Oral Language Acquisition in Preschool Children Who are Deaf and Hard-of-Hearing

Abstract

The purpose of this study was to compare developmental trajectories of oral language acquisition of children who are deaf and hard of hearing (DHH) and children with typical hearing across the preschool years. Thirty children who are DHH who use amplification and spoken language and 31 children with typical hearing completed an early language and literacy assessment battery every six months from age 4 to age 6. The developmental trajectories of each group’s language skills were examined via growth curve analysis. Oral language skills were lower for children who are DHH than for children with typical hearing at study entry. For vocabulary, children who are DHH demonstrated growth over the two years but did not close the gap in performance over time. For morphosyntax, specifically verb tense marking, children who are DHH demonstrated growth over preschool, becoming more adult-like in their productions.

Introduction

Despite advancements in amplification technology, as a group, children who are deaf and hard of hearing (DHH) who use amplification and are developing spoken language, continue to perform lower on measures of oral language than children with typical hearing. These deficits are evident across multiple areas of spoken language, including vocabulary (Lund, 2016; Pittman, Lewis, Hoover, & Stelmachowicz, 2005; Wake, Poulakis, Hughes, Carey-Sargeant, & Rickards, 2005; Werfel, 2017) and morphosyntax (McGuckian & Henry, 2007; Nittrouer, Sansom, Low, Rice, & Caldwell-Tarr, 2014; Werfel & Douglas, 2017; Werfel, 2018). Language deficits in children who are DHH are evident by the preschool years (Easterbrooks, Lederberg, Miller, Bergeron, & McDonald Connor, 2008; Moeller, Tomblin, Yoshinaga-Itano, Connor, & Jerger, 2007; Niparko, Tobey, Thal, et al., 2010; Tomblin et al., 2015; Werfel, 2017). The purpose of the current study was to compare growth in early oral language skills, specifically vocabulary and morphosyntax, of children who are DHH and children with typical hearing over multiple time points. Describing the early trajectory of these foundational oral language skills is a first step to determine areas in which early interventions have the potential to be maximally impactful for children who are DHH.

Vocabulary Acquisition in Children Who are DHH

Vocabulary deficits of children who are DHH have been reported across a range of semantic skills, including vocabulary breadth and depth, as well as word learning skills. First, children who are DHH have less vocabulary breadth than their peers with typical hearing. For example, Wake and Poulakis (2004) and Wake et al. (2005) reported that children who are DHH performed below average on measures of receptive vocabulary in a large Australian study. Lund’s (2016) meta-analysis of vocabulary performance in children who are DHH confirmed these findings and extended them to expressive vocabulary. When compared to performance of children with typical hearing, children who are DHH had a mean difference across studies of −11 standard score points for expressive vocabulary and −20 standard score points for receptive vocabulary measures. In addition to widely documented weaknesses in vocabulary breadth for children who are DHH, deficits are also evident in vocabulary depth, or how well the words are known, compared to children with typical hearing (Walker, Redfern, & Oleson, 2019). Walker et al. (2019) reported that these deficits in vocabulary depth compared to children with typical hearing were even more pervasive than deficits in vocabulary breadth. This finding shows yet another weakness associated with later literacy skills for children who are DHH; stronger vocabulary depth is associated with better reading comprehension (Ouellette, 2006).

Children who are DHH also have demonstrated deficits in word learning skills. Lund and Schuele (2017) and Robertson, von Hapsburg, and Hay (2017) have reported that children who are DHH aged three and under showed decreased ability to learn new words compared to children with typical hearing. Walker and McGregor (2013) likewise found that preschool children who are DHH learn new words at a slower rate than children with typical hearing. These findings also have been reported for toddlers, as well as school-aged children who are DHH (Moeller, Osberger, & Eccarius, 1986; Nott, Cowan, Brown, & Wigglesworth, 2009; Pittman et al., 2005).

In terms of developmental trajectories of vocabulary skills, findings have been mixed across studies as to whether children who are DHH acquire vocabulary at the same rate as children with typical hearing. Expressive vocabulary deficits in children who are DHH can emerge as early as the time of children’s first words (Moeller et al., 2007). During the very early period of vocabulary acquisition in Moeller, Hoover, et al. (2007), the rate of expressive vocabulary growth was similar to that of children with typical hearing, and these children did not catch up to their same-age peers by the study’s end at 24 months. There were no differences in level or growth of receptive vocabulary reported by mothers in Moeller, Hoover, et al. (2007). Likewise, Werfel (2017) reported that preschool children who are DHH performed worse on measures of expressive and receptive vocabulary at age 4 compared to their typical hearing peers. Their rate of growth over the next six months was similar to their peers with typical hearing and was insufficient to close the gap; despite growth, they were unable to make the dramatic gains necessary to catch up. Other researchers, however, have reported slower than typical growth in vocabulary for children who are DHH (Easterbrooks et al., 2008), consistent with Walker and McGregor (2013) who reported that children who are DHH learn new words more slowly than children with typical hearing.

