Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2017 Mar 1.
Published in final edited form as: J Autism Dev Disord. 2016 Mar;46(3):1013–1024. doi: 10.1007/s10803-015-2647-7

Early Predictors of Growth in Diversity of Key Consonants Used in Communication in Initially Preverbal Children with Autism Spectrum Disorder

Paul J Yoder 1, Tiffany Woynaroski 1, Linda Watson 2, Elizabeth Gardner 1, Cassandra R Newsom 1, Bahar Keceli-Kaysili 3
PMCID: PMC4747804  NIHMSID: NIHMS740892  PMID: 26603885

Abstract

Diversity of key consonants used in communication (DKCC) is a value-added predictor of expressive language growth in initially preverbal children with autism spectrum disorder (ASD). Studying the predictors of DKCC growth in young children with ASD might inform treatment of this under-studied aspect of prelinguistic development. Eighty-seven initially preverbal preschoolers with ASD and their parents were observed at five measurement periods. In this longitudinal correlational investigation, we found that child intentional communication acts and parent linguistic responses to child leads predicted DKCC growth, after controlling for two other predictors and two background variables. As predicted, receptive vocabulary mediated the association between the value-added predictors and endpoint DKCC.

Keywords: vocal communication, consonant inventory, predictors, autism, intentional communication, parent linguistic responses

Attaining useful speech by 5 years of age predicts occupational and social outcomes in individuals with autism spectrum disorder (ASD; e.g., Billstedt, Gillberg, & Gillberg, 2007; Eisenberg, 1956; Howlin, Mawhood, & Rutter, 2000; Kobayashi, Murata, & Yoshinaga, 1992; Venter, Lord, & Schopler, 1992). BLINDED (2015) identified four variables with incremental validity in predicting (i.e., added value in explaining) the development of useful speech in initially preverbal preschoolers with ASD. Diversity of key consonants used in communication (DKCC) was one of these value-added predictors.

Rationale for Studying Diversity of Key Consonants Used in Communication

One next step in this line of research is to identify the factors that have incremental validity in predicting DKCC. DKCC is the least studied of the predictors of useful speech. The BLINDED et al. (2015) study replicated an earlier finding that DKCC predicted later expressive language in children with autism who began the study in the early stages of language learning (Wetherby, Watt, Morgan, & Shumway, 2007). Other measures of diversity in consonant use have been shown to predict “useful speech” or spoken language in previous studies involving preschoolers with ASD (Schoen, Paul, & Chawarska, 2011) and children at risk for ASD (Paul, Fuerst, Ramsay, Chawarska, & Klin, 2011). DKCC was called “consonant inventory in communication acts” in our previous report (BLINDED et al., 2015) and in the Wetherby et al. (2007) study. The variable label has been changed here to avoid confusion with consonant inventory variables in the broader literature.

Given its value-added status as a predictor of useful speech growth in young children with ASD, it is surprising that we do not yet know how to facilitate, or whether we can facilitate, DKCC in initially preverbal children with ASD. Identifying the value-added predictors of DKCC growth can shed light on potential mechanisms by which DKCC growth occurs in children with ASD and help us think more precisely about potential reasons that children with ASD vary in DKCC growth. Perhaps most importantly, identifying the value-added predictors of DKCC growth can inform potential treatment targets. Future intervention research might then test whether targeting the identified predictors of DKCC yields highly generalized DKCC growth in children with ASD.

Theoretical Support for Four Potential Predictors of DKCC

Stoel-Gammon (2011) has articulated a theory that implicates four potential predictors of DKCC. The tenets of Stoel-Gammon's theory, as they relate to each of the four potential predictors, are as follows. The vocal tracts of immature speakers are different from adults, and young vocalizers’ control over the muscles used to produce speech is less than the control of adults. Thus, we expect some aspects of motor control, such as motor imitation, to be a predictor of DKCC. However, motor imitation also requires attention to others’ models. Thus, interaction with others (e.g., adults) is an important part of the theory. During interactions with adults, immature speakers hear words for the objects that match the foci of the young vocalizers’ attention and communication. Parent linguistic input may facilitate growth of consonant use in vocal communication, in part, because it helps children notice and emulate the range of sounds that adults use to communicate about objects or events in their environment and/or because children try to say words that have been modeled by adults. However, parent linguistic input would not be beneficial unless children attend to it. Therefore, we expect both parent linguistic responses to child leads and attention to child-directed speech to be predictors of DKCC. The intent to communicate is necessary to use consonants to communicate. Therefore, we expect intentional communication to be a predictor of DKCC.

We call this theory a “transactional theory of speech sound development” because it proposes that not only child factors, but also parent responses to children's leads, will best account for individual differences in DKCC growth in initially preverbal preschoolers with ASD. Like other applications of the transactional theory, it is assumed that parents and children affect each other in ways that change over time. The sequence delineated in the previous paragraph is a simplified version of the bidirectional influence between parents and children that likely contributes to DKCC growth.

Given our interest in the transactional theory of speech development, the DKCC's exclusive focus on vocal communication is critical. We know from the developmental literature that mothers are more likely to interpret their babies’ vocalizations as communicative (BLINDED, 1988) and to respond with linguistic input (West & Rheingold, 1978) when the vocalizations are directed to the mother than when the vocalizations are undirected or directed to an object only. The special role that consonants play is highlighted by the finding that mothers tend to interpret consonant-vowel vocalizations as language-oriented and to respond more to vocalizations with a consonant than to vowel-only vocalizations (Gros-Louis, West, Goldstein, & King, 2006).

Empirical Support for the Potential Predictors of DKCC

Of the four potential predictors of DKCC outlined above (motor imitation, attention to child directed speech, parent linguistic responses, and intentional communication), only the first two have empirical support as predictors of later DKCC in preverbal children with ASD. BLINDED (2012) found that motor imitation and attention to child-directed speech were correlates of later DKCC in initially preverbal children with ASD. No research has been conducted to test whether intentional communication or parent linguistic responses to child leads predict growth in DKCC in children with ASD. Additionally, the effect of the intercorrelation of the four potential predictors on the value-added status of predictors of DKCC has not been studied. Further, predictors of the growth of DKCC have not been studied.

