|Home | About | Journals | Submit | Contact Us | Français|
The aims of this article are twofold: (a) to offer a set of recommended measures that can be used for evaluating the efficacy of interventions that target spoken language acquisition as part of treatment research studies or for use in applied settings and (b) to propose and define a common terminology for describing levels of spoken language ability in the expressive modality and to set benchmarks for determining a child’s language level in order to establish a framework for comparing outcomes across intervention studies.
The National Institute on Deafness and Other Communication Disorders assembled a group of researchers with interests and experience in the study of language development and disorders in young children with autism spectrum disorders. The group worked for 18 months through a series of conference calls and correspondence, culminating in a meeting held in December 2007 to achieve consensus on these aims.
The authors recommend moving away from using the term functional speech, replacing it with a developmental framework. Rather, they recommend multiple sources of information to define language phases, including natural language samples, parent report, and standardized measures. They also provide guidelines and objective criteria for defining children’s spoken language expression in three major phases that correspond to developmental levels between 12 and 48 months of age.
Autism is a neurodevelopmental disorder characterized by primary impairments in social interactions, communication, and repetitive and stereotyped behaviors (American Psychiatric Association, 2000). In addition, autism often results in significant disability, including intellectual deficits and language and adaptive behavior deficits, as well as problem behaviors. It is now recognized that classic autism is part of a spectrum of related disorders that includes pervasive developmental disorder, not otherwise specified (PDD-NOS), and Asperger syndrome; this set of diagnoses, collectively, is referred to here as autism spectrum disorders (ASDs). Outcomes for children with ASD represent a broad continuum, with only a small percentage achieving independence and full employment as adults (Howlin, Goode, Hutton, & Rutter, 2004). ASDs are no longer thought to be rare. Current reports indicate that 1 in every 150 children in the United States will receive an ASD diagnosis (Bertrand et al., 2001; Kuehn, 2007; Yeargin-Allsopp et al., 2003).
Children with ASD have long been known to respond to interventions that target specific skills and behaviors (National Research Council [NRC], 2001), and numerous studies have demonstrated the positive effects of early intervention on language development for the majority of children with ASD (Dawson & Osterling, 1997; Koegel & Koegel, 1988; Lovaas, 1987; Rogers, 2005; Rogers & Vismara, 2008), with some, though sparse, evidence of long-lasting benefit. The fact that language development can be positively affected by early treatment has tremendous potential significance, because the emergence of spoken language is one of the most important variables predicting better outcomes in later childhood and adulthood (Gillberg & Steffenburg, 1987; Howlin et al., 2004; Venter, Lord, & Schopler, 1992). Thus, given the role of language acquisition in shaping long-term outcomes, it has become important to identify the most successful strategies for facilitating language acquisition in young children with ASD, who uniformly demonstrate significant delays in at least some aspects of language and communicative development, especially in the domain of pragmatics (Tager-Flusberg, Paul, & Lord, 2005).
Although various intervention approaches teach and measure language acquisition in different ways, depending on the philosophy and underlying theory of the approach (see Rogers, 2005, for a review), consumers of this literature must be able to compare language outcomes from different treatment approaches. Despite the numerous published language outcome studies of early intervention in ASD (Rogers, 2005), it is not possible to compare language outcomes across reports, because of the lack of uniform measurement approaches to assessing language skills and the lack of uniform terminology for describing language outcomes in ASD. Many intervention programs for children with ASD aim to facilitate the development of functional speech. However, because there has never been consensus on the definition of functional speech, it is impossible to compare the longer term efficacy of different treatment programs. In this article, we offer an alternative framework for describing spoken language acquisition in children with ASD. The proposal described here replaces the arbitrary singular categorical distinction encompassed by the terminology of functional speech with a framework that captures the continuous developmental processes that underlie language acquisition.
In December 2006, the National Institute on Deafness and Other Communication Disorders (NIDCD) assembled a group of experts in language disorders and language acquisition in young children with ASD to address these issues.1 Over the next year, the group worked together through a series of conference calls and correspondence, with their efforts culminating in a meeting held in December 2007. This article summarizes the group’s recommendations relating to our primary goal of providing benchmarks for defining the acquisition of spoken language in the expressive modality in young children with ASD.