Morphosyntax Acquisition in Children Who are DHH

Deficits in morphosyntax also have been reported for children who are DHH. First, children who are DHH have lower mean lengths of utterance than their peers with typical hearing and use fewer bound morphemes (Koehlinger, Van Horne, & Moeller, 2013; Nicholas & Geers, 2018; Nittrouer et al., 2014; Werfel, 2018). Additionally, children who are DHH exhibit a higher rate of errors in their production of morphosyntax features; these errors include errors of omission and errors of commission (Werfel & Douglas, 2017; Werfel, 2018). That is, their morphosyntactic errors are not limited to errors of absence but also include incorrect productions. Finally, not only do children who are DHH use fewer morphosyntax features and exhibit a greater proportion of errors when they do, they appear to acquire morphosyntax features in a different developmental order than children with typical hearing (McGuckian & Henry, 2007; Moeller, Tomblin, et al., 2007; Werfel & Douglas, 2017). Moeller et al. (2010) argued that delayed access to auditory information due to a period without amplification combines with ongoing decreased audibility even with amplification to result in protracted development of morphosyntactic skills in children who are DHH. In English, much morphosyntactic information is stored in word-final high frequency sounds, such as third person singular (she runs) and possessive (the cat’s), which are of noted difficulty perceptually for children who are DHH. Werfel (2018) additionally reported that children who are DHH produced more errors in verb tense than children with typical hearing, beyond verb tense marked by high-frequency phonemes (e.g., irregular third person singular).

Motivation for the Present Study

Much of the existing body of work on oral language in preschool children who are DHH is cross-sectional or includes only two time points during the preschool years (Ching & Dillon, 2013; Easterbrooks et al., 2008) and, thus, does not provide information on the development of oral language skills across multiple time points during the preschool years. Exceptions include the Outcomes of Children with Hearing Loss study (Tomblin et al., 2015), which included yearly assessments during preschool for children who are hard of hearing, and the Childhood Development at Cochlear Implantation study (Niparko et al., 2010), which included six-month interval testing following cochlear implantation through age 5.

In the Outcomes of Children with Hearing Loss study (Tomblin et al., 2015), children who are hard of hearing, defined as a better-ear pure tone average between 25 and 75 dB HL, were followed longitudinally, including across the preschool years. During preschool, children were assessed one time per year, and standardized measures of receptive and expressive language were administered. Findings indicated that children who are hard of hearing had lower overall spoken language performance than peers with typical hearing, and the magnitude of group differences approached 1 SD. Additionally, Tomblin et al. (2015) reported that children who are hard of hearing had more vulnerability in morphosyntax than vocabulary; however, growth of these skills was not measured separately over time.

In the Childhood Development at Cochlear Implantation study (Niparko et al., 2010), children with severe to profound hearing loss who use cochlear implants were followed longitudinally for three years after cochlear implantation. Therefore, only some children were in the preschool years across all testing sessions; many were younger or older than preschool age for some or all time points. Assessment occurred at 6-month intervals during the first year, and annually in years 2 and 3 of the study, and standardized measures of receptive and expressive language were administered. Findings indicated that children with cochlear implants had lower receptive and expressive language than children with typical hearing over time, and the gap between groups appeared to increase over the course of three years.

Our study differs in three important ways from this work: (a) we report vocabulary and morphosyntax separately, based on findings of Tomblin et al. that morphosyntax is more affected by hearing loss than vocabulary for children who are hard of hearing, (b) we include children who are hard of hearing who use hearing aids and children with profound hearing loss who use cochlear implants, and (c) we include testing intervals at 6 months throughout the preschool years. In the present study, we followed preschool children who are DHH and preschool children with typical hearing from age four to six, assessing at six-month intervals. We examined the development of multiple oral language skills: expressive vocabulary, receptive vocabulary, and morphosyntax, over the preschool year to compare developmental trajectories of children who are DHH and children with typical hearing. Our aim was to compare developmental trajectories of language skills of children who are DHH and children with typical hearing across the preschool years, with a specific, separate focus on (a) vocabulary and (b) morphosyntax.

Method

All study procedures were approved by the [University of South Carolina] Institutional Review Board.

Participants

Participants were 30 preschool children who are DHH (18 boys) and 31 preschool children with typical hearing (CTH; 10 boys). Children who are DHH were recruited through social media (e.g., parent groups for children who are DHH), speech-language pathology and audiology clinics, and listening and spoken language preschool programs throughout the US. Children with typical hearing were recruited through community preschools, pediatricians’ offices, and social media.

All participants reported English as the language spoken in the home at least 70% of the time, had no additional diagnoses known to affect language and literacy skills, such as autism or Down syndrome, and had nonverbal intelligence within the average range, as measured by the Primary Test of Nonverbal intelligence (Ehrler & McGhee, 2008). Children who are DHH scored lower on measures of omnibus spoken language and articulation compared to children with typical hearing as measured by the Test of Early Language Development—Fourth Edition (Hresko et al., 2018) and the Arizona Articulation and Phonology Scale—Fourth Edition (Fudala & Stegall, 2017), respectively. Chi square analyses revealed that the DHH and CTH groups did not differ by distribution of ethnicity, race, or speech perception skill, measured by the Early Speech Perception Test (ESP; Moog et al., 2012). Group comparisons of age, English use in home, maternal education, nonverbal intelligence, and speech and language performance are displayed in Table 1. The majority of children in each group identified as White (DHH: 80%, CTH: 86%; p > .05) and not Hispanic or Latino (DHH: 16%, CTH: 3%; p > .05). Only one child in each group had speech perception skills lower than Consistent Word Identification (DHH: 3%, CTH: 3%; p > .05), the highest score possible on the ESP.

Table 1.

