Media Exposure and Language Experience: Examining Associations from Home Observations in Mexican Immigrant Families in the US

Abstract

The current exploratory study describes exposure to digital media in young children from Mexican immigrant homes and its association with language input and output. Using multiple recordings of children’s home environments, we report on the rate (i.e., percentage of total recording time), language (Spanish or English), and type (adult- or child-directed programming) of auditory media exposure in toddlers under three years of age (N = 30; M_age = 20;3). We also examine total adult words and adult-child conversational turns, as indicators of child language input, and the rate of child language vocalizations as a measure of early language development. Findings showed that digital media comprised approximately 14% of the child language environment that families selected to record, with wide variability observed. Children were more likely to be exposed to media in Spanish than English and adult-directed than child-directed programming. Children’s general media exposure was negatively associated with the amount of children’s vocalizations and conversational turns but not the quantity of adult words in the environment, suggesting that the relation between media exposure and child language development is likely not mediated by a general decrease in adult input in Mexican immigrant homes. Instead, media exposure may decrease opportunities for children to engage in conversation and practice language expression, both critical mechanisms for successful language acquisition. Selection of child-directed programming may encourage child vocalizations but is less likely to be in Spanish in these homes, which may reduce opportunities for engagement with Spanish-dominant adults. Together these findings provide a window into the nature of media exposure in children from Mexican immigrant homes and into the associations between media and language input and output. Directions for future research are discussed.

1.1. Introduction

The first three years of life are a remarkable period of cognitive development. In particular, children’s early language skills develop at rapid pace. By 12 months, children have said their first word, embarking on a vocabulary explosion over the next year that ultimately supports two- and three-word utterances into age 3. All basic models of language acquisition recognize the importance of the language learning environment. For infants and toddlers specifically, it is well-accepted that the quantity and quality of the language input coming from caregivers directly impacts the rate of early vocabulary acquisition in children, and this has been shown for children of different races, ethnicities, and language backgrounds (see Head Zauche et al., 2016 for a review). Yet, other people are not the only significant source of information in the environment for the young infant decoding language.

Indeed, over the past several decades children have become increasingly exposed to media consumed through electronic devices (i.e., digital media), from radios to television (TV), to computers, and more recently mobile phones. Even for infants and toddlers, the rate of media exposure is significant. In a recent population-based survey in the United States (US), the number of children between 2 and 24 months of age that regularly watched TV, movies, and videos increased between 3 and 24 months of age (Zimmerman et al., 2007). Specifically, 40% of 3-month-olds regularly watched TV, movies, or videos, whereas the proportion rose to 90% by 24 months of age. In the US, parents report that children (including Hispanic children) watch approximately 1 to 3 hours of media a day by age 2 (Duch et al., 2013; Radesky et al., 2014; Zimmerman et al., 2009). The nature of media use is also changing drastically over time even in infancy and toddlerhood, with some parent estimates reporting rates of mobile media use among 2 to 4-year-olds increased from 39% to 80% between 2011 and 2013 (Canadian Pediatric Society, 2017). As such, it is important to understand how characteristics of media use impact children’s language learning. Of interest in the present study are characteristics such as the rate with which children experience media, the types of media programming to which they are exposed, and, the language(s) used during media exposure. Further, given the complex and culturally-informed ecologies of media use and limited attention to children from diverse backgrounds, this exploration is particularly needed in racially, ethnically, and linguistically diverse populations using observational methods to understand media exposure as it occurs in the homes of young children. Describing media use in specific communities can help to tailor family recommendations related to providing a supportive developmental context. The present study seeks to answer this call and add to the growing body of literature by examining the auditory experiences of children during media exposure and its relation to language input and output for the growing population of Spanish-learning toddlers in Mexican immigrant homes specifically. Though Mexican and Spanish-speaking families have likely been included as participants in extant literature, no peer-reviewed observational study exists to describe the role of media in the language learning environment of Spanish-learning infants and toddlers in the US specifically.

1.1.1. Media Exposure as a Cultural Process

Scholars studying media use consistently appeal to ecological systems theory to point out the intricate relationship between several micro, meso, and macro level factors influencing children’s access (Radesky et al., 2014). Within this theoretical framework, children’s families exert the most proximal influence on decisions around media use. Extant literature proposes a “family media ecology” that influences how and when media is used (Levine et al., 2017; Linebarger & Vaala, 2010). This family media ecology is further influenced at the macro level by the ever evolving and complex culture(s) with which families identify (e.g., race, ethnicity, socioeconomic status). Indeed, many scholars have constructed media use as a cultural process (e.g., Acerbi, 2016) governed by beliefs, practices, and values just as all other forms of information transmission. In other words, similar to variation in approaches to supporting child language development, families vary in their approaches to their child’s media consumption, and this variation appears to be related, in part, to families’ cultural backgrounds. Yet, the majority of research examining media exposure has focused primarily on White, English-speaking families. When families of other races, ethnicities, and languages are included in studies, non-White and non-English-speaking participants are part of the larger group sample and differences are typically not examined further.

Given the complex ecological framework influencing digital media exposure in young children, it is no surprise that the few studies that do examine differences as a function of race and ethnicity consistently report meaningful differences (Rideout et al., 2011; Kabali et al., 2015; Radesky et al., 2014; Radesky et al., 2016). For example, in a study of children between 0 and 6 years of age in 2006, Hispanic¹ children were more likely to spend time consuming media than their White counterparts (Rideout et al., 2011). Similarly, 39% of Hispanic children (0–6 years) had a TV in their room, compared to 27% of White children. Hispanic parents were also more than twice as likely to say their child had lunch or dinner in front of the TV compared to White parents. Moreover, speaking to the complexity of culture, differences have been found in perspectives on media consumption for families who share an overarching ethnic background (i.e., Hispanic families from varied home countries). For example, some parents who are Mexican immigrants to the US have reported using TV to support toddlers’ language acquisition generally, and their English learning in particular (Cycyk and Hammer, 2020). Other families of Latinx descent in the US have expressed concern about TV limiting the language development of their young children (Cycyk and Huerta, 2020). These noted differences within the Latinx community suggest that exploration of children’s media consumption and its impact on early language development benefits from continued and careful specification of family ecologies. The present study therefore seeks to build on this work by precisely characterizing the media consumption of Spanish-learning toddlers from Mexican immigrant families in the US specifically through direct multi-day observations of the home environment. These families represent an important subgroup of the Latinx community.