Rationale for Considering Additional Background Variables in Models of DKCC Growth

Ruling out covarying variables that provide less compelling explanations for predicted associations improves the clinical value of expected correlational findings. This is particularly true if the covarying variables are less malleable than the theoretically-motivated potential predictors. Level of cognitive impairment and degree of autism symptomatology are among the most salient child background variables that could account for our predicted associations. Thus, these background variables need to be considered (i.e., controlled) when testing whether more theoretically-motivated predictors account for growth in DKCC in our sample.

Why Receptive Vocabulary Might be a Mediator for the Prediction of DKCC

Although tested in a correlational design, motivating theories for the prediction that receptive vocabulary will mediate the association between value-added predictors and later DKCC are stated in causal terms. Two paths of influence motivate the prediction. The first path of influence is quantified by the association between the predictor (e.g., early parent linguistic responses) and the mediator (i.e., midpoint receptive vocabulary). The second is quantified by the association between midpoint receptive vocabulary and endpoint DKCC. The transactional theory of speech development posits both of these pathways.

The first path has already been empirically established for all four putative predictors of DKCC growth. Past work has demonstrated that parent linguistic responses to child leads are associated with later receptive language in children with ASD who are in the early stages of language development (Haebig, McDuffie, & Weismer, 2013a; 2013b; BLINDED et al., 2015). Studies have additionally shown links for early attention to child directed speech and intentional communication with later receptive language in this population (BLINDED et al., 2015). Motor imitation ability has specifically been identified as a replicated predictor of productive language in children with ASD (Charman, Baron-Cohen, Swettenham, Baird, Drew, & Cox, 2003; BLINDED et al., 2015), but production and reception are strongly related in children with ASD (BLINDED, in press).

The second path of influence was predicted because, as children develop, there might be an increasing probability that instances of consonant use in communication acts are manifestations of children attempting to say words they understand. Children's prelinguistic vocal patterns in place and manner of articulation of consonants appear to be carried forward to first words (Stoel-Gammon & Cooper, 1984; Vihman, Macken, Miller, Simmons & Miller, 1985). If children attempt to say the words they understand prior to their ability to make themselves understood, it would manifest as the production of a variety of consonants in what appear to be prelinguistic communication acts. That is, it is proposed that one link for the above-indicated continuity is through receptive vocabulary. A larger receptive vocabulary means more words with varying consonants that the child has available to say.

Research Questions

Two research questions were examined:

  1. Controlling for level of cognitive impairment and autism symptomatology, which of the four potential predictors add value to explaining the variability in growth of DKCC in initially preverbal children with ASD?

  2. Are the associations between value-added predictors and later DKCC mediated through receptive vocabulary?

Methods

Participants

The 87 children (71 male and 16 female) participating in the study were between 24 and 48 months chronological age and had a clinical diagnosis of autism or PDD/NOS. If children had an existing diagnosis of autism or PDD/NOS through licensed and experienced community providers, their diagnoses were confirmed using the revised diagnostic algorithm on ADOS module I (Gotham, Risi, Pickles, & Lord, 2007), which was administered by research staff who were research reliable on this instrument. Children who did not enter the study with a previous diagnosis were assessed and diagnosed by a licensed clinician on the research team who was independently research reliable on the ADOS and was experienced with evaluating young children with autism spectrum disorder. Research diagnoses were based on best clinical judgment that the data from the ADOS and a clinical interview met criteria for autism or PDD/NOS in the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition-Text Revision (American Psychiatric Association, 2000). With one exception, children met the autism spectrum cut-off using the ADOS algorithms revised for improved diagnostic validity (Gotham, et al., 2007). One child who was diagnosed by community clinicians as having PDD/NOS scored under the autism spectrum disorder cut off on the ADOS, but was also judged to have PDD/NOS by the licensed examiner on the research team. Ninety-five percent of the participants met criteria for autism, and the remaining met criteria for PDD/NOS.

Participants, at the time of enrollment, were reported to say no more than 20 different words according to parent report on the MacArthur-Bates Communicative Development Inventories: Words and Gestures checklist (MB-CDI; Fenson et al., 2003) and produced no more than five different word roots during a 15-min language sample. We excluded children with severe sensory or motor impairments, identified progressive neurological disorders, and identified genetic syndromes.

Based on parent report, ethnic distribution for participants was 5% Hispanic and 95% non-Hispanic. According to parent report, the racial distribution of the children was 75% White, 18% Black/African American, 6% Asian, and 1% Native American or Alaska Native. Primary caregivers’ self-reported levels of formal education were 5% some high school education, but did not graduate; 22% high school diploma or equivalent; 24% one to two years of college or technical school education; 32% three to four years of college or technical school education; and 17% some graduate or professional school. Additional descriptive information on participants is provided in Table 1.

Table 1.

Description of Participant Characteristics at Time 1

M SD Min Max
Chronological age in months 34.7 7.2 20.4 47.9
MSEL early learning composite 50.9 4.1 <50 122
Mental age in months 12.1 4.7 3.75 26.5
Developmental ratio .36 .15 .17 .75
MB-CDI words understood 75.8 85.4 0 385
MB-CDI words said 3.7 5.0 0 18
UCS number of different words .7 1.2 0 5
ADOS social affect and restricted and repetitive behavior total 22.6 3.8 6a 28
Open in a new tab

Note. MSEL = Mullen Scales of Early Learning; Early Learning Composite reflects standard scores; Mental age = mean age equivalent across Visual Reception, Fine Motor, Receptive Language, and Expressive Language subtests of the MSEL; Developmental ratio = mental age/chronological age; MB-CDI = MacArthur-Bates Communicative Development Inventories: Words and Gestures checklist; UCS = Unstructured communication sample with examiner. ADOS = Autism Diagnostic Observation Schedule.

a

Only 1 child scored 6, the next to lowest score was 15.

