Lexical and Morphosyntactic Profiles of Autistic Youth With Minimal or Low Spoken Language Skills

Abstract

Purpose:

Autistic youth who are minimally or low verbal are underrepresented in research leaving little to no evidence base for supporting them and their families. To date, few studies have examined the types of words and word combinations these individuals use. The purpose of this study was to take a strengths-based approach to outline descriptive profiles of autistic youth who use few words and elucidate the lexical and morphosyntactic features of their spoken language.

Method:

We analyzed language samples from 49 autistic youth ages 6–21 years who used fewer than 200 words. Systematic Analysis of Language Transcripts was used to investigate the relationship between number of different words (NDW) and proportion of nouns and verbs (vs. other word classes), mean length of utterance in morphemes (MLUm), and the frequency of early developing morphosyntactic structures. We used linear regression to quantify the relationship between NDW and lexical and morphosyntactic features.

Results:

Proportion of nouns and verbs produced did not increase significantly in those with higher NDW. Conversely, MLUm and the frequency of early developing morphosyntactic structures increased significantly in those with higher NDW.

Conclusions:

Youth with higher NDW did not produce more nouns and verbs, suggesting lexical profiles that are not aligned with spoken vocabulary level. Youth with higher NDW had higher MLUm and more early morphosyntactic forms, suggesting that morphosyntactic profiles align with spoken vocabulary level. We discuss the implications for improving clinical services related to spoken language.

The diagnosis of autism spectrum disorder (ASD) is characterized by difficulties with social communication and the presence of restricted and repetitive behaviors and/or interests (American Psychiatric Association, 2013). Although challenges with structural language are not considered diagnostic for ASD according to the Diagnostic and Statistical Manual of Mental Disorders (5th ed.; American Psychiatric Association, 2013), there is a wide range of language trajectories across the autism spectrum (Anderson et al., 2007). Whereas some individuals present without structural language deficits, 30% of individuals use few spoken words or phrases (often referred to as “minimally verbal” [MV]) into the school years and into adulthood (Kjelgaard & Tager-Flusberg, 2001; Tager-Flusberg & Kasari, 2013). Language in autistic individuals is one of the strongest predictors of positive long-term social and adaptive outcomes (e.g., Howlin et al., 2004, 2014). Since useful speech by the age of 5 years has driven research with autistic children, Pickett et al. (2009) reported that some autistic children start talking after the age of 7 years and as late as the age of 13 years, which highlights the importance of understanding spoken language development in older MV autistic youth.

MV is typically defined as a small number of spoken words and fixed phrases (Tager-Flusberg & Kasari, 2013). Definitions of MV vary in the specific number of words (e.g., 1–10 words: Schreibman & Stahmer, 2014; 25 or fewer words: Yoder & Layton, 1988; fewer than 20 intelligible words: Chenausky et al., 2016; see also Koegel et al., 2020, for a systematic review), but most reference number of spoken words as a central criterion. In previous literature, the number of spoken words criterion is sometimes based on number of different words (NDW) in a 20-min language sample, but other times, it is determined by a 5-min clinical judgment, as in the context of the Autism Diagnostic Observation Schedule (ADOS) assessment. Another common definition is anchored to the Autism Diagnostic Observation Schedule–Second Edition (ADOS-2; Lord et al., 2012) with “minimally verbal” defined as those who were administered Module 1 of the ADOS-2 and “low verbal” to refer to individuals who were administered Module 2 of the ADOS-2 (see Bal et al., 2016; Barokova et al., 2021; Chenausky et al., 2019; Schwartz et al., 2020a, 2020b, for uses of the ADOS-2 aligned definition). Module 1 is assigned to individuals who have no speech up to and including the use of a few simple phrases (Lord et al., 2012). Module 2 is assigned to individuals who are using phrase speech but are not verbally fluent (Lord et al., 2012). Low verbal individuals who have a wider vocabulary and more combinatorial language are likely on a continuum of language ability with MV individuals. Here, we take a strengths-based approach and recognize that language development represents a continuum of features in which each skill builds upon the next in this population of autistic youth.

Autistic children have expressive language development profiles that differ from neurotypical development and when comparing across the range of language ability (Tek et al., 2014). The linguistic features of autistic youth who may have somewhat larger vocabularies but are not verbally fluent (low verbal) may be a more relevant “zone of proximal development” (Vygotsky, 1978) for MV autistic youth than language-matched 12- to 24-month-old neurotypically developing infants at the same stage of expressive language development (see Tager-Flusberg, 1999, for similar recommendations on approaches to studying language development in autistic children). In terms of clinical practice, understanding the spoken language of low-verbal youth will help clinicians individualize their assessments and treatment programs for language for MV youth.