1.1.2. Media Use and the Language Learning Environment

In addition to characterizing digital media exposure in the home, researchers have sought to examine the impact of digital media on the language learning environment of young children. In particular, a negative link between media exposure and language development has been shown in infants and toddlers prior to age 3 primarily using global measures of child language, including within Hispanic children (Duch et al., 2013; Masur et al., 2016; Anderson & Hanson, 2017; Bittman et al., 2012; Chassiakos et al., 2016; Lin et al., 2015; Moon et al., 2019; van den Heuvel et al., 2019). For example, Lin and colleagues (2015) found that toddlers who presented with a language delay (as determined from a global measure of child development) reportedly watched more than 2 times the amount of television than their peers with typical language development. Although this finding has not been replicated in all contexts (Lin, 2020; Patterson, 2002), increased media exposure is generally associated with decreased child vocabulary in studies using direct observational methods rather than relying only on parent report (e.g., Christakis et al., 2009). Although these studies do not demonstrate causation, we can speculate as to the mechanisms underlying this association. To understand the possible mechanisms, it is important to first setup the framework by which children learn language.

Social-pragmatic views of language development posit that children learn their earliest words in the context of social interactions with more competent communicators (i.e., their caregivers). Within these culturally-influenced interactions, children engage in inferring meaning from caregivers while practicing production of newly acquired words and structures, which prompts feedback from interactional partners that further supports language learning (Bruner, 1983; Tomasello, 1992; 2000; 2009). Accordingly, caregiver input and child output play complementary and dynamic roles in language development. A wide body of research has demonstrated the importance of the quantity of caregiver input and the quality of the caregiver-child interaction to child language development prior to age 3 (Head Zauche et al., 2016). A number of input characteristics are thought to represent quantity and quality, with the sheer number of words heard and caregiver-child conversational turns experienced by children, respectively, being among the most commonly studied measures. The frequency with which children use their language has also been found to impact language outcomes over time, such that children who communicate more frequently with words and babbles (i.e., vocalizations) at younger ages demonstrate stronger expressive language skills at older ages (Brady et al., 2004; Calandrella & Wilcox, 2000; McCathren et al., 1999; Whitehurst et al., 1991).

Given this information, it becomes clear how media use could theoretically negatively impact early language learning. Specifically, media use may compete with children’s opportunities for rich social interactions with caregivers that support language development, displacing other parent-child activities that are more enriching to child language development (Nathanson & Rasmussen, 2011). In a series of studies, researchers showed that digital media changes the nature of parent-child interactions. Parents spend less time interacting with their infants and toddlers and end up talking less overall in the presence of media, regardless of whether the media is the focus of their attention or simply in the background (Anderson & Hanson, 2017; Kirkorian et al., 2009). In one population-based study of English speakers specifically, parents recorded their child’s environment when children were between 2 and 48 months of age. Researchers found that for every hour of TV consumed by children, parents spoke 770 fewer words to their child (Christakis et al., 2009). Similarly, for every hour of TV, the number of parent-child conversational turns as well as children’s vocalizations were reduced. These effects seem to extend to mobile devices, such that each additional 30-minute increase in daily mobile media device use is associated with increased odds of parents reporting that their child experienced delays in expressive speech (van den Heuval et al., 2019). The observable decreases in adult input and child output associated with concurrent media use may also reflect children’s language experiences when media is not present. In one study, the higher the use of TV at home, the less parents talked to their children when tested in the laboratory environment without the presence of the TV (Pempek et al., 2011). Taken together, these research studies suggest that the negative language outcomes reported may be due to a general decrease in total adult input and opportunities for caregiver-child engagement that may also be associated with higher frequencies of media use, even though the exact causal nature of the effect is unclear.

In addition to displacing richer social learning conditions necessary for language learning, the nature of digital media itself may hinder learning more broadly. The goal for a lot of media, even when targeted to children, is to encourage prolonged consumption through behavioral reinforcement (Canadian Pediatric Society, 2017). Few programs exist to promote language stimulation specifically. In one study, researchers carefully examined the content of DVDs marketed for infants (Goodrich, 2008). In general, TV for infants relies heavily on perceptually salient features that promote visual engagement. These features include use of a rapid pace, camera cuts, and visual and auditory special effects, among others. Similar engagement techniques are used for mobile media (Canadian Pediatric Society, 2017). Though these characteristics promote visual engagement, they are not the best learning conditions for language comprehension to take place. In fact, research has shown that toddlers learn significantly fewer new words through television when compared to in-person human interaction, regardless of whether programming is geared toward children (Krcmar et al., 2007). In addition, higher media consumption is associated with poorer sleep patterns (e.g., LeBourgeous et al., 2017). Although the reason for this association with sleep is also not well understood, given the importance of sleep for general cognitive abilities and language learning specifically (e.g., Dionne et al., 2011), it is possible that this too serves a mechanism underlying the negative links between early digital media use and language delay reported in some studies.

Given the landscape of digital media in infants and toddlers and the possible negative consequences on language development, several governing agencies around the world have provided general guiding principles for parents (e.g., Bozzola et al., 2018; Canadian Pediatric Society, 2017; Council on Communications and Media, 2016; World Health Organization, 2019). These principles reiterate the importance of limiting overall screen time. They also specify that not all screen time is equal; active engagement with co-viewing partners and purposeful selection of media may overcome some of the limitations of digital media. Indeed, it seems that a consistent pattern in this research is that for media to be successful in promoting learning of any kind, then it must critically engage children. In one study, mothers who talked with their 14-month-olds while watching educational TV programs promoted stronger language outcomes than mothers who did not engage in co-viewing educational TV (Mendelsohn et al., 2010). In another study, co-viewing media increased the number of new words used per utterance when parents spoke to their children about the programming, another important indicator of input quality, despite the fact that parents spoke fewer different words overall (Lavigne et al., 2015). Yet, recent population-based surveys suggest that joint media engagement is not as common as individual use for young infants and toddlers (Vaala et al., 2015). Not all parents favor actively engaging their child during media exposure. For example, a study of 39 parents of young children of Latinx descent found that half of the parents accepted co-viewing as a strategy to support language development while consuming media (Cycyk and Huerta, 2020). The other half of parents expressed disagreement with this strategy due to limitations on their time and their belief that co-viewing would decrease children’s enjoyment of media. Further, it seems that joint media engagement (co-viewing) varies greatly and decreases with age (Anderson & Hanson, 2017).

