Gonzalez-Barrero et al., 2021; Quirk et al., 2024; Read et al., 2021).

Parents' greater code-switching during word explanations, irrespective of their dominant reading language, illustrates how parents may transcend their own language preferences and attune their communicative behaviors in ways that scaffold children's understanding and learning. A similar pattern was reported by Quirk et al. (2024) in a context where both languages had a similar majority status. Interestingly, we found no difference in children's code-switching between word explanations and labeling. Indeed, preschool-aged DLLs have been shown to code-switch primarily to fill lexical gaps rather than for other pragmatic functions, such as emphasis (Halpin and Melzi, 2021). The difference in code-switching that we observed for parents, but not for children, suggest that parents showed more systematic patterns of code-switching across types of talk during their reading interactions.

We also observed notable patterns in dyads' overall code-switching. Overall, code-switching was common, as most parents and children actively used both of their languages to varying degrees. Children in dyads who primarily read in their L1 alternated languages more than children in dyads who primarily read in their L2. This greater use of Norwegian is unsurprising, given their frequent exposure to Norwegian through preschool with peers and teachers. Similarly, Halpin and Melzi (2021) found that children who were Spanish-dominant (L1) code-switched more than those who were English-dominant (L2) when reading with a bilingual researcher. For parents, this difference in overall code-switching between dyads who primarily read in their L1 compared to those who primarily read in their L2 was not significant, albeit in the same direction as that observed for children.

4.1 Limitations and future directions

The present study has several limitations. We categorized dyads into two groups, whereas other studies have included a third “balanced” group (e.g., Avelar et al., 2025; Mancilla-Martinez and Lesaux, 2011; Read et al., 2021). Including such a group may acknowledge the heterogeneity of bilinguals' language use, whereas our categorization may have overlooked some of these nuances. However, only seven dyads in our dataset would have fitted into the balanced group, which was too small for a meaningful comparison. Moreover, because few dyads used word explanations, we restricted the analysis of code-switching differences to this subset in order to examine within-dyad differences between word explanations and labeling. This limitation underscores the need for future research to test these findings with larger samples and in diverse contexts. We did not differentiate between inter-sentential and intra-sentential code-switching in our analysis, as we aimed for a broad exploration of code-switching patterns consistent with prior research (e.g., Kaushanskaya and Crespo, 2019; Read et al., 2021). Future research could extend this by examining how distinct code-switching behaviors vary across talk about words.

In the present study, parents' proficiency in their two languages was not assessed. Thus, we cannot determine whether the language they primarily used for reading or whether their code-switching patterns were related to parents' language proficiency in their L1 and L2. Future research could examine this. It is also worth noting that most children read with their mothers during our reading observation. However, we instructed parents to read with their child as they normally would at home. Thus, the fact that the majority of children read with their mothers likely reflects the dyads' usual reading practices. Nevertheless, it would be worthwhile for future research to explore potential differences between mother-child and father-child dyads using more controlled designs or larger samples that permit such comparisons, or by observing the same child reading with each parent separately.

Moreover, nearly all dyads engaged in some word-focused talk, most commonly through labeling and less frequently through word explanations. This general emphasis on words during shared reading aligns with a large body of literature emphasizing reading as a context for supporting children's vocabulary development (Ninio and Bruner, 1978; Noble et al., 2018; Rowe and Snow, 2020). Given the overall purpose of reading wordless picture books is to construct a narrative from the illustrations, dyads naturally engaged in talk beyond labeling and word explanations. While the present study specifically focused on talk that directed attention to words, parents' greater use of code-switching during word explanations might extend to other elaborative talk such as narrative talk, where dyads make past and future connections across the book and relate the story to their own experiences (Rowe, 2013). For instance, during narrative talk, parents might engage in more code-switching compared to labeling because making connections across the book and relating the story to the children's own experiences might require more elaboration than simply labeling. Thus, future research could explore these possibilities as well as potential variation across dominant reading languages.

Finally, although maternal education was included as an indicator of family SES and controlled for in the present study, our aim was not to examine potential family SES variations. Future research could include a broader measure of family SES to explore nuances in bilingual parent-child dyads' reading practices, for instance, by comparing dyads from different SES backgrounds. Relatedly, future studies could examine the role of language status, such as parents' perception of language status, as language-minority groups likely differ in how they perceive the language status of their L1 and L2.

5 Conclusion

This study provides new insight into how bilingual parent–child dyads naturally use their two languages during shared reading of a wordless book at home. Dyads who primarily read in their L1 engaged in more labeling, but not word explanations, compared to dyads who primarily read in their L2. Furthermore, parents code-switched more during word explanations than during labeling, irrespective of their dominant reading language, suggesting that parents' code-switching might be more relevant in more elaborative interactions. These findings provide insight into how bilingual parent-child dyads use their languages during shared reading and when children are most frequently exposed to code-switching across two types of talk that may support word understanding, thereby contributing to our understanding of how bilingual acquisition takes place in a natural context during early childhood. A broader implication of our findings is that parents should recognize that drawing on both of their languages during shared reading can function as a valuable dialogic resource that can help scaffold children's understanding in elaborative interactions.

Generative AI statement

The author(s) declared that Generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Pui Fong Kan, University of Colorado Boulder, United States

Reviewed by: Kirsten Read, Santa Clara University, United States

María Jesús Lorenzo-Modia, University of A Coruña, Spain