Language Environments and Spoken Language Development of Children With Hearing Loss

Abstract

This study examined the relationships between adult language input and child language production in regard to the quantity and diversity of spoken language, as well as children’s knowledge of basic concepts and vocabulary. The quantity and diversity of language provided by teachers and parents were related to children’s language output and knowledge. Language ENvironment Analysis technology audio-recorded the language environments of 26 preschool children with hearing loss over 2 days. The language samples were analyzed for quantity (adult word count, child vocalization count, and conversational turn count) and diversity (lexical diversity, syntactical complexity, and clausal complexity) of language. Results indicated a relationship between adult language input and child language production, but only in regard to the quantity of language. Significant differences between the teachers and parents were reported in regard to the diversity of adult language input. These results suggest that the language input provided by adults across environments (school versus home) is considerably different and warrants further investigation.

Child language development is a heavily researched topic, yet specific influences (e.g., quantity and diversity of the language environment) on language development are less known, particularly among those with hearing loss. For children with typical hearing, some studies have found that the amount of talk to which children are exposed relates to their linguistic and academic outcomes (e.g., Hart & Risley, 1995, 2003), whereas others have found that the density or diversity of the language exposure has a stronger relationship to child language development (e.g., Hoff, 2003; Huttenlocher et al., 2007; Rowe, 2012; Vasilyeva et al., 2008). The diversity of language, defined by Huttenlocher et al. (2010) as “the variety of words, phrases, and clauses” (p. 344) produced by both adults and children, may be more relevant to child language development than simple quantitative measures. For young children with hearing loss who use listening and spoken language, both the quantity and diversity of adult language input are often emphasized in family-centered intervention (e.g., Cole & Flexer, 2016), but the specific ways that each influences language development for this population have not yet been clearly defined in the literature.

In the present study, researchers analyzed language samples for the following components of language diversity: lexical, or the number of different words; syntactical, or the number of syntactical elements (e.g., adverbs, adjectives, and prepositional phrases); and clausal, or the number of clausal combinations (e.g., coordination, objective relative clause, object of the main clause, etc.). Findings from other studies have suggested that the more diverse the adult language model, in terms of lexicon (Hoff & Naigles, 2002; Pan et al., 2005) or syntax and clauses (Huttenlocher et al., 2010; Vasilyeva et al., 2008), the more likely the child is to internalize richer language and utilize it for communicative intent. Hoff and Naigles (2002), for example, reported that the lexical diversity and syntactical complexity appeared to account for the variation in children’s expressive vocabularies. However, questions remain regarding which characteristics of language exposure that have the strongest relationships with, or influence on, receptive language outcomes in children, particularly those with hearing loss.

Theoretical Framework

The theoretical framework for this study is nested in social interactionist theory (Bruner, 1975, 1977; Vygotsky, 1986), which emphasizes the role that environmental as well as biological factors play in language development. Infants and children are recognized as social beings who acquire language in service of their needs to communicate. Consequently, language acquisition occurs as a result of the natural interaction between children and their environment, more specifically, their parents or caregivers. The relevance of interactionist theory for this study is the importance it places on the home and the cultural environment in early childhood language acquisition, where the child-directed speech produced by the primary caregivers provides the information children need to “bootstrap” their progress in language acquisition (Höhle, 2009).

Vocabulary Development for Children With Hearing Loss

Although advances in technology, such as digital hearing aids and cochlear implants, have played an important role in the language development of children with hearing loss, there remains a significant gap between children with hearing loss and their hearing peers (Cole & Flexer, 2016; Cupples et al., 2018; Tomblin et al., 2015; Trezek et al., 2010; Wang et al., 2008). Research on this gap has often focused on the fact that children with hearing loss regardless of communication modality tend to have smaller vocabularies than their hearing peers (Lund, 2016; Paul, 2001; Vohr et al., 2011; Werfel, 2017), struggling with the acquisition of function words in particular (Wang et al., 2008). Although children with hearing loss continue to demonstrate vocabulary delays (Lund, 2016), early use of technology can accelerate vocabulary development for children with hearing loss. Early age at cochlear implantation relates to better language outcomes (Cupples et al., 2018; Ching, 2015; Ching et al., 2017; Duchesne & Marschark, 2019) and a faster rate of vocabulary acquisition (Connor et al., 2006; El-Hakim et al., 2001). After accounting for early implantation or hearing aid fitting, increased doses of early intervention have been demonstrated to result in higher language outcomes (Geers et al., 2019).

Basic Concept Development for Children With Hearing Loss

Research has indicated that basic concepts are developed as part of a child’s vocabulary and that they are an important aspect of everyday language, particularly for directions and academic language (Boehm, 1967; Bracken & Cato, 1986; Kaufman, 1978; Steinbauer & Heller, 1978). Basic concepts are frequently embedded in conversation where reference is made to location (e.g., top, left), number (e.g., more, less), quantity (e.g., empty, full), description (e.g., big, little), and time (e.g., before, after). Though function words are often excluded from assessments of vocabulary, they are an aspect of vocabulary that children with hearing loss struggle to acquire (Harrington et al., 2009). In spoken language, function words that represent basic concepts are typically unstressed, which could pose difficulty for children with hearing loss who may not be able to perceive them auditorily.

There are a limited number of studies that examine basic concept development in children with hearing loss; the few studies that have been published reported a significant delay for these children as compared to their typically developing counterparts (Bowers & Schwarz, 2013; Bracken & Cato, 1986; Davis, 1974; Harrington et al., 2009). Davis (1974) found children with hearing loss performed considerably lower on basic concepts tasks than their peers with typical hearing. No significant differences in basic concept knowledge were found between the older and younger children with hearing loss, suggesting that these children struggled to develop and improve their knowledge over time. Bracken and Cato (1986) found children with hearing loss, on average, performed two standard deviations below those who were identified as typically developing. Similarly, Harrington et al. (2009) examined the relationship between early child factors and school readiness skills among children with hearing loss who attended a listening and spoken language program and used amplification in the form of cochlear implants and/or hearing aids, and found that performance on the basic concepts assessments was largely based on children’s receptive vocabulary knowledge.