Moreover, deliberate selection of the programming that children view may further support the potential for digital media to promote learning. Digital media in and of itself is not detrimental to development. A growing body of work has made great strides in describing the ways in which media can be an especially helpful tool in promoting learning outcomes (e.g., Barr, 2013; Barr & Wyss, 2008). Some educational programming directed to children specifically, such as Dora the Explorer, has been found to positively affect children’s language development (Linebarger & Walker, 2005). Indeed, the impact of viewing Sesame Street in particular in early childhood has consistently shown long-term benefits to children’s learning (Fisch, 2004). In contrast, programming directed at adults without a curricular focus negatively impacts children’s development and increases risk for language delay (Barr et al., 2010; Chonchaiya & Pruksananonda, 2008). These differences in child outcomes may be related to differences in children’s opportunities for communicating during specific programming: mothers have reported interacting more frequently with their infants when viewing child-directed content than content not geared toward children (Mendelsohn et al., 2008) and such co-viewing may be particularly facilitative for language development and cognitive outcomes broadly (e.g., Anderson & Hanson, 2017). Yet, based on parent report, most children watch YouTube or Netflix primarily, and smaller proportions watch educational programs or play early-learning apps (Kabali et al, 2015). In one of the only studies to examine Spanish-speakers of Hispanic descent specifically, Duch et al. (2013) found that child-directed programming was more commonly reported than adult-directed programming among infants and toddlers, and that child-directed but not adult-directed programming negatively influenced parent report on a screener of child communication. These authors hypothesized that Spanish-speaking parents were challenged to interact with their children in ways that support language learning during child-directed media viewing because these programs were in English. Thus, the type of media programming (i.e., child- versus adult-directed), and potentially, the language of media exposure, and the ways in which these characteristics of digital media may impact opportunities for language development in children from varied backgrounds deserves further attention. Importantly, most studies to date examining the impact of digital media on child development have focused on the amount of exposure rather than the characteristics of the content of the digital media to which children are exposed and often rely on parent report to capture the language and media in the environment (Kostyrka-Allchorne et al., 2017).

1.1.3. The Present Study

The present exploratory study seeks to build on the existing literature on digital media use and language learning in children in two ways. First, we will increase the diversity of family ecologies that are examined with a focus on a particular ethnic subgroup: children of Mexican immigrants in the US exposed to Spanish. Although many population-based studies on the topic of child media use include children of different races and ethnicities, the overwhelming focus is on English-learning and White families. Indeed, non-English learners specifically have been excluded from other observational studies of media use among infants and toddlers (e.g., Christakis et al., 2009; Zimmerman et al., 2009). It is conceivable that different racial, ethnic, and language groups may expose young infants and toddlers to digital media in varying amounts across multiple languages and sources. Such differences are important for devising recommendations and interventions that are precise, equitable, and responsive to the needs of diverse families to mitigate the potential negative impact of media use on child language development. The context of Mexican immigrant homes in the US is especially interesting because while the primary home language is often Spanish (Gonzalez-Barrera & Lopez, 2013), the majority language in the US is English. These families often make both explicit and implicit decisions about the language environment of the home, and children’s exposure to Spanish and English in Mexican immigrant contexts is variable (Cycyk and Hammer, 2020).

Second, the present study will increase the precision used to characterize children’s media use in these homes. Within extant literature, the primary method of data collection is parent report of child behavior related to media use (e.g., parent interviews, questionnaires). Although parent report is an efficient and reliable tool, it also presents limitations for ecological validity. More recently, technology has allowed researchers to capture children’s media use in their home environments directly (e.g., Christakis et al., 2009). This method also allows us to add to the scant literature describing the content of children’s media exposure and to contribute to understanding the language of media in multilingual homes, which has not yet been explored.

The present study has four primary research questions. First, what is the rate of digital media exposure for Spanish-learning infants in Mexican immigrant families? Second, what is the language of digital media exposure for Spanish-learning infants in Mexican immigrant families? Third, we ask about the type of media exposure: what proportion of media in the auditory environment is adult- vs. child-directed? Fourth, what is the association between rate, type, and language of media exposure with child vocalizations and the quantity of adult words and adult-child conversational turns in the auditory environment in Mexican immigrant homes? With this question, we sought to characterize the associations between several factors in the larger auditory language learning environment to further understand the role of digital media in child language development. In the present study, we employ multiple naturalistic audio recordings from children to serve as home observations of digital media exposure (i.e., rate, language, type) and to estimate language input (i.e., number of adult words and adult-child conversational turns) and child language output (i.e., vocalizations). The number of child vocalizations recorded serves as the indicator of child language abilities in this study because this variable (a) is operationalized to include both early word productions and intentional communicative babbles used by children in infancy and toddlerhood; and, (b) is associated with future expressive language development (McCathren et al., 1999; Whitehurst et al., 1991). Further, it is language input and child output that together interact to support language development as described previously. Examining both language input and child output therefore helps us understand how the rate, type, and language of media in the environment is associated with fundamental aspects of language acquisition.

We expected that all children will be exposed to digital media in their homes at variable rates. Given prior research findings with infants and toddlers based on day-long recordings and parent report (Duch et al., 2013; Radesky et al., 2014; Zimmerman et al., 2009), we hypothesized that an average of 8 to 13% of children’s audio environment recorded in the present study will be characterized as media exposure but with substantial variability across children. Moreover, we predicted that Spanish will be the primary language of the media these children consume and there will be large variability in children’s access to English language media. While we are not aware of prior research that has directly examined the language of television in Mexican immigrant homes, observational studies examining the languages heard by young children from caregivers in similar contexts lend support to our hypothesis (Marchman et al., 2017; Wood et al., 2016). For example, Marchman et al. (2017) found that 3-year-old children of Mexican immigrant parents were exposed to nearly 18 times more words in Spanish than English during naturalistic home recordings of the audio environment; however, all children were also exposed to some words in English. In addition, we hypothesized that children would have significant exposure to both adult- and child-directed media types (Barr et al., 2010; Mendelsohn et al., 2008) but more exposure to child-directed programming (Duch et al., 2013).

Previous findings generally show that greater amounts of TV are linked to poorer language skills among young children (Kostyrka-Allchorne et al., 2017); thus, we anticipated that children’s rate of digital media exposure will be negatively associated with the frequency of child vocalizations. Although the findings have been mixed (Barr et al., 2010; Chonchaiya & Pruksananonda, 2008), Duch et al. (2013) found that child-directed programming was negatively associated reports of child language in Hispanic homes specifically. As such, we hypothesized a negative association between the proportion of child-directed programming and child vocalizations in this sample. We further anticipated that child language input will be negatively associated with overall rate of media exposure as well as frequency of exposure to adult-directed media in particular (Christakis et al., 2009; Mendelsohn et al., 2008). As no prior research has directly explored the language of television in multilingual homes, we did not have a hypothesis for the links between the language of television, child language input, or child language output.