Design

This study used a longitudinal correlational design with five measurement periods, each of which was separated by approximately 4 months. The dependent variable, DKCC, was measured at every measurement period. Motor imitation, attention to child-directed speech, a component variable for intentional communication and both background variables were measured at Time 1, providing a 16-month interval between these variables and estimated level of DKCC at the study endpoint. Parent linguistic responses and one of the component variables for intentional communication were measured at Time 2 to reduce the burden on families at Time 1. The interval between Time 2 and Time 5 was 12 months. The potential mediator, receptive vocabulary, was measured at Time 3 because mediation analysis assumes the mediator is measured after the predictors (i.e., value-added predictors of DKCC growth), but before the dependent variable. In the tests of mediated relations, Time 5 DKCC growth was used as the dependent variable to meet the assumption that the outcome be measured after the mediator. Table 2 provides a summary of the constructs, procedures, measurement periods, and variables used to address the research questions.

Table 2.

Constructs, Procedures, Untransformed Component Variables, and Analyzed Variables

Construct Procedures/periods Untransformed component variables Analyzed variable
Receptive vocabulary MB-CDI @ T3 Number of words understood only + number of words understood and said Log 10-transformed sum
Intentional communication UCS @ T1 Number of intentional communication acts Square root-transformed
ESCS @ T2 Number of communication acts summed across pragmatic functions average z score
Attention during child-directed speech (ACDS) ACDS @ T1 % of the total time that CDS “vignettes” were presented that the child was looking to the presentation window Untransformed score
Motor imitation MIS @ T1 Total raw score Log 10-transformed average z score
NVOA @ T1 Total raw score
Parent linguistic responses PCFP @ T2 Number of 5-second intervals with child's attentional lead followed by adult utterance about child's referent Average z score
PCS @ T2 Number of 5-second intervals with child attention or communication lead followed by adult utterance about child's referent
Diversity of key consonants used in communication CSBS @ T1-T5 Subscale 11 weighted raw score Untransformed scale score
Level of cognitive impairment MSEL @ T1 Average age equivalency across Visual Reception, Fine Motor, Receptive Language, and Expressive Language subscales/chronological age Untransformed developmental ratio
Autism symptomatology ADOS module I @ T1 Diagnostic algorithm score Reflected log 10-transformed score
Open in a new tab

Note. CSBS = Communication and Symbolic Behavior Scales- Developmental Profile Behavior Sample, MB-CDI = MacArthur-Bates Communication Development Inventory, ESCS = Early Social Communication Scales, UCS = Unstructured communication sample with examiner, ACDS = Attention during child directed speech procedure, MIS = Motor Imitation Scale, NVOA = Nonverbal Volitional Oral Abilities subscale, PCFP = Parent-child free play, PCS = Parent-child snack, MSEL = Mullen Scale of Early Learning, ADOS = Autism Diagnostic Observation Schedule.

Procedures and Variables

A brief description of all procedures relevant to this study is provided here. A more detailed description of the procedures is available in BLINDED et al. (2015). Unless otherwise stated, all coded variables were derived by observing recorded sessions. We measured each putative predictor in two contexts and, when the component variables from the two measurement contexts were sufficiently intercorrelated, aggregated across them. Doing so increases the stability, and thus the potential validity of the estimate for a predictor, particularly when children are in the earliest stages of development (Sandbank & Yoder, 2014). Further support for, and detail regarding, the aggregated measures is presented in the Results section. No putative predictors were measured from the same procedure as DKCC to avoid associations due to shared measurement method variance (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003).

Interobserver reliability was estimated for all coded variables on a random sample of at least 20% of the sessions from all relevant measurement periods. Reliability observers coded independently from the primary coder. The primary coder did not know which sessions would be selected for reliability coding. The reliability estimate used was an absolute intraclass correlation coefficient (ICC) from a two-way random model.

Measure and metric for DKCC (the dependent variable)

DKCC was measured using the Communication and Symbolic Behavior Scales - Developmental Profile Behavior Sample (CSBS; Wetherby & Prizant, 2002) at all five measurement periods. This structured communication sample was designed for use with children who have a functional communication age between 6 months and 24 months. The authors of the scale indicate this developmental span often corresponds to a chronological age range of approximately 6 months to 6 years in children with ASD.

The metric for DKCC was the weighted raw score for Subscale 11, derived according to the CSBS manual. Subscale 11 inventories a child's production of 13 select consonants (i.e., m, n, b, p, d, t, g, k, y, w, l, s, sh) in communication acts (i.e., vocalizations directed to an adult). These 13 consonants were selected for coding in Subscale 11 because they are early-emerging and/or because they can be coded reliably even in young children (Wetherby & Prizant, 2002). However, some of these consonants are relatively later-occurring (e.g., l, s, sh). Including later-occurring consonants in the count reduces the probability of ceiling effects in the developmental period studied (i.e., the transition to linguistic communication). Cognates (i.e., pairs of consonants that are articulated in the same place along the vocal tract) that differ only in terms of voicing (i.e., d versus t, b versus p, and g versus k) are not credited separately because some young children do not consistently distinguish between voiced and voiceless cognates and because collapsing across cognate members increases the reliability of the measure (Wetherby & Prizant, 2002). Thus, the maximum raw score that could be achieved by a child on Subscale 11 is 10. The weighted raw score was the raw score multiplied by 2, making the possible maximum score 20. The interobserver reliability for DKCC was .95 at Time 1, .96 at Time 2, .95 at Time 3, .95 at Time 4, and .90 at Time 5.