Demographic information for participants by group

Variable	DHH mean (SD) range	CTH mean (SD) range	t	p	d
Age at study entry (months)	52.27 (4.41) 45–62	49.87 (4.34) 45–61	−2.138	.037	0.55
English use in home (percent)	94.44 (12.20) 70–100	97.43 (7.72) 70–100	1.117	.269	−0.29
Maternal education (years)	15.78 (2.72) 12–22	17.23 (2.20) 12–22	2.274	.027	−0.59
Nonverbal intelligence*	106.73 (16.68) 76–137	114.77 (11.42) 89–145	2.190	.033	−0.56
Omnibus spoken language*	89.60 (22.69) 41–125	115.42 (1.93) 88–133	5.632	<.001	−1.45
Speech production*	86.30 (11.33) 55–108	94.35 (9.21) 80–114	3.052	.003	−0.77

Note: DHH = children who are deaf and hard of hearing; CTH = children with typical hearing; ^*standard score.

All children who are DHH had a diagnosis of permanent bilateral hearing loss by a certified audiologist, used amplification, and were developing primarily spoken language. Amplification type varied for children who are DHH: 16 children used at least one cochlear implant (12 bilateral, 4 bimodal), 12 children used bilateral hearing aids, and 2 children used bone-anchored hearing aids. Degree of hearing loss ranged from moderate to profound, and all children who are DHH were receiving speech-language services, per parent report. Table 2 displays additional audiologic information for the DHH group. Children with typical hearing passed a bilateral hearing screening at 20 dB HL in a quiet, but not sound-treated, room.

Table 2.

Audiologic information of DHH group

Variable	Mean (SD)	Range	Median
Age at identification	7.67 (11.62)	0–36	1.50
Age at first hearing aid	11.97 (11.70)	1.5–36	5.00
Age at first implant (n = 16)	21.50 (11.91)	9–44	16.50

Note: DHH = children who are deaf and hard of hearing; all ages reported in months. Age at identification was collected via parent response to “age at which your child’s hearing loss was identified.”

Measures

Of interest in the present study were measures of oral language. See Figure 1 for the longitudinal study timeline.

Figure 1.

Longitudinal study design.

Vocabulary

Expressive vocabulary was assessed using the Expressive One-Word Picture Vocabulary Test—Fourth Edition (EOWPVT; Brownell, 2011). The examiner presented children with a color drawing of an image and asked, “What is this?” Standard scores (i.e., mean of 100 with SD of 15 points) were calculated following test manual instructions. Raw scores were used in study analyses. Test–retest reliability reported in the test manual is .93–.97 and is based on an average 19-day gap between testing sessions. The test–retest reliability observed within the present participant sample was based on the six-month gap between testing sessions and ranged from .79–.89.

Receptive vocabulary was assessed using the Peabody Picture Vocabulary Test-Fourth Edition or Fifth Edition (PPVT; Dunn, 2019; Dunn & Dunn, 2007), depending on the date of testing. Children viewed a page with four colored pictures and were instructed to point to the picture that matched the word provided by the examiner. Standard scores were calculated following test manual instructions. Raw scores were used in study analyses. Test–retest reliability reported in the test manual is .89 for the PPVT-5 and .91–.94 for the PPVT-4. Sample test–retest reliability for the PPVT was .82–89, again with a six-month gap between each test administration.

Morphosyntax

Morphosyntax, specifically verb tense marking, was assessed using the Rice/Wexler Test of Early Grammatical Impairment (TEGI; Rice & Wexler, 2001). Participants completed the TEGI screener, which consists of the past tense and the third person singular subtests.

The TEGI past tense subtest was used to assess children’s use of past tense marking. Children were presented with two pictures. In the first, a child is performing an action (e.g., painting) and in the second, the same child is shown after the action has been completed. The examiner described the first picture with the present progressive form of the target verb (e.g., painting); the child was then asked to describe the second picture. The subtest includes items aimed to elicit 10 regular and 8 irregular past tense verbs. Child responses for regular verbs were marked as correct (e.g., He painted), incorrect (e.g., He paint), unscorable (e.g., He was painting), or no response. For irregular verbs, child responses were marked as correct (e.g., She ate), overregularization (e.g., She eated cookies), unscorable (e.g., She is eating), or no response. The score for the Past Tense probe was determined by adding up the child’s correct and overregularized productions and dividing by the sum of correct, overregularized, and incorrect productions to determine the percentage of responses that contain past-tense marking. Test–retest reliability reported in the test manual is .82.

The TEGI third person singular subtest was used to assess children’s use of third person singular marking. Children viewed one picture of a person (e.g., pilot). The examiner said, for example, “Here is a pilot. Tell me what a pilot does.” The child’s response had to include both a subject and a verb. If it did not, the prompting hierarchy in the test manual was utilized. This subtest includes 10 items. Child responses were scored as correct (e.g., She flies planes), incorrect (e.g., She fly helicopters), unscorable (e.g., She is flying a plane), or no response. The third person singular probe score was determined by dividing the number of correct productions by the sum of the child’s correct and incorrect productions to determine the percentage of responses that contain third person singular marking. Test–retest reliability reported in the test manual is .92.

The morphosyntax score was the TEGI Screener Probe Score, which is an average of the past tense and third person singular probe scores. Children with final consonant deletion of /s/ and /z/ or /t/ and /d/ as determined by the TEGI phonological probe were not administered this measure: this resulted in 26 total exclusions across all 5 times: 8 at Time 1, 6 at Time 2, and 4 each at Times 3, 4, and 5. For the present sample, test–retest reliability for the TEGI Screener Probe Score ranged from .78–.94 between each six-month testing window.