1.2. Method

1.2.1. Participants

Thirty children from Mexican immigrant households participated. The children averaged 19.73 months of age (range = 15–27 months). Half were female. All children were born in the United States, and all of their parents were born in Mexico. All families had two parents living in the home (see Table 1 for demographic information about mothers and fathers). The median number of total people living in the home was 5, broken down by 3 adults and 2 children (M_Total = 6.586, SD = 2.693; M_Adults= 3.310, SD = 1.365; M_Children = 3.276, SD = 2.016). Fifty-six percent of the children’s mothers and 73% of the children’s fathers did not complete high school. On average, families had a Hollingshead Index raw score of 15.638, (SD = 4.851) reflecting that the sample as a whole was relatively low SES (Hollingshead, 1975). The primary language spoken in all homes was Spanish, and all mothers and fathers reported being native speakers of Spanish. Parents reported no concerns with their child’s development. Sixty percent of mothers, and 60% of fathers used only Spanish with their child. Approximately 3% of mothers used English and Spanish an equal amount of time when speaking with their child. Parents reported that children heard Spanish from 96% of household members. The majority of household members spoke Spanish almost exclusively to children (54.9%), with 68.8% using more Spanish than English overall. Every home had a television, and 89% of parents reported that media viewing was a daily activity for their child.

Table 1.

Demographic characteristics of children’s mothers and fathers.

	Mothers		Fathers
	%	M (SD)	%	M (SD)
Age (years)		29.17 (4.61)		30.70 (4.94)
Employed	36.70		96.70
Educational level	53.30
Some schooling	36.70		76.70
High school diploma	3.30		13.30
Vocational Training	--		6.70
Some college but no degree	3.30		3.30
Associate’s degree	3.30		--
Bachelor’s degree			--
Language to child All Spanish	60.00		56.70
More Spanish	33.30		43.30
English and Spanish equally	6.70		--

1.2.2. Measures and Procedures

Below we describe the measures and procedures for the independent and dependent variables of interest as well as for the collection of demographic data in the present study. Table 2 provides a brief description of these variables by research question.

Table 2.

Description of study variables by research question.

RQ	Variable of Interest	Description
1	Rate of digital media exposure	Percentage of total LENA audio recorded (hours:minutes:seconds) categorized as TV/ES
2	Language of digital media exposure*	Percentage of total minutes of TV/ES coded as occurring in (a) Spanish, and (b) English
3	Type of digital media exposure*	Percentage of total minutes of TV/ES coded as (a) adult-directed, and (b) child-directed
4	Associations between media exposure (rate, language*, and type*), child vocalizations, and the quantity of adult words and adult-child conversational turns	• Rate of digital media exposure (as described above) • Language of digital media exposure (as described above) • Type of digital media exposure (as described above) • LENA estimated average count of child vocalization (CVC) per hour • LENA estimated average count of adult words (AWC) per hour • LENA estimated average count of conversational turns (CTC) per hour

Note. RQ = Research Question; LENA = Language ENvironment Analysis; TV/ES = television and electronic sounds.

^*Includes only 5-minute segments in which TV/ES represented at least 50% of the total audio categorized for that segment (i.e., 2.5 minutes)

1.2.2.1. Demographic Questionnaire

The Center for Early Childhood Education – Dual Language Learners (CECER-DLL) Child and Family Questionnaire was used to gather demographic information on children and their families. This questionnaire was developed and validated with Spanish-speaking Latinx families of young children (Castro et al., 2020; Hammer et al., 2020). Mothers were provided with a written copy of the questionnaire to complete in Spanish or English or were read the questions aloud if they preferred. A total of 44 questions asked about child and caregiver demographics, child language experiences, and household characteristics.

1.2.2.2. Home Audio Recordings

The Language ENvironment Analysis Digital Language Processor (LENA DLP; LENA ProSystem, 2012) was used in the present study to audio record the home environments of young children. The LENA DLP is a small recorder that houses a microphone to capture up to 16 continuous hours of the child’s vocalizations as well as the auditory environment. It was specifically designed to capture the language experiences of children between ages 2 and 48 months across a full day. The DLP sits within the pocket of a small vest worn by the child, such that the microphone is on the child’s chest and just below their face. The DLP is then connected to a computer with LENA software to provide automated data on the child’s audio environment. This software automatically sorts the audio by recording and by audio input using acoustic features and computer algorithms to estimate characteristics of the child’s language environment.

The LENA and its audio recognition algorithms were originally developed from day-long recordings of English monolingual speakers; however, its use has been expanded successfully to multilingual contexts, including contexts in which Spanish is spoken (e.g., Jackson & Callender, 2014; Marchman et al., 2017; Weisleder & Fernald, 2013; Wood et al., 2016). Relevant to the current study, agreement between segmentation of LENA automated data and human coding of English language recordings in the initial LENA validation study was found to be 82% for adult speech, 76% for child vocalizations, and 71% for television and electronic sounds (Xu et al., 2009). While the LENA has not yet been extensively validated in Spanish-speaking contexts, Weisleder and Fernald (2013) found that the LENA estimate of the number of adult words used in Spanish-speaking homes was strongly correlated with the estimates of human transcribers. In general, there is moderate to good agreement between human coders and the LENA software’s identification of most features of adult and child speech and television in English and other language contexts (Christakis et al., 2009; Cristia et al., 2020a; 2020b Zimmerman et al., 2009), particularly when there is little competing noise. More recent research has found that LENA’s automated identification of electronic sounds is sometimes mistaken for overlapping speech and therefore may be underestimated (Bulgarelli & Bergelson, 2020). Similarly, LENA’s conversational turns can be overestimated in environments with many speakers, though accuracy generally improves for children between 14 and 24 months of age as compared to 6 to 13 months of age (Ferjan Ramírez et al., 2021). Indeed, generally speaking, the LENA automation also appears to be most accurate for adult word and child vocalization counts as compared to adult-child conversational turn counts (Cristia et al., 2020a; 2020b). We used relative estimates (i.e., proportions) in analyses to overcome limitations of over- and under-estimation as recommended by researchers (i.e., Bulgarelli & Bergelson, 2020). See section 1.2.2.3. for further details.

Following informed consent, families were instructed to complete recordings across three typical days for their child. To capture typical variability across weekends and weekdays, mothers were instructed to record a minimum of 4 hours on two weekdays and 8 hours on one weekend day on a single LENA DLP. While contiguous recordings of a single day are recommended per the intent of the LENA developers (Gilkerson & Richards, 2020), prior work has noted that LENA estimates collected from separate days are highly correlated (Weisleder & Fernald, 2013). Families in the current study were encouraged to record only while the family was at home. Families were also provided with a written log to document the days, times, activities, and participants recorded. The log was available in Spanish and English and included check boxes for common activities (e.g., naptime, mealtime, watching TV) and participants (e.g., mother, father, sibling) as well as options to write in activities/participants that did not appear on the list available on the log.

