Evaluating Teacher Language Within General and Special Education Classrooms Serving Elementary Students with Autism

Examining the language that teachers use with their students with autism is complicated by differences in the operationalization of teacher language as a construct across educational settings and populations. Hence, we draw from the early childhood, general education, and special education literature to guide our conceptualization of teacher language in relation to students with autism. To begin, many studies evaluating teacher language have been carried out in early childhood general education classrooms, with investigations conceptualizing and measuring teacher language as an aspect of broader interaction quality as outlined on the Classroom Assessment Scoring System (CLASS; Pianta et al., 2008). The CLASS is a widely-used observation scale for quantifying teacher-student interactions in the classroom by measuring the quality of teachers’ feedback, content, and questions to students. High quality teacher-student interactions, in which teachers provide explicit and genuine feedback to students, model academically rich vocabulary, and ask open-ended questions to encourage critical thinking have been associated with positive student outcomes such as academic achievement and social emotional development (Burchinal et al., 2008; Hamre & Pianta, 2005, 2007). This literature base, although broad, is rigorous as it utilizes large samples, quantitative and consistent measures of interaction quality, and a range of student outcome measures (O’Connor, 2010; Pianta et al., 2008).

Other studies examining teacher language have also used observational methods to capture the amount and type of language that teachers direct to their students. This method for measuring teacher language quantifies specific features of talk, such as the frequency of open-ended questions, close-ended questions, language models, and directives. However, conceptualization of and tools for measuring teacher language in this manner have varied across studies and environments. In a few studies within early childhood settings, teacher language has been categorized by pragmatic function, such as language used to encourage continued interaction (open- and close-ended questions), to provide information (language models), and to direct behavior (directives; DeWitt & Hohenstein, 2010; Gast et al., 2010; Walsh & Rose, 2013). Studies that have categorized teacher language by pragmatic function have been descriptive in nature, documenting differences in the talk that teachers use across educational contexts. For example, Gast et al. (2010) found that teachers most frequently provided information during unstructured activities, such as snack and free-play, and encouraged interaction and directed behavior during academically-based activities.

Across settings, studies have also conceptualized teacher language as use of open-ended language to encourage interaction and close-ended language to elicit specific information (Walsh, 2002; Westgate & Hughes, 1997). These studies have primarily described the language environment, examining the relation between features of teachers’ talk and students’ communication (Connor et al., 2020; Jadallah et al., 2011; Liu, 2008; Sadler & Mogford-Bevan, 1997; Sparapani et al., 2020). Studies have found that teachers’ use of open-ended language, such as asking open-ended questions and making contextual statements is associated with student participation (e.g., Milburn et al., 2014), generative or interactive talk (Connor et al., 2020; Duke et al., 2011; Reznitskaya et al., 2009), and on-topic initiations (Mercer, 1996; Walsh, 2002). Open-ended language has also been linked with higher order thinking and academic achievement (Burchinal et al., 2008; Connor et al., 2020).

In contrast, teachers’ use of close-ended language, such as asking yes/no, choice, and simple “wh” questions, using expectant pauses and fill-in strategies, and directing student behavior have been associated with less generative talk and fewer instances of on-topic initiations (Milburn et al., 2014; Sadler & Mogford-Bevan, 1997). While close-ended language may help to scaffold interactions (Mirenda & Donnellan, 1986), this type of talk often elicits fixed or constrained responses, potentially limiting opportunities for students to contribute new ideas and think critically about a given topic (Milburn et al., 2014; Turnbull et al., 2013; Walsh, 2002). Similarly, studies have ssuggested that teachers’ use of directives may limit student engagement within activities and the overall quality of the interaction (de Kruif et al., 2000; Liu, 2008; McWilliam et al., 2002; Williford et al, 2017).

Sparapani et al. (2020) used systematic observational methods to quantify the amount and type of language that interventionists used with elementary students with autism during a reading intervention. Although this study was conducted outside the classroom, the authors operationalized four categories of teacher language drawn from the research literature: responsive language, open-ended questions, close-ended questions, and directives. The authors found that interventionists’ use of open-ended questions was associated with students’ generative responding, while close-ended questions and directives were associated with less frequent generative responding and initiating. These findings are consistent with the literature outlined above linking features of teachers’ talk to students’ contributions, providing some evidence that current conceptualizations of the teacher language construct may extend to students with autism. However, understanding the dimensionality of teacher language for students with autism, and how and whether specific features of talk should be measured separately or together, would provide a more consistent method for conceptualizing and assessing the construct in future studies.

This study is informed by dynamic systems and bioecological theories (Bronfenbrenner & Morris, 2006; Sameroff, 2009; Yoshikawa & Hsueh, 2001), which highlight learning as a dynamic and transactional process between students and their environment. This framework posits that learning is a dynamic and transactional process involving multiple sources of influence that work together to shape child development over time, with a core emphasis on the bidirectional interplay between the child and his/her environment (Connor, 2016; Sameroff, 2009). We can apply this framework to illustrate the dynamic and transactional interplay between students and their environment by studying how teachers talk with their students, and how individual students (with a range of cognitive, language, and social abilities) influence the amount and types of talk that teachers use. That is, while studies suggest that teacher language influences and shapes students’ participation and development (e.g., Burchinal et al., 2008; Mashburn et al., 2008), we posit that the individual characteristics that students bring with them into the classroom influence the language teachers use with them, which in theory, shapes their classroom learning experiences and development over time.

There is an emerging literature base outlining the effect of specific student characteristics on teacher language within classrooms serving students with and without autism. Studies have documented differences in the frequency of open- versus close-ended questions based on children’s age, with teachers using fewer open-ended questions with younger typically developing children (Girolametto & Weitzman, 2002). Teachers’ perceptions of problematic behavior have also been found to influence the type of talk they use with their students, with studies suggesting that teachers use higher rates of directives to comply with students they perceive as “misbehaving” (Dobbs & Arnold, 2009; Dobbs et al., 2004; Koenen et al., 2019; Partee et al., 2020). Similarly, studies examining students with autism within preschool and clinical settings have suggested that autism symptom severity, receptive and expressive language ability, cognitive functioning, and perceived problematic behavior affect the amount and/or type of language teachers use in educational settings (e.g., Dykstra et al., 2013). More specifically, studies have shown that teachers use less language overall with students who exhibit more severe autism symptoms and/or cognitive and language impairment (Dykstra et al., 2013; Irvin et al., 2013, 2015). Teachers also tend to be less responsive with students who exhibit challenging behaviors and those with co-occurring cognitive and language impairment (Keen et al., 2005; Qian, 2018). For example, Irvin et al. (2015) examined the association between adult talk and student characteristics during center activities in 73 children with autism participating in inclusive preschool settings and found that adults used more language to manage behavior with students who exhibited more severe autism symptoms. Similarly, Sparapani et al. (2020) examined interactions between elementary students with autism and their interventionists during a reading intervention and found that interventionists used more language to direct behavior (e.g., sit down, don’t touch that, etc.) than all other types of language with their students who exhibited more severe autism symptoms and limited expressive language skills. The interventionists within the study also asked fewer questions overall, none of which were open-ended, to students with limited expressive language abilities.

This emerging body of literature highlights the importance of teacher language on student learning experiences and the need for continued research, particularly with studies carried out in elementary classrooms serving students with autism. Taken together, these studies suggest immense variability in the language that students with autism experience in the classroom, and these differences may be, in part, due to individual student characteristics (Bronfenbrenner & Morris, 2006; Connor, 2016; Dykstra et al., 2013). It is possible that the types of language that teachers use, such as open-ended questions and language models, may afford different learning opportunities for different students (Hestenes et al., 2004; Irvin et al., 2013; Sparapani et al., 2020). However, better understanding the learning opportunities that varying types of teacher language afford is an area of future research.

Previous studies examining teacher language have largely conceptualized the construct by measuring key aspects of language associated with the quality of teacher-student interactions or by categorizing teacher language into varying dimensions, such as pragmatic function or open- verse close-ended statements. Structural equation modeling (SEM) has been utilized in larger studies examining teacher-student interaction quality (e.g., Hamre et al., 2013, 2014); however, to our knowledge, no study has utilized SEM to examine the latent structure of teacher language in classrooms serving students with autism or whether differences in the language environment exist between general and special education settings. Thus, this study utilized SEM (multi-group confirmatory factor analysis) to characterize dimensions of teacher language as outlined in the literature, evaluate similarities and differences in the dimensions of teacher language between general and special education settings, and examine the predictive association between student characteristics and teacher language within a large sample of students with autism and their educators.

This study included participants who were recruited between 2010 and 2014 for the Classroom Social Communication Emotion Regulation Transactional Support (SCERTS) Intervention project (CSI; Morgan et al., 2018), a randomized controlled intervention trial evaluating the efficacy of the SCERTS Model as a teacher-mediated intervention for learners with autism. Teachers and students were recruited from 54 schools across eight school districts in FL, GA, and San Diego, CA. Data for the present study include classroom video observations recorded at the beginning of the school year, prior to the start of intervention during Years 1–3 of the CSI project. See Morgan et al., 2018 for a complete description of participant recruitment. This study was approved by the Institutional Review Board at Florida State University as well as the review boards for participating school districts.

Teachers and their paraprofessionals from 126 kindergarten–second grade general (n = 71) and special education (n = 55) classrooms participated in the study. While lead teachers are primarily responsible for students’ attainment of curricular standards, language provided by all members of each student’s educational team was analyzed for this study. This approach was considered to be more ecologically valid as all adults in the classroom may interact with students throughout the school day (e.g., Irwin et al., 2015). Therefore, the focus of this study was on the language environment that each student experienced rather than individual teachers’ language.

Students who participated in the larger trial met the following criteria: (1) enrollment in kindergarten, first, or second grade at the beginning of the school year in either general or special education classrooms; (2) confirmation of a clinical or educational diagnosis of Autistic Disorder, PDD-NOS, or Asperger Syndrome as defined by the DSM-IV-TR (American Psychiatric Association [APA], 2000) prior to the start of the study; and (3) no co-occurring severe motor delay/impairment, dual sensory impairment, or history of traumatic brain injury.

Participants in this study (N = 194) included 78 students whose primary placement was in general education classrooms and 116 students with primary placements in special education classrooms. However, during the baseline video collection, nine students were in an educational setting that differed from their primary classroom placement (i.e., student primarily receives services within a special education setting but spends a percentage of time in a general education classroom). This resulted in a total of 87 students participating in general education classrooms and 107 students participating in special education classrooms during baseline. Twenty-nine percent of the students (n = 57) were in kindergarten, 41% (n = 80) in first grade, and 29% (n = 57) in second grade. The sample was primarily male, consistent with the observed 4:1 sex ratio for individuals with autism (Baio et al., 2018). See Table 1 for student demographic information.

As part of the larger CSI project (Morgan et al., 2018), students completed a battery of diagnostic and developmental measures at the beginning of the school year that examined: (1) autism symptomology; (2) intellectual functioning; (3) adaptive functioning; (4) receptive and expressive vocabulary; (5) social impairment; and (6) internalizing and externalizing behaviors. Descriptive statistics on each of the standardized measures for students in general and special education classrooms are presented in Table 2. Each month, for the duration of the study, trained videographers also collected a continuous 60-min video observation of each student participant in their classroom. See Morgan et al., 2018 for a complete description of the video observation process. This study specifically examined video observations collected during baseline, prior to the start of intervention as part of the larger CSI project.

The sample of students showed marked variability in intellectual functioning (M = 73.51, SD = 20.25) and overall adaptive behavior (M = 74.10, SD = 11.25), with 59% (n = 114) of the students exhibiting ABIQ scores at or above 70. Expressive vocabulary (based on sample z-scores), on average (M = 98.41, SD = 14.31), was higher than receptive language (M = 74.02, SD = 23.54), which is consistent with research examining language functioning in individuals with autism (Charman et al., 2003; Howlin et al., 2004). We observed significant differences between classroom setting for developmental characteristics but not for autism symptom severity (ADOS-2), severity of social impairment (SRS), and presence of problem behaviors (TRF).

Teachers in special education classrooms used more verbal bids overall (M = 66.55; SD = 37.10) compared to teachers in general education classrooms (M = 45.63; SD = 26.64), with a higher proportion of their bids directed toward individual students (M = 76.26%; SD = 23.53) compared to groups of students (M = 23.53%; SD = 24.10). Teachers in general education classrooms demonstrated the opposite pattern, with 59.16% (SD = 30.57) of their bids directed toward groups of students. Special education teachers also used a higher proportion of language models (M = 13.07%; SD = 10.55) compared to general education teachers (M = 6.47%; SD = 6.09). Across both classroom settings, teachers used relatively more directives and close-ended questions than all other types of talk and very few open-ended questions, fill-ins, and indirect requests. See Table 3.

Pearson product-moment correlations among the types of teacher talk and standardized measures were estimated across classroom settings (Table 4). Overall, we observed small to moderate positive, significant correlations between the teacher language categories in both general and special education classrooms. Within general education classrooms, teachers’ use of open-ended questions was positively associated with students’ expressive vocabulary (r = .23), and directives and indirect requests were positively associated with autism symptom severity (r = .22; r = .24). Teachers’ use of close-ended questions and fill-ins was negatively associated with students’ cognitive functioning (r = − .31), receptive vocabulary (r = − .28), and expressive vocabulary (r = − .35). However, these relations looked different in special education classrooms, such that, teachers were more likely to ask open-ended questions with their students who exhibited less severe autism symptoms (r = − .26), and they used more directive language with students who exhibited less developed expressive vocabulary (r = − .36), receptive vocabulary (r = − .31), and cognitive abilities (r = − .36). Similar to general education classrooms, teachers used more directive language with their students with more severe autism symptoms (r = .21).

To begin, we found that teachers in special education classrooms used relatively more language overall than teachers in general education classrooms, with much of their talk (69%) consisting of directives and close-ended questions that were directed at the individual student. Because we analyzed the language environment that each student experienced, we captured, at times, both teachers and paraprofessionals directing language at a student simultaneously within special education classrooms. This high frequency of competing verbal bids directed at individual students may be potentially concerning given the core communication challenges that differentiate students with autism (APA, 2000). That is, too much language or different features of talk directed simultaneously may overwhelm students with autism rather than afford them with rich language learning opportunities. Future research that includes measures of teacher language and student participation is needed, however, to better understand how teachers might best match their language to meet their students’ needs.

In addition, studies suggest that teachers’ use of directive language tends to be intrusive as it is often used to control, stop, or redirect student behavior (de Kruif et al., 2000; McWilliam et al., 2002; Williford et al, 2017). Increased directive language has been found to have detrimental effects on classroom active engagement for students with autism, jeopardizing the quality of the interaction and further intensifying the presence of challenging behaviors (Keen et al., 2005). Close-ended questions have also been found to hinder student participation and generativity within an exchange (e.g., Milburn et al., 2014). By framing questions in a close-ended manner, teachers may be providing more structure to interactions, yet they might also be limiting opportunities for their students think critically about the content and generate new, creative ideas and responses (Milburn et al., 2014)—which could potentially impact their developmental and educational outcomes (Connor et al., 2020).

Although we documented more language overall in special education classrooms, teachers in general education classrooms used a greater proportion of directives and close-ended questions with their students than all other language categories, consisting of over 60% of their talk. Teachers in both settings used relatively fewer language models and fill-ins—features of talk that help to scaffold interactions (Battaglia & Mcdonald, 2016). They also rarely asked students open-ended questions. Yet, open-ended questions have consistently been associated with active engagement and academic growth (Connor et al., 2020; Milburn et al., 2014) as well as increased student initiations and generative talk (Duke et al, 2011; Walsh, 2002) in students with and without autism. Taken together, these data suggest that the language environment that students with autism experience in classrooms, may be less than optimal—potentially limiting their opportunities to engage in rich exchanges that support learning and development.

We tested three models of teacher language drawn from a large, comprehensive literature base in order to understand the dimensionality of talk teachers directed toward their students with autism in general and special education classrooms. Although previous studies have conceptualized teacher language by pragmatic function (e.g., Walsh & Rose, 2013) or by teachers’ use of open-versus close-ended language (e.g., Walsh, 2002), we found that the construct was best represented by a unidimensional factor consisting of six features of talk (open-ended questions, language models, close-ended questions, directives, indirect requests, fill-ins). Our observation of a single underlying teacher language factor may be unique to students with autism, considering we observed high frequencies of directives and close-ended questions yet very low frequencies of other types of talk (i.e., open-ended questions; fill-ins). This pattern of talk could have impacted the overall dimensionality of the construct. In addition, we found that the unidimensional model of teacher language was consistent across general and special education classrooms, but the degree to which specific features of talk contributed to the latent factor varied between the settings. This provides evidence that the types of talk that teachers use with their students with autism manifest differently between general and special education classrooms. This finding also coincides with and helps to explain the differences we documented in the association between student characteristics and teacher language across the two settings.

Our data suggest that the characteristics of learners with autism are related to specific features of their teachers’ talk, yet these relations differ between general and special education classrooms. We documented that teachers’ use of directives within general and special education classrooms were related to autism symptom severity, suggesting that teachers used more directives with their students who exhibited more severe autism symptoms. In special education classrooms, teachers also used more directives with their students who exhibited greater cognitive and vocabulary impairment, and more open-ended questions with students who exhibited less severe autism symptoms. Whereas, teachers in general education classrooms asked more open-ended questions of their students who exhibited stronger expressive vocabulary skills. Finally, using SEM we found that teachers in general education classrooms used less language overall with their students who exhibited limited receptive vocabulary, yet this was not the case within special education classrooms. These findings are consistent with previous studies (e.g., Dykstra et al., 2013) conducted in preschool and clinical settings, providing evidence that students with autism experience differences in the amount and type of language they receive in classrooms, and this is, in part, due to their autism symptoms, expressive and receptive vocabulary, and cognitive skills. Although outside the scope of this study, there may be a number of reasons why teachers use different types of talk with their individual students. Nevertheless, these data might suggest a need for teachers to include scaffolds, modifications, materials, and/or other adaptations into classroom activities rather than rely on oral language, such as the use of directives and/or close-ended questions, for students with limited language and lower cognitive skills. However, this warrants further investigation.

One of the primary contributions that this study makes to the existing literature is the sampling and analysis of classroom video observations at the student level, collecting one video observation for every participating student. By examining teacher language within general and special education classrooms, this study provides a snapshot of each student’s unique experience within his/her classroom. Hence, the large and heterogeneous sample of students with autism supports the overall generalizability of our findings to the larger population of students with autism in elementary classrooms. In addition, a broad range of reliable and valid standardized measures were used to characterize the sample, including “gold standard” diagnostic measures to confirm autism. Evaluation of teacher language was carried out using systematic observational methods, allowing us to capture nuances of talk during a continuous sampling of classroom activities. Inter-rater agreement between observers was calculated for each of the teacher language categories and indicated good agreement overall. Finally, the use of advanced methodological approaches, including a latent approach, further contributes to the richness and novelty of these data relative to the extant literature examining teacher language in classrooms serving students with autism.

This study has a few notable limitations. We examined teacher language within a systematically selected 15-min sample of classrooms activities. However, future research is needed to understand how representative or reflective the sampled time is of the entire 60-min video observation—an area of need within systematic observational measurement more broadly (Yoder et al., 2018). In addition, we used systematic observational methods and continuous sampling to examine each instance of teacher language, which provided a rich depiction of the classroom language environment. However, this detailed analysis was time-consuming and may have limited utility outside the research laboratory. Efforts focusing on using empirical evidence from studies such as this one to develop practical tools that gauge teacher language within classrooms could greatly advance the field. Although this study included a large sample of students with autism in general and special education classrooms, the sample size is considered fairly small for the analytical methods that were used. Therefore, we were unable to account for the nesting nature of the data when running the multi-group CFA. Future studies, which include students with and without autism, are needed to fully grasp how these findings can be applied to educational settings for elementary students with autism more broadly.

Furthermore, differences between classroom types regarding the proportion of students who are racially/ethnically diverse may have contributed some bias to the results and should be fully explored in future studies to better understand the intersectionality between classroom placement, student characteristics, and racial/ethnic background. Finally, we did not collect information on teacher/student ratio within general and special education classrooms or similarities and differences between teachers’ instructional approaches or teaching philosophies across educational settings. Hence, future studies are needed to understand the impact of adult support and varying evidence-based interventions (i.e., such as positive behavioral supports, ABA techniques, and cooperative learning opportunities) on the language environment.

By examining the factor structure of teacher language, this study provides initial empirical support for quantifying and measuring the amount and type of talk that teachers use with their students with autism in general and special education classrooms. Hence, this comprehensive examination of the classroom language environment provides a foundation for understanding specific features of teachers’ talk that may relate to (or hinder) student development and learning—an area of research that could help inform best practices for learners with autism. Findings from this study may also help to improve the consistency and accuracy of evaluating teacher language within classrooms. However, future research linking teacher language to student outcomes as well as examining the utility of a unidimensional model of teacher language across populations is needed.

In addition, the differences in teacher language that we documented between general and special education classrooms may have important implications for learners with autism. One potential concern is that the language environment within special education classrooms may not adequately prepare students for the linguistic and social pragmatic directives within general education classrooms (e.g., amount and types of talk). That is, inconsistent talk between general and special education classrooms may create an instructional barrier for learners with autism who transition between settings, as consistency between environments may provide the predictability needed to support generalization. While this is only speculation, our findings suggest that the language environment is important to consider when evaluating teachers and students within educational settings.

Finally, the links that we observed between teacher language and student characteristics (i.e., more directives used with students who exhibit more severe autism symptoms) suggest that students with more severe autism symptoms and those with co-occurring cognitive and/or vocabulary impairment may have limited access to a rich language environment that supports their learning and development. Hence, these findings raise a potential concern related to equity and highlight the need for continued research that focuses on understanding and evaluating educational practices for learners with autism. Efforts related to identifying and evaluating salient features of teachers’ talk that support student engagement and learning will offer insight into effective school-based practices while ensuring that all students with autism, including those with more severe impairment, are presented with equitable learning opportunities across educational settings.