Introduction
There is evidence of substantial heterogeneity in the course of early spoken language development in children on the autism spectrum, with a large minority entering school with little or no functional speech, often despite intervention (Brignell et al.,
2018; Norrelgen et al.,
2015). These children are variously described in the research literature as being prelinguistic, non-verbal, or minimally verbal (Koegel et al.,
2020), or as having complex communication needs (Light et al.,
2019). Irrespective of the term used, the impacts of children having spoken language difficulties can be far reaching, and include negative effects on learning and participation during childhood and into adult life (Cummins et al.,
2020). Accordingly, it is essential that factors that may influence children’s spoken language outcomes are identified, with the view to tailoring interventions to children’s individual strengths and needs. It is important, however, to highlight at the outset that spoken language is just one mode of communication used alongside other means (e.g., gesture) and, like all people, children on the autism spectrum should be supported and encouraged to use a range of modes of their choosing, including augmentative and alternative communication (AAC) systems. Nevertheless, spoken language is highly effective in helping people communicate in spontaneous and flexible ways for a variety of purposes, including expressing thoughts, emotions, wants, and needs in everyday interactions and settings, and thus was the focus of this study.
A range of factors have been identified as possible predictors of spoken language outcomes in later childhood for children on the autism spectrum including broad developmental characteristics (e.g., chronological age, autism characteristics, non-verbal cognition, and receptive language), more specific social-cognitive (e.g., joint attention, play, imitation) and linguistic capacities (e.g., phonetic inventory), and social-contextual factors (e.g., parental social-economic status) (Chenausky et al.,
2018; Mouga et al.,
2020; Thurm et al.,
2007,
2015; Wodka et al.,
2013). However, as noted by Pecukonis et al. (
2019) in a review of 21 studies examining social-communication factors, inconsistencies in participant characterisation, research design, predictor and outcome construct selection and measurement, and analytic approach make it difficult to draw conclusions that can directly inform clinical practice. One way to enhance the relevance of findings is to evaluate factors that clinicians themselves believe to be relevant to children’s spoken language outcomes and embedding these in research within community settings.
While inherently challenging for controlled design, embedding research in community settings offers the opportunity to build an empirical evidence base with direct relevance to clinical practice and rapid translational potential. Toward this end, (Trembath et al.,
2021) undertook a qualitative study to delineate clinician-proposed predictors of spoken language outcomes for children on the autism spectrum. The sample of 14 speech pathologists, each working in early intervention settings, together identified 183 factors, including a range of child autism-specific and broader developmental characteristics (e.g., cognitive ability), specific social-cognitive factors (e.g., prelinguistic skills), and the presence of co-occurring conditions. Drawing on these clinical insights and the published research evidence more broadly, we designed the current quantitative study to examine the potential predictive value of selected factors for spoken language outcomes of children on the autism spectrum who had minimal spoken language at the point of intake into community-based early intervention programs.
The current study was a clinical-research collaboration between clinical staff working in seven community-based early intervention centres and researchers across six universities, developed in response to the clinician-identified need to better understand and support communication development in children enrolling at the centres with minimal spoken language. All aspects of the study methodology, including selection of factors that would be examined, were designed in accordance with an evidence-based practice framework, which combines the best available research evidence with evidence derived from clinical practice, along with the preferences and priorities of fully informed clients (Sackett et al., 1996). Accordingly, nine factors were selected through discussion involving the clinical representatives of each early intervention centre and the researchers in the team, that involved consideration of the following criteria: (a) identified relevance by speech pathologists engaged in community practice (Trembath et al.,
2021), (b) clear theoretical relevance to spoken language development, (c) existing empirical evidence for a potential association with spoken language outcomes in children on the autism spectrum, and (d) capacity for feasible measurement during semi-structured play-based assessment undertaken in the context of community-based early intervention settings.
Three broad developmental factors—child
chronological age, autism characteristics, and
receptive language—were included given their identified clinical relevance at entry to early intervention programs and frequent inclusion in published empirical research on spoken language outcomes for children on the autism spectrum. While the latter has yielded inconsistent evidence of predictive value (possibly due to differences in study sample and design characteristics; Pecukonis et al.,
2019), strong theoretical rationales for each of these as potential determinants of spoken language outcome (e.g., Mundy & Neal,
2000; Towle et al.,
2020) further justifies their inclusion in ongoing research.
Two social-cognitive and social-communicative factors were selected.
Functional use of objects reflects an underlying organization of thought and behaviour enabling children to learn from interaction with objects in their environment (Lyytinen et al.,
1997,
1999). Empirically, this has been shown to predict change in children’s communication skills, including in longitudinal naturalistic observational studies (e.g., Poon et al.,
2012), and to moderate intervention outcomes (e.g., Yoder & Stone,
2006). Children’s
rate of communicative acts was selected based on theory and evidence that the propensity to direct communicative vocalizations, verbalizations, or gestures towards others is a strong predictor of both later verbal and nonverbal outcomes (e.g., Plumb & Wetherby,
2013; Shumway & Wetherby,
2009). Clinically, this propensity for children to communicate with others and the way they engage with objects (e.g., toys, equipment, learning materials) is directly relevant to their learning and participation in early childhood education settings.