Procedure

Participants completed language and early literacy assessment batteries at six-month intervals, beginning at approximately four years of age, through age six. Study participation occurred over one or two assessment sessions at each time point, depending on behavior and attention. Total testing time for each time point was approximately 1.5–2 h, and testing sessions took place individually in a quiet room at the child’s school, home, or a local library. Parents were permitted to be present if desired but were instructed to refrain from participating in the testing. Reminders were provided as needed. Order of administration was prerandomized for each participant to control for order effects.

Assessments were administered by trained examiners who were members of the Written Language Lab. Examiners included the lab director, speech-language pathologists, and graduate students in speech-language pathology. Examiners completed the following training procedures prior to conducting study assessments: First, examiners read the administration chapter of test manuals for each assessment. Next, examiners passed a knowledge test on the administration of each measure with 100% accuracy. After passing the knowledge test, examiners observed via video recording or live assessment session the administration of each measure by an already-trained examiner. Examiners then practiced test administration with another lab member until comfortable with procedures and administered each assessment with 0 errors to a certified speech-language pathologist.

Reliability

All tests were double-scored item-by-item for accuracy, and any disagreements were resolved by discussion. Disagreements of scoring were rare, and generally consisted of incorrect calculation of basals. Therefore, final scores represent 100% scoring agreement. Study data were managed by REDCap electronic data capture tools (Harris et al., 2009) hosted at the University of South Carolina.

Analysis

Children’s scores on each of the language measures were z-scored based on the full sample means and standard deviations to maximize the interpretability of the findings. Z-scoring allowed zero, for each measure, to be equivalent to the average score children received on that measure. A value of −1 was equivalent to a score one standard deviation below the sample average and a value of +1 was equivalent to a score one standard deviation above the sample average.

Children’s scores on the z-scored expressive and receptive vocabulary measures were averaged as a composite to minimize the impact of any missing data. There were three instances where a child had a score for one vocabulary measure but not the other. In these cases, the composite vocabulary measure included only the measure that had been administered.

Missing data

Missing data were examined by group and time point (see Tables 3 and 4), revealing an overall missing data rate of 18.69%. Data were examined for patterns in missingness based on all measures included in the present dataset, including group membership and performance on background measures. One pattern was identified in the missing data, with age of enrollment serving as the primary driver of missingness. Children who enrolled later in the study (e.g., at Time 2 or Time 3) automatically had missing data for the earlier time points. No additional evidence was observed to suggest that missing data might follow a different pattern compared to the modeled results. Consequently, data was treated as missing at random.

Table 3.

Descriptive statistics of study variables by group across assessment points

Variable	Children who are DHH			Children with typical hearing			Tests for group differences
	n	Mean	SD	n	Mean	SD	t	p	d
Expressive vocabulary
Time 1	16	94.93	15.83	20	117.15	11.03	4.899	<.001	−1.63
Time 2	26	102.32	16.52	28	120.61	10.19	4.907	<.001	−1.33
Time 3	30	101.00	18.17	31	114.42	9.28	3.614	.001	−0.93
Time 4	27	100.11	17.22	26	115.65	9.85	4.108	<.001	−1.11
Time 5	21	98.57	14.03	22	115.91	11.08	4.508	<.001	−1.37
Receptive vocabulary
Time 1	16	91.40	16.85	21	115.90	13.42	4.857	<.001	−1.61
Time 2	26	98.38	17.77	28	118.04	11.79	4.761	<.001	−1.30
Time 3	30	97.60	17.86	31	113.77	1.09	4.336	<.001	−1.11
Time 4	26	97.93	15.75	26	115.38	10.37	4.747	<.001	−1.31
Time 5	20	94.15	13.71	22	114.18	9.76	5.493	<.001	−1.68
Morphosyntax
Time 1	9	14.22	17.68	20	73.30	23.55	6.697	<.001	−2.84
Time 2	21	49.85	29.11	27	82.67	22.07	4.445	<.001	−1.27
Time 3	26	50.69	33.37	31	90.82	8.10	5.982	<.001	−1.65
Time 4	24	61.92	30.46	25	91.06	9.52	4.565	<.001	−1.29
Time 5	18	62.44	34.73	22	89.75	11.12	3.204	.004	−1.06

Note: DHH = deaf and hard of hearing. Vocabulary scores are standard scores, and morphosyntax scores are percent correct.

Table 4.

Descriptive statistics of Z-scores for focus measures

Variable	Children who are DHH			Children with typical hearing			Full sample
	n	Mean	SD	n	Mean	SD	n	Mean	SD
Vocabulary
Time 1	16	−1.34	0.90	21	−0.39	0.55	37	−0.80	0.86
Time 2	26	−0.76	0.86	28	0.22	0.54	54	−0.25	0.86
Time 3	30	−0.42	0.94	31	0.38	0.53	61	−0.02	0.85
Time 4	27	−0.12	0.94	26	0.80	0.49	53	0.33	0.88
Time 5	21	0.12	0.88	22	1.08	0.51	43	0.61	0.86
Morphosyntax
Time 1	9	−1.64	0.94	20	0.05	0.80	29	−0.48	1.15
Time 2	21	−0.75	0.96	27	0.47	0.52	48	−0.06	0.95
Time 3	26	−0.71	1.11	31	0.64	0.27	57	0.02	1.02
Time 4	24	−0.35	1.04	25	0.65	0.32	49	0.16	0.91
Time 5	18	−0.32	1.18	22	0.61	0.36	40	0.19	0.95

Note: DHH = deaf and hard of hearing.