Benchmarks for Spoken Language Development in Children on the Autism Spectrum

Tager-Flusberg et al. (2009) outlined spoken language development benchmarks for autistic children, a framework that has since been frequently adopted in studies of spoken language in autistic children, with some including children with a small number of spoken words (e.g., Biller & Johnson, 2019; Haebig et al., 2021; Kover et al., 2014, 2016). Based on current best evidence, they defined criteria to assign autistic children to four stages: (a) preverbal, (b) first words, (c) word combinations, and (d) sentences. These benchmarks outline the minimum criteria for each stage across the domains of phonology, vocabulary, grammar, and pragmatics. Of interest to this study are those vocabulary and grammar benchmarks that are determined by natural language sampling. In the preverbal phase, which corresponds to 6–12 months in neurotypical development, autistic children primarily use prelinguistic vocalizations and gestures to indicate communicative intent. In the first words stage, which corresponds to 12–18 months in neurotypical development, autistic children use at least five different words (nonimitated, spontaneous word root types) and 20 word tokens. In the word combinations stage, which corresponds to 18–30 months in neurotypical development, autistic children use at least 30 different words (nonimitated, spontaneous word root types) and have a mean length of utterance (MLU) of 1.8. In the sentences stage, which corresponds to 30–48 months in neurotypical development, autistic children have a minimum of 92 different words (nonimitated, spontaneous word root types) in 65 utterances and an MLU of 3.

Noun–verb combinations are a cornerstone of early language development (e.g., “see train”; Braine & Bowerman, 1976). Young neurotypically developing children begin to combine words into phrases when they have about 50–60 different words, including both nouns and verbs (Brown, 1973; Dromi, 1987; Nelson, 1981a). Having both nouns and verbs is fundamental to the development of combinatorial language in which nouns and verbs can be combined to form proto-sentences and acquire grammatical structures. Using a communicative development index (based on parent report), Bates et al. (1994) reported that for typically developing children with NDW below 50, nouns comprised around 38% of their vocabularies, whereas verbs comprised around 7%. In typically developing children with 51–100 different words, nouns comprised around 50% of their vocabularies, whereas verbs increased to just 10%. In neurotypical children with 101–200 different words, nouns comprised around 57% of their vocabularies, whereas verbs increased to 15%. At this early stage of lexical development, the proportion of children's vocabularies composed of nouns and verbs grows, but more so for nouns. In terms of clinical practice, it is essential to understand the use of nouns and verbs in the spoken language production of minimally or low verbal autistic youth to best support their language development into the combinatorial stage.

As vocabulary development proceeds, early morphosyntax tends to be acquired in a series of stages (Brown, 1973). When MLU is around 2, corresponding to 27–30 months in neurotypical development, early morphosyntax emerges beginning with the regular plural –s (e.g., “dogs”), present progressive –ing (e.g., “is playing”), and the prepositions in and on (Stage II in Brown, 1973). In Stage III, corresponding to 31–34 months in neurotypical development, the irregular past tense (e.g., “got”), possessive –‘s (e.g., “uncle's car”), and uncontractible copula (e.g., “This is yummy”) emerge. In Stage IV, corresponding to 35–40 months in neurotypical development, articles (e.g., “the”), regular past tense (e.g., “walked”), and regular third-person present tense (e.g., “She plays the piano”) emerge. In Stage V, corresponding to 41–46+ months in neurotypical development, irregular third-person uncontractible auxiliary (e.g., “She was running”), contractible copula (e.g., He's here), and contractible auxiliary (e.g., “It's jumping”) structures emerge. Of interest to this study is the relationship between NDW and these lexical and morphosyntactic features since number of words has played a central role in defining this 30% of the autism spectrum.

Lexical Development in Autistic Children

In early lexical development, children generalize words beyond a specific referent to develop knowledge of syntactic classes: The words “dog” and “train” are both in the class of nouns because they both follow the article “the.” Rescorla and Safyer (2013) examined lexical composition in young autistic children compared to sex- and age-matched children with neurotypical development using the Language Development Survey. They found that autistic children had smaller vocabularies, but there were no significant differences in the grammatical category (e.g., noun, verb) of the words that autistic children and neurotypical children knew. Mixed results have been found, however, using the MacArthur–Bates Communicative Development Inventories. Charman et al. (2003) and Luyster et al. (2007) found no difference in noun and verb usage between autistic and neurotypical children, but Haebig et al. (2021) reported that young preverbal and MV autistic children aged 5 years and younger used more verbs compared to neurotypical children.

Although the research findings' differences between young autistic and neurotypical children are mixed, previous research has not examined lexical composition of nouns and verbs in older youth with little spoken language. Binger et al. (2020) highlighted the need to focus language programs on both verb and noun phrases to teach early structural language to children with complex communication needs. Here, we investigated the relationship between NDW and the use of nouns and verbs compared to other syntactic classes.

Morphosyntactic Development in Autistic Children

As with other dimensions of structural language, not all autistic children experience challenges with expressive grammar (e.g., Kjelgaard & Tager-Flusberg, 2001; Sukenik & Friedmann, 2018). Some studies have found less complex morphosyntax, as measured by MLU, in autistic children compared to sex- and age-matched neurotypical children (Bacon et al., 2019; Eigsti et al., 2007; Park et al., 2012), whereas others have found no differences (Tager-Flusberg et al., 1990; Walenski et al., 2014).