Bowers and Schwarz (2013) examined the effects of intervention on the development of basic concepts in four children with hearing loss. These children attended a residential preschool program for the deaf that delivers instruction using simultaneous communication, or the use of spoken English and manual communication. The results indicated that all four participants performed below the expected average during baseline; however, they all improved in their basic concepts during intervention. This suggests that even though the basic concept knowledge of children with hearing loss might be delayed compared to their typically developing peers, direct instruction/intervention can mitigate those delays.

These studies have indicated that children with hearing loss experience delays in regard to their acquisition and development of vocabulary as well as basic concepts. Questions remain regarding the specific characteristics or variables that influence this development among children with hearing loss. Given this gap or delay of vocabulary development among children with and without hearing loss, the answer may lie in the language environments or more specifically, the language input.

Language Input for Children With Hearing Loss

Quantity of the Language Environment

Several studies have considered how language exposure and its impact on language acquisition may look different for children with hearing loss than for children with typical hearing. Lederberg and Everhart (2000) found differences between the language environments of children with typical hearing and those with hearing loss, reporting that, for children with hearing loss, a mother’s language input was related to the child’s linguistic abilities. Similarly, DesJardin and Eisenberg (2007) investigated the language models of hearing mothers with their children with cochlear implants, finding that mothers’ involvement and self-efficacy were related to their use of facilitative language techniques, and the quantity of linguistic input was positively related to child’s language development. Although Vohr et al. (2013) did not report a significant relationship among the quantity of language input and child language outcomes, the authors suggested that there was a “significant association between the language environment and child speech and language skills at school age” (p. 67).

A specific aspect of quantity—the number of conversational turns—has been found to be strongly correlated with scores on standardized language assessments, stressing the importance of conversational interaction in regard to language development for children with hearing loss (Ambrose et al., 2014; Aragon & Yoshinaga-Itano, 2012) and those with typical hearing (Romeo et al., 2018). VanDam et al. (2012) were among the first to use Language ENvironment Analysis (LENA) technology to compare and examine the language environments of children with and without hearing loss. All families reported using spoken English at home; those who were identified with hearing loss were reported to use hearing aids as their amplification devices. Among the children with hearing loss, the degree of loss was related to the number of adult words and conversational turns to which they were exposed. In other words, children with more severe hearing loss were exposed to lower levels of adult words and conversational turns.

Most recently, Rufsvold et al. (2018) investigated the impact of quantitative aspects of adult language input, as measured by the LENA technology, on child language output, as well as the child’s knowledge of vocabulary and basic concepts. Rufsvold and colleagues examined the home and school language environments of children with hearing loss who used listening and spoken language, as well as children with typical hearing. The authors’ preliminary analysis revealed that the language environments (as measured by the number of adult words and conversational turns) of the children with hearing loss (n = 30) and those who were typically developing (n = 11) were similar. However, among the participants with hearing loss, type of amplification related to the child’s knowledge of basic concepts, with those with hearing aids performing better than their peers with cochlear implants. Furthermore, the numbers of adult words to which the child was exposed to were highly correlated across weekday and weekend days, indicating that the children were consistently exposed to similar linguistic environments in both school and home settings. In contrast to the findings of previous studies, however, the quantity of adult language and conversational turns was not related to the children’s language outcomes as measured by vocabulary or basic concepts. Although the findings cannot be explained by the quantity of language exposure, this calls for further examination of other aspects of the language environment, such as the diversity of the language exposure provided by adults, and their relationship to the child’s language development.

Diversity of Language Environments

Research on language input for children with and without hearing loss began with a focus on the ways in which the quantity of the input affected vocabulary development. Researchers then began to include measures of syntactical diversity and interaction. Current research in the field, in line with the present investigation, tends toward a more nuanced exploration of the ways in which lexical and syntactical diversity intersect with the pragmatics of linguistic interactions in various contexts to influence children’s development and use of language. Although diversity of language exposure appears to be a factor in language acquisition among children with typical hearing and those with hearing loss, few studies examine the diversity of language exposure across settings for those with hearing loss. Past studies have found the quantity and diversity of language input and development to relate to a variety of demographic factors: age (e.g., Rowe, 2012), type of hearing loss (e.g., Trezek et al., 2010; Wang et al., 2008), type of amplification (e.g., Yoshinaga-Itano et al., 2010), additional disabilities (e.g., Thiemann-Bourque et al., 2014; Warren et al., 2010), and maternal education (e.g., Hoff, 2003).

The quantity and diversity of language input may also relate to the type of environment in which interactions between adults and children take place. Soderstrom and Wittebolle (2013) compared the linguistic environments found in the homes and daycare settings of those who were typically developing. Specifically, the authors found that the two environments were similar in regard to the amount of talk; however, type of activity (e.g., storytime and organized playtime) had a stronger relationship to the quantitative measures of the language environments than did the time of the day or type of environment. Such findings warrant further investigation of the effects that different environments may have on language input for children with hearing loss.

Significance of the Study

The present study employs LENA technology to explore the ways in which the quantity and diversity of adult language input can shape language development among children who have hearing loss. Specifically, this study aims to add to the growing body of literature by examining aspects of the quantity and diversity of language across environments to help identify factors that facilitate language growth. By critically examining the linguistic environments of children with hearing loss, this study has the potential to influence parent education, early intervention practices, school-based instruction, and adult–child interactions. In so doing, this study may yield information that could help close the gap between children with hearing loss and their peers with typical hearing in terms of spoken language development.

Research Questions

• (1)

  What demographic characteristics (i.e., age, type of hearing loss, type of amplification, presence of additional disability, and maternal education) of the participants are related to the diversity of adult input, diversity of child language, child’s vocabulary, and child’s understanding of basic concepts?