Modeling

Children’s composite vocabulary and morphosyntax scores were examined descriptively and plotted to visualize growth over time by group. Growth models were specified separately for each outcome measure in a mixed-effect regression framework (McNeish & Matta, 2018) using restricted maximum likelihood estimation within the lme4 package (Bates, Maechler, Bolker, & Walker, 2015) in the R environment (R Core Team, 2020). Mixed effect models allow for inclusion of both fixed and random effects, where fixed effects reflect consistent estimates across individuals (similar to linear regression estimates) and random effects reflect variation within and between individuals included in the sample. This framework is useful for investigating changes in children’s assessment scores over time while accounting for within-child variation. This facilitates more precise examination of general growth trends.

For both the vocabulary and morphosyntax models, time was included as a fixed effect and centered at the first assessment time point (i.e., approximately age four for the participants), so that a one-unit increase in time was equivalent to a six-month interval aligned with the assessment windows. Group was also entered as a fixed effect, with children who are DHH designated as the reference group. Both models were examined for evidence of nonlinear growth (i.e., a quadratic term) and interaction effects between linear growth and group. Interactions reflect differences in growth by group. A significant interaction term would indicate that children who are DHH exhibit different growth trajectories, with either larger or smaller changes over time, in their language scores compared to children with typical hearing.

Model fit was examined through: (1) evaluation of variance parameters for evidence of misfit, such as negative variance estimates; (2) model alignment with assumptions including residual normality, homogeneity of variance, and linearity; and (3) examination of model estimates, p-values, and R² values, which were assessed in relation to the theoretical framework of the present study. The specific models identified based on this approach were:

Results

Descriptive statistics for the vocabulary scores and morphosyntax TEGI Screener Probe score are provided by time point in Tables 3 and 4. No evidence of non-normality (i.e., distributional skew or kurtosis, outliers, ceiling/floor effects) was observed in the participants’ vocabulary composite scores (Table 4), which were computed from students’ z-scored expressive and receptive vocabulary. Contrastively, descriptive examination of participants’ morphosyntax scores over time revealed evidence of a ceiling effect, as some students scored the maximum for morphosyntax prior to the last assessment window. Some heteroscedascity by group was also observed. This was indicated both by plots of students’ morphosyntax scores by group and by a significant Levene’s test: F(1, 221) = 1.81, p = .001. More variability was observed in the morphosyntax scores of children who are DHH compared to those with typical hearing.

Examination of children’s growth in vocabulary scores over time revealed significant positive, linear growth over the course of the two years of the study (see Table 5). On average, both children who are DHH and those with typical hearing grew approximately .41 standard deviations (95% CI [.38, .44], p < .001) every six months. However, children with typical hearing scored .91 standard deviations (95% CI [.55, 1.28], p < .001) higher than children who are DHH at time 1. No evidence of change in this gap was observed, as there were no significant interactions between group and time (see Table 6). Figure 2 depicts the fitted values from the model. The solid line represents the predicted vocabulary scores for children with typical hearing, and the dotted line represents predicted vocabulary scores for children who are DHH. The shading indicates the standard errors for the model estimates. Time point and group membership accounted for approximately 45.1% of the variation in children’s vocabulary scores.

Table 5.

Growth model predicting vocabulary over time: main effects

Main effects model: fixed effects
Predictors	Estimates	Conf. Int (95%)	p-value
Intercept	−1.30	−1.57 to −1.03	<.001
Time (centered at time 1)	0.41	0.38–0.44	<.001
Group (CTH)	0.91	0.55–1.28	<.001
Random effects
n  Children = 61 Child	σ2 = .07
Observations = 248	τ00 = .51 Child
	ICC = .87
	Marginal R2/conditional R2 = .451/.931

Note. The intercept may be interpreted as the number of standard deviations below the sample mean a child who is deaf or hard of hearing would be predicted to score on vocabulary at Time 1 (approximately 4 years of age). CTH = children with typical hearing. The marginal R² indicates the variance in children’s vocabulary scores accounted for only by the fixed effects (i.e., time, group), whereas the conditional R² accounts for both fixed and random effects.

n = number of level-two clusters in the dataset (i.e., number of children who completed vocabulary assessments).

Observations = total number of datapoints in the dataset. Each child contributed between 1 and 5 time points.

σ² = within-child residual variance

τ₀₀ = between-child variance

ICC = intraclass correlation coefficient, reflecting the proportion variance in vocabulary scores attributable to child-specific clustering or nesting. Bolded values are significant at p < .001.

Table 6.

Growth model predicting vocabulary over time: interaction model

Interaction model: fixed effects
Predictors	Estimates	Conf. int (95%)	p-value
Intercept	−1.33	−1.61 to −1.06	<.001
Time (centered at Time 1)	0.42	0.39–0.46	<.001
Group (CTH)	0.97	0.59–1.36	<.001
Interaction: time × group	−.03	−.09 to .02	.270
Random effects
n  Children = 61 Child	σ2 = .07
Observations = 248	τ00 = .51 Child
	ICC = .87
	Marginal R2/conditional R2 = .451/.931

Note. The intercept may be interpreted as the number of standard deviations below the sample mean a child who is deaf or hard of hearing would be predicted to score on vocabulary at Time 1 (approximately 4 years of age). CTH = children with typical hearing. The marginal R² indicates the variance in children’s vocabulary scores accounted for only by the fixed effects (i.e., time, group, interaction term), whereas the conditional R² accounts for both fixed and random effects.

n = number of level-two clusters in the dataset (i.e., number of children who completed vocabulary assessments).