Finally, four factors specifically concerning child language development were selected. Children’s
range of communicative functions was selected based on longstanding evidence that children on the autism spectrum show a relative reduction in the pragmatic use of language for socially-motivated versus more instrumental functions (Wetherby,
1986; Wetherby & Prutting,
1984), thereby reducing the range of opportunities for spoken language to be used.
Symbolic word learning—the ability to infer associations between spoken object labels, pictures of objects, and actual objects in the environment without explicit teaching—was selected given it is fundamental to communication development (Allen & Lewis,
2015), combined with emerging empirical evidence to suggest it may be impaired in some children on the autism spectrum (Allen & Lewis,
2015; Hartley & Allen,
2015; Rose et al.,
2020).
Phonetic inventory—the number of different speech sounds produced—is a core aspect of structural language development in children but impaired in at least some children on the autism spectrum (Saul & Norbury,
2020; Yoder et al.,
2015). The
ratio of speech to non-speech vocalizations was selected on the conceptual basis that it reflects children’s capacity to produce intelligible—and thus easily interpreted—communicative acts, and emerging evidence of a possible association with greater language development among children receiving intervention (e.g., Plumb & Wetherby,
2013; Trembath et al.,
2019).
Discussion
Our first aim in this study was to identify the proportion of children who entered the community-based early learning centres with minimal spoken language, who progressed to develop spoken language. The results are somewhat sobering, in that we observed only a small increase in the number of children (7 of 61 children) below Phase 3 at Time 1 who then went on to meet criteria for Phase 3 at Time 2. Only two children moved from below age-expected phase level to age-expected phase level over the course of the study, while three children were assessed to have regressed from Time 1 to Time 2. These findings are consistent with previous research (e.g., Norrelgen et al.,
2015; Rose et al.,
2020; Wodka et al.,
2013) in demonstrating variability in outcomes, including promising change for some children, but substantial ongoing communication needs for many more children on the autism spectrum. From a clinical perspective, the variability in children’s outcomes over the study period should serve to support parents and clinicians in calling for individualised approaches to promoting the communication development of children on the autism spectrum who commence early intervention with minimal verbal language. For the children who were assessed to have regressed, the reasons for this are unclear, but could conceivably reflect aspects of the child’s development including co-occurring conditions, or could be an artefact of testing (e.g., variability in engagement in tasks at the two timepoints). Further research is needed to explore the variability in children’s trajectories through measurement at multiple timepoints.
The magnitude of change identified in the current cohort of children should be contextualised for several reasons. First, the Spoken Language Benchmarks require children to meet the criteria for a given phase level across all four language domains (phonology, vocabulary, grammar, and pragmatics). However, children on the autism spectrum often demonstrate uneven profiles of language development (Ellawadi & Ellis Weismer,
2015); hence, a lack of change in phase level may mask more subtle increases (or decrease) in skills in a particular domain. Second, the time period between assessments was relatively short, particularly for children who entered the early intervention programs part-way into the year, meaning that the measurement may not have been sensitive to change. Third, the phases of the Spoken Language Benchmarks are sequential, but the nature and level of skill acquisition required to move from one phase to the next varies. To illustrate, grammar is only assessed from Phase 2 onwards. Furthermore, the Spoken Language Benchmarks are based on the phases of acquisition in neurotypical children, which may not be linear for children on the autism spectrum. Fourth, it is expected that a multitude of factors influenced children’s progression (or lack thereof), including the characteristics of any interventions received. In this regard, and while acknowledging the limitation, the close examination of within cohort changes has the potential to shed light on clinically relevant factors that may influence children’s spoken language outcomes.
Our second aim was to examine nine factors that clinicians proposed may account for differences in spoken language acquisition amongst preschool aged children who enter early intervention services with minimal verbal language. The model comprising seven factors (functional use of objects and ratio of speech to non-speech behaviours excluded due to multicollinearity) was significant and explained 64% of the variance in spoken language outcomes (i.e., acquisition of Phase 3 spoken language) with high sensitivity indicating the relevance of the predictors. The direction of effect for each variable was consistent with hypotheses posed. Younger chronological age, higher receptive language scores, and fewer autism characteristics were associated with increases in spoken language. Not surprisingly, but importantly, the findings point to the relevance of clinical insight when it comes to better understanding the heterogenous developmental profiles and outcomes for children on the autism spectrum (citation withheld for blind review). With these considerations in mind, the findings point to the importance of clinicians applying, and sharing with colleagues, their clinical insights when making recommendations for interventions for children on the autism spectrum within an EBP framework. Furthermore, the findings should serve to encourage researchers to collaborate with clinicians in developing and testing hypotheses.