Measures and metric for intentional communication (a potential predictor of DKCC)

Intentional communication was measured in an unstructured communication sample at Time 1 (UCS) and in the Early Social Communication Scales at Time 2 (ESCS; Mundy et al., 2003). The UCS is a 15-min unstructured sample in which the examiner follows the child's lead in playing with a standard set of developmentally appropriate toys. The examiner uses topic-following comments and questions, and avoids presenting directives when the child is already productively engaged with an object or activity. The number of intentional child communication acts was coded from the UCS using a timed-event behavior sampling method. Intentional communication acts in the UCS were defined as: (a) nonconventional gestures, non-word vocalizations, or imitative symbols (signs or words) that occurred with coordinated attention to an object and an adult; (b) conventional gestures with attention to the adult; or (c) spoken word and American Sign Language approximations. The ICC for intentional communication in the UCS at Time 1 was .88.

The ESCS was used in addition to the UCS to increase the number and structure of sampling opportunities for intentional communication. The ESCS is a structured procedure designed to motivate young children to communicate for the purpose of regulating the behavior of another person, socially interacting with another person, or directing the other person's attention to an object or event. The number of intentional communication acts (regardless of pragmatic function) was coded for the ESCS using event behavior sampling. For this procedure, intentional communication acts were defined in accordance with the ESCS manual, and included child gestures, vocalizations, and/or verbalizations that were directed to an adult and that served an identifiable communicative function. The ICC for intentional communication from the ESCS at Time 2 was .97. The metric for intentional communication that was used in analyses was an aggregate of the number of intentional communication acts produced across the UCS and ESCS samples.

Measure and metric for attention during child-directed speech (ACDS; a potential predictor of DKCC)

ACDS was measured using a procedure from Watson, Baranek, Roberts, David, and Perryman (2010) at Time 1. In this procedure, the child is seated at a table facing a puppet theater that contains a window in which all stimuli are presented. Three 1-min child-directed speech (CDS) vignettes were presented. These were a video of a woman reading a children's picture book, a brief live puppet show delivered by a research assistant, and a video of a woman playing with and describing a novel toy. All speakers were adult females who used vocal intensity, pitch, and duration consistent with characteristics of natural child-directed speech. The ACDS media files were coded using a timed-event behavior sampling method to quantify the duration of child looking at the CDS stimuli presented in the puppet theater window, or child not looking at the CDS stimuli presented in the puppet theater window. The metric for ACDS was the proportion of seconds in which CDS vignettes were present that the child looked at CDS stimuli. The ICC for this variable was .99.

Measures and metric for motor imitation (a potential predictor of DKCC)

Motor imitation was measured using the Motor Imitation Scale (MIS; Stone, Ousley, & Littleford, 1997) and the Nonverbal Volitional Oral Abilities Scale (NVOA; adapted from Amato & Slavin, 1998) at Time 1. The MIS consists of 16 items involving single-step motor imitation acts, eight involving body movements only and eight involving actions with objects. Each item is scored in situ as 0, 1, or 2 points on the basis of the quality and accuracy of the imitation. Points were summed across all 16 MIS items to derive the MIS total score. In the NVOA, the participant is prompted to imitate 11 oral motor movements, such as tongue lateralization, blowing, and puckering lips, as demonstrated by the examiner. Each item is scored as 0, 1, or 2 points on the basis of similarity to the model. Points were summed across all 11 items to derive the NVOA total score. The metric for motor imitation that was used in analyses was an aggregate of total raw scores across the MIS and NVOA.

Measures and metric for parent linguistic responses (a potential predictor of DKCC)

Parent linguistic responses to child leads were measured in a 15-min parent-child free play (PCFP) and a 10-min parent-child snack session (PCS). In the PCFP, the parent was provided with a standard set of developmentally appropriate toys and instructed, “Play as you would at home if you had no interruptions and had time to play with your child.” The child and parent were free to position themselves as they chose throughout the sample. In the PCS, the parent was provided with a 4 oz. cup, a pitcher of juice, and several single-bite cookies, crackers, or parent-provided snack and was told, “We want to see how your child communicates during snack times. Just interact with him as you would at home if you wanted to elicit his communication.” The parent and child were seated at a table throughout the PCS.

A 5-s partial interval behavior sampling method was used to code each codable interval in the PCFP for child attention leads (i.e., the child touching or looking at an object) and parent linguistic responses to child attention leads (i.e., parent talking about the object referenced by the child lead, the action referenced by the child lead, or both). The PCS was coded similarly, with two exceptions. In addition to child attention leads and adult responses to child attention leads, child communication leads (see UCS section for the definition of intentional communication) and adult linguistic responses to child communication leads were coded. The PCFP could not be reliably coded for child communication leads (and thus adult responses to child communication leads) because the free positioning of the parent-child dyad during the PCFP sample prevented the reliable use of child gaze to adult's face to judge presence or absence of attention to the adult. Thus, parent linguistic responses were to child attention (PCFP and PCS) or communication (PCS only) leads. The ICC for parent linguistic responses to child leads was .98 for the PCFP and .98 for the PCS procedures. The metric for parent linguistic responses that was used in analyses was an aggregate of the number of linguistic response raw scores across the PCFP and PCS.

Measure and metric for receptive vocabulary (a potential mediator of DKCC growth)

Receptive vocabulary was measured using the MB-CDI (Fenson et al., 2003). Parents were asked to check a list of early vocabulary items to indicate which words their child “understands only” and “understands and says.” The sum of the raw number of words understood only + the raw number of words both understood and said (i.e., total number of words understood) was used as the metric for receptive vocabulary in the mediation analyses.

Measure and metric for level of cognitive impairment (a controlled covariate of DKCC growth)

Level of cognitive impairment was measured using the Mullen Scales of Early Learning (MSEL; Mullen, 1995) at Time 1. We used developmental ratio (i.e., mental age divided by chronological age), rather than the standard score (i.e., the Early Learning Composite score), as the index of cognitive impairment because the majority of participants had the lowest possible standard score of 49. Thus, using the developmental ratio produced more variability in cognitive levels than did the standard scores. Mental age was the average age equivalency score from four MSEL subscales: Visual Reception, Fine Motor, Receptive Language, and Expressive Language.

Measure and metric for autism symptomatology (a controlled covariate of DKCC growth)

Autism symptomatology was measured using the ADOS Module 1 Social Affect and Restricted and Repetitive Behavior Total (Gotham, Risi, Pickles, & Lord, 2007) at Time 1. The algorithm score was reflected (i.e., the maximum score + 1 was subtracted from the original score so that adaptive scores were high) to allow for necessary transformations to this variable and to aid interpretation.

Data Analysis Decisions

A summary of data analysis decisions most relevant to the present report is provided here. More detailed rationale for data analysis decisions are provided in BLINDED et al. (2015). In preliminary analyses, we aggregated variables, transformed variables that were not normally distributed, and imputed missing data points. We confirmed that all component variables that we intended to aggregate were not only theoretically, but also empirically related, as evidenced by intercorrelation ≥ .40. Aggregates were then formed by averaging z-transformed component variable scores. All variables to be utilized in analyses that had univariate skewness > |.8| or kurtosis > |3.0| were transformed in accordance with Tabachnick & Fidell (2001). Missing data were multiply imputed (Enders, 2010).

In primary analyses, growth curve modeling was used to quantify growth of DKCC because parameters from growth curves provide more precise estimates of change than alternatives when five or more measurement periods are used (Maxwell, 1998). Time in Study was centered at Time 5 so the intercept would be interpretable as Time 5 DKCC outcome. In the mediation analyses, we used the Time 5-centered intercept of DKCC growth as the dependent variable. A mediated relation is tested for significance by examining whether the product of the two unstandardized coefficients for the associations comprising the indirect relation has a confidence interval that excludes zero (Hayes, 2013). Table 2 provides a summary of the constructs, procedures, measurement periods, and variables used to address the research questions.

Results

Preliminary Analysis

Details of the preliminary analysis results, including the multiple imputation procedure used, are in BLINDED et al. (2015). Briefly, all planned aggregate variables met the empirical criterion for aggregation. Several variables were transformed to address extreme skewness or kurtosis. The untransformed component variables and variables used in final analyses, after aggregation and transformation, are summarized in Table 2. An expectation maximization method and 40 imputations using all continuous observed variables were used to impute missing data. Depending on the variable, potential predictors had between 0% and 33% missing data.

Growth curve modeling showed that DKCC grew in a simple linear fashion, and that there was much variability in DKCC growth. A repeated measures ANOVA indicated a simple linear Time effect, F(1,62) = 34.9, p < .001, and a simple linear growth trajectory best fit the data. The unconditional growth model indicated that, on average, children incremented the number of key consonants they used in communication acts about every 6.9 months. The statistically significant fixed effects indicated that the average DKCC at Time 5 and the average rate of DKCC growth across the study period were different from zero, both p values < .001. Significant random effects suggested that there was significant among-participant variability to be explained in the DKCC outcome at Time 5 and in the rate of DKCC growth across the study period, both p values < .001. See Table 3 for descriptive statistics on DKCC at all time periods.

Table 3.

Means and 95% Confidence Intervals for Diversity of Key Consonants Used in Communication by Period

95% Confidence interval
Measurement period Mean Lower bound Upper bound
1 5.6 4.3 7.0
2 6.2 4.9 7.5
3 7.4 5.9 8.9
4 8.7 7.1 10.3
5 10.1 8.5 11.2
Open in a new tab

Note. Scores displayed in this table are weighted raw scores (i.e., raw scores for production of up to 13 consonants in up to ten cognate categories multiplied by 2), derived in accordance with the CSBS manual instructions for this subscale. Thus, the possible max score for this subscale is 20.

All four potential predictors were significant zero-order correlates of DKCC growth. Table 4 indicates the proportion of explainable variance (pseudo-R square) in DKCC growth accounted for by each predictor. Table 5 indicates the intercorrelations among the predictors and background variables. Intentional communication was significantly associated with motor imitation and ACDS. Number of parent linguistic responses was nonsignificantly associated with the other three predictors. Cognitive impairment was significantly associated with all four predictors and with autism symptomatology, which was associated with three of the four predictors of DKCC growth. The intercorrelations among the predictors and between the predictors and the background variables needed to be statistically controlled to identify which of these variables had added value in explaining DKCC growth.

Table 4.

Pseudo-R Squared Values for Significant Zero-order Associations of Potential Predictors with Intercept or Slope of Growth in Diversity of Key Consonants Used in Communication

Growth parameter for change in DKCC
Predictors T5-centered intercept Linear slope
Intentional communication .28 .13
Motor imitation .14 ns
Attention during child-directed speech .07 ns
Parent linguistic responses .07 .09
Open in a new tab

Note. Pseudo R Squared = (Growth parameter's random coefficient from the unconditional model's - growth parameter's random coefficient from the model with predictor)/Growth parameter's random coefficient from the unconditional model.

Table 5.

Intercorrelation of Background Variables and Significant Zero-order Correlates of at Least one of the Growth Parameters for Change in Diversity of Key Consonants Used in Communication

Variables ACDS Motor imitation Intentional communication Level of cognitive impairment Reflected autism symptomatology
Parent linguistic responses .12 .14 −.09 .32* .24*
ACDS .15 .27* .49** .27*
Motor imitation .40** .35** .09
Intentional communication .42** .30**
Level of cognitive impairment .57*
Open in a new tab

Note. ACDS = attention during child-directed speech.

*

p < .05.

**

p < .01.

Primary Analyses

In the growth curve model with all four predictors and the two background variables, only parent linguistic responses and intentional child communication were value-added predictors of DKCC growth. As shown in Table 6, the model with only these two value-added predictors accounted for medium to large amounts of explainable variance (i.e., pseudo-R squared values) in the intercept and slope, respectively. The total model accounted for a large amount of explainable variance in the growth of DKCC.

Table 6.