Observations = total number of datapoints in the dataset. Each child contributed between 1 and 5 time points.

σ² = within-child residual variance

τ₀₀ = between-child variance

ICC = intraclass correlation coefficient, reflecting the proportion variance in vocabulary scores attributable to child-specific clustering or nesting. Bolded values are significant at p < .001.

Figure 2.

Vocabulary performance over time by group.

Given the heteroscedasticity observed between the groups of children based on their hearing status, growth models for morphosyntax were estimated twice following recommendations from Field and Wilcox (2017). After the model was specified in the lme4 package, the model was re-specified in the robustlmm package (Koller, 2016). The primary findings from the fixed effects were highly consistent across the two estimation approaches, indicating stability in the specific estimates of interest. We report the results from the robust analysis in Tables 7 and 8. Overall, children’s scores increased between each time point, though overall growth slowed over the course of the five assessment windows (Linear Growth Est. = .41, with Quadratic = −.05, p = .01). Specifically, the participants’ morphosyntax scores increased approximately .36 standard deviations between the first and second time points, but between the second and third time points, the average increase was only .26 standard deviations. This slowing of growth over time continued across the five time points.

Table 7.

Growth model predicting morphosyntax over time: main effects (robust)

Main effects model: fixed effects
Estimate	Estimates	Conf. int (95%)	p-value
Intercept	−1.31	−1.61 to −1.00	<.001
Time (centered at Time 1)	0.41	0.25–0.58	<.001
Time: quadratic	−.05	−.09 to −.01	.010
Group (CTH)	1.28	0.91–1.64	<.001
Random effects
n  Children = 60 Child	σ2 = .19
Observations = 223	τ00 = .43 Child
	ICC = .69
	Marginal R2/conditional R2 = .419/.817

Note. The intercept may be interpreted as the number of standard deviations below the sample mean a child who is deaf or hard of hearing would be predicted to score on morphosyntax at Time 1 (approximately 4 years of age). CTH = children with typical hearing. The marginal R² indicates the variance in children’s morphosyntax scores accounted for only by the fixed effects (i.e., linear time, quadratic time, group), whereas the conditional R² accounts for both fixed and random effects.

n = number of level-two clusters in the dataset (i.e., number of children who completed morphosyntax assessments).

Observations = total number of datapoints in the dataset. Each child contributed between 1 and 5 time points.

σ² = within-child residual variance

τ₀₀ = between-child variance

ICC = intraclass correlation coefficient, reflecting the proportion variance in morphosyntax scores attributable to child-specific clustering or nesting.

Table 8.

Growth model predicting morphosyntax over time: interaction model (robust)

Interaction model: fixed effects
Estimate	Estimates	Conf. int (95%)	p-value
Intercept	−1.80	−2.12 to −1.49	<.001
Time (centered at Time 1)	0.69	0.53–.85	<.001
Time: quadratic	−.06	−.09 to −.03	<.001
Group (CTH)	1.99	1.59–2.39	<.001
Interaction: time × group	−0.35	−0.44 to −.26	<.001
Random effects
n  Children = 60 Child	σ2 = .14
Observations = 223	τ00 = .41 Child
	ICC = .74
	Marginal R2/conditional R2 = .467/.861

Note. The intercept may be interpreted as the number of standard deviations below the sample mean a child who is deaf or hard of hearing would be predicted to score on morphosyntax at Time 1 (approximately 4 years of age). CTH = children with typical hearing. The marginal R² indicates the variance in children’s morphosyntax scores accounted for only by the fixed effects (i.e., time, group, interaction term), whereas the conditional R² accounts for both fixed and random effects.

n = number of level-two clusters in the dataset (i.e., number of children who completed morphosyntax assessments).

Observations = total number of datapoints in the dataset. Each child contributed between 1 and 5 time points.

σ² = within-child residual variance

τ₀₀ = between-child variance

ICC = intraclass correlation coefficient, reflecting the proportion variance in morphosyntax scores attributable to child-specific clustering or nesting.

Main effects modeling revealed that, on average, children with typical hearing scored 1.28 standard deviations (95% CI [.91, 1.68], p < .001) higher than children who are DHH at the first time point (see Table 7). However, an interaction was noted between time and group membership, suggesting that the morphosyntax scores of children who are DHH increased at a higher rate over time compared to those of children with typical hearing. This is evidenced by the significant interaction term shown in Table 8: −.35, 95% CI [−.44, −.26], p < .001. This finding is likely attributable in part to a ceiling effect on the measure, which children with typical hearing tended to reach sooner than did the children who are DHH, consistent with theoretical and data-based expectations of verb morphology production across preschool. The results are depicted in Figure 3, again with separate lines representing the predicted scores for children with typical hearing and those for children who are DHH. The curved lines reflect the significant slowing of growth over time (i.e., the quadratic term), and the shading indicates the standard errors of the estimates.

Figure 3.

Morphosyntax performance over time by group.