Differences in expressive grammar have also been reported at the morpheme level. Bartolucci et al. (1980) found that school-age autistic children were more likely to omit certain morphemes compared to age-matched neurotypical children, particularly articles, auxiliary and copula forms of BE, past tense, third-person singular, and progressives. Using natural language sampling, Wittke et al. (2017) categorized autistic children into three subgroups: those with “normal language” (LN; 63%), those with marked difficulty in grammatical production but relatively intact vocabulary (GI; 21%), and those with more “globally low” language abilities (LI; 16%). Whereas grammatical errors distinguished the GI group from the LN group, the LI group showed lower MLUs than the other two groups, as would be predicted. A longitudinal study by Tek et al. (2014) found that autistic toddlers who had higher verbal abilities initially had similar growth trajectories to neurotypical children, acquiring nouns, verbs, and many morphosyntactic forms at a similar rate. However, autistic children with lower verbal skills had relatively flat trajectories. Information about morphosyntactic development in young autistic and neurotypical children may inform the investigation of minimally and low verbal autistic youth who are at a similar developmental stage.

Current Study

This study aimed to explore lexical and morphosyntactic profiles of autistic youth ages 6–21 years who have 1–200 different words, spanning minimally to low language levels using language sampling to address the following research questions:

1. What stages of language development (Tager-Flusberg et al., 2009) characterize autistic youth who were administered ADOS Modules 1 and 2 based on vocabulary and grammar criteria in a language sample taken from the first 20 min of the ADOS assessment?

2. What is the relationship between NDW and other lexical and morphosyntactic features?

   1. Do those with higher NDW use more nouns and verbs compared to those with lower NDW?

   2. Do those with higher NDW have a higher MLUm compared to those with lower NDW?

   3. Do those with higher NDW use early developing morphosyntactic structures more frequently compared to those with lower NDW?

Method

Participants

Participants included 49 individuals (11 female) who were administered an ADOS-2 Module 1 (n = 33) or ADOS-2 Module 2 (n = 16), all of whom had participated in prior studies at the Center for Autism Research Excellence at Boston University. Participants were between the ages of 6 and 21 years (M = 12.49, SD = 4.14). English was the primary language spoken to participants in the home. The studies from which the data were drawn were approved by the institutional review board at Boston University before enrolling participants, and parents of participants gave informed written permission for their children to participate.

All participants had at least one intelligible spontaneous spoken word and no more than 200 (including all unique, free unbound morphemes; M = 51.92, SD = 61.73). We chose an upper cutoff of 200 because one participant assigned to Module 2 of the ADOS-2 was an outlier with NDW above 300. Below, we discuss in more depth how NDW was derived.

ASD Diagnosis

Participants had a community diagnosis of ASD confirmed by the ADOS-2 (Lord et al., 2012) for children up to 12 years of age (N = 23) or the Adapted ADOS (A-ADOS; Bal et al., 2020) for those over the age of 12 years (N = 26). The ADOS-2 and A-ADOS assessments were administered by members of the lab who had attained research reliability. The ADOS-2 has diagnostic algorithms for two domains: Social Affect (SA) and Restricted and Repetitive Behaviors (RBB). The total ADOS-2 score (sum of SA and RBB) is then standardized into the calibrated severity score (CSS) based on age and language level, which provides a continuous measure of overall ASD symptom severity less influenced by child age and language ability than raw totals (Gotham et al., 2007, 2009).

Nonverbal IQ

Participants were drawn from three studies in which two different nonverbal IQ measures were administered: the Leiter International Performance Scale–Third Edition (N = 39; Roid et al., 2013) and the Raven's Coloured Progressive Matrices (N = 10; Raven et al., 1998). Raw scores on these measures were converted into standard scores based on age-normed values. The SA, RRB, CSS, and nonverbal IQ scores are reported in Table 1, and Table 2 summarizes participant demographic characteristics.

Table 1.

Participant characteristics.

Characteristic	M	SD	Range
Chronological age	12.49	4.14	6–21
ADOS-2 Overall calibrated severity score	7.45	1.39	4–10
ADOS-2 Social Affect severity score	7.02	1.44	4–10
ADOS Restricted and Repetitive Behaviors severity score	8.35	1.56	5–10
Nonverbal IQ standard score	68.49	20.61	30–115

Note. ADOS-2 = Autism Diagnostic Observation Schedule–Second Edition; ADOS = Autism Diagnostic Observation Schedule.

Table 2.

Participant demographics.

Sex, race, ethnicity	n	%
Male	38	78%
Non-Hispanic	33	82.5%
Hispanic	3	7.5%
Prefer not to respond	4	10%
Data not available (not included in percentages above)	9	—
White	26	65%
African American or Black	0	0%
Asian	6	15%
Native Hawaiian or Other Pacific Islander	1	2.5%
More than one race	4	10%
Prefer not to respond	3	7.5%
Data not available (not included in percentages above)	9	—