Despite the overall significance of the model, none of the seven predictors was individually significant, which demonstrates the interconnected nature of the skills examined. However, children’s
rate of communicative acts at Time 1 showed a large effect size and warrants consideration. From a clinical perspective, the finding appears to reinforce the importance of creating learning environments that foster children’s motivation to spontaneously engage in communicative acts as the foundation for spoken language and broader communication development (Paul,
2008). In making this observation, we note that a child’s rate of communicative acts will rarely be independent of a range of other factors, such as the number, quality, and consistency of communication opportunities provided by the communication partner and their responses to the child’s communicative attempts. Nevertheless, and importantly, the variable was derived from an existing clinical assessment tool (CSBS-DP) in which children are offered a consistent set of communicative temptations (i.e., situations designed to be enticing to communicate to access activities or materials). The presentation of communication opportunities in this manner may help to differentiate the relative influence of the child’s own propensity to communicate from the influence of the communication partners’ behaviour. Furthermore, it appears that a clinically feasible, play-based task may not only help to characterise children’s communication as part of assessment of skills and goal setting, but also to predict which children are likely to experience a more favourable outcome. From a research perspective, there is currently concerted effort to identify variables that may help to match and personalise interventions for individual children on the autism spectrum (e.g., Lin et al.,
2021; Mouga et al.,
2020; Pecukonis et al.,
2019). Our findings of the potential relevance of the rate of communicative acts, which are consistent with those reported in earlier studies (e.g., Plumb & Wetherby,
2013; Shumway & Wetherby,
2009), suggest it could be a worthy candidate for inclusion in algorithms designed to support clinical decision-making.
Limitations and Future Directions
The findings of this study should be considered in the context of several limitations, each of which also has implications for future research. First, this was a prospective cohort study with children attending community-based early intervention centres, not a clinical trial. There was no control group and children were exposed to different interventions. Accordingly, the analysis related to children’s change within programs, without attention to intervention effects. There is a need for more consistent examination and reporting of spoken language outcomes, and factors that may moderate these, in the context of clinical trials where the influence of the interventions can also be assessed.
Second, in selecting the putative predictors and corresponding measurement tools, priority was given to clinically relevant tools and, where possible, those already commonly used in practice. Therefore, for example, the rate of communicative acts was calculated on the basis of the three standardised trials rather than a more comprehensive measure across the assessment session. In research of this kind, there is often a balance to be struck between the use of clinically feasible measures and those with potentially more precision, but that are not readily amenable to clinical application. We would encourage, where possible, greater consistency in the measurement of variables of interest across studies, including clinician-derived data from standardised tools. A broader consideration is that factors were selected for examination in this study via a clinically-focused, as opposed to data-driven, process. Two key considerations are (a) that a different team of clinicians and researchers may have prioritised a different set of factors; and (b) relatedly, the adoption of a data-driven approach (e.g., principal component analysis) would allow for replication of the process by which variables were selected.
Third, the assessments were usually completed in one session, and by research assistants who were not familiar communication partners. These methods, which were selected to help increase rigor, may nevertheless pose challenges in attempts to conduct comprehensive assessments of the communication skills of children with minimal spoken language. Communication between children and adults is co-constructed, and so it is possible that children would have demonstrated different repertoires of skills if interacting with a familiar adult. Furthermore, in practice, clinicians collect information over time, including via dynamic assessment, which focuses less on the child’s existing skills and more on the child’s development of skills when provided with new learning opportunities. Future studies could include examination of children’s skills with both familiar and unfamiliar communication partners, thereby helping to account for the important influence of partners on children’s communication, learning, and participation; as well as examination of changes in children’s skills when provided with new learning opportunities, such as through the provision of AAC.
Fourth, the current study focused on spoken language outcomes. However, we acknowledge that aspects of children’s development do not occur in isolation. For example, communication may be considered within the context of broader cognitive development and any developmental delay in fine motor skills may impact children’s performance in assessment tasks. A further consideration is the critical question of whether the spoken language changes observed, and factors predicting these, may be unique to children on the autism spectrum, or instead are relevant to a broader group of children with neurodevelopmental conditions, such as intellectual disability and developmental language disorder. As stated at the outset, children on the autism spectrum should be supported and encouraged to use a range of communication modes of their choosing, which may include AAC. It is important that future studies examine the potential moderating effects of AAC on children’s spoken language outcomes, as well as changes in children’s broader communication including the use of AAC.
Finally, consideration should be given to two aspects of the sample. First, the lack of detailed demographic characteristic information is a clear limitation of the study. Relevant factors to consider include children’s race, culture, socio-economic status, home language environment (e.g., mono-lingual, bi/multi-lingual), and co-occurring conditions (e.g., epilepsy, intellectual disability). This information would have been helpful in contextualising the study, including describing the participants and interpreting the findings. Second, the relatively small sample size, which, although not uncommon in clinical research (e.g., Pecukonis et al.,
2019; Saul & Norbury,
2020; Yoder et al.,
2015), limits both power and generalization of the findings. In terms of power, findings could inform a priori calculations for future larger studies into whether predictors (especially the rate of communicative acts, which appears most promising) are indeed unique predictors. Further, larger more diverse samples would allow more fine-grained analysis of potential subgroups/profiles in terms of response to intervention and stepwise changes to better inform intervention efforts moving into the future. Such research may be achieved through greater international collaboration efforts and multi-site pooling of standardised protocols to advance the field.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.