As a further assessment of the robustness of the results, we conducted a post-hoc sensitivity analysis of the model estimates. As shown in Table 1, the groups exhibited significant differences in maternal education and nonverbal IQ as measured by the PTONI (Ehrler & McGhee, 2008), which could influence language development. The observed differences and interactions reported above remained after controlling for maternal education and nonverbal IQ. These results are provided as Supplementary Material.

Discussion

The primary aims of this study were to compare the developmental trajectories of preschool vocabulary and morphosyntax skills for children who are DHH and children with typical hearing. We present three main findings. First, deficits across oral language skills are already present for children who are DHH by age 4. Consistent with findings reported in Author (XXXX), the children who are DHH had lower performance across vocabulary and morphosyntax skills at the outset of this study. Second, children who are DHH made similar gains as children with typical hearing in vocabulary. In the growth models for vocabulary, the main effect of time was positive and significant. Third, children who are DHH made substantial gains in morphosyntax across the preschool years, but even by school entry their performance in verb marking was lower than their same-age peers. In addition to positive and significant growth, a significant interaction effect was present for morphosyntax, but it is likely a result of ceiling effects on the measure for children with typical hearing.

Vocabulary Trajectories across the Preschool Years for Children Who are DHH

At all time points, the group means for expressive and receptive vocabulary were within the average range for children who are DHH relative to the normative sample for the respective assessments; however, the mean scores were generally in the low range of average. It might be tempting to consider that a score within the average range is indicative of a child who are DHH “catching up” in their vocabulary skills, and several previous reports have concluded that mean scores within the average range represent normalized language skills (Baldassari et al., 2009; Nicholas & Geers, 2007). The group mean on these measures for the children with typical hearing in our sample, conversely, were generally above the average range relative to the normative sample, and, consistent with much prior work (Lund, 2016; Wake et al., 2005), there were significant differences between groups at all times. These differences remained significant even after accounting for small group differences in nonverbal IQ and maternal education, likely reflecting a real difference in vocabulary skills between children who are DHH relative to children with typical hearing. Importantly, our findings add to a growing body of research that suggests that norm-referenced test scores within the average range are not sufficient evidence of similar performance to peers with typical hearing.

Our findings that children who are DHH appear to make similar gains in vocabulary across preschool as do children with typical hearing, extend the findings of Moeller, Hoover, et al. (2007), who reported similar growth in children’s vocabulary from 10 to 16 months. Taken together, these findings suggest that very early vocabulary development is vitally important for children who are DHH, particularly given that all children who are DHH in our sample were enrolled in speech-language services. Previous work had yielded mixed findings in terms of rate of growth, with multiple groups reporting slower growth and impaired word learning skills (Easterbrooks et al., 2008; Walker & McGregor, 2013). In our study, however, both groups in this study made gains over time, and the interaction model indicated no difference in gains between the groups. The standard scores at each time for each of the vocabulary measures also were relatively stable, indicating that children were making age-appropriate progress between each visit. This effect of “one year’s growth in one year’s time” phenomenon is that, during preschool, children who are DHH are not falling further behind on the vocabulary skills measured in this study, expressive and receptive vocabulary depth, across their preschool years.

When considering that children who are DHH are already behind in vocabulary acquisition by age four and that all children who are DHH in this study were enrolled in speech-language services, however, our findings also indicate that children who are DHH are not making sufficient gains to close the gap in their vocabulary performance that appears to emerge as early as first words (Moeller, Hoover, et al., 2007) prior to school entry. Given the importance of preschool oral language for later literacy achievement (NELP, 2008), we conclude that vocabulary intervention, including not just a focus on adding words to the lexicon but also on how to learn new words, is an important target in early intervention and preschool for children who are DHH, and that current practice is not sufficient in catching these children up by school entry.

Morphosyntax Trajectories across the Preschool Years for Children Who are DHH

To our knowledge, this is the first report of developmental trajectories of morphosyntax of children with hearing loss across the preschool years. Importantly, in the area of morphosyntax, although children who are DHH make substantial gains during the preschool years toward adult-like verb morphology production, they still perform significantly below peers with typical hearing at age 6, consist with Tomblin et al.’s (2015) findings that morphosyntax represented an area of greater difficulty for children with hearing loss than vocabulary and extending this finding to elucidate the developmental course. The difference between groups at age 4 was almost two standard deviations, but by age 6 this difference had decreased to just over half a standard deviation. We propose that this finding represents a delay in acquisition of early morphosyntax skills over preschool for children who are DHH. In our comparison group, consistent with a large body of work documenting morphosyntax performance in children with typical hearing (Rice, 2004), performance on verb tense marking at age 4 was already near adult-like and by age 6, children with typical hearing had very consistent marking of grammatical tense. In contrast, our group of children who are DHH, consistent with the group of children with specific language impairment in Rice (2004), exhibited growth over the preschool years but scored below benchmark at each time point. Previous work (Koehlinger et al., 2013; Werfel, 2018) reported morphosyntax deficits in spontaneous language for children who are DHH, and our results extend these findings to an elicited task that creates an obligatory context for particular types of verb tense marking.

Our findings also indicated that linguistic skills play a role beyond speech production skills in these delays in morphosyntax production for children who are DHH. On the morphosyntax measure used in this study, the TEGI, children must pass a speech production screening to ensure that they produce the phonemes of the morphological markers (i.e., /s/, /z/, /t/, /d/) in word final position to complete the morphosyntax subtests. At no time point did more than eight children not complete the measure based on this criterion. By age 5, no more than four children were excluded from completing the measure on the basis of speech production. It is also important to emphasize that the percent correct scores are based only on the productions of those children who pass the speech production screening.