Families made between 2 and 13 different recordings on days and at times they preferred. An average of 15.12 hours of total recording time per child (SD = 3.01 hours; range = 6.90 – 24.58 hours) were captured for the present study. Prior studies using the LENA with Spanish-speaking families have also included recordings from multiple days and recordings of similarly varying durations and contexts (Jackson & Callender, 2014; Marchman et al., 2017; Weisleder & Fernald, 2013; Wood et al., 2016). While families mainly recorded in their homes, some families also used the LENA outside of the home (e.g., during car travel, at the playground). A total of 5 families (17%) completed the recording logs in full for all days of recordings while 12 families (40%) returned partially completed logs for some days they recorded. Thirteen families (43%) did not return the recording logs. The recordings took place between 2014 and 2015.

1.2.2.3. Data Processing

Several key variables were extracted from the audio environment recorded by the LENA DLP using the LENA Advanced Data Extractor software (LENA Research Foundation, 2012). See Table 2 for an overview. Any recordings less than 1 hour in length were eliminated. While the accuracy of the LENA automated estimates improves as the length of the recording increases, Xu et al. (2009) specify that 1 continuous hour of recording will return relatively accurate automated estimates. LENA software categorized the child’s overall audio environment in 5-minute increments. The software categorized each 5-minute segment into the following: TV & ES (television and other digital and electronic sound exposure), meaningful (clear human input that occurs less than 6 feet away from the child), distant (human input that occurs more than 6 feet away), noise (any noise that is not categorized as speech or electronic noise), and lastly, silence & background (LENA Research Foundation, 2015). The software also provides information on total hours/minutes/seconds of each input category and the proportion of these categories by total recording time in comparison to other auditory input. For the purposes of our study, the LENA-estimated percentage of TV & ES audio data across the recording length (i.e., total duration of audio denoted as TV/ES divided by total LENA recording duration in hours:minutes:seconds) served as children’s rate of auditory media exposure, a key variable of interest. Note that rate of media exposure in this study does not represent the proportion with which children were exposed to TV/ES in a single day from waking to sleeping but rather the proportion of the LENA recordings submitted by families that were characterized as TV/ES.

Next, we quantified the language (Spanish vs. English) and type (child- vs. adult-directed) of auditory media exposure. To do so, each 5-minute increment in which television and/or electronic sounds represented 50% (i.e., 2.5 minutes) or greater of the recording were selected for further analysis. Specifically, for each selected 5-minute segment, trained Spanish-English bilingual research assistants listened and assigned a language (Spanish or English) and media type (adult- vs. child-directed). To distinguish between adult- and child-directed programming, coders attended to language, music, and context of the media audio recording. Audio with complex language typically indicated that the program was directed to an adult audience, whereas simple sentences and words are used in programming to children (Goodrich, 2008). Similarly, child-directed programming typically has music, infant-directed speech, and distinct sounds that support the topics being talked about and introduced. Further, the events, characters, and story line of the program itself often allowed coders to identify whether the media was intended for adults or children. In a small number of cases, it was possible to note the specific program that children were viewing (e.g., Thomas and Friends). In instances where the media program switched languages exactly half way through, the audio file was categorized as 50% English and 50% Spanish. The same logic was applied if the media program type switched halfway through the 5-minute segment. As with overall media exposure, percentages were calculated to determine the amount of media exposure that was in Spanish and English, as well as the amount that was adult- vs. child-directed for each participant. More specifically, the total number of minutes of TV/ES coded for each language and each type of programming were divided by the total number of minutes of TV/ES coded regardless of language or type of programming. A random subset of 20% of the data was coded by a second independent research assistant and showed acceptable reliability (Cohen’s k: 85% and 89% agreement for media type and language, respectively).

Lastly, we extracted data about the language environment. Specifically details about adult and child speech as well as conversational turns were extracted from the full length of the recordings of the auditory environment (not including recordings less than 1 hour in length). Specifically, LENA software quantifies the total number of adult words in the environment (Adult Word Count; AWC), representing caregiver input quantity in this study, and further delineates AWC by whether the speaker was an adult male or an adult female. LENA also estimates number of child vocalizations (CVC), defined as a segment of speech spoken by the child wearing the LENA DLP before and after a pause that lasts at least 300 milliseconds (LENA Research Foundation, 2015). Crying and vegetative sounds are not included in CVC. This variable represents child language output in the current study. AWC and CVC capture all of the adult and child vocalizations, respectively, and therefore make no distinction between direct conversation or overheard speech. As such, we also sought to quantify moments of interactional engagement between adults and children. Toward this end we extracted the conversational turns (CTC) variable which is defined by the LENA as a child vocalization that precedes or follows an adult word with no more than 5 seconds between the child and adult speech (LENA Research Foundation, 2015). CTC should be considered a proxy estimate of language input quality in our study rather than a direct, absolute measure of the frequency of adult-child engagement. Indeed, LENA only uses temporal information about adult and child speech to estimate CTC. This means that adult speech overheard by children that coincides at random with a child vocalization can be erroneously categorized leading to overestimation compared to manual coding (Ferjan Ramírez et al., 2021). To manage differences in recording lengths and overcome some of the over- and under-estimation limitations of LENA, relative measures (i.e., rates) for AWC, CVC, and CTC by hour were computed by dividing the child’s total AWC, CVC, and CTC provided by the LENA software for the recordings analyzed by the total number of hours and minutes of the recording length for each child across each child’s multiple LENA recordings. These hourly rate calculations for AWC, CVC, and CTC were used as key variables of interest in the present study.

1.2.3. Analysis Plan

Taken together, the home environment recordings allowed us to extract variables on children’s media exposure as well as quantity of adult words (spoken by males and females), child vocalizations, and proxy estimates of adult-child conversational turns. Recall that our first, second, and third research questions concerned describing the rate, language, and type of children’s auditory media exposure (child- vs. adult-directed), respectively. As such, we calculated descriptives to quantify auditory media exposure overall and in Spanish and English across adult- and child-directed programming. T-tests also allowed for a comparison of Spanish vs. English media exposure, and adult- vs. child-directed programming exposure.

The fourth research question concerned the relation between media exposure (rate, type, and language) and measures of child language input and output (AWC, CVC, and CTC). This question sought to examine the role of media exposure in promoting language outcomes and how this was related to the language environment. First, preliminary analyses examined child age, maternal education, child sex, and recording length to determine whether demographic and recording variables covaried with the key variables of interest given prior work (Bittman et al., 2012; Zimmerman & Christakis, 2005; Zimmerman et al., 2007). Then, we established the significant associations between the rate of media exposure and measures of child language input and output. Next, we conducted follow-up partial correlations to determine the relative contribution of media type (adult- vs. child-directed programming) and language (Spanish vs. English) on the language environment after controlling for the rate of media exposure. To help interpret the correlations, power analyses (1-β) were conducted. Next, media exposure and language input (AWC, CTC) and output variables (CVC) were entered into a correlation matrix to assess their associations.