Language Samples

Language samples were drawn from administrations of the ADOS-2—a measure administered across all participants that provides a useful context for natural language sampling (Condouris et al., 2003; Tager-Flusberg et al., 2009). The ADOS-2 involves a variety of open-ended activities like playing bubbles, eating a snack, assembling puzzles, using miniature toy objects, describing pictures, and having a birthday party for a doll. It is designed to elicit social interaction and communication by providing opportunities to initiate and respond in a semistructured context (Lord et al., 2012). The Adapted ADOS-2 includes tasks and materials that are slightly modified to be developmentally appropriate for assessing older MV individuals while being comparable to the ADOS-2 (Bal et al., 2020). For example, the Adapted ADOS-2 construction task presents a more complicated puzzle, and the “tell a story from a book” task uses a more age-appropriate book. Rather than make-believe play (ADOS-2), the Adapted ADOS-2 presents an interactive soccer game. Similarly, rather than a birthday party for a doll (ADOS-2), the Adapted ADOS-2 presents a picnic (for a more detailed comparison of the ADOS-2 and Adapted ADOS-2, see Bal et al., 2020; see Table 1). The ADOS-2 sessions were audio- and video-recorded. As the ADOS-2 precludes coding for nonspeech gestures and augmentative and alternative communication (AAC; Lord et al., 2012), no participants used AAC during the sessions.

Transcription and Coding

Language samples were transcribed in accordance with Systematic Analysis of Language Transcripts (SALT; Miller & Iglesias, 2012) procedures. Transcriber training consisted of 10 hr of self-paced online SALT modules covering transcription format, utterance segmentation, conventions, and practice samples. Self-paced SALT training was followed by supervised practice with the first or third authors who are speech-language pathologists with training and experience in SALT until greater than 80% accuracy was achieved on utterance segmentation and morpheme identification. All transcripts were second-passed by trained transcribers. Words were transcribed using standard orthography, and common phrases with co-occurring words that were spoken without pauses between them (e.g., “alldone,” “nothankyou,” “allgone,” “cleanup,” “gimme,” “kinda”) were transcribed as one word following Tager-Flusberg and Anderson (1991). Utterances were segmented into C-units. Mazes and pauses and bound morphemes were marked. A second trained transcriber reviewed all transcripts to proof the transcription. Transcription proofing involved reviewing the initial transcript while viewing the video of the ADOS-2 administration. Discrepancies were resolved by consensus in accordance with SALT conventions (Miller & Iglesias, 2012). In the rare case that a consensus could not be reached, the word or utterance in question was marked as unintelligible. Although ADOS-2 administration is typically 30–45 min in length, for all analyses, we cut off the transcripts at the 20-min mark to align with the Tager-Flusberg et al. (2009) recommended sample length for vocabulary and grammar benchmarks.

All analyses included only spontaneous utterances. Utterances were coded as spontaneous or stereotyped as part of a previous project (La Valle et al., 2020). In the previous project, a subset of 20 transcripts was coded for reliability. Cohen's kappa was computed to determine intercoder reliability. There was substantial agreement for spontaneous/stereotyped utterances, κ = .816 (95% CI [0.789, 0.843], p < .0005; La Valle et al., 2020). Following La Valle et al. (2020) and Tager-Flusberg and Anderson (1991), stereotyped utterances were defined as repetitions, scripted recitations, neologisms, and idiosyncratic language. Repetitions that were complete or partial repetitions of utterances used within the past five utterances of the participant or the examiner were coded as stereotyped. Intercoder agreement for spontaneous versus stereotyped utterance coding was 95.5%. Although stereotyped language is common in autistic children and particularly individuals who are MV (La Valle et al., 2020) and it may serve communicative functions (e.g., Prizant & Duchan, 1981; Sterponi & Shankey, 2014), our goal was to investigate lexical and morphosyntactic profiles of participants' spontaneous spoken language.

Language Development Stages

We adopted Tager-Flusberg et al.'s (2009) minimum criteria for stages of vocabulary and grammar based on a 20-min language sample. We tallied the number of individuals who met the minimum criteria for the first words stage (five different word root types and 20 word tokens) and the word combinations stage (30 different word root types and MLU = 1.8). Those who did not meet criteria for first words were categorized as preverbal.

NDW

NDW is a direct measure of vocabulary diversity. In SALT, it is calculated by the number of unique, free (i.e., unbound) morpheme types produced in the sample. For example, “play,” “play/ed,” and “play/ing” would be treated as one word root (play) occurring 3 times. The bound morpheme “/ed” does not factor into the calculation of NDW. Only words located in the main body of the utterance (excluding mazes) are counted in the calculation of NDW (Miller et al., 2015). The analysis set excluded mazes, unintelligible words, and stereotyped words and utterances to include only intelligible, spontaneously produced words. NDW was calculated based on the first 20 min of the language sample.

Nouns and Verbs

We used the SALT analysis of grammatical categories function to tally the number of nouns, verbs, and total words in spontaneous utterances. We then calculated the combined proportion of nouns and verbs over the total number of words.

MLU in Morphemes

MLU in morphemes (MLUm), a direct measure of morphosyntactic complexity, was obtained via SALT. MLUm is calculated by dividing the total number of words (including morphemes) by the total number of utterances in a language sample (Brown, 1973). For example, the word “play/ed” would count as a length of two morphemes: “play” and “/ed.” The analysis set for MLUm was all verbal, complete, intelligible, spontaneous utterances.