• (2)

  Are the quantity and diversity of adult language related to the quantity and diversity of the child’s language as well as their knowledge of vocabulary and basic concepts?

• (3)

  Is there a difference in teacher or parent input in regard to the quantity and diversity of language?

Method

Participants

The participants comprised 26 monolingual preschool children with permanent hearing loss whose ages ranged from 36 months to 59 months (M = 47.69, SD = 7.883). A total of 22 participants in this study were a subset of the participants with hearing loss in Rufsvold et al. (2018) study; four new participants were added to this subset. The majority of the participants utilized cochlear implants (53.8%), some had unilateral or bilateral hearing aids (42.6%), and one used no hearing technology at the time of the study (3.3%). Four of the participants had diagnosed additional disabilities (15.3%), whereas the rest had no additional disabilities (Table 1). The participants were recruited from highly regarded and nationally recognized private programs in three states (New York, California, and Missouri), which focused on developing listening and spoken language. The educators at the schools, many of whom were certified Listening and Spoken Language Specialists or were in the process of earning the credential, were certified and licensed teachers of the deaf and hard of hearing according to state standards. All of the children were from families that identified themselves as monolingual English speakers, and all families used listening and spoken language to communicate with their children.

Table 1

Demographics of the participants

Characteristics		Participants (n = 26)
Age (months)	Mean age (SD)	47.69 (7.883)
Sex % (n)	Male	50% (13)
	Female	50% (13)
Highest education: mother	Some college	26.9% (7)
	Associate degree	11.5% (3)
	Bachelor’s degree	30.8% (8)
	Master’s degree	15.4% (4)
	Doctoral degree	3.8% (1)
	Unknown	11.5% (3)
Type of hearing loss	Conductive	7.7% (2)
	Mixed	7.7% (2)
	Sensorineural	84.6% (22)
Type of amplification	None	3.8% (1)
	Cochlear implant(s)	53.8% (14)
	Hearing aid(s)	42.3% (11)
Presence of additional disability	Yes	15.4% (4)
Disability diagnosis	Waardenburg syndrome	3.8% (1)
	Chromosomal deletion	3.8% (1)
	CHARGE syndrome	3.8% (1)
	Dev. delay (unspecified)	3.8% (1)
	No	84.6% (22)

Procedures

Institutional Review Board approval was obtained through Teachers College, Columbia University and all participating staff and parents provided written informed consent. Prior to collecting the language samples and testing, the parents filled out a demographic questionnaire. The researchers administered the two formal assessments, Peabody Picture Vocabulary Test-4 (PPVT-4; Dunn & Dunn, 2007) and Boehm-3 Preschool Test of Basic Concepts (BTBC-3; Boehm, 2001), to the children during the school day at their individual schools. LENA technology was used to audio-record 32 hr of language samples over a two-day period (a weekday for 16 hr and a weekend day for 16 hr) from each child. This length of time was determined by the parameters of the LENA devices; each device can only capture 16 hr of language before turning off on its own. Therefore, parents were instructed to turn the device on in the morning and then to leave the device to power off when it became full. The recordings during the weekday consisted of approximately 6–6.5 hr of the participants’ preschool environment, whereas the rest of the day was home environment; the weekend recording captured only their home language environment. These audio recordings were used to produce reports of the quantity aspects of language input, and the language samples were transcribed for analyzing the diversity of language input. All data (formal assessments and LENA recordings) were collected within a time span of 1–3 weeks.

Data Collection

Quantity data from LENA: Adult word count (AWC), child vocalization count (CVC), and conversational turn count (CTC). The LENA system was used to collect and analyze the language environments, as previous studies and analysis have found LENA technology to be reliable and accurate (e.g., Ambrose et al., 2014; Dykstra et al., 2012). Xu et al. (2009) investigated the reliability of the LENA system within children’s natural home environments; they reported that the correlation between human coders and the LENA was r = .91, p < .01, and child vocalizations were accurately reported 75% of the time. The LENA system consists of a small digital language processor inserted into the pocket of a vest worn by a child participant. The digital language processor has the capacity to record up to 16 hr of continuous audio and can then be connected to the LENA Pro computer software to download the raw audio. Using automatic speech processing, the software produces reports of the data pertaining to AWC, CVC, CTC, and amount of exposure to electronic media such as the TV and radio.

For this study, AWC and CTC were used as estimates of adult language input. LENA’s algorithm estimates the number of adult words spoken around the child as AWC. LENA defines CTC as an estimate of the number of conversational turns taken by the child with an adult conversation partner; each instance in which an adult responded within 5 s of the child’s vocalization, or the child responded within 5 s of an adult’s utterance, is counted by the LENA Pro software as a conversational turn (Gilkerson & Richards, 2009). CVC was used as an estimate of child vocalization, or output. In estimating CVC, LENA’s algorithm discriminates between “key child speech” and “key child nonspeech” (Xu, Yapanel, & Gray, 2009, p. 10). LENA identifies key child speech, or CVC, to include words, babbles, and pre-speech communicative vocalizations, such as growls and squeals. In contrast, key child nonspeech was considered fixed signals (e.g., scream, cry, or laugh) or vegetative sounds, such as breathing or burping, and was not included in the CVC estimate.

The researchers estimated the number of waking hours for each child by examining the hours of CTC. Wake time was considered to begin at the hour of the first conversational turns (when the child awoke) and ended following the hour of the final conversational turns (when the child went to sleep). The number of hours between these two points of time were counted and recorded as wake time. CTC was used for this calculation because it requires interaction between the child and an adult, ensuring the child was awake during this period of recording. This wake time was used to create a proportion score for AWC, CTC, and CVC. For example, the AWC was divided by the duration of wake time for a given day (weekday or weekend). Using these proportions in subsequent analyses mitigated the effects of variability in waking hours and LENA wear time across participants.