1.3. Results

To begin, preliminary analyses examined how media exposure in the auditory environment varied as a function of key demographic variables, including child age, maternal education, child sex, and recording length. Results showed that the rate of media exposure was not correlated with age and not correlated with recording length. An ANOVA with media exposure as the dependent variable and sex and maternal education as independent variables showed no significant main effects and no interactions. Given that these variables did not significantly covary with media exposure they were dropped from further analyses.

Recall that our first question concerned characterizing the rate of digital media in children’s auditory environment. Table 3 presents descriptive statistics. The audio recordings of all children who participated included exposure to media. As a group, TV and other sources of auditory media made up an average of approximately 14% (SD = 11.16%) of the total auditory environment that was recorded regardless of language or type of programming. This approximates 8 minutes of auditory media exposure for every 1 hour that was recorded on average. However, there was a wide amount of variability in the proportion of digital media in the auditory environment (range = 0.4 – 51.31% of total recording time).

Table 3.

Descriptive information for variables of interest.

	M	SD	Range
Media Variables
Proportion of media exposure	13.70%	11.16%	0.4 – 42.35%
Spanish media	67.30%	35.53%	0 – 100%
English media	28.74%	33.74%	0 – 100%
Adult-directed media	65.60%	27.27%	0 – 100%
Child-directed media	30.59%	27.71%	0 – 100%
Language Variables (per hour)
Adult word count	740.99	407.32	176.92 – 1876.31
Child vocalizations	116.74	53.23	10.13 – 228.10
Adult-child conversational turns	26.34	17.29	1.00 – 72.57

Our second research question asked about the language of children’s media exposure. Both English and Spanish were heard in the audio recordings. In general, Spanish (M = 67%, SD = 36%) was more common than English (M = 29%, SD = 34%). All but two children heard media in Spanish, and all but seven children heard media in English. A paired samples t-test revealed that the difference between digital media exposure in Spanish and English was significant (t(23) = 2.779, p = .0107). See Figure 1. Spanish and English media exposure did not correlate with the overall rate of media represented in children’s LENA recordings.

Figure 1.

Bar graph depicting type of programming and language of digital media exposure in the auditory environment.

Note. * denotes significant difference at p<.05.

The third research question concerned the type of media exposure in children’s auditory environment. As a group, when children were exposed to media, they were more likely to be exposed to programming directed to adults (M = 65%, SD = 27%, range = 0–100%) relative to programming directed to children (M = 31%, SD = 28%, range = 0–100%). All but two children had exposure to adult-directed media, and all but two children had exposure to child-directed media. A paired samples t-test showed that the difference in the rate of adult- vs. child-directed media exposure was significant (t(23) = 3.21, p = .004). Refer to Figure 1. The proportion of media that was adult- and child-directed did not correlate with the rate of media exposure overall. Examples of adult media included news, music, and telenovelas (soap operas) while examples of child media included cartoons like Thomas and Friends, Chowder, and Daniel Tiger’s Neighborhood.

The fourth research question of interest in the present study examined the association between the rate, type, and language of media exposure with language in the auditory environment and with child language output in toddlers from Mexican immigrant homes in the US (see Table 3 for descriptives). We ran correlations to examine how media exposure was associated with the language environment. Rate of media exposure was negatively correlated with the rate of adult-child conversational turns (r = −.41, p = .037, 1-β = .719) and child vocalizations (r = −.473, p = .015, 1-β = .84) as recorded in the auditory environment (see Table 4 for correlation matrix). However, rate of media exposure was not correlated with the rate of adult words (AWC). This association was also not significant when examining word counts from male vs. female input, respectively. Given the association between rate of media exposure with conversational turns and child vocalizations, and given that they covary (r = .783, p < .001, 1-β = .99) a follow-up partial correlation was conducted to determine whether either variable explained more variance after controlling for shared variance. The correlation between rate of media exposure and conversational turns controlling for child vocalizations was not significant. Similarly, the correlation between rate of media exposure and child vocalizations controlling for conversational turns was also not significant (all n.s. p’s > .368). These results suggest that conversational turns and child vocalizations contribute shared variance to the association with rate of digital media exposure in the auditory environment. Because some studies remove nap times from their recordings based on family logs (e.g., Weisleder & Fernald, 2013), we examined whether our results would hold when nap times were excluded. Recall that only a subset of families returned logs. As such, we evaluated whether the pattern of associations for children for whom naps could be excluded based on information available in the logs was consistent with the larger sample given that the significant reduction in sample size would preclude strong conclusions from parametric tests. Indeed, the pattern of results held for the 17 children for whom nap logs were available, such that media exposure was negatively associated with the rate of adult-child conversational turns and child vocalizations in the recordings after removing segments from the recordings based on parent-reported nap times.

Table 4.

Correlation matrix for characteristics of media exposure and language variables of interest.

	1	2	3	4	5	6	7	8
1. Rate of media exposure	--
2. Adult-directed media	r = − .12, p = .576	--
3. Child-directed media	r = .138, p = .52	r = − .889, p < .001	--
4. Spanish media	r = − .175, p = .24	r = .662, p < .001	r = − .501, p = .013	--
5. English media	r = .212, p = .318	r = − .618, p < .01	r = .635, p < .001	r = − .926 p <.001	--
6. Adult word count	r = − .237, p = .243	r = .338, p = .107	r = − .261, p = .21	r = .248, p = .24	r = .194, p = .36	--
7. Conversational Turns	r = .41, p = .037	r = .136, p = .527	r = .0006 p = .99	r = .225, p = .291	r = − .125, p = .559	r = .736, p < .001	--
8. Child vocalizations	r = − .473, p = .015	r = − .151, p = .481	r = .308, p = .143	r = .248, p = .24	r = − .133, p = .536	r = .272, p = .154	r = .783, p < .001	--

Note. Bold denotes p < .05.

An additional analysis with the full dataset was conducted to further characterize the influence of media type (adult- vs. child-directed) and language (Spanish vs. English) after controlling for the overall rate of media exposure. First, partial correlations were conducted to examine the association between the percent of media that was child-directed and child vocalizations and adult-child conversational turns. These two language input and output variables were examined given their significant correlation with overall rate of media exposure as described previously. Results showed that child-directed media was positively correlated with children’s vocalizations after controlling for overall rate of media exposure; this finding was marginally significant (r = .41, p = .050). However, child-directed media exposure was not associated with conversational turns. Second, partial correlations with adult-directed media were examined controlling once more for overall rate of media exposure. Adult-directed media did not significantly correlate with child vocalizations or with adult-child conversational turns. Together these results suggest that child-directed media specifically may further modify the language environment by providing opportunities for child vocalization specifically, which in turn may increase conversational turns indirectly.