Early Developing Morphosyntax

We examined the use of Brown's (1973) 14 early developing morphosyntactic forms (in order of emergence in typical development): present progressive –ing, prepositions in/on, plural –s, irregular past tense, possessive –s, uncontractible copula, articles a/the, past tense –ed, third-person singular –s, third-person irregular, uncontractible auxiliary, contractible copula, and contractible auxiliary. We used SALT to tally the use of inflectional morphemes from the bound morpheme table and the use of the prepositions in and on from the word root table. We hand-calculated uncontractible copulas and uncontractible auxiliaries, as they are not available for automatic extraction from SALT. Five transcripts were reviewed by a second researcher for uncontractible copulas and uncontractible auxiliaries, which resulted in 97% agreement.

Statistical Analyses

We conducted two simple linear regressions to assess how well NDW predicted two dependent variables: (a) proportion of nouns and verbs and (b) a composite morphosyntax variable including MLUm and frequency per minute of early developing morphosyntactic forms. We did not include age or nonverbal IQ in these models because hierarchical leave-one-out regression (Vehtari et al., 2017) starting with age and nonverbal IQ showed that these variables did not contribute significantly to the model variance over the dependent variables of interest. We log-transformed NDW, as it was not normally distributed. The analyses with log-transformed NDW, however, did not differ from those with nontransformed NDW, so we do not report log-transformed NDW here. Along similar lines, we converted NDW to NDW rate—NDW divided by the total number of utterances (Greenhalgh & Strong, 2001; Mills et al., 2013)—to account for variation in the amount of speech production among our participants, but the results did not differ from those with NDW. Nonparametric Spearman correlations among age, nonverbal IQ, NDW, MLUm, frequency of early developing morphemes, and proportion of nouns and verbs are shown in Table 3. Due to the high correlation between MLUm and frequency of use of early developing morphemes (and to reduce the likelihood of family-wise error), in the statistical analysis, we created a composite morphosyntax variable by converting MLUm and morpheme frequency to z scores and then summing them. NDW was highly correlated with MLUm and the frequency of early developing morphemes but not the proportion of nouns and verbs (out of the total number of words used).

Table 3.

Spearman correlations among variables.

Variable	Age	NVIQ	NDW	MLUm	FreqMorph	PropNV
Age	1	−.603***	.052	−.096	−.059	.078
NVIQ	−.603***	1	.317	.436**	.350*	.048
NDW	.052	.317*	1	.905***	.876***	.276
MLUm	−.096	.436**	.905***	1	.888***	.299
FreqMorph	−.059	.350*	.876***	.888***	1	.157
PropNV	.078	.048	.276	.299*	.157	1

Note. Entries above the diagonal are adjusted for multiple tests using the false discovery rate. NVIQ = nonverbal intelligence quotient; NDW = number of different words; MLUm = mean length of utterance in morphemes; FreqMorph = frequency of morpheme use; PropNV = proportion of nouns and verbs.

^* p < .05.

^** p < .01.

^*** p < .001.

Results

Language Development Stages

The first aim was to understand what stages of language development (Tager-Flusberg et al., 2009) characterize autistic youth who are minimally or low verbal. As Table 4 shows, nearly half of the individuals in our sample were categorized as preverbal with fewer than five different words and 20 word tokens. These individuals did not meet the vocabulary criterion for the first words stage, and there are no grammar criteria for the first words stage. We found that 39% were in the first words stage with NDW greater than 5 (and more than 20 word tokens) but fewer than 30 (no grammar criteria apply at the first words stage). We found that 16% were in the word combinations stage with NDW of 30 or higher and MLUm greater than 1.8. As expected, none of the individuals in our sample were in the sentences stage.

Table 4.

Participants categorized by stages of spoken language development.

Developmental stage	Vocabulary criterion	Met vocabulary criterion n (%)	Grammar criterion	Met grammar criterion n (%)	n (%)
Preverbal	None	N/A	None	N/A	22 (45%)
First words	NDW = 5 and 20 tokens	19 (39%)	None	N/A	19 (39%)
Word combinations	NDW = 30	11 (22%)	MLU = 1.8	18 (37%)	8 (16%)
Sentences	NDW = 92 in 65 utterances	0 (0%)	MLU = 3.0	4 (8%)	0 (0%)
Total					49 (100%)

Note. NDW = number of different words; MLU = mean length of utterance; N/A = not applicable.

Close to half of the participants did not meet the vocabulary criterion for first words, so we further broke down the number of individuals in the preverbal stage according to which first word criterion they did not meet—NDW or word tokens. Table 5 shows that 64% of the sample did not meet either of the criteria. A smaller percentage of the sample (18%) met one criterion or the other.

Table 5.

Number of different words (NDW) and word tokens in preverbal participants.