Diversity data from LENA. To examine the diversity of the adult and children’s language, 30 min of weekday snacktime with teachers and 30 min of weekend dinnertime with parents/family were transcribed for each participant. The length of time was selected to provide consistency between a mealtime during the day (snacktime) and typical evening meal (dinnertime), both of which were reported by most teachers and parents to last for 30 min. Professional transcribers transcribed each audio clip using word-processing software. Because the conversations were to be coded for the diversity of language and not specific phonetic features of speech, broad orthographic transcription was used. The transcriptions were completed to distinguish adult speakers and the participating child, as well as other children present. The transcriber transcribed each adult’s and child’s utterances verbatim using standard spelling. For example, a sample of the transcription might look like this: “Adult: What should we eat for dinner? Target child: I want lots of pizza!” Unintelligible utterances were noted as such on the transcription and consequently not used in the coding.

All adult utterances as well as the participating child’s utterances were analyzed using the coding mechanism described in Huttenlocher et al. (2010), as described in the following three sub-sections: lexical diversity, syntactical complexity, and clausal complexity. It should be noted that Huttenlocher et al. (2010) examined “constituent diversity,” which was defined as the optional components (words or phrases) within a clause (p. 348). For example, if a child responded “green” to the question: “What color is your shirt?”, it would not be coded as an adjective because it is not optional. Meanwhile, in the response “dark green,” dark would be coded as an adverb, but green would not be coded. However, given the language delays and language development of children with hearing loss compared to their hearing peers, coding only optional elements would result in significantly low occurrences for each element, which would misrepresent the language diversity among the children’s utterances. Therefore, this study coded all elements, optional and not, of utterances made by adults and children.

Lexical diversity.  Huttenlocher et al. (2010) defined lexical diversity as the number of different word types used by adults and children. Proper names and nicknames were treated as one word type, including variations such as Ben/Benjamin. Proper names that consisted of more than one word, such as BeautyAndTheBeast, were counted as one word type. All inflected types were treated as the same type (e.g., run/runs/running = one type). Word types that had irregular inflectional morphology were treated as the same type (e.g., buy/bought). Words that had different derivational morphology were treated as different word types (e.g., slow/slowly). To code the transcripts for lexical diversity, the coders noted the number of different words per utterance for the adult and child. Those numbers were compiled to establish a total “score” of lexical diversity for both adult utterances and child utterances.

Syntactical complexity. Seven different forms were associated with syntactical complexity (Huttenlocher et al., 2010). Adjectives (e.g., white rabbit, blue car) and adverbs modifying verbs (e.g., walk slowly, throw hard), and adverbs modifying other adjectives (e.g., very cool) were each treated as one type of syntactical complexity. Four different types of phrases were also coded: (1) prepositional phrases (e.g., In the evening, Sarah is coming over), (2) noun phrases that occurred with no preposition and outside of argument positions (e.g., Last night we went to the store), (3) possessives (e.g., Mom’s purse), and (4) quantifiers (units) for mass nouns (e.g., a sip of soda). To code the transcripts for syntactical diversity, the research team devised a color-coding strategy to account for the different components of syntax. Each syntactical element was assigned a specific color (e.g., blue for adjectives) in which coders highlighted corresponding words or phrases. The total number of color-coded elements was used to represent syntactical complexity.

Clausal complexity. Clausal complexity was measured according to the various ways clauses were combined within utterances. Clauses are defined as follows, and each were treated as one type: (1) coordination in which two clauses are joined by and, or, or but (e.g., He went to school and studied), (2) adjunct clause that precedes the main clause (e.g., After you finish getting ready, pick up your toys), (3) adjunct clause that follows the main clause (e.g., Pick up your toys after you finish getting ready), (4) subject relative clause that modifies the subject of the main clause (e.g., My friend that you went to school with is getting married), (5) objective relative clause that modifies the object of the main clause (e.g., I found the toy I want), (6) the subject of the main clause (e.g., Finger-painting is fun), and (7) the object of the main clause (e.g., They said it’s right here) (Huttenlocher et al., 2010). The research team again applied a color-coding strategy, with each type of clausal combination assigned a specific color (e.g., red for coordination). The total number of color-coded clausal combinations was used to represent clausal complexity.

Reliability. A second researcher double-checked each full transcription. If discrepancies were found, the two researchers resolved the conflicts by mutual agreement. Approximately 25% of the recording was independently coded by a second researcher. Inter-rater reliability was calculated per transcript for each of the diversity variables (lexical, syntactical, and clausal). Averages of the inter-rater reliability were calculated across the three diversity variables and were reported to be 93% for lexical complexity, 80% for syntactical complexity, and 94% for clausal complexity. The range of the percentages for inter-rater reliability was between 70 and 99%, with an outlier of 59%. This particular outlier was calculated for a transcript that had very few adult and child utterances; therefore, the few discrepancies between the coders were emphasized in the percentages. The conflicts in coding this transcript were resolved between the two coders with mutual agreement.

Vocabulary Measure. The PPVT-4 is a norm-referenced, untimed receptive vocabulary assessment, which comprises two parallel forms with 228 items each that are divided into 19 sets. These sets increase in difficulty throughout the assessments so that examiners can easily identify and administer the sets that are appropriate for the child’s vocabulary knowledge. An easel, which consists of four-color pictures arranged on each page, is shown to the child as the examiner prompts them with phrases such as “Show me _____”, “Point to _____”, or “Where is____?”. The content that is referenced in this assessment includes but is not limited to: body parts, emotions, household objects, people, vehicles, and so on. Internal consistency was reported to fall between .94 and .95 on both forms. The reliability found using alternate form was reported to be very high, as the mean for both forms was .89 (Dunn & Dunn, 2007). This measure has been widely used in research pertaining to language development of children. The percentile rank was used as a variable in the present study.

Basic Concepts Measure. The BTBC-3 is a norm-referenced assessment that measures the child’s understanding of concepts related to qualities of individuals (tall, angry, small), spatial relationships (under, top, on), time (before, after), and quantity (more, few). As the examiner administers the assessment, the child is presented with an easel that consists of four pictures on each page. The child is then prompted with a verbal direction such as “Point to the box that is empty”.