Next, we conducted parallel exploratory analyses to assess the role of media in Spanish compared to English. The first set of partial correlations examined the relative association between child vocalizations and adult-child conversational turns against the percent of media segments that were identified as primarily providing Spanish input. Results showed that Spanish media was positively associated with child vocalizations after controlling for adult-child conversational turns but this was not significant (r = .194, p >.38; see scatter plots as Supplementary Materials). Similarly, there was a positive but non-significant association between Spanish media and conversational turns after controlling for child vocalizations (r = .168, p > .40). For English media, the reverse pattern was observed, such that a negative association emerged with child vocalizations and conversational turns, though these partial correlations were also non-significant.

Lastly, the correlation matrix revealed associations between the rate, type (adult- vs. child-directed), and language (Spanish vs. English) of media. See Table 4. Results showed a negative correlation between the proportion of media that was in Spanish and the proportion that was child-directed (r = −.501, p = .013), whereas there was a positive correlation between child-directed programming and English media (r = .635, p < .001). Conversely, for adult-directed media, results showed a positive correlation with Spanish media (r = .662, p < .001) whereas there was negative correlation with English media (r = −.618, p < .01). Further, the rate of media in the environment did not correlate with the frequency of adult-directed or child-directed media specifically, nor did it correlate with the frequency of media in Spanish or English.

1.4. Discussion

The present study characterizes the nature of media exposure in the auditory environment of young Spanish-learning toddlers in Mexican immigrant homes. In addition, the present study contributes to extant research by describing the amount of digital media exposure across adult- and child-directed programming and Spanish and English given that these children were growing up in multilingual homes. Further, in an approach unique to this study, direct observations of the language environment were employed to examine the association between the rate, content type, and language of digital media and children’s language input and output. Taken together, the findings of this exploratory study provide important information on how children from Spanish-speaking Mexican immigrant homes in the US experience digital media in their unique media ecologies and how this experience may be impacting the context of their early language learning.

1.4.1. Characteristics of Children’s Media Exposure

With respect to the rate with which children from this background were exposed to digital media, we observed that a wide range of digital media exposure across children’s recordings of the auditory environment, as expected and consistent with a large body of research. All children were exposed to media at home, reiterating a consistent finding in the literature: that digital media exposure is ubiquitous in the lives of children from an early age (Kabali et al., 2015; Zimmerman et al., 2007). In the present study, on average, children were exposed to 8 minutes of digital media in the auditory environment for every hour that families recorded, and the amount of media exposure in children’s environment did not predict how much of it was directed to adults vs. children or in Spanish vs. English. As families were not instructed to record a single day from children’s wake to sleep times and also self-selected when to record, we cannot extrapolate hourly averages to determine children’s cumulative daily media exposure. Nevertheless, the rate of media exposure across discrete recordings provided by the present families is relatively consistent with prior research using day-long recordings and parent report of child media use over 24 hours (Duch et al., 2013; Christakis et al., 2009; Zimmerman et al., 2009). At the time that these data were collected, the recommendation from the American Academy of Pediatrics was against media exposure prior to age 2 (Council on Communications and Media, 2013). The present study reiterates the need to create clinical recommendations about child media use that are informed by different linguistic and cultural considerations. While completely eliminating digital media does not appear to be reasonable for families from this background (as all children experienced some media in their homes), perhaps endeavors to support families to increase opportunities for child and caregiver engagement are useful. We return to discuss the precise association between media exposure and the language environment in section 1.4.2 below.

We further sought to characterize the content of children’s media exposure by the type and language of programming. In contrast to Duch et al. (2013) and our hypothesis, children in this study were exposed to more adult-directed than child-directed programming. This may be due to methodological differences. While Duch and colleagues similarly focused on Spanish-speaking Hispanic families from lower income backgrounds, exposure to type of programming was based solely on parent report of the prior 24 hours. Parents may have been apt to report that their children experienced more child-directed programming if this was perceived as the socially desirable response. Additionally, parents may not have had accurate recall of media use, or they may have been more likely to report media that children are actively attending to (like child-directed media) rather than media in the background (like adult-directed media), whereas devices like the LENA are likely capturing both. Thus, we assert that direct observation of the media environment may be a more reliable method for examining the type of programing to which young children are exposed in the auditory environment, especially when families are able to record a full day (unlike in the present study), whereas parent report may be useful for describing parent perceptions or capturing beliefs and practices for foreground media use with children. As the current body of work exploring the developmental implications of the content of children’s media exposure in particular is limited (Kostyrka-Allchorne et al., 2017), we hope that the methodological approaches for characterizing media content described here are useful to the field. Given the aforementioned alignment of our results with the observational studies completed by Christakis et al. (2009) and Zimmerman et al. (2009) but the lack of consistency with the parent report findings of Duch et al. (2013), it is possible that these respective methods provide unique information on media exposure. Direct observation coupled with parent report may together offer a more precise characterization of media exposure in children’s auditory environment and the decisions that govern media experiences.

Regarding the language of media exposure, the use of English and Spanish in the digital media children heard in their environment unsurprisingly mirrored their overall language input. Spanish was the primary language in the home (96% of household members reportedly spoke Spanish to children), and relatedly, the primary language of children’s media exposure as hypothesized (67% of children’s media was in Spanish). Yet, our analysis suggests that these children as a group received more English through digital media than they do from the people in their environment, and that media in English was also likely to be child-directed programming whereas greater adult-directed programming predicted greater media exposure in Spanish. The Spanish dominant parents in this study may have deliberately exposed their children to more English through child-directed digital media due to the belief that this would support children’s English language acquisition (Cycyk and Hammer, 2020). Indeed, results showed that the amount of media that was child-directed was positively associated with children’s rate of vocalizations (in contrast to our hypothesis), but not with conversational turns, after controlling for the effects of overall rate of media exposure. This suggests that child-directed programming provoked children to practice their language skills, perhaps in English, in the absence of adult engagement. A growing body of literature has shown that parents monitor home media exposure with specific learning outcomes in mind (e.g., Chiong & Shuler, 2010; Takeuchi, 2011; Sonck et al., 2013; Böcking & Böcking, 2009); however, it is presently unclear how media use may benefit acquisition of English in toddlers primarily exposed to Spanish, if at all, and none of the results from this study demonstrate causation. Generally, toddler-aged children are simply much less successful learning new words via television than human interaction (Krcmar et al., 2007).

Our exploratory analyses of media language and input point to the possibility that English media exposure may ultimately reduce opportunities for this crucial interaction for language learning. Though child-directed media may increase child vocalizations, exposure to child-directed media was not associated with adult-child conversational turns suggesting that the present group of primarily Spanish-speaking parents did not engage in co-viewing child-directed media that was largely in English. Given the small sample size in the present study, future studies should investigate whether exposure to media in the primary language of the home may be more beneficial to child language development in multilingual contexts (regardless of media type). Perhaps caregivers are more likely to view media in their primary language alongside their child, thereby naturally increasing opportunities for caregiver-child interactions even when this media is adult-focused content. In contrast, when children are exposed to child-directed programming in English, they may be more likely to view this media alone or with other children whose language input may be qualitatively different than that of adults (and is not captured with the LENA conversational turn count). This supports the hypothesis offered by Duch et al. (2013) to explain why child-directed media (primarily in English) but not adult-directed media (primarily in Spanish) was negatively linked to parent report of child language skills in Spanish-speaking homes. Together, these findings suggest that attention should be paid to how parents select the language of digital media in multilingual contexts and the impact of these decisions on child language environments and outcomes in the multiple languages to which children are exposed. This information may support tailored suggestions for media co-viewing in such contexts.

1.4.2. Impact of Digital Media on Child Language Input and Output

Lastly, the present study sought to examine the associations between the characteristics of digital media and the language learning environment and child language abilities. While a number of studies have shown a negative association between digital media exposure and child language outcomes prior to age 3, few studies have used direct observational methods. In one study including only English speakers, researchers showed a negative association between the frequency of digital media and all language input and output variables through audio recordings with LENA as used in the present study (Christakis et al., 2009). Our results extend these findings to Spanish-speaking Mexican immigrant families as anticipated, though not all associations are replicated. Specifically, we showed that children’s rate of digital media exposure was not associated with adult word counts. This lack of association with adult words held for both child-directed and adult-directed programming. That is, digital media (whether designed for children or adults) and overall adult speech are not associated for these children. Instead, rate of digital media use was negatively associated with our proxy measure of adult’s direct engagement with the child, much like Christakis et al. (2009): as the frequency of digital media increases, the number of adult-conversational turns decreases. In contrast to our hypothesis, the specific type of programming was not associated with children’s frequency of conversing with adults. Taken together, this suggests that the digital media to which children in this context are exposed in general is not displacing the overall quantity of adult words in the auditory environment but rather children’s opportunities for back-and-forth interactions with adults specifically that support language learning (an indicator of input quality; Head Zauche et al., 2016). While both quantity and quality of child input matter to child language development, quality may be particularly important, especially as children age (Romeo et al., 2018; Rowe 2012; Zimmerman et al., 2009). This is consistent with the findings of Zimmerman et al. (2009), for example, which showed that adult-child conversational turns fully mediated the relationship between child language abilities and television exposure and adult words.

What explains the distinction between association with media and the quantity of adult words in the environment versus adult-child conversational turns? The LENA does not differentiate adult words that are overheard by the child from words that are directed to the child – all near and clear adult words captured by the LENA are included in the adult word count. Moreover, counts of household numbers showed that adults outnumbered children in these homes. In homes where adults outnumber children, for example, young children may often hear conversations between adults, and these conversations may take place when media is also present. However, conversational turns can include a subset of adult words – child-directed speech specifically – and therefore may offer a proxy as to the child’s opportunities for language engagement with caregivers. In fact, the moments of active conversation that produced turn-taking likely overlapped with the moments of increased child vocalization, as these variables were highly correlated in the present study. In other words, children were more likely to practice their language output when actively receiving conversational input from their adult caregivers (or when simply overhearing conversations between adults; Ferjan Ramírez et al., 2021). Consistent with social-pragmatic theories of language development (Bruner, 1983; Tomasello, 1992; 2000; 2009), children need to practice their emerging language productions with the scaffold of adult feedback readily available in conversational exchanges. These opportunities may be the very moments of interaction that media may be displacing to explain the negative links to language outcomes. Because overheard talk is less supportive of early language development than talk directed at young Spanish-learning children (Weisleder & Fernald, 2013), this finding warrants further exploration. This may represent an important cultural difference in family media ecologies that can be examined by continuing to use direct observational methods and by including families that are not White and who speak languages in addition to English from a variety of socioeconomic backgrounds. If replicated in future research, recommendations for how to enhance conversational interactions between children and adults during media use may be appropriate for families from Spanish-speaking Mexican immigrant backgrounds.

1.4.1. Limitations and Future Directions

The current study does have some limitations. First, the sample size of 30 children was relatively small, and only families from lower-income Mexican immigrant backgrounds in the US were included. This research should therefore be replicated with larger samples of children from varied Spanish-speaking backgrounds before generalizing the current findings to families of Latinx descent across the US or internationally. To build confidence in this line of work, additional rigorous validation studies of the LENA in Spanish-speaking contexts are needed, especially as related to investigations of media use in larger households and adult-child conversational turn counts in general given that these variables influence the reliability of automated determinations of the child language environment (Bulgarelli & Bergelson, 2020; Cristia et al. 2020a; 2020b; Ferjan Ramírez et al., 2021). Second, families independently selected the days and times for recording their child’s home language environment. As such, some families may have chosen to record segments that represented particular features of child media use, and the total recording length varied across families. While an average of 15 hours of recording across multiple days was adequate for this exploratory study, future research should strive for consistency in obtaining continuous day-long recordings to use the LENA as originally designed and to achieve the greatest possible reliability of the LENA estimates. This will allow us to best understand the frequency of children’s media use in particular as well as the relative amount of exposure to Spanish versus English media. Third, the nature of the LENA audio recording makes it nearly impossible to determine whether children were exposed to media as passive participants, as may be the case when media is in the background of their learning contexts, or as active participants, as may be the case when media is in the foreground of their learning contexts. Similarly, it was not possible to precisely discern when children were co-viewing with their family members (either adults or other children). It is likely that background versus foreground media exposure enact different effects on children’s language experiences (Napier, 2014) and that co-viewing alters the language environment (Mendelsohn et al., 2010; Wiecha et al., 2001); future research should test this possibility with young children from Spanish-speaking backgrounds, potentially by coupling video recordings and/or parent report with naturalistic audio recordings. Finally, children’s caregivers were not directly asked about their beliefs and practices regarding media exposure in early childhood. As such, conclusions cannot be drawn about why children may have had more or less exposure to media, why specific programming was favored in the home, or why the language of media was selected. Future research should be completed to determine caregivers’ decision-making process in order to best tailor recommendations for media consumption in the homes of infants and toddlers.

Highlights.

• Home audio recordings of toddlers in Spanish-speaking Mexican-immigrant homes were collected.

• Findings show significant media exposure in the auditory environments of toddlers.

• Children experienced more adult- than child-directed and more Spanish than English media.

• Media exposure was negatively associated with child vocalizations and conversations.

• Language of media exposure may play an important role in multilingual homes.