NDW	Word tokens	Number (%) of preverbal individuals
< 5	< 20	14 (64)
< 5	≥ 20	4 (18)
≥ 5	< 20	4 (18)
Total preverbal individuals		22 (100)

Nouns and Verbs

The second research question was: What is the relationship between NDW and other lexical and morphosyntactic features in minimally or low verbal autistic youth—do those with higher NDW use more nouns and verbs? Figure 1 shows the proportion of nouns and verbs (out of total words produced) by NDW. The vocabulary composition of participants with higher NDW (compared to lower) did not consist of significantly more nouns and verbs. In the linear regression, NDW was not a significant predictor of proportion of nouns and verbs (β = .00, SE = 0.00, t = 0.93 p = .36). Those in our sample who used fewer than 50 words produced an average of 22.6% nouns and 12.3% verbs. Those with 51–100 words produced an average of 22.4% nouns and 12.5% verbs. Those with 101–200 words produced an average of 22.9% nouns and 13.1% verbs. If we contrast these findings with those of Bates et al. (1994), though they were based on parent report, young neurotypical children in their sample who had fewer than 50 words produced an average of 38% nouns and 7% verbs. Those with 51–100 words produced an average of 50% nouns and 10% verbs, and those with 101–200 words produced an average of 57% nouns and 15% verbs. Although different methods were used, there is a difference in the relationship between NDW and the rate of increase in the proportion of the vocabulary that is composed of nouns and verbs.

Figure 1.

Proportion of nouns and verbs out of total words produced by number of different words (NDW; gray shading indicates bootstrapped 95% confidence interval).

Since the proportion of nouns and verbs did not increase significantly in those with higher NDW, we further investigated what kinds of words other than nouns and verbs were being produced, focusing on those with NDW below 10. In individuals with NDW 1–5 (N = 17), the other words were “yes” (or “yeah” or “yup”), “no,” “please,” “ok,” “great,” “alldone,” “more,” “big,” “green,” and “two,” with “yes” and “no” being the most common words across individuals. For those with NDW 6–10 (N = 6), the other words were “yes” (or “yeah”), “no,” “what,” “this,” “ok,” “zoom,” “more,” “my,” “please,” “happy,” “sad,” and “off.” Although a full analysis of the words other than nouns and verbs used by minimally or low verbal (MLV) individuals is out of the scope of the current study, it warrants further investigation.

NDW and MLUm

The relationship between NDW and MLUm is shown in Figure 2. NDW significantly predicted the composite morphosyntax variable, which included MLUm and frequency of early developing morphosyntactic forms (see below; β = .04, SE = 0.00, t = 14.73, p < .001). Those participants with higher NDW had significantly more morphosyntactically complex utterances.

Figure 2.

Number of different words (NDW) and mean length of utterance in morphemes (MLUm; gray indicates bootstrapped 95% confidence interval).

Early Morphosyntactic Forms

Figure 3 shows that the frequency of use of early developing morphosyntactic forms increased in those with higher NDW. As mentioned above, NDW significantly predicted the composite morphosyntax variable.

Figure 3.

Frequency per minute of use of early developing morphosyntactic items and number of different words (NDW; gray shading indicates bootstrapped 95% confidence interval).

As a descriptive follow-up on the production of early developing morphosyntax, we plotted the frequency of use per minute for each morphosyntactic form individually for each participant, as shown in Figure 4. Vertical lines delineate the stage of language development (Tager-Flusberg et al., 2009) based on the NDW criterion for each stage. The morphosyntactic structures on the y-axis are listed in order of acquisition in neurotypical language development. Very few morphosyntactic forms were produced in the first words stage. The use of a multiple of early morphosyntactic forms increased for those in the word combinations stage. The relationship between NDW, MLUm, and the use of early developing morphosyntactic forms is summarized in Table 6, which will be discussed further in the Clinical Implications section below.

Figure 4.

Frequency of use of morphosyntactic items by participant number of different words (NDW; arranged by order of acquisition from bottom to top).

Table 6.

Number of different words (NDW) by 20-word increments and associated mean length of utterance in morphemes (MLUm) and use of early developing morphosyntactic forms.

NDW	MLUm	Morphosyntactic forms (produced at least 2 times)
1–20	Approximately 1.3	None
21–40	Approximately 1.5	None
41–60	Approximately 1.75	None
61–80	Approximately 2.1	Present progressive –ing, regular plural –s, uncontractible copula, articles, contractible copula, contractible auxiliary

Discussion

The purpose of this study was to explore the lexical and morphosyntactic profiles of minimally and low verbal autistic youth on the autism spectrum and discuss the implications for supporting these individuals and their families in the domain of language. Of particular interest was the relationship between NDW and the use of nouns and verbs, MLUm, and the production of early developing morphosyntactic forms. Results suggest potential differences in lexical compared to morphosyntactic domains. Although youth with higher NDW also had higher MLUm and greater frequency of use of early developing morphosyntactic forms, those with higher NDW did not produce significantly more nouns and verbs as a proportion of their total vocabulary. As these autistic youth have been neglected in autism research (Tager-Flusberg & Kasari, 2013) and the evidence base for supporting them in the area of language is scant (Brignell et al., 2018), we discuss the clinical implications in more detail below.

Stages of Language Development

Nearly half of participants (45%) were preverbal, having NDW below 5 and/or producing fewer than 20 word tokens. Of the remaining half, 39% were in the first words stage, and 16% were in the word combinations stage. This result suggests that these autistic youth have heterogeneous profiles of lexical and morphosyntactic skills. Taken together, these results suggest that fine-grained profiles of spoken language are important to understand the full range of skills in autistic youth with limited language to tailor individualized language programs (see also Trembath et al., 2020). The approach in this article was to go beyond benchmarks and stages to focus on lexical and morphosyntactic development as they related to increases in NDW. This approach may yield more fine-grained understanding of lexical and morphosyntactic features as they relate to vocabulary development.