The working group set the following two major objectives:
As such, this article is addressed primarily to researchers; however, practitioners and other consumers are also relevant audiences. For researchers in early autism intervention who may come from a wide range of theoretical backgrounds and practices, our goal is to provide common terminology and a suggested approach to defining language abilities before, during, and after treatment. The varying measurement approaches used in language intervention research require different levels of financial and human resources and expertise. In addition, researchers have differing aims and hypotheses that may require specialized descriptions of language acquisition of their participants. Thus, we propose a measurement approach that may be applied in a “bare bones” fashion (e.g., relying on direct assessments and parent reports) as well as a more elaborated measurement system (e.g., adding in measures derived from natural language samples), covering the full range of language domains that could be included in treatment programs. By proposing these guidelines, we hope to move beyond the ambiguously defined treatment goal of functional speech to a more standardized approach, using common measures and common definitions that will allow comparison of outcomes across studies.
For practitioners, the proposed measures and benchmarks presented here provide a framework for describing the language progress of their clients during treatment. By providing a common framework, we hope to facilitate the assessment process for clinicians, allowing them to measure their clients’ language gains in relation to the research literature. Thus, we aim to enhance the relationship between treatment research and clinical practice in the field of language intervention in ASD.
Our final target group of readers includes parents, early intervention professionals, and others who work to extract evidence of progress, be it research effectiveness or clinical efficacy, from clinical reports and research articles that use language measures to chart change in children with ASD. Clearly defined benchmarks of speech and language development will aid families, early childhood educators, and others who turn to the language research literature to understand language growth in young children with ASD.
Some additional comments are in order. First, we focus here exclusively on the development of spoken language through the preschool years, omitting consideration of measures and benchmarks for defining preverbal communicative skills. Although the working group recognizes that sophisticated language skills take many forms, including both verbal and nonverbal means for effective communication, and that children with ASD continue to make important advances in language well into the school years, we selected these constraints because outcome studies are uniform in the predictive power of spoken language (i.e., speaking in sentences that serve a variety of functions; e.g., Paul & Cohen, 1984) by age 5 years for individuals with ASD (e.g., Howlin et al., 2004; Venter et al., 1992). Second, we limited our focus to the development of expressive language skills in children with ASD because most intervention studies target expressive language as the primary outcome and also because expressive language is more reliably assessed, especially in children with ASD (cf. Tager-Flusberg, 2000). Third, we have limited our recommendations for measures and benchmarks to English, in part because almost all current studies have focused on English-speaking children with ASD. We hope, however, that the overall framework and guidelines presented here can be readily translated into other languages with some modifications.
In order to capture the spoken language and communicative abilities of young children with ASD and to avoid sampling effects, assessments in this domain should include measures derived from multiple sources. These sources should ideally include (a) natural language samples, (b) parent report, and (c) direct standardized assessment.
Natural language samples (NLSs) that are collected in different communicative contexts provide excellent measures of a child’s expressive language abilities, including phonological repertoire, lexical and grammatical knowledge, and pragmatic/communicative skills; the last (i.e., pragmatic/communicative skills) are especially difficult to measure using other types of assessment. An NLS may be collected during either experimenter/clinician–child or mother–child interactions. Contexts during which an NLS may be collected include the administration of the Autism Diagnostic Observation Schedule (ADOS; Lord et al., 2000), the Communication and Symbolic Behavior Scales (CSBS; Wetherby & Prizant, 2002), the Early Social and Communication Scales (ESCS; Mundy et al. 1996; Siebert, Hogan, & Mundy, 1982), or equivalent contexts that include social communicative presses. The specific context should be determined based on the goals of the assessment. For example, if a primary outcome measure of a treatment program includes the functional use of specifically targeted forms, then adequate sampling of a range of different communicative contexts (e.g., contexts for requesting, protesting, sharing) would be needed.