Similar to vocabulary, our morphosyntax findings indicate that children who are DHH do not make sufficient gains across the preschool years to catch up to their peers with typical hearing by school entry despite enrollment in speech-language services. Importantly, previous findings indicate that children who are DHH exhibit not just delays but differences in morphosyntax acquisition compared to children with typical hearing (McGuckian & Henry, 2007; Werfel & Douglas, 2017; Werfel, 2018). Considered together, we further conclude that morphosyntax intervention is an additional important target in early intervention and preschool for children who are DHH.

Clinical Implications

In both linguistic skills measured in the present study, children who are DHH performed lower than children with typical hearing by age 4 and gaps in performance were present at each assessment point through age 6. Thus, our findings have important implications not only for preschool intervention, but also for early intervention. Previous work in vocabulary has pointed to deficits across many vocabulary skills, including the number of words known, the depth of knowledge of words, and the ability to learn new words (Lund, 2016; Lund & Schuele, 2017; Walker et al., 2019). Early interventionists should consider targeting each of these aspects of early vocabulary knowledge when planning intervention for children who are DHH. Direct instruction has been shown to be a more effective vocabulary intervention method than more implicit methods for children who are DHH (Lund & Douglas, 2016). Previous work in morphosyntax has pointed to not only delays, but differences in developmental acquisition of morphological markers for children who are DHH (McGuckian & Henry, 2007; Werfel & Douglas, 2017). Early interventionists should consider teaching techniques such as explicit instruction (Richels, Schwartz, Bobzien, & Raver, 2016) and auditory bombardment and grammatical recasting (Encinas & Plante, 2016) as they target morphosyntax for young children who are DHH. Additional considerations for oral language intervention for children who are DHH may be drawn from other foundational literacy skills, such as repetition, considering task difficulty, and strategically introducing words based on acoustic properties (Werfel & Reynolds, 2019). Finally, professionals should work to ensure consistent auditory access for young children who are DHH who are developing spoken language, as improved audibility is associated with increased oral language skills (Tomblin et al., 2015).

Limitations

These findings should be considered with the following limitations in mind. First, sample matching across groups of children who are DHH and children with typical hearing was imperfect. Within the present study, we observed differences in nonverbal IQ scores, maternal education, and age of enrollment. Although overall findings were robust to measured group differences, it is important to continue to investigate the extent to which sample selection influences findings among children who are DHH. Additionally, the representativeness of race and ethnicity may not match all communities, and these potential differences should be kept in mind when interpreting findings. In terms of services received, although all children who are DHH were receiving SLP services at study enrollment, we did not collect information on the specific training and expertise of the service providers. Ongoing effort to understand the unique experiences and characteristics of this population are essential.

Another consideration within the present work is the presence of missing data, most notably for the morphosyntax measure. To be able to complete the morphosyntax measure reliably, children were required to demonstrate that they could accurately produce phonemes corresponding to the morphosyntax features of interest. Children who did not meet these criteria were not administered the morphosyntax measure, resulting in missing data. As previously described, data was treated as missing at random given that (1) no child had more than three time points of missing data on any measure, and (2) no patterns emerged to suggest that the missing data would substantially alter the primary model results. However, it is worth recognizing that this missing data, combined with the ceiling effect observed on the morphosyntax measure, may have hidden additional patterns in the data. Further work, investigating longitudinal morphosyntax growth at more complex levels, is needed to understand what patterns may be missed as students progress beyond the ceiling of the included measure.

This work was conducted with a small sample of participants and was designed specifically to allow for a two-group comparison of basic forms of growth over five time points. Correspondingly, the study was not sufficiently powered to facilitate examination of more complex forms of growth, further group interactions, or specific predictors of students’ vocabulary and morphosyntax development over time (e.g., socioeconomic factors). Additional research is needed not only to replicate the findings, but also to extend these efforts to better understand how language develops among children with hearing loss.

Conclusion

In this study, we found that children who are DHH perform lower than children with typical hearing on measures of vocabulary and morphosyntax by age 4. For vocabulary, rates of growth did not differ across groups, with children who are DHH making age-appropriate gains over the course of preschool; however, they continued to perform below the children with typical hearing at school entry. There was no closing of the vocabulary gap. Importantly, the group mean of children who are DHH on standardized measures of vocabulary was within the average range; therefore, our findings add to a growing body of research that suggests that norm-referenced test scores within the average range are not sufficient evidence of similar performance to peers with typical hearing. For morphosyntax, children who are DHH exhibited growth toward adult-like production across the preschool years, whereas productions of children with typical hearing were already largely adult-like at age 4. This effect likely represents a delay in morphosyntax acquisition for children who are DHH. Importantly, children who are DHH continued to perform about one standard deviation below their peers with typical hearing, who as a group had mastered verb tense production at school entry. Although children who are DHH exhibit growth in oral language skills over the course of the preschool years, this growth does not appear to be sufficient to result in “catching up” in either vocabulary or morphosyntax before school entry.

Note: Krystal Werfel and Gabriella Reynolds were affiliated with University of South Carolina at the time of data collection and analysis.

Funding

The National Institutes of Health [R03DC014535 to K.W.]. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Conflict of Interest

No conflict of interest was reported.