Reliability for BTBC-3 was conducted using internal consistency, standard error of measurement, and test–retest reliability. The internal consistency reliability was high, as the coefficient alphas were reported to range from .85 to .92. The standard error of measurement was reported to range from 2.08 to 2.88, which indicated low variability. The test–retest reliability coefficients ranged from .9 to .94. The reported high reliability indicated that this was an acceptable measure of language comprehension, specifically in regard to children’s understanding of basic concepts for the age range of participants included in this study (Boehm, 2001). The percentile rank was used as a variable in the present study.

Data Analysis

Research Question 1. Research question 1 identified the demographic characteristics that related significantly to the diversity of adult language input or to child’s diversity of language output, vocabulary, or basic concepts knowledge. The diversity of adult language input comprised six separate variables: the number of syntactical elements (syntactical complexity), multi-clause constructions (clausal complexity), and the number of unique lexical units (lexical diversity) in a mealtime sample, each measured on the weekday and on the weekend. The diversity of child language was measured in the same way, using six variables to represent weekday and weekend syntactical and clausal complexity and lexical diversity. Percentile rank on the PPVT-4 and BTBC-3 represented receptive vocabulary and basic concepts, respectively.

The relation between age (defined as age in the months at the first recording) and the language variables was examined using Pearson correlations. Type of hearing loss, type of amplification, and maternal education were coded as categorical variables and were examined using one-way analyses of variance (ANOVAs). Tukey’s honestly significant difference (HSD) correction was applied to all planned pairwise post hoc comparisons among three or more groups. With the exception of one participant with unilateral hearing loss, participants had the same type of hearing loss in both ears. Type of hearing loss in the right and left ears was thus combined into one composite variable, with the child with unilateral loss included in the sensorineural group based on the type of hearing loss in the poorer ear. Amplification in the right and left ears was combined into a composite variable to represent bilateral amplification. Children with one or two cochlear implants were considered to be cochlear implant users, whereas children with one or two hearing aids were considered to be hearing aid users. No participants used bimodal technology (one cochlear implant and one hearing aid). Because only one participant did not use any amplification at the time of the study, this case was removed for the analysis. The presence of additional disabilities was examined using independent-samples t-tests with equal variances not assumed. Table 1 presents the concomitant diagnoses of the participants with disabilities, as well as the group sizes for the categorical variables. Maternal education level was also coded categorically according to the highest level of education completed by the mother at the time of the first recording, from some college through doctoral degree. Only one mother had a doctoral degree, so this case was removed for post hoc analysis.

It must be noted that group sizes were unequal for several categorical variables, particularly maternal education, type of hearing loss, and the presence of an additional disability. As noted above, groups with only one case were excluded for post hoc analyses. Although recruitment took place through a variety of programs across the country in order to gather data from a diverse sample, the participants were not equally distributed across groups. Although unequal group sizes can contribute to violations of ANOVA’s assumption of homogeneity of variance across groups, ANOVA has also been found to relatively robust to moderate differences in group variances, such as those in this study (Blanca et al., 2017; Tomarken & Serlin, 1986).

Effect sizes were calculated and included for the data analysis. Effect sizes that were reported for ANOVA were calculated by SPSS as eta-squared values, whereas the effect sizes for presence of additional disabilities, and research question 3 were calculated using a specific effect size calculator (Buchanan et al., n.d.). Additionally, effect sizes for correlations (age and research question 2) were calculated manually since these were not reported using SPSS.

Research Question 2. Research question 2 examined correlations between the quantity and diversity of adult and child language and the child’s understanding of vocabulary and basic concepts. For this research question, which concerned general diversity, composite diversity variables were created by adding the lexical, clausal, and syntactical numbers for each adult and child on the weekday and again on the weekend. This resulted in four diversity composite variables: adult diversity weekday, adult diversity weekend, child diversity weekday, and child diversity weekend. The quantity of the adult language comprised the weekend and weekday AWC and CTC, whereas the quantity of the child language comprised weekday and weekend CVC and weekend and weekday mean length of utterance (MLU). Because these variables were continuous, Pearson correlations were performed. A Bonferroni correction was applied to the alpha criterion (.05/14 = .0036) to control for multiple comparisons among the 14 variables within this research question.

Research Question 3. Paired-samples t-tests were performed to investigate differences between weekday and weekend input in regard to AWC, CTC, lexical diversity, and clausal and syntactical complexity.

Results

Research Question 1: What demographic characteristics of the participants are related to the diversity of adult input, diversity of child language, child’s vocabulary, and child’s understanding of basic concepts?

Age

The Bonferroni correction for the alpha value was .05/15 = .0033. Strong positive correlations were found for the following variables: age and child lexical diversity weekday, r(26) = .613, p = .001, r² = .376; age and child syntactical complexity weekday, r(26) = .599, p = .001, r² = .358; age and child clausal complexity weekday, r(26) = .598, p = .001; age and child clausal complexity weekend, r(26) = .601, p = .001, r² = .361.

Type of Hearing Loss

A difference was found in child lexical diversity weekend based on type of hearing loss, F(2,23) = 5.005, p = .016, η² = .303. Post hoc analysis using Tukey’s HSD correction revealed that children with conductive hearing loss (M = 184.50, SD = 40.305) had higher lexical diversity than those with mixed hearing loss (M = 45.00, SD = 56.569) or sensorineural hearing loss (M = 67.50, SD = 52.323). A significant difference was also found in percentile rank on the PPVT-4, F(2,23) = 4.509, p = .022, η² = .282. Post hoc analysis with Tukey’s HSD found children with conductive hearing loss (M = 85.50, SD = 12.021) performed significantly higher than those with mixed hearing loss (M = 17.15, SD = 23.830) and those with sensorineural hearing loss (M = 28.46, SD = 27.36). Additionally, a significant difference was found for child clausal complexity weekend based on type of hearing loss, F(2,23) = 7.059, p = .004, η² = .380. Post hoc analysis with Tukey’s HSD found children with conductive hearing loss (M = 19.50, SD = 7.778) combined significantly more clauses than those with mixed hearing loss (M = 1.00, SD = 1.414) and those with sensorineural hearing loss (M = 5.09, SD = 5.528).