Typically, natural language samples will be at least 30 min in length to provide adequate time and opportunity to sample a sufficient number and range of utterances. For children with ASD, one may need to concatenate several short language samples to obtain 30 min of language behavior. Following the collection of an NLS (see Miller & Chapman, 2000, for a discussion of methods), the data must be transcribed and coded to derive useful measures of the child’s language. The particular level of transcription (e.g., phonetic, lexical, inclusion of adult language) will again depend on the specific focus of the assessment. Transcription and analyses can be supported by computer-based software, including the widely used Systematic Analysis of Language Transcripts (SALT;Miller & Chapman, 2008), Child Language Data Exchange System (CHILDES; MacWhinney, 2000), Lingquest (Mordecai & Palin, 1982), or Computerized Profiling (Long & Fey, 2004).
Parent report measures, administered in questionnaire or interview format, can provide useful information about a child’s language skills that may not be observed in a laboratory or clinic setting. The most widely used measure is the MacArthur-Bates Communicative Development Inventory (MCDI; Fenson et al., 1993; 2007). The MCDI can be used to assess children’s expressive vocabulary and grammatical knowledge between the ages of 8 and 42 months. Although there are concerns that some parents may over- or underreport their child’s language repertoire, parent report instruments have generally been shown to provide valid assessments of young children’s language as measured by evidence that early predictors of language also predict MCDI productive raw scores in children with ASD (Charman et al., 2003; Luyster, Qui, Lopez, & Lord, 2007). There is also evidence that MDCI scores are highly correlated with other measures of language in children with autism (Luyster, Kadlec, Connolly, Carter, & Tager-Flusberg, 2008).
Direct assessment of a child’s language skills should be accomplished using standardized tests that have good psychometric properties, with particular attention paid to the reliability and validity of the measures that are derived from such tests for children with ASD. Standardized tests can be used to assess expressive language skills in phonological, lexical, grammatical, and pragmatic domains of language. We note, however, that few standardized assessment instruments provide opportunities for assessing language skills aside from basic naming ability in children younger than 24 months of age. In addition, most elicited production tests have very few items during this early language period, which means that age equivalency or standard scores can change dramatically with a difference of only one or two raw score points.
Many children in the process of acquiring language use imitation and repetition of spoken language, especially during the early stages, to serve some functional communicative goals. Echolalia and stereotyped language, consisting of scripts heard in previous contexts repeated in a noncommunicative way, are atypical imitation behaviors that are part of the symptom pattern of ASD (Kanner, 1946; Prizant, 1983). During the early stages of language acquisition, it may be difficult to discriminate typical from atypical verbal repetition in young children, and there are no clear criteria for defining delayed echolalia (Prizant & Duchan, 1981). Nevertheless, when characterizing the complexity of children’s language, we recommend that echolalic (and imitative) language be omitted from analyses, as well as from speech samples used to classify children according to the benchmarks described later in this article.
We take as our starting point a developmental approach in which we benchmark criteria for the acquisition of spoken language and recommend measures for expressive language at different development levels. For each level, we provide approximate age ranges, although these ranges should be viewed as overlapping and not necessarily definitive. A developmental perspective provides a conceptual framework to guide intervention and evaluation of children with ASD, ensuring that researchers and clinicians strategically plan to target key language milestones within language intervention programs for children with ASD. Within a developmental framework, we identify five key phases of expressive language acquisition:
Children in this phase communicate using preverbal intentional communication through vocal (babble) and gestural means. This phase generally covers the age range of 6–12 months in typically developing children. As noted previously, we have not included measures or benchmarks for this developmental phase as it is outside the scope of our goals.
Children in this phase use nonimitated spontaneous single words referentially and symbolically to communicate about objects and events, including those outside the immediate context. At least some of their speech is intelligible and incorporates the most frequent consonant sounds heard in typical babble (Oller, 2000; Stoel-Gammon, 1998). Children in this phase use speech with a variety of people in different settings to serve several functions, including, but not limited to, labeling, requesting, and commenting on (directing joint attention to) some objects or activities. This phase generally covers the age range of 12–18 months in typically developing children.3
Children in this phase have a vocabulary that is rapidly increasing in size and includes a variety of parts of speech (nouns, verbs, descriptors). They are able to combine words creatively to refer to objects and events. Two- and three-word combinations are used for several different communicative functions. This phase generally covers the age range of 18–30 months in typically developing children.