Lexical Development

Lexical development is a known area of difficulty for many autistic children (Arunachalam & Luyster, 2016) as most of them acquire their first words later than their neurotypical peers (Charman et al., 2003; Luyster et al., 2007). Our results showed that there is no significant increase in the combined use of nouns and verbs as a proportion of the total words produced by autistic youth with limited spoken language. This contrasts with existing knowledge of neurotypical language development trajectories showing increases in the proportion of vocabulary composed of nouns and verbs, although these observations are descriptive rather than based on statistical inference. While Bates et al. (1994) data were based on the MacArthur–Bates Communicative Development Inventories and ours came from language samples, the contrast in the lexical composition of nouns and verbs for neurotypical children and autistic youth at the same NDW level should be explored further as recent evidence suggests that verb diversity in the preschool years is predictive of later language outcomes for autistic children (LeGrand et al., 2021).

Morphosyntactic Development

The morphosyntactic profiles of minimally or low verbal autistic youth have not previously been investigated in depth. Wittke et al. (2017) reported that “low verbal” youth had lower MLUs than higher verbal youth with autism and language impairment, as would be predicted. Ours is the first study, to the best of our knowledge, to examine the in-depth morphosyntactic profiles along a continuum of NDW. Those with higher NDW also had higher MLUm and more frequent use of early developing morphosyntactic forms, which suggests the development of NDW and MLUm is aligned.

While this study is cross-sectional, rather than longitudinal, NDW serves as the cornerstone of language development for autistic children. In this study, the use of early developing morphosyntactic forms emerged in those whose NDW was around 60 and MLUm was around 2, which aligns with the early observations of Brown (1973). This result suggests that the relationship between MLU and the use of early developing morphosyntactic forms in autistic youth may be significantly delayed but not atypical relative to the level of vocabulary development based on NDW.

The use of individual morphosyntactic forms in minimally and low verbal autistic youth is somewhat different than the neurotypical order of acquisition described by Brown (1973). Given that the acquisition of morphosyntactic forms is subject to individual variation in neurotypical development (Bates et al., 1995; Bretherton et al., 1983; Nelson, 1981b) and autism spectrum (Kjelgaard & Tager-Flusberg, 2001), we do not aim to suggest that the acquisition of individual morphological forms in different orders suggests atypical morphosyntactic profiles or trajectories. Rather, our aim is to bring to light evidence on the use of early developing morphological forms and the associated level of vocabulary development to improve clinical practice.

The results of this study showed that morphosyntactic forms were first used in individuals who had NDW around 60. Those forms were the present progressive –ing, plural –s, uncontractible copula, articles, contractible copula, and contractible auxiliary (vs. present progressive –ing, plural –s, and the prepositions in and on as described by Brown, 1973). To summarize, in minimally and low verbal autistic youth, the use of early developing morphosyntactic forms in those with NDW of 60 and MLUm of 2 is aligned with our knowledge of neurotypical language development. Of the morphosyntactic forms that are first used by individuals with NDW of 60 and MLUm of 2, the present progressive –ing and the regular plural –s align with the observations of Brown (1973). The other forms, however, were observed by Brown to emerge at much later stages in neurotypical development, namely, uncontractible copula, articles, contractible copula, and contractible auxiliary.

More generally, we found that lexical profiles did not align with existing knowledge of neurotypical early lexical development (based on Bates et al., 1994)—the proportion of nouns and verbs in the vocabularies of MLV youth did not increase as NDW increased. On the other hand, morphosyntactic profiles were aligned with neurotypical early language development—MLUm and the use of specific early developing morphosyntactic structures increased as NDW increased. While the use of specific morphosyntactic structures did not align with the predictions of Brown (1973), the development of specific morphosyntactic structures is known to involve significant individual variation even in neurotypical development (e.g., Nelson, 1981b). Taken together, these results suggest that despite having a small number of spoken words overall, challenges in vocabulary development compared to morphosyntactic development may play a bigger role in language profiles of these autistic youth. Given this finding, it would be worthwhile for future research to investigate whether a syntactic bootstrapping approach to teaching vocabulary would produce better outcomes compared to a single-word approach to vocabulary learning, which is common for individuals with a small number of spoken words (see, e.g., Brignell et al., 2018; Koegel et al., 2020). Syntactic bootstrapping, the use of syntax–semantics links to acquire the meanings of unfamiliar verbs (Gleitman, 1990; Naigles, 1990), however, requires knowledge of some nouns and comprehension of basic linguistic structures, such as [noun] is [verb]-ing (Gleitman et al., 2005; see also Arunachalam & Luyster, 2016, for an overview). Although this study has shown that some individuals produce linguistic structures like this, it is an open question whether they also comprehend them, which is an essential component of learning through syntactic bootstrapping.

Clinical Implications

Nearly half (45%) of participants in this study were in the preverbal stage, having NDW below 5 and/or producing fewer than 20 word tokens. In clinical practice with preverbal individuals, assessing aspects of speech and communication other than lexical and morphosyntactic forms may take precedence. Targets such as speech motor ability (Chenausky et al., 2019), general imitation ability (Pecukonis et al., 2019), gesture (La Valle et al., 2021), nonspeech vocalizations, and eye gaze (McDaniel & Schuele, 2021) have established links to expressive communication in MV individuals (see Kasari et al., 2013, for an overview).

In Table 6, we outlined the morphosyntactic features in increments of 20 different words. As Table 6 shows, MLUm increases steadily as NDW increases. The use of individual morphosyntactic forms, however, emerged when individuals had NDW around 60. In terms of language targets, these findings suggest that a word-combining approach to increasing vocabulary may be appropriate (as those with NDW between 1 and 20 have MLUm around 1.3) and effective, but more evidence is needed to explore this possibility. More specifically, language targets that combine nouns and verbs and incorporate early developing morphosyntactic forms that were used first (in terms of NDW) by minimally and low verbal youth may be appropriate, but more research is needed to evaluate how this finding might translate to clinical outcomes. Since the present progressive –ing and the (un)contractible copula were among the first used (in terms of NDW), intervention targets might include “Noun is Verb-ing or Noun's Verb-ing.” Similarly, regular plural –s and articles were also among the first forms to be used. Intervention targets that combine the noun with an article or a plural may be more appropriate than a single-word approach and would also avoid modeling telegraphic language (e.g., “mommy eat”), as telegraphic language use is related to lower lexical diversity in children on the autism spectrum (Venker et al., 2015).

Limitations and Future Work

This study has a few limitations to note. First, the participants are not an ethnically and racially diverse sample. Second, this study is cross-sectional rather than longitudinal. A longitudinal study of lexical and morphosyntactic development would provide stronger evidence for these findings. Third, while we discussed the clinical implications of these findings, clinical research studies are needed to follow up and evaluate more rigorously the points discussed here. Fourth, the use of the ADOS-2 assessment as a language sampling context may underestimate spoken language abilities as the assessment is examiner driven and designed to probe various skills relevant to the diagnosis of autism (cf. Kover et al., 2014; Sturman et al., 2022). Similarly, participants could have engaged in different activities during the 20-min sample, if, for example, one participant completed the ADOS-2 in 20 min while another took 45 min to complete all activities. Finally, this study did not include the use of AAC. Future studies should examine all modalities of communication within this population to develop a more comprehensive understanding of expressive communication skills in context with others. Future studies can provide full descriptive profiles of expressive communication in autistic youth with limited language skills to develop individualized language targets across different communication modalities to support their language and communication development.

Conclusions

The purpose of this study was to outline descriptive profiles of autistic youth and elucidate the lexical and morphosyntactic features of their spoken language using language sampling. The morphosyntactic features that we examined, MLUm and the use of early developing morphosyntactic forms, both aligned with our knowledge of neurotypical language development. MLUm was significantly higher in those with higher NDW. Similarly, the use of early developing morphosyntactic forms occurred in those whose MLUm was around 2. On the other hand, lexical composition of nouns and verbs did not align with neurotypical language trajectories. The proportion of their vocabularies that consisted of nouns and verbs was not significantly higher in those with higher NDW. These analyses provided an in-depth view of the kinds of words and word combinations minimally and low verbal autistic youth are using. In terms of clinical practice, these findings imply that targeting noun–verb combinations may improve vocabulary development given that morphosyntactic abilities appear aligned with neurotypical development at similar NDW and MLUm.

Author Contributions

Lindsay K. Butler: Conceptualization (Equal), Data curation (Lead), Formal analysis (Lead), Visualization (Lead), Writing – original draft (Lead), Writing – review & editing (Equal). Lue Shen: Conceptualization (Supporting), Data curation (Supporting), Validation (Lead), Visualization (Supporting), Writing – review & editing (Equal). Karen V. Chenausky: Data curation (Supporting), Investigation (Supporting), Writing – review & editing (Equal). Chelsea La Valle: Data curation (Supporting), Investigation (Lead), Writing – review & editing (Equal). Sophie Schwartz: Investigation (Supporting), Funding acquisition (Lead), Writing – review & editing (Equal). Helen Tager-Flusberg: Conceptualization (Equal), Funding acquisition (Lead), Investigation (Lead), Resources (Lead), Supervision (Lead), Writing – review & editing (Equal).

Data Availability Statement

The data reported in this study came from three separate projects. Some of the ADOS-2 and nonverbal intelligence quotient scores reported in this study are available through the National Database for Autism Research. The full set of data (including the remaining ADOS-2 and nonverbal intelligence scores along with the natural language sample measures) is available upon reasonable request from the first author in consultation with the final author.

Funding Statement

This research was supported by National Institute on Deafness and Other Communication Disorders (NIDCD) Grants P50DC13029 (principal investigator [PI]: Helen Tager-Flusberg), P50DC018006 (PI: Helen Tager-Flusberg), and K99R00DC017490 (PI: Karen V. Chenausky) as well as Autism Speaks Grant 10085 (PI: Sophie Schwartz). Lindsay K. Butler was supported as a National Institutes of Health trainee on NIDCD Grant T32DC013017 (PI: Christopher A. Moore).