Pseudo-R Squared Change for Value-added Predictors of Growth of Diversity of Key Consonants Used in Communication by Linear Growth Parameter

Growth parameter for change in DKCC
Model T5-centered intercept Linear slope
Intentional communication .33*** .17**
Parent linguistic responses .12** .13**
Total model .37*** .24**
Open in a new tab
**

p < .01.

***

p < .001.

We used the structural equation from the final model of DKCC growth to compute the estimated DKCC at Time 1 and Time 5 for hypothetical participants who were −1 SD from the mean, at the mean, and +1 SD from the mean on the two value-added predictors, then plotted the three resulting growth trajectories in Figure 1. As shown in the figure, even the children with relatively low numbers of intentional communication and parent linguistic responses showed positive growth in DKCC. However, the average rate of DKCC growth was much faster for children who entered the study with relatively more frequent intentional communication and parent linguistic responses. The structural equation for the final model is provided in the notes section of Figure 1.

Figure 1.

Figure 1

Open in a new tab

Growth of diversity of key consonants used in communication (CSBS Subscale 11 weighted raw score) as a function of three values on the value-added predictors. The structural equation used to generate the illustrated trajectories was:

eDKCC = −1.55-10(Time)+2.11(PLR)+9.71(COMM)+.13(TIME*PRL)+.33(TIME*COMM).

In above formula, estimated DKCC is “eDKCC”, T5-centered time is “TIME,” parent linguistic responses is “PLR,” and intentional communication is “COMM.”

Both of the significant value-added predictors were related to the Time 5 estimated level of DKCC (i.e., intercept of Time-5-centered DKCC) through receptive vocabulary at Time 3. These mediational models are illustrated in Figure 2. The kappa square values (i.e., an effect size metric for indirect effects) for the indirect effects of parent linguistic responses and intentional communication predicting Time-5-centered DKCC intercept through Time 3 receptive language were .36 and .44, respectively. These are large effect sizes. Both indirect effects had confidence intervals that excluded zero, meaning that the associations between the value-added predictors and DKCC were significantly reduced after controlling for receptive vocabulary. These results confirmed the predicted mediated associations (Hayes, 2009).

Figure 2.

Figure 2

Open in a new tab

Results of simple mediation models for the value-added predictors of T5-centered intercept for the growth curve of DKCC (i.e., diversity of key consonants used in communication).

Discussion

This study was conducted to identify the value-added predictors of an under-studied predictor of useful speech in initially preverbal children with ASD: DKCC. Of the four potential predictors and two background variables, only children's intentional communication and parents’ linguistic responses to children's attention and communication leads added value to the prediction of growth in DKCC. Variation in midpoint receptive vocabulary, at least in part, mediated the associations between these predictors and endpoint variation in DKCC. Within the context afforded by a correlational design, the mediational model findings are consistent with an interpretation that receptive vocabulary is partly responsible for the associations between the value-added predictors and endpoint DKCC.

Three weaknesses are apparent in this study. First, like all other correlational studies, we cannot rule out alternative explanations for the detected associations. Additionally, we examined only one aspect of vocal communication: diversity of selected consonant use. Finally, we examined only four potential predictors of DKCC.

Seven strengths are apparent in this study. First, selecting preverbal or nonverbal children with ASD and observing them for 16 months allowed predicting growth of DKCC from the period before many of the children were talking through a period when many of the children acquired their early spoken vocabularies. Second, imputing missing data enabled use of all participants and minimized the bias that likely would have resulted from other methods of handling missing data (Enders, 2010). Third, using multiple potential predictors and two background variables in the same statistical model allowed us to rule out the possibility that covariation with the other variables in the statistical model explained the associations between value-added predictors and growth of DKCC. Fourth, using growth curve modeling over five measurement periods enabled a better estimate of change in DKCC than is produced by other methods of quantifying change (Maxwell, 1998). Fifth, when justified and available, two measures of several of the predictors were used to improve the stability, and thus the potential validity of estimates for these emerging skills (BLINDED, 2014). Sixth, all variables were derived from different procedures, preventing shared measurement method variance from explaining the associations. Finally, because theory suggested that receptive vocabulary might help explain the associations between value-added predictors and growth of DKCC, we were able to predict and confirm that these associations were, at least in part, mediated through receptive vocabulary. Had we not tested these simple mediational models, we would have missed the important role receptive vocabulary might play in understanding why children's intentional communication and parent linguistic responses predict growth in DKCC.

These findings lend empirical support to the transactional theory of speech sound development. We confirmed that one parent factor (linguistic input) and two child factors (intentional communication and receptive vocabulary) suggested by Stoel-Gammon (2011) contribute in a dynamic manner to growth in vocal communication development (i.e., DKCC) in children with ASD. As indicated in the introduction, most of the prior work motivated by the transactional theory of speech development has focused on typically developing infants and their caregivers. One such report also detailed a complex interplay between one form of child communication (specifically, vocal communication), parental responses, and vocabulary as it relates to increased vocal complexity in typically developing infants (Gros-Louis, West, & King, 2014). We are hopeful that future work across laboratories will increase our understanding of how parent and child factors impact vocal and verbal development in various populations in the early stages of language development.

Very little study of the predictors of vocal development in initially preverbal children with ASD has been undertaken to date. In one of the only such studies to our knowledge, two child factors that are seemingly consistent with Stoel-Gammon's (2011) theory of phonological development, motor imitation and attention to child-directed speech, were identified as predictors of DKCC in preverbal children with ASD (Patten et al., 2012). The findings from the Patten et al. (2012) study were the result of an earlier analysis of the current study's participants. It differed from the current analyses in the following ways: (a) it used an endpoint analysis of DKCC, (b) the DKCC metric was derived only for Time 1 and Time 3, and (c) only a subset of the current study's predictors were examined. The present study shows that motor imitation and ACDS are significantly correlated with intentional communication and, when entered into the same model, become nonsignificant predictors of DKCC growth. Thus, if the current study's findings are replicated, they suggest a need to place higher weights on intentional communication, parent linguistic responses, and receptive vocabulary than on motor imitation as potential goals for treatment of DKCC in preverbal children with ASD.

Because of the paucity of data on predicting DKCC growth in children with ASD, the findings of the current study require replication. The proposed causal chain indicated in the transactional model of speech sound development can be most rigorously tested in a treatment study that uses an internally-valid experimental research design. In such a study, parent linguistic responses and intentional communication would be treated, with receptive vocabulary as a short-term goal and DKCC growth as a longer-term goal for children with ASD. The results of simple mediation analyses would have to show an indirect effect of treatment on DKCC growth through receptive vocabulary. Confirmation of such a mediation relation in a well-controlled treatment study would increase our confidence that targeting child intentional communication and parent linguistic responses produces early effects on children's receptive vocabulary, which translate to gains in DKCC growth, possibly because children begin to try to produce the words that they have come to understand through transactions with their adult communication partner.

To our knowledge, this is the first study to identify intentional communication and parent linguistic responses as value-added predictors of DKCC growth in preverbal children with ASD. These value-added predictors were found to be indirectly related to DKCC through receptive vocabulary. It is hoped that this correlational research will motivate experimental treatment studies to test whether these associations are causal.

Acknowledgments

This research was funded by National Institute for Deafness and other Communication Disorders (NIDCD R01 DC006893) and supported by the National Institute for Child Health and Disorders through the Vanderbilt Kennedy Center (P30HD15052) and the Carolina Institute for Developmental Disabilities (P30HD03110). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. We are very grateful to our wonderful staff (Nicole Thompson, Paula McIntyre, Ariel Schwartz, Tricia Paulley, Kristen Fite, Maura Tourian, Ann Firestine, Lucy Stefani, Olivia Fairchild, Amanda Haskins, Danielle Kopkin, and Kathleen Berry) and the families who trust us with their precious children.