Children in this phase combine words into clausal structures, or sentences, and use some morphological markers such as plurals, prepositions, and some verb endings. Their vocabulary is sufficiently large to serve their communicative needs in everyday situations. They communicate a wide range of functions in different settings with both familiar and unfamiliar people. The portion of this phase relevant for the proposed benchmarks defined here corresponds to typically developing children between the ages of 30–48 months.
By the end of the preschool years, typically developing children have large and rich vocabularies that they use to communicate a wide range of topics (including abstract or hypothetical ideas) using complex grammatical constructions (e.g., relative clauses, sentential complements, anaphora) in different discourse contexts (e.g., conversation, narrative). We do not include either measures or benchmarks for this developmental phase (excluding measures not designed for children below the age of 48 months), as our focus is primarily on younger children with ASD. (For further reading, see Hoff & Schatz, 2007; Menn & Bernstein Ratner, 2000.)
Table 1 provides a summary of our proposed benchmarks that define the key developmental phases for spoken language expression (First Words, Word Combinations, Sentences) across the different domains of language, with examples of how each type of measure can be used to assess children’s level of language use. As noted earlier, our objective in presenting this frame-work of benchmarks in each language domain at different developmental phases is explicitly designed to move away from the commonly used term functional speech as the outcome goal for intervention studies.
The benchmarks presented in Table 1 can be used for multiple purposes: (a) to evaluate whether a child meets criteria for achieving the various language phases in the context of treatment research; (b) as measures to be incorporated into intervention studies; or (c) as a means for monitoring a child’s progress in ongoing community treatment. Although we present our benchmarks in each of the developing phases of language, it is important to keep in mind that these phases are dynamic and overlapping periods that, in reality, have no clear boundaries.
Some treatment studies include goals to advance a child’s language to a particular level. For example, in studies that begin with very young or preverbal children (i.e., children who do not meet the criteria for being in the First Words phase), the goal might be to provide interventions that lead the child into becoming “verbal”— which might then be defined as the First Words level. Other studies might have a more flexible goal of advancing children to the next level within a prescribed treatment period, or to chart language gains based on continuous measures (e.g., number of different consonants, words, or communication functions). Across all intervention studies, criteria for defining each language phase will facilitate the comparison of different treatment studies that may have different designs or measures.
For each language phase, we defined the minimum criteria for evaluating a child’s language level: In order for a child to be considered to be at a particular level of expressive language functioning, the child’s measured language must meet at least one of the defined minimum benchmarks in every language domain that defines that phase. This stringent approach recognizes the comprehensive developmental approach to language acquisition in children with ASD that we have proposed, one which encompasses all aspects of language used to communicate effectively with others in everyday life.
Although each phase contains benchmarks for all language domains based on how language develops in typically developing children, we recognize that, particularly in children with ASD, there is likely to be asynchrony across different language domains (e.g., vocabulary development may be significantly more advanced than pragmatics). This will result in a mixed phase profile for many children. A child might meet minimum criteria for one phase in all domains and may also meet criteria for the more advanced level in one or two domains assessed. Researchers or clinicians may choose to describe a child’s language separately for each language domain in place of the criteria defining the language phase.
This phase represents the emergence of spoken language covering the age range of 12–18 months in typically developing children. The benchmarks targeting this phase are placed at the 15-month age-equivalent level. To conclude that a child has reached the First Words phase, he or she must meet the following criteria within each of the following domains:
The phase covers the age range of 18–30 months in typically developing children. The benchmarks targeting this phase are placed at the 24-month age-equivalent level. The following criteria define meeting the benchmarks for this phase:
This phase covers the broad age range from 30 to 48 months in typically developing children. The benchmarks are targeted to the 36-month age-equivalent level. The following criteria need to be met for this phase:
This report represents the consensus of our working group based on discussions carried out over the course of 18 months. We offer here the following summary and conclusions.
We recommend a move away from using the term functional speech as a goal for intervention research and practice, replacing it with a developmental framework. We recognize that the impetus for the use of the term came from studies suggesting that achieving functional speech by age 5 years is an important prognostic indicator in children with ASD. Nevertheless, it is not clear from the literature what definitions earlier studies relied on, though the descriptions in these studies suggest that children with optimal outcomes were able to speak in full sentences serving a range of communicative functions (Paul & Cohen, 1984). In our view, given the significant changes in the age of diagnosis and the increased access to early intensive intervention, it is time to reopen the question of the timing and role of language acquisition as key prognostic indicators in ASD.
In evaluating treatment outcomes, we depend on objective measures, but we recognize that the measures available to us are imperfect. This is particularly evident when assessing the earliest phases of language in the emergence of words, grammatical combinations, or the pragmatic uses of communication for which few if any standardized direct assessments are available for children under the age of 2. To address these limitations, we encourage the use of measures derived from natural language samples and parent report. We recognize that the collection, transcription, and coding of natural language samples involves increased labor costs in research and clinical settings. However, we believe this cost cannot be avoided if we are to ensure that the data gathered have the highest degree of validity possible.
We provide objective criteria for defining children’s expressive language development in order to provide guidance to researchers and clinicians who assess language in young children with ASD. These may be used to guide intervention research as well as treatment offered in clinical settings. The use of benchmarks based on typical development for charting children’s progress reflects findings that language development in early ASD generally follows a similar developmental pathway as in other children (Tager-Flusberg et al., 2005). Using benchmarks from typical development also draws attention to those typical language milestones that should be targeted by early intervention programs. These definitions and benchmarks will allow comparisons of outcomes across different studies.
Finally, we set ourselves a practical goal: to provide a common vocabulary for discussing language acquisition with a wide interdisciplinary professional and lay audience. The terms selected for the benchmarks are intended to be transparent, reflecting important language features that define them. For each benchmark, we have provided definitions for behavior that can be objectively assessed by a broad range of early intervention professionals.
The framework we have developed here should be expanded in several ways, by (a) incorporating benchmarks for identifying a range of preverbal communication skills; (b) developing valid and reliable measures of language comprehension for children with ASD; (c) adapting the framework for assessing children who communicate using AAC systems; and (d) evaluating the relative merits of different types of measures for children with ASD. Further research is needed to address these important issues; nevertheless, we hope that the concepts and recommendations presented in this article will enhance early intervention research targeting spoken language development in ASD and will provide clinical professionals with the ability to extract and clearly define important information about treatment effectiveness in their work.
Support for the preparation of this report was provided by the National Institute on Deafness and Other Communication Disorders and by Autism Speaks. We are especially grateful to Andy Shih and Alycia Halladay of Autism Speaks for their support of this project.
1The group was co-chaired by Helen Tager-Flusberg, Sally Rogers, and Judith Cooper (NIDCD). Members included Rebecca Landa, Catherine Lord, Rhea Paul, Mabel Rice, Carol Stoel-Gammon, Amy Wetherby, and Paul Yoder.
2We have limited our recommendations on measures and benchmarks to spoken language, although we recognize that many children with ASD who do not speak can successfully acquire some expressive language skills using augmentative or alternative communication (AAC) systems such as vocal output devices or manual signing. We have not included a detailed presentation of how our framework might apply to interventions that target AAC systems, as there are no clear guidelines available for how to measure nonspoken language skills that are comparable to those available for spoken language.
3Children with ASD will often have a very small spoken vocabulary used primarily to regulate others’ behavior; however, unless the criteria for the definition of the First Words phase (as specified in Table 1) are met, they should be considered to be in the Preverbal Communication phase.
Helen Tager-Flusberg, Boston University School of Medicine, Boston, MA.
Sally Rogers, University of California, Davis/Medical Investigation of Neurodevelopmental Disorders (MIND) Institute, Sacramento, CA.
Judith Cooper, National Institute on Deafness and Other Communication Disorders, Bethesda, MD.
Rebecca Landa, Kennedy Krieger Institute, Johns Hopkins School of Medicine, Baltimore, MD.
Catherine Lord, University of Michigan, Ann Arbor.
Rhea Paul, Yale Child Study Center, New Haven, CT.
Mabel Rice, University of Kansas, Lawrence.
Carol Stoel-Gammon, University of Washington, Seattle, WA.
Amy Wetherby, Florida State University, Tallahassee.
Paul Yoder, Vanderbilt University, Nashville, TN.