Type of Amplification

There were significant differences based on type of amplification for the following variables: child clausal diversity weekday, F(1,23) = 7.348, p = .012; η² = .242; BTBC-3 percentile, F(1,23) = 6.599, p = .017, η² = .223; and PPVT-4 percentile, F(1,23) = 4.495, p = .045, η² = .163. Hearing aid users (M = 8.18, SD = 8.589) had greater clausal diversity on the weekday than did children with cochlear implants (M = 1.71, SD = 2.301). Hearing aid users also performed higher on the BTBC-3 (M = 45.45, SD = 36.101) and PPVT-4 (M = 42.07, SD = 32.804) than did cochlear implant users (BTBC-3: M = 16.21, SD = 20.238; PPVT-4: M = 19.65, SD = 19.657).

Presence of Additional Disabilities

There were statistically significant differences in children’s performance on the BTBC-3, t(22.499) = 2.955, p = .007, g = .73, d = .76 (equal variances not assumed) among those with and without additional disabilities. Individuals with no additional disability (M = 34.91, SD = 34.07) performed higher on the BTBC-3 than individuals with additional disabilities (M = 10.75, SD = 7.5).

Maternal Education Level

Statistically significant differences among the following variables were revealed: child lexical diversity weekend, F(4,18) = 3.214, p = .037, η² = .417 and child syntactical complexity weekend, F(4,18) = 6.763, p = .002, η² = .600.

Post hoc analysis revealed the following results: children with mothers that obtained master’s degrees (M = 116.5, SD = 86.087) had higher lexical diversity than those with bachelor’s degrees (M = 61.63, SD = 38.075), associate degrees (M = 49.33, SD = 20.551), and some college (M = 65.29, SD = 50.763) on the weekend. Additionally, children with mothers that obtained master’s degrees (M = 45, SD = 39.657) had higher syntactical complexity than those with bachelor’s degrees (M = 28.63, SD = 26.349), associate degrees (M = 13, SD = 8.185), and some college (M = 30, SD = 21.4) on the weekend.

The following demographics were not found to be significant in diversity of adult input, diversity of child language, child’s vocabulary, and child’s understanding of basic concepts: gender, degree of hearing loss, aided pure tone average, paternal education, and income.

Research Question 2: Are the quantity and diversity of adult language related to the quantity and diversity of child’s language as well as their knowledge of vocabulary and basic concepts?

The data indicated that CTC weekday was strongly positively related to CVC weekday, r(26) = .849, p < .001, r² = .721 and CTC weekend was strongly positively related to CVC weekend, r(26) = .717, p < .001, r² = .514 (Table 2). This indicates that the number of conversational turns on a given day relates to the number of vocalizations made by the child. In other words, as one variable (CTC or CVC) increases, the other does as well. Because these data were collected concurrently, whether children who vocalized more elicited more turn-taking behavior from adults or whether engaging in conversational turns with adults led children to vocalize more could not be determined. There were no significant relationships among the adult language variables and the child language diversity variables (Table 3). Additionally, the data indicated that there were no significant relationships between the adult language variables and the child’s knowledge of basic concepts and vocabulary (Table 4).

Table 2

Correlations among quantity and diversity of adult language related to the quantity of the child’s language

	1. AD Week day	2. AD Week end	3. AWC Week day	4. CTC Week day	5. AWC Week end	6. CTC Week end	7. CVC Week day	8. CVC Week end	9. MLU Week day	10. MLU Week end
1	1
2	.242	1
3	.172	.187	1
4	−.077	−.006	.685**	1
5	−.026	.354	.241	.236	1
6	.201	.408†	.414†	.351	.738**	1
7	−.097	−.008	.450†	.849**	.140	.216	1
8	.208	.215	.411†	.392†	.228	.717**	.453†	1
9	.076	−.174	.365	.274	−.221	−.042	.423†	.365	1
10	.198	−.068	.499†	.231	−.108	.128	.298	.370	.661**	1

Notes. AD = adult diversity; AWC = adult word count; CTC = conversational turn count; CVC = child vocalization count; MLU = mean length of utterance.

^*p < .05;

^**p < .001.

^†Correlations with p-values that are <.05 but greater than the Bonferroni correction.

Table 3

Correlations among quantity and diversity of adult language related to the diversity of the child’s language

	1. AD Week day	2. AD Week end	3. AWC Week day	4. CTC Week day	5. AWC Week end	6. CTC Week end	7. CD WEEK DAY	8. CD WEEK END
1	1
2	.242	1
3	.172	.187	1
4	−.077	−.006	.685**	1
5	−.026	.354	.241	.236	1
6	.201	.408†	.414†	.351	.738**	1
7	.192	−.160	.496†	.439†	−.142	.101	1
8	.329	.210	.481†	.202	.032	.272	.612**	1

Notes. AD = adult diversity; AWC = adult word count; CTC = conversational turn count; CD = child diversity.

^*p < .05;

^**p < .001.

^†Correlations with p-values that are <.05 but greater than the Bonferroni correction.

Table 4

Correlations among quantity and diversity of adult language related to the child’s knowledge of vocabulary and basic concepts

	1. AD Week day	2. AD Week end	3. AWC Week day	4. CTC Week day	5. AWC Week end	6. CTC Week end	7. BTBC	8. PPVT
1	1
2	.242	1
3	.172	.187	1
4	−.077	−.006	.685**	1
5	−.026	.354	.241	.236	1
6	.201	.408†	.414†	.351	.738**	1
7	.364	.036	.265	.112	.160	.313	1
8	.317	.001	.307	.062	−.022	.247	.838**	1

Notes. AD = adult diversity; AWC = adult word count; CTC = conversational turn count; BTBC = Boehm Test of Basic Concepts; PPVT = Peabody Picture Vocabulary Test.

^*p < .05;

^**p < .001.

^†Correlations with p-values that are <.05 but greater than the Bonferroni correction.

Research Question 3: Is there a difference in teacher or parent input in regard to the quantity and diversity of language?

A paired-samples t-test revealed that there was a statistically significant difference between adult lexical diversity weekday (teacher; M = 326.81, SD = 63.23) and adult lexical diversity weekend (parent; M = 233.00, SD = 105.56), t(25) = 4.223, p < .001, d_diff = 2.22; adult syntactical complexity weekday (teacher; M = 269.27, SD = 102.64) and adult syntactical complexity weekend (parent; M = 153.88, SD = 114.45), t(25) = 4.559, p < .001, d_diff = 9.77; adult clausal complexity weekday (teacher; M = 72.92, SD = 33.83) and adult clausal complexity weekend (parent; M = 45.27, SD = 38.63), t(25) = 2.860, p = .008, d_diff = 5.76. Meanwhile, a statistical analysis found that there were no statistically significant differences between AWC weekday and AWC weekend or CTC weekday and CTC weekend.

Discussion

Overview of the Results

The results of this study indicated that several demographic factors (i.e., age, type of hearing loss, type of amplification, the presence of an additional disability, and maternal education level) are related to the diversity of adult and child language and knowledge of basic concepts and vocabulary. Additionally, the quantity of language input (as measured by number of conservational turns) was positively related to the quantity of the child’s language (number of child vocalizations) on both the weekend and weekdays. Relationships between the diversity of adult language and diversity of child language were not observed. Moreover, significant differences in the diversity of adult language provided by teachers and parents were found.

Language Input and Vocabulary Development

The results from this study illuminate the complex interplay between language input and language development, building upon previous literature and highlighting several areas in which this interplay may be unique for children with hearing loss who use listening and spoken language. Several demographic factors related significantly to the quantity and diversity of adult and child language. Children of mothers with higher levels of education were found to employ more lexically diverse language on the weekends than were children of mothers with lower education levels. Interestingly, in contrast to the findings of previous studies (e.g., Hart & Risley, 1995; Hoff, 2003; Hoff-Ginsberg, 1991; Huttenlocher et al., 2010; Pan et al., 2005; Vasilyeva et al., 2008), significant differences in the diversity of adult language on the basis of maternal education level were not found. Although the language environments of children in families with advanced levels of education were not more complex in this sample, the children’s own utterances were more complex. It is possible that the intensive intervention the families received in their preschool programs helped to mitigate the effects of parental education level on the home language environment. Because the data in this study were cross-sectional, it is also possible that past differences in parental language input, which were not measured in this study, led to the present observed differences in child language.

A significant positive correlation was found between children’s ages and the length and diversity of their utterances, indicating, not surprisingly, that the language output of the participants increased in complexity as they grew older. However, in contrast to the changes over time found by Huttenlocher et al. (2010), the diversity of language input provided to the children with hearing loss in this study did not appear to be greater for older children than for younger ones. It is possible that the specialized intervention received by the children and their families, or the relatively restricted range in ages of the participants, contributed to this finding; these ideas warrant further investigation.

Differences were also found in the language diversity and vocabulary demonstrated by children based on their use of hearing technology. Children who used hearing aids produced longer utterances with more clausal diversity on weekdays than did their peers with cochlear implants. The hearing aid users also performed better on the BTBC-3 and the PPVT-4. As previous studies suggested, this gap in receptive vocabulary knowledge may be attributed to greater delays in listening age for cochlear implant users or to differences in the ways different devices process the language environment (Fagan & Pisoni, 2010; Rufsvold et al., 2018). It is also possible that degree of hearing loss, which is generally more profound in children who receive cochlear implants, contributed to the greater language and vocabulary delays observed in these children (Connor et al., 2006), though significant differences based on degree of hearing loss alone were not found in this sample.

It is not surprising that presence of additional disabilities was correlated with the child’s language, but interestingly, this demographic factor only related to performance on the BTBC-3. Additional disabilities may influence the perception of language, including receptive processing of the basic concepts that even children who are typically developing struggle to acquire (Bracken & Cato, 1986; Davis, 1974; Harrington et al., 2009). Though further investigation of these results is needed, that the diversity of language experienced and used by children in this study did not differ significantly based on the presence of concomitant disabilities suggests that early, individualized intervention might help to mitigate the potential effects of hearing loss and other conditions on children’s language development.

Strong positive correlations were found between the number of vocalizations emitted by a child and the number of conversational turns to which he or she was exposed on both weekdays and weekends. These results support the strong body of literature on the relationship between conversational turns and a child’s language and extend them to children with hearing loss who use listening and spoken language (see reviews in Wang et al., 2017). Conversational turns represent engagement between children and caregivers, or the back-and-forth conversation that facilitates language interaction, joint attention and, consequently, language development. The directional nature of this relationship between this engagement and child vocalizations remains unclear in this cross-sectional study. Conversational engagement may have led to more vocalizations by the child. It is also possible that, the more the child vocalized around the adult, the more the adult engaged and responded with language. Interestingly, the diversity and quantity of the adult language were not related to children’s knowledge of the basic concepts and vocabulary. Further investigation into this relationship, particularly into the possible influence of the quality of language input on longitudinal outcomes for children with hearing loss, is needed.

Most interestingly, though the quantity of adult language input positively related to the quantity of child language production, there was no relationship between the diversity of adult language input and the child’s language, as has been found in previous studies with typically developing children (e.g., Huttenlocher et al., 2010). It is possible that children with hearing loss require different types of diversity in adult language input than do their typically developing peers. It is also possible that our adopted definition of diversity of language from Huttenlocher et al. (2010)—that is, lexical diversity, syntactical complexity, and clausal complexity—might have been too limited to capture a holistic view of language diversity. Maybe, instead of the language diversity/complexity, it is the content of the adult language input, such as its sensitivity and responsiveness to the child (e.g., Hirsh-Pasek et al., 2015), that really matters. The effect of the diversity and content of adult language input may also take time to manifest, becoming more obvious as the child ages. Leigh et al. (2011), for example, found a longitudinal component of the impact of maternal sensitivity and responsiveness on later language development in children.

Differences Between Language Environments

Additionally, this study examined differences in the language environments experienced by children on weekdays and weekends. Results indicated that the diversity of language the children were exposed to was significantly different during interactions with teachers on weekdays and interactions with parents on weekends. Similar to the findings in Soderstrom and Wittebolle’s (2013) study, the language environments at school and at home were quantitatively similar; however, the diversity and type of the language provided differed significantly. Children were exposed to fewer unique words, fewer combinations of clauses, and fewer advanced syntactical elements (e.g., adjectives, adverbs, and prepositional phrases) during dinnertime at home than they were during snacktime at school. This difference in diversity may have been due to differences in the structure of mealtimes at home and at school; although snacktime with teachers tended to be structured and routine, dinnertime for the families varied in terms of location and type of engagement. Opportunities to provide diverse language were presented at each mealtime, but the structure of that time appeared to influence the language environment.

Furthermore, this study calls attention to discrepancies in the diversity of the language provided by teachers and parents. Though there was no significant difference between the amount (quantity) of language input for teachers (i.e., weekday) versus parents (i.e., weekend), there was a statistically significant difference in terms of the diversity of the language they provided. Although teachers used the same amount of language as the parents, teachers used much more diverse and complex language. Family-centered early intervention in listening and spoken language, however, emphasizes the importance of rich linguistic input in routine, unstructured settings like the home (Cole & Flexer, 2016). Even though the parents in this study appeared to have accomplished the goal of providing frequent linguistic input, they were not able to match the diversity of the input provided by teachers and other professionals. Parent coaching should therefore emphasize both the quantity of input and specific aspects (e.g., facilitative language techniques, and diverse use of syntax and clauses) of parental language that have been related to outcomes for children with and without hearing loss in the longitudinal literature (Cruz et al., 2013; Hoff & Naigles, 2002).

Aragon and Yoshinaga-Itano (2012) proposed that creating a “super language learning environment” (p. 350) is necessary to close the gap between children with hearing loss and their typically developing peers. The results of this study suggest that, in addition to—or perhaps a precursor to—a “super language learning environment,” consistency across environments in terms of quantity and diversity of language input may be necessary. This consistency may provide a greater number of opportunities for children with hearing loss to perceive and attend to language components that could affect their language development.

Limitations of the Study

The first limitation of this study was the low number of participants (N = 26), a persistent issue in research pertaining to children with hearing loss. Given that children with hearing loss form a low-incidence population, it can be a challenge to recruit children and families that fit the inclusion criteria. Furthermore, the sample of children with hearing loss included in this study were children who were monolingual English speakers, which by nature excludes a substantial number of potential participants that identify as culturally and linguistically diverse. This study should be replicated on a larger scale that is more representative of the larger population of children with hearing loss. Additionally, due to the nature of the LENA technology and its data collection procedures, only spoken language from the adults and children was recorded and examined. Nonverbal components of language, such as facial expressions, body language, gestures, and eye contact, may also be crucial to language perception and development.

Although this study measured many demographic variables, others that may have had an effect on the outcomes were not collected. These include the demographics of the classroom teachers, specifically pertaining to credentials and years of experience; the listening age and/or the age of implantation, as applicable, of the child; and information regarding intervention received (e.g., frequency, duration, individual versus group settings, and family-centered sessions). Additional information about the parents, such as their knowledge of listening and spoken language strategies, might also partially explain how adults’ language influences child language development.

The limitations of correlational data analysis must also be acknowledged, and the results interpreted with caution. The nature of correlational research allows for identifications of associations and relationships between two or more variables but it does not allow researchers to note causations (Asamoah, 2014). For example, while we cannot assert that the number of conversational turns elicits more child vocalization, we can confidently purport that there is an association between the two. This association is significant to language environments and warrants further investigation. Moreover, given that we cannot identify a causal relationship in this study, the results are not conclusive, emphasizing the need for more studies pertaining to the effects of language environments on child language development, particularly among those with hearing loss.

Conclusion

This study contributes to the growing body of work on the quantity and diversity of language input and highlights factors that may influence the language development of young children with hearing loss who use listening and spoken language. The number of conversational turns in which the children engaged on both weekdays and weekends related significantly to the amount of language they used expressively, emphasizing the importance of linguistic engagement between children and adults. Although the quantity of language input was consistent across settings, differences were found in the diversity of language to which children were exposed at school and at home. This suggests that parents of children with hearing loss would benefit from increased support in enriching the syntax, clauses, and lexicon they use with their children, especially during routine activities, like mealtimes.

Given the small sample size in this study, future studies should replicate its method on a larger scale, perhaps incorporating a means to examine nonverbal aspects of language, such as body language, facial expressions, and gestures. LENA provides the means to measure the spoken language environments of all children, including those with hearing loss; therefore, we anticipate continued use of this technology with children with hearing loss with supplementary use of other technology (e.g., video recordings) to capture nonverbal communication. Future studies might also investigate the diversity of adult language in terms of sensitivity, responsiveness, or the use of conversation techniques. This may also include observational field notes, analysis of conversational topics, as well as communication breakdown and repair to further examine interactional and socio-linguistic variables. Finally, future comparisons of teacher and parent language environments should account for teacher characteristics, such as years of experience and certification (including the Listening and Spoken Language Specialist credential). By incorporating the results of this and future studies, parent-centered intervention and professional development offered to educators can improve the quantity and diversity of language input provided to children with hearing loss and begin to close the language gap between these children and their peers with typical hearing.

Conflicts of interest

No conflicts of interest were reported.