References

  1. Amato J, Slavin D. A preliminary investigation of oromotor function in young verbal and nonverbal children with autism. Infant-Toddler Intervention: The Transdisciplinary Journal. 1998;8(2):175–184. [Google Scholar]
  2. American Psychiatric Association . Diagnostic and Statistical Manual of Mental Disorders-IV-TR. APA; Washington, DC: 2000. [Google Scholar]
  3. Billstedt E, Carina Gillberg I, Gillberg C. Autism in adults: symptom patterns and early childhood predictors. Use of the DISCO in a community sample followed from childhood. Journal of Child Psychology and Psychiatry. 2007;48(11):1102–1110. doi: 10.1111/j.1469-7610.2007.01774.x. doi: 10.1111/j.1469-7610.2007.01774.x. [DOI] [PubMed] [Google Scholar]
  4. BLINDED The role of supported joint engagement and parent utterances in language and social communication development in children with Autism Spectrum Disorders. Journal of Autism and Developmental Disorders. 2014;44:2162–2174. doi: 10.1007/s10803-014-2092-z. NIHMSID: 613125. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Charman T, Baron-Cohen S, Swettenham J, Baird G, Drew A, Cox A. Predicting language outcome in infants with autism and pervasive developmental disorder. International Journal of Language & Communication Disorders. 2003;38(3):265–285. doi: 10.1080/136820310000104830. [DOI] [PubMed] [Google Scholar]
  6. Eisenberg L. The autistic child in adolescence. The American Journal of Psychiatry. 1956;112:607–612. doi: 10.1176/ajp.112.8.607. [DOI] [PubMed] [Google Scholar]
  7. Enders C. Applied missing data analysis. Guilford; New York: 2010. [Google Scholar]
  8. Fenson L, Dale P, Reznick J, Thal D, Bates E, Hartung J, Reilly J. MacArthur communicative development inventories: User's guide and technical manual. Paul H. Brookes; Baltimore, MD: 2003. [Google Scholar]
  9. Gotham K, Risi S, Pickles A, Lord C. The Autism Diagnostic Observation Schedule: Revised algorithms for improved diagnostic validity. Journal of Autism and Developmental Disorders. 2007;37(4):613–627. doi: 10.1007/s10803-006-0280-1. doi: 10.1007/s10803-006-0280-1. [DOI] [PubMed] [Google Scholar]
  10. Gros-Louis J, West MJ, Goldstein MH, King AP. Mothers provide differential feedback to infants' prelinguistic sounds. International Journal of Behavioral Development. 2006;30(6):509–516. doi: 10.1177/0165025406071914. [Google Scholar]
  11. Gros-Louis J, West MJ, King AP. Maternal responsiveness and the development of directed vocalizing in social interactions. Infancy. 2014;19(4):385–408. [Google Scholar]
  12. Hayes AF. Beyond Baron and Kenny: Statistical mediation analysis in the new millennium. Communication Monographs. 2009;76(4):408–420. doi: 10.1080/03637750903310360. [Google Scholar]
  13. Haebig E, McDuffie A, Weismer SE. The contribution of two categories of parent verbal responsiveness to later language for toddlers and preschoolers on the autism spectrum. American Journal of Speech-Language Pathology. 2013a;22(1):57–70. doi: 10.1044/1058-0360(2012/11-0004). [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Haebig E, McDuffie A, Weismer SE. Brief report: Parent verbal responsiveness and language development in toddlers on the autism spectrum. Journal of Autism and Developmental Disorders. 2013b;43(9):2218–2227. doi: 10.1007/s10803-013-1763-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Howlin P, Mawhood L, Rutter M. Autism and developmental receptive language disorder—A follow-up comparison in early adult life. II: Social, behavioural, and psychiatric outcomes. Journal of Child Psychology and Psychiatry. 2000;41(5):561–578. doi: 10.1111/1469-7610.00643. [DOI] [PubMed] [Google Scholar]
  16. Kobayashi R, Murata T, Yoshinaga K. A follow-up study of 201 children with autism in Kyushu and Yamaguchi areas, Japan. Journal of Autism and Developmental Disorders. 1992;22(3):395–411. doi: 10.1007/BF01048242. [DOI] [PubMed] [Google Scholar]
  17. Lord C, Risi S, Lambrecht L, Cook EH, Jr., Leventhal BL, DiLavore PC, Rutter M. The autism diagnostic observation schedule-generic: a standard measure of social and communication deficits associated with the spectrum of autism. Journal of Autism and Developmental Disorders. 2000;30(3):205–223. [PubMed] [Google Scholar]
  18. Maxwell SE. Longitudinal designs in randomized group comparisons: When will intermediate observations increase statistical power? Psychological Methods. 1998;3(3):275–290. [Google Scholar]
  19. BLINDED Types of parent verbal responsiveness that predict language in young children with autism spectrum disorder. Journal of Speech, Language, and Hearing Research. 2010;53(4):1026–1039. doi: 10.1044/1092-4388(2009/09-0023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Mullen E. Mullen Scales of Early Learning. American Guidance Service; Circle Pines, MN: 1995. [Google Scholar]
  21. Mundy P, Delgado C, Block J, Venezia M, Hogan A, Seibert J. Early social communication scales. University of Miami; Coral Gables, FL: 2003. [Google Scholar]
  22. BLINDED Motor behaviors and associations with later consonant inventory in nonverbal children with ASD.. Paper presented at the International Meeting for Autism Research; Toronto, CA.. 2012. [Google Scholar]
  23. Paul R, Fuerst Y, Ramsay G, Chawarska K, Klin A. Out of the mouths of babes: Vocal production in infant siblings of children with ASD. Journal of Child Psychology and Psychiatry. 2011;52(5):588–598. doi: 10.1111/j.1469-7610.2010.02332.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Podsakoff PM, MacKenzie SB, Lee JY, Podsakoff NP. Common method biases in behavioral research: a critical review of the literature and recommended remedies. Journal of Applied Psychology. 2003;88(5):879. doi: 10.1037/0021-9010.88.5.879. [DOI] [PubMed] [Google Scholar]
  25. BLINDED Measuring representative communication in 3-year-olds with intellectual disabilities. Topics in Early Childhood Special Education. 2014 doi: 10.1177/0271121414528052. doi: 10.1177/0271121414528052. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Schoen E, Paul R, Chawarska K. Phonology and vocal behavior in toddlers with autism spectrum disorders. Autism Research. 2011;4(3):177–188. doi: 10.1002/aur.183. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Siller M, Sigman M. The behaviors of parents of children with autism predict the subsequent development of their children's communication. Journal of Autism and Developmental Disorders. 2002;32(2):77–89. doi: 10.1023/a:1014884404276. [DOI] [PubMed] [Google Scholar]
  28. Stoel-Gammon C. Relationships between lexical and phonological development in young children*. Journal of Child Language. 2011;38(1):1–34. doi: 10.1017/S0305000910000425. doi: doi:10.1017/S0305000910000425. [DOI] [PubMed] [Google Scholar]
  29. Stoel-Gammon C, Cooper J. Patterns of early lexical and phonological development. Journal of Child Language. 1984;11:247–71. doi: 10.1017/s0305000900005766. [DOI] [PubMed] [Google Scholar]
  30. Stone WL, Ousley OY, Littleford CD. Motor Imitation in Young Children with Autism: What's the Object? Journal of Abnormal Child Psychology. 1997;25(6):475–485. doi: 10.1023/a:1022685731726. doi: 10.1023/a:1022685731726. [DOI] [PubMed] [Google Scholar]
  31. Tabachnick B, Fidell L. Using multivariate statistics. 4th ed. Allyn and Bacon; Boston, MA: 2001. [Google Scholar]
  32. Venter A, Lord C, Schopler E. A follow-up study of high-functioning autistic children. Journal of Child Psychology and Psychiatry. 1992;33(3):489–507. doi: 10.1111/j.1469-7610.1992.tb00887.x. [DOI] [PubMed] [Google Scholar]
  33. Vihman M, Macken M, Miller R, Simmons H, Miller J. From babbling to speech: a re-assessment of the continuity issue. Language. 1985;61:397–445. [Google Scholar]
  34. Watson LR, Baranek GT, Roberts JE, David FJ, Perryman TY. Behavioral and physiological responses to child-directed speech as predictors of communication outcomes in children with autism spectrum disorders. Journal of Speech, Language, and Hearing Research. 2010;53(4):1052–1064. doi: 10.1044/1092-4388(2009/09-0096). doi: 10.1044/1092-4388(2009/09-0096) [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. West MJ, Rheingold H. Infant stimulation of maternal instruction. Infant Behavior and Development. 1978;1:205–215. [Google Scholar]
  36. Wetherby A, Prizant BM. Communication and symbolic behavior scales developmental profile- first normed edition. Paul H. Brookes; Baltimore: 2002. [Google Scholar]
  37. Wetherby A, Watt N, Morgan L, Shumway S. Social communication profiles of children with autism spectrum disorders late in the second year of life. Journal of Autism and Developmental Disorders. 2007;37(5):960–975. doi: 10.1007/s10803-006-0237-4. doi: 10.1007/s10803-006-0237-4. [DOI] [PubMed] [Google Scholar]
  38. BLINDED Atypical cross-modal profiles and longitudinal associations between vocabulary scores in initially low verbal children with ASD. Autism Research. doi: 10.1002/aur.1516. in press. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. BLINDED Mothers' attributions of communication to prelinguistic behavior of infants with developmental delays and mental retardation. American Journal of Mental Retardation. 1988;93(1):36–43. [PubMed] [Google Scholar]
  40. BLINDED Value-added predictors of expressive and receptive language growth in initially nonverbal preschoolers with autism spectrum disorders. Journal of Autism and Developmental Disorders. 2015;45(5):1254–1270. doi: 10.1007/s10803-014-2286-4. [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES