Teachers’ oral language use may be an important factor in student achievement, particularly for students who struggle with language, learning, and behavior. This study examined features of teacher talk during whole-class instruction in 14 general education (GE) and 14 self-contained special education (SE) elementary classrooms that included students with or at risk for emotional and behavioral disorders. Across settings, 74% of teachers’ utterances contained vagueness markers that may hinder comprehension. Within teachers, the quantity, complexity, and clarity of oral language tended to remain stable across lessons, regardless of lesson content. Teacher-rated severity of behavior did not differ by setting, but students in self-contained SE classrooms had significantly lower language and reading skills than their counterparts in GE settings. Analyses of multilevel models revealed no significant differences in form or content of teacher talk between groups of teachers across settings (GE or SE) or grade levels (K–2, 3–4).
Classroom communication plays a critical role in children’s performance in schools (Thatcher, Fletcher, & Decker, 2008). A great deal of research has focused on students’ language development in relation to school performance; however, communication involves creating shared meaning between speakers and listeners. Therefore, it is also important to consider teachers’ language use in forming instructional messages. Although it is likely that all learners benefit from clear communication, it is imperative that teachers use clear, concise, and concrete language when delivering instruction to students who struggle to meet the academic and behavioral demands of the classroom (Archer & Hughes, 2011; Engelmann & Carnine, 1991).
A rich body of research from diverse academic disciplines has described specific features of teacher talk in general education (GE) settings across grade levels from preschool to college; however, few studies of teacher talk have included special education (SE) students. Although researchers have hypothesized that modifying teachers’ oral language use may be an effective means of preventing classroom conflict and problem behavior (Harrison, Gunter, Reed, & Lee, 1996), no known studies have analyzed teachers’ instructional language in classrooms including children with severe and chronic behavioral difficulties. This study extends the literature by observing and quantifying how both GE and SE teachers encoded instructional messages in a sample of 28 elementary school classrooms that included students with or at risk for emotional and behavioral disorders (EBD).
Bloom and Lahey (1978) described a linguistic model composed of three basic components: form, content, and use. The form, or structure, of spoken language includes basic elements of speech affecting production of words and sounds, including fluency, rate, amount of talk, and syntactic complexity. Content references semantics, or the linguistic units (words, phrases) used to convey meaning. The use of language references pragmatics, or the function of language in discourse, conversation, instruction, or social contexts.
A great deal of research has provided important information about how teachers use classroom discourse to accomplish the work of instruction (Howe & Abedin, 2013). For example, Sinclair and Coulthard (1975) observed that lessons often conform to a sequence of instruction identified as an initiation-response-feedback or initiation-response-evaluation (IRE) format. In this sequence, a teacher initiates an interaction, a student responds briefly, and the teacher gives feedback or evaluates the quality of the response. Many researchers have extended these findings, for example, by examining teachers’ use of questions or specific forms of feedback and evaluation (e.g., Cazden, 2001; Mehan, 1979). Researchers also have used both qualitative and quantitative methodologies to analyze ways in which classroom discussions, discourse, and dialogue influence students’ learning outcomes (Reznitskaya & Gregory, 2013). Finally, researchers have examined ways in which teachers’ verbal behaviors (e.g., praise statements, positive and negative feedback, opportunities to respond to instruction) function as reinforcers for student behavior (Partin, Robertson, Maggin, Oliver, & Wehby, 2010).
This study, however, focuses only on the form and content of teacher talk, which also has a rich supporting body of research. Since Rosenshine’s (1971) initial review of the literature of effective teaching behaviors, researchers have continued to assert that teachers’ use of instructional language is a critical component of effective instruction for all students, but particularly those who struggle to learn (Archer & Hughes, 2011; Brophy, 1988; Engelmann & Carnine, 1991, 2011). Indeed, a wealth of intervention research on explicit instruction, based on the analysis of behavior, communication, and content, has produced large effects for struggling learners (Engelmann & Carnine, 2011). Known by various names (e.g., systematic, explicit, or direct instruction), such systems emphasize that teachers must use language that is characterized by “consistent, unambiguous wording and terminology” (Archer & Hughes, 2011, p. 2). These authors also recommend that the “complexity of your speech (e.g., vocabulary, sentence structure) should depend on students’ receptive vocabulary, to reduce possible confusion” (p. 2).
Despite recommendations for practitioners to use clear communication (e.g., Chilcoat, 1987), empirical evidence supporting precise guidelines for such practices is somewhat lacking. Criticisms of research on instructional clarity have been that (a) the constructs of teacher talk and clarity have been poorly defined, (b) student outcomes have been primarily affective (e.g., perceptions of teacher effectiveness) rather than cognitive (understanding or retention of lesson content), and (c) measures have included high-inference variables that require observers to judge whether the outcomes of interest were present or absent (Titsworth, Mazer, Goodboy, Bolkan, & Myers, 2015). Some studies on instructional communication have addressed these shortcomings by examining low-inference or readily observable teaching and learning behaviors (Titsworth et al., 2015). This study examines low-inference dependent variables adopted from several areas of inquiry related to classroom communication. We have grouped these variables within the following four constructs: quantity, or the amount of teacher talk; complexity of grammatical structures; content, or the words and phrases used to convey meaning; and clarity, defined as use of linguistic features known to interfere with listening comprehension.
Amount of talk may be defined in terms of number, length, or proportion of words, utterances, or turns in a given sample. An utterance is a unit of speech that represents a complete thought (Miller & Iglesias, 2010), from a single word to a complex sentence (although sentence commonly references written language and utterance spoken language, these terms will be used interchangeably). A turn is composed of all consecutive utterances by a single speaker.
It is well established that teachers talk far more than students (Sinclair & Coulthard, 1975), particularly as grade levels increase (Sturm & Nelson, 1997). Still, the ideal amount of teacher talk is unknown (Nelson, 1985): Teachers who talk too little may fail to communicate lesson content, but teachers who talk too much may overwhelm students’ capacity for listening comprehension (Gruenewald & Pollak, 1990). Teachers are encouraged to speak concisely, which may indicate greater mastery of subject matter than a tendency to ramble. For example, Carlsen (1993) reported that new teachers tended to speak more often and for longer durations when delivering unfamiliar lesson content than when they had more complete understanding of the topic, indicating that less teacher talk may reflect greater confidence and content knowledge.
Speaking rate also may affect comprehension. Montgomery (2004) reported that children with specific language impairment (SLI) benefited from slow rates of speech (4.4 syllables per second), whereas altering input rate had no effect on sentence comprehension in children with typically developing language skills. Rate is also a component of speech known to affect comprehension in students with learning disabilities (Bradlow, Kraus, & Hayes, 2003).
The syntactic or grammatical structures of speaker utterances often are quantified in terms of the complexity of clausal structures (Masterson, Davies, & Masterson, 2006). Fragments, or utterances containing a word or phrase without a verb (e.g., Okay; To the store), are very common in spoken language but rarely found in written texts. Simple sentences consist of an independent clause without any subordinate, coordinate, or embedded clauses (e.g., She went home), even if the subject is understood (e.g., Go home). Complex sentences contain an independent clause plus any dependent clauses; therefore, they are longer and more difficult to comprehend than simple sentences and contain more embedded grammatical forms (Sturm & Nelson, 1997). Even if students have acquired the necessary vocabulary, when controlling for sentence length, grammatical complexity has been demonstrated to negatively affect oral language comprehension for children with SLI (Montgomery, Evans, & Gillam, 2009). Although logically the complexity of teacher talk increases across grade levels, studies of this phenomenon have produced conflicting results (Kean, 1967; Sturm & Nelson, 1997).
The semantic content of teacher talk references the vocabulary used to convey the meaning of a message. A mismatch between teacher talk and student comprehension may exist if teachers use words students do not understand. Quantifying teachers’ vocabulary has been approached in various ways. These include measures of lexical diversity and lexical density, which provide estimates of the ease or difficulty of comprehending a message (Bradac, Desmond, & Murdock, 1977). Researchers also have quantified difficulty of vocabulary words in terms of whether a child is highly likely (familiar, common, frequently used words) or unlikely (unfamiliar, rare, sophisticated words) to encounter a given word (Dickinson & Porsche, 2011; Horst, 2010). Prior studies have indicated that in early elementary school, teachers typically use concrete vocabulary (e.g., pencil, window) to reference the immediate context. In higher grades, teachers’ lexical references become more abstract, complex, and context-free (e.g., honesty, freedom; Lazar, Warr-Leeper, Nicholson, & Johnson, 1989).
Student learning suffers when teachers use language characterized by “poor organization, intrusion of extraneous material, vague terminology, hemming and hawing due to poor preparation, or related problems that make it difficult for them to follow the argument or understand clearly what is being said” (Brophy, 1988, p. 245). Conversely, teachers who express themselves clearly use language that is concrete, explicit, fluent, and error free. Researchers have identified several variables that increase clarity of teacher talk and have demonstrated that student achievement is negatively affected by teachers’ use of unclear language, particularly by students with low language proficiency (Ernst-Slavit & Mason, 2011). The following six variables are considered to indicate vagueness and have been shown to affect comprehension and achievement in both typically developing and struggling learners. In addition, we subcategorize these vagueness markers as examples of abstract (figurative, mental state, and ambiguous) or disfluent (mazes, errors, and abandoned utterances) oral language use.
This form of abstract language, in which words or phrases have multiple or nonliteral meanings, may be particularly problematic for students with SLI (Nippold, 1991) or EBD (Mack & Warr-Leeper, 1992). Students with limited vocabularies not only know fewer words than same-age peers but also tend to have narrower representations of the words they do know (Beck & McKeown, 2007). Students with only concrete representations of vocabulary words or little exposure to idioms and expressions may interpret figurative language literally: If given the instruction check your paper, a student may write a check mark on the paper rather than make sure the answers are correct (Lazar et al., 1989). Idioms, expressions, and colloquialisms such as play it by ear or true blue often are meaningless when taken literally and can leave students in the dark. Phrasal verbs, in which the meaning of the verb changes when paired with a preposition (e.g., hang on, hang up, hang out, hang around, hang in; cut off, cut in, cut above, cut down, cut up) are used frequently in English but may pose problems for students with language deficits. Irony and sarcasm also are forms of nonliteral language in which the meaning of the intended message is the opposite of what is actually stated. Humor in the form of word play (puns) is a demonstration of the confusion that can arise from multiple-meaning words. Teachers’ use of these linguistic forms appears to be variable across grade levels (Lazar et al., 1989).
Abstract language also includes verbs that refer to internal states, such as desire (want, need, appreciate), cognition (imagine, judge, know, remember), emotion (hope, care, doubt), and perception (appear, feel). Relative to typically developing children, children with SLI have been shown to perform poorly on tasks assessing comprehension and use of mental state verbs (Spanoudis, Natsopolous, & Panayiotou, 2007). Mental state verbs, particularly those representing beliefs (think, believe, hope), are relevant to developing a theory of mind (false belief understanding, perspective taking) and are among the most difficult words for children to acquire (Papafragou, Cassidy, & Gleitman, 2007).
When language is ambiguous, the speaker’s meaning is open to interpretation. Ambiguity is a quality of speech that detracts from the meaning the speaker attempts to convey (Snyder et al., 1991) and has a negative effect on comprehension in school-age children (Bugental, Lyon, Lin, McGrath, & Bimbela, 1999; Crossan & Olson, 1969; Hiller, Fisher, & Kress 1969). Ambiguity has been described as words or phrases that indicate the speaker lacks confidence or knowledge, as demonstrated by equivocating, approximating, hedging, or bluffing (pretty much, maybe, probably, I guess; Hiller et al., 1969); decreased specificity of content or context (the thing, some kind of, all that; Smith, 1980); ambiguous referents (e.g., a pronoun without its noun referent; Chilcoat, 1987; Masterson et al., 2006); or hesitations that indicate the speaker’s lack of confidence (Bugental et al., 1999). Ambiguity also includes cloze statements in which the teacher asks an open-ended or fill-in-the-blank type of question, expecting a specific answer when in fact a range of responses would be logical (squirrels do what?).
Even short, simple sentences using concrete and familiar vocabulary may become incomprehensible if not delivered smoothly or fluently. Disfluency occurs when speakers become lost in linguistic mazes, or patterns of speech that do not make sense semantically. Mazes include fillers (um, uh, like), false starts, hesitations, revisions, or reformulations. These are thought to reflect linguistic uncertainty (Loban, 1963) and may occur when the speaker is “expressing an idea that is abstract, complicated, or not yet fully developed” (Fiestas, Bedore, Pena, & Nagy, 2005, p. 731). In an experimental study with typically developing elementary-age students, Bugental et al. (1999) demonstrated that ambiguous and disfluent teacher talk resulted in decreased student attention and increased errors on academic tasks. This effect was stronger for older (ages 9 and 10 years) than younger (ages 7 and 8 years) children. The authors suggested that attentional disengagement likely contributes to performance deficits but that teachers may attribute the child’s academic difficulties to behavioral issues such as inattention or lack of effort.
Grammatical errors, such as incorrect subject-verb agreement or tense marking, also reflect disfluency and may negatively affect comprehension. Forney and Smith (1979) demonstrated that fourth graders performed better on a researcher-developed measure after listening to grammatically correct lectures in comparison to lectures with errors such as redundant phrases, misplaced clauses, and double negatives. Speakers may even abandon an utterance entirely. Listeners may have difficulty following the train of thought of a speaker who switches topics midsentence or fails to complete sentences.
Scholars also have suggested that adults’ language use may be an important factor in children’s problem behavior, particularly for children with EBD and low language skills (Donahue, Cole, & Hartas, 1994; Harrison et al., 1996). Although the relationship has not been investigated empirically, we derive support for this hypothesis from the literature describing comorbidity of language and behavioral disorders and teacher-student interactions. We also discuss the theoretical model that undergirds the study.
From a very early age, language and behavioral development are closely intertwined. For example, preschool-age children often have tantrums to express wants and needs (I want it; help me), then typically learn to “use their words” to replace those behaviors with verbal communication as they mature. Longitudinal studies have confirmed that co-occurrence can be identified at an early age, and difficulties are likely to persist into adulthood (Benasich, Curtiss, & Tallal, 1993; Johnson, Beitchman, & Brownlie, 2010).
Precise interactions of language, learning, and behavioral development are not well understood (Hinshaw, 1992); however, studies consistently have revealed strong associations among these constructs (Benner, Nelson, & Epstein, 2002; Bornstein, Hahn, & Suwalsky, 2013; Tomblin, Zhang, Buckwalter, & Catts, 2000). Language impairment (LI; used here as an umbrella term for below-average language skills) is thought to disrupt self-talk, or the internal dialogue that is central to developing emotion regulation (Cohen & Mendez, 2009; Fujiki, Brinton, & Clarke, 2002), and affect labeling, perspective taking (Lieberman et al., 2007), and social cognition skills such as negotiation and conflict resolution (Im-Bolter & Cohen, 2007). Lack of proficiency in these regulatory skills may contribute to the well-documented learning and behavioral difficulties of students with LI (Tomblin et al., 2000) as well as the language learning and difficulties of students with EBD (Reid, Gonzalez, Nordness, Trout, & Epstein, 2004).
Despite decades of research demonstrating the comorbidity of language and behavior problems, language deficits are commonly overlooked in children whose problem behavior is an obvious and immediate concern. In a recent meta-analysis, Hollo, Wehby, and Oliver (2014) synthesized 22 research reports in which participants were school-age children with formally identified EBD but no history of LI prior to study enrollment. Results indicated that regardless of setting, program type, or reason for evaluation, 81% of students had mild to severe LI. These difficulties likely contribute to a host of lifelong negative outcomes for students with and at risk for EBD, including school dropout, substance abuse, unemployment, and incarceration (Bradley, Doolittle, & Bartolotta, 2008; Wagner, Kutash, Duchnowski, Epstein, & Sumi, 2005).
To reverse this negative trajectory, teachers must be able to deliver instruction and classroom management procedures using language that is accessible to students. Effective instruction that increases student academic engagement has long been cited as an intervention for preventing problem behavior (Wehby, Symons, Canale, & Go, 1998). More recently, students’ communication competence has been called a “keystone skill” (Ducharme & Shecter, 2011) that provides access to a host of positive outcomes. Although teachers’ communication skills are considered an important element of effective instruction (Archer & Hughes, 2011; Engelmann & Carnine, 1991), few studies have addressed the relationship between adult language and child behavior (e.g., Williams & Forehand, 1984).
Despite the lack of empirical research examining links between teacher talk and student behavior, a classroom-based ecological model (Conroy, Sutherland, Haydon, Stormont, & Harmon, 2009) provides a cogent explanation for the hypothesized role of teacher talk in student performance. From this perspective, problem behavior may be viewed as a form of functional communication: When children are unable to manage the language demands of the classroom, they may resort to challenging behavior to gain access to preferred items, or activities, or attention, or to escape aversive environments, demands, activities, or attention (Ducharme & Shecter, 2011). Because it is likely that the majority of school-age children with or at risk for LI or EBD have functional but weak language skills (Benner et al., 2002; Hollo et al., 2014), they too may engage in challenging behavior when presented with high-level linguistic demands, such as understanding academic content, expressing thoughts and emotions, and resolving interpersonal conflicts.
According to this model, teachers and children both continually shape—and are shaped by—the environment (Sutherland & Oswald, 2005). Reinforcement is embedded within reciprocal social transactions, which in turn influence larger classroom ecologies (Conroy et al., 2009). One assumption of the ecological model is that disturbance or dysfunction is “the result of a mismatch between an individual’s skills and knowledge, and the environmental demands” (Burns, 2011, p. 134) within a given system. Because teachers may overlook LI in children with EBD (Hollo et al., 2014), teachers also may have inaccurate expectations of students’ language abilities. Consequently, this mismatch may contribute to occurrences of problem behavior (Harrison et al., 1996; Kevan, 2003). That is, teachers’ use of complex, sophisticated, abstract, or vague language may increase the difficulty of task demands as well as the probability that students will engage in antisocial escape or avoidance behaviors. In turn, teachers learn to avoid verbal interactions with students.
This cycle of coercive interactions in which escape from aversive stimuli shapes and maintains both teacher and student behaviors has been called a negative reinforcement trap. Such transactions are well documented in EBD classrooms and contribute to negative atmosphere, reduced instructional interactions, and poor student outcomes (Sutherland & Oswald, 2005; Wehby et al., 1998). Because most teacher-student interactions are verbally mediated, it is reasonable that teachers who use explicit and accessible language might be more adept at promoting student compliance and preventing problem behaviors than teachers who speak less clearly. Conversely, when communication fails, verbally mediated interventions may be ineffective or even have countertherapeutic effects (Javorsky, 1995). That is, if adults send unclear messages, children may respond in ways that increase conflict and decrease instruction (Donahue et al., 1994; Harrison et al., 1996; Kevan, 2003).
Ample evidence supports the importance of teacher talk in relation to student academic achievement. The likelihood that students with and at risk for EBD perform below grade level and have unidentified language deficits increases the importance of using clear instructional language. Effective communication may benefit both students and teachers by reducing conflicts that contribute to high rates of student dropout (Wagner et al., 2005) and teacher burnout and attrition (Billingsley, 2004) in EBD classrooms. However, despite the strong theoretical link between teacher talk and student behavior, research to date has not included students who exhibit severe, chronic, problem behavior, whether currently identified or at risk for receiving SE services. This study addresses this gap in the literature by examining teachers’ use of oral language in classrooms including students with or at risk for EBD. Our analysis begins by thoroughly describing the characteristics of the teachers and students sampled.
Several elements of teacher talk are known to impede comprehension and are hypothesized to increase problem behavior, yet the extent to which teachers engage in these verbal behaviors has not been established. The long-term goal of this novel line of research is to develop interventions to improve classroom communication to affect change in academic and behavioral outcomes. As a precursor to such experimental studies, we first examined current practices in a convenience sample of GE and SE elementary classrooms that included children exhibiting severe and chronic problem behavior. The primary objective was simply to assess whether in fact teacher talk commonly included elements that may interfere with comprehension by examining relevant teacher talk variables in the sample as a whole. Therefore, we use descriptive statistics to report observed outcomes within each of the four constructs of teacher talk in the sample as a whole.
Researchers have noted that modifying teachers’ language practices has proven to be remarkably difficult (Dickinson, 2011). To select teacher talk targets for future interventions, it is important to understand what elements of instructional language may be more or less amenable to change. It is possible that relatively flexible language features may be more malleable than highly stable or habitual ways of speaking. Therefore, our next objective was to conduct exploratory analyses that contribute to our understanding of variability and stability of teacher talk within individuals.
Our final objective was to explore between-group differences in teacher talk. Some studies have demonstrated that adults adjusted the way they spoke to accommodate the ability of listeners with known language and learning difficulties in day care (Girolametto, Hoaken, Weitzman, & van Lieshout, 2000), college (DePaulo & Coleman, 1986; Owen, 1996), and clinical (DeThorne & Channell, 2007) settings. Therefore, we applied our coding system to analyzing the form and content of instructional talk between groups of GE and SE teachers. In addition, studies have demonstrated that features of teacher talk become increasingly complex (Sturm & Nelson, 1997) or opaque (Bugental et al., 1999; Lazar et al., 1989) as students mature and presumably develop higher level language skills. Consequently, we also analyzed differences in teacher talk between lower (K–2) and upper grade (3–4) elementary classrooms.
Table 1. Sample Demographics
| GE | SE | |||
|---|---|---|---|---|
| n | n | χ2 | p | |
| Teachers (N = 28): | ||||
| Grade level: | ||||
| K–2 | 7 | 7 | ||
| 3–4 | 7 | 7 | 0 | 1.0 |
| Gender: | ||||
| Male | 3 | 3 | ||
| Female | 11 | 11 | 0 | 1.0 |
| Ethnicity: | ||||
| African American | 3 | 5 | ||
| Hispanic | 1 | 3 | ||
| White | 10 | 6 | 2.5 | .29 |
| Site: | ||||
| Tennessee | 12 | 10 | ||
| Minnesota | 2 | 4 | .85 | .36 |
| Years teaching: | ||||
| 0–9 | 8 | 6 | ||
| 10–30 | 6 | 8 | .57 | .45 |
| Degree: | ||||
| Bachelor’s | 7 | 6 | ||
| Master’s | 7 | 8 | .14 | .71 |
| Students (N = 124): | ||||
| Grade level: | ||||
| K–2 | 10 | 52 | ||
| 3–4 | 17 | 45 | 1.34 | .25 |
| Gender: | ||||
| Male | 18 | 79 | ||
| Female | 9 | 18 | 1.08 | .30 |
| Ethnicity: | ||||
| African American | 17 | 59 | ||
| Hispanic | 2 | 2 | ||
| White | 7 | 26 | ||
| Other | 1 | 2 | 1.88 | .60 |
| Site: | ||||
| Tennessee | 26 | 79 | ||
| Minnesota | 1 | 18 | 2.14 | .14 |
| SES (% FRL) | 92 | 87 |
Note. GE = general education; SE = special education; FRL = eligible for free/reduced-price lunch, proxy for socioeconomic status (SES).
Table 2. Students’ Academic and Behavioral Performance
| GE | SE | ||||
|---|---|---|---|---|---|
| n, Mean (SD) | n, Mean (SD) | t (df) | p | d | |
| WJ-III: | |||||
| Oral language | 26, 97.4 (8.2) | 86, 88.6 (12.1) | 3.13 (110) | <.001 | .85 |
| Broad reading | 23, 91.3 (1.2) | 83, 76.6 (16.4) | 4.08 (104) | <.001 | 1.08 |
| TRF: | |||||
| Total problem behavior | 27, 71.0 (9.9) | 95, 68.5 (9.5) | 1.14 (118) | .24 | .25 |
Note. GE = general education; SE = special education; WJ-III = Woodcock Johnson III Tests of Achievement; TRF = Teacher Report Form. N for reported standard scores varies due to missing data in the original study.
A subgroup of teachers from the larger study used self-monitoring to assess the frequency of praise statements and opportunities to respond during whole-class instruction. Teachers were given minicassette recorders to capture and self-evaluate their instruction during whole-group lessons: They simply turned on the recorder and let it run for the duration of instruction. No restrictions were placed on the frequency or duration of recordings, and content areas (e.g., math, language arts, science) were allowed to vary. These extant recordings provided data for our study. First, we randomly selected three recordings for each SE teacher and confirmed that each represented at least 10 consecutive minutes of whole-group teacher-led instruction (e.g., the lesson did not end or transition to independent work after a brief whole-class introduction or did not contain long passages of reading aloud from text). Through this procedure, we identified 14 SE teachers with at least three eligible language samples during whole-class lessons. Next, we repeated the procedure with GE teachers and identified a pool of 19 teachers with eligible samples. Finally, to form equal groups, we retained only the 14 GE teachers who matched the SE group on both gender and grade level. The self-contained SE classes were composed of children in multiple grades; therefore, we included two grade bands (K–2 and 3–4) for matching and for data analysis.
Research assistants (RAs) and the first author transcribed each sample according to conventions outlined by the authors of the Systematic Analysis of Language Transcripts software (SALT; Miller & Iglesias, 2010). The transcription process required multiple passes through each recording.
The first step was segmenting continuous streams of teacher talk into turns, defined as all consecutive utterances spoken by the teacher. A teacher turn ended when the teacher allowed a student to talk. If a student spoke but the teacher ignored it (e.g., an unacknowledged callout), the teacher turn did not end. Child utterances often were inaudible and therefore were not transcribed. A blank speaker line was used to represent children’s speaking turns, regardless of duration. Instances of text reading were noted in the transcript but not coded or transcribed.
Next, each teacher turn was segmented into utterances, defined as modified communication units, or c-units. Miller and Iglesias (2010) defined a c-unit as an utterance that contains an independent or main clause and all its dependent (subordinate) clauses. It represents a complete thought and cannot be divided without altering its meaning. Segmenting in this study differed from SALT conventions in one respect: If clauses with coordinating conjunctions (and, but) were not separated by a pause or change in intonation and communicated a connected thought, these were counted as 1 c-unit. This decision was made because parsing utterances as students heard them was more logical than segmenting by grammatical rules in this instance.
Table 3. Outcome Variables: Definitions and Sample Means (SD)
| Variable Name | Definition | Label | Mean | SD |
|---|---|---|---|---|
| Quantity: | ||||
| Total words | Number of total teacher words; includes mazes; omits partial words | TWa | 3,061 | 706 |
| Total utterances | Total number of utterances, including abandoned and interrupted | TUa | 522 | 106 |
| Utterance length | Mean length of utterances (c-units) in words | MLUwb | 5.81 | .87 |
| Turn length | Mean length of turns in words | TLWb | 24.69 | 7.13 |
| Turn length | Mean length of turns in utterances | TLUb | 4.19 | .86 |
| Percentage teacher talk | Ratio of child turns to teacher utterances | PTTc | .80 | .03 |
| Complexity: | ||||
| Subordination index | Ratio or total clauses to total c-units | SIc | 1.15 | .18 |
| Fragment | Proportion of utterances without a clause | FRAGc | .315 | .07 |
| Simple | Proportion of utterances with a single clause (main or dependent) | SIMPc | .412 | .06 |
| Complex | Proportion of utterances with two or more clauses | COMPc | .271 | .06 |
| Content: | ||||
| Lexical diversity | Number of different words | NDWb | 257 | 37.2 |
| Lexical density | Lexical density, defined as content words/total words | LEXc | .48 | .02 |
| Type–token ratio | Number of different words (types)/total words (tokens) | TTRc | .27 | .04 |
| Academic words | % of words matching a list of 550 cross-subject academic words | AWc | .01 | .00 |
| Frequency 1 | % of words within the 1,000 most frequently used words in English | FRE1c | .87 | .02 |
| Frequency 2 | % of words within the 1,001–2,000 most frequently used words | FRE2c | .07 | .15 |
| Clarity: | ||||
| Figurative | Nonliteral words or phrases (e.g., idioms, irony, puns, phrasal verbs) | FIGa | 76 | 21 |
| Mental state | Verbs of perception, cognition, desire/ judgment, affect/emotion | MSVa | 87 | 30 |
| Ambiguous | Unclear content (e.g., pronouns without referents, cloze statements) | AMBa | 62 | 24 |
| Errors | Incorrect use of grammar or inaccurate content | ERRa | 22 | 13 |
| Mazes | Number of hesitations, false starts, part words, or fillers | MZa | 115 | 66 |
| Abandoned | Number of utterances begun but never finished | ABa | 13 | 8 |
| Abstract | Aggregate of FIG, MSV, AMB | ABS | .44 | .09 |
| Disfluent | Aggregate of ERR, MZ, AB | DIS | .29 | .14 |
| Vague | Aggregate of all clarity codes (abstract and disfluent) | VAGUE | .74 | .20 |
Table 3 provides a brief definition of each variable. Information about how these variables were developed, defined, and quantified is provided.
Variables related to quantity were calculated automatically in SALT (Miller & Iglesias, 2010). Amount of talk included total number of words, utterances, and turns. For each 10-minute sample, rate was calculated as words per minute (WPM). Pauses or child turns were untimed; therefore, WPM was the number of words spoken by teachers only. The ratio of teacher-to-child language was computed as number of teacher utterances divided by total utterances.
Following prior studies of teacher talk in K–8 settings (e.g., Masterson et al., 2006; Sturm & Nelson, 1997), complexity was analyzed in two ways. Each utterance was coded for number of clauses. First, utterances were classified as fragments (no clause), simple (one clause), and complex (two or more clauses). Second, SALT computed a subordination index, or the ratio of clauses to c-units (utterances). Following SALT conventions for subordination index analysis, unintelligible or nonverbal utterances (vocalizations, e.g., hmm) were excluded from the analysis. Contrary to most subordination index analyses, however, incomplete and abandoned utterances were variables of interest in this study and were retained in the analysis set. Similarly, grammatically incomplete responses to another speaker (elliptical sentences) are commonly excluded from the subordination index in SALT’s databases but were of interest here.
Perhaps the most commonly used measure of lexical diversity is the type–token ratio (TTR). The numerator of the ratio is number of different words, or types, divided by the number of total words in the sample, or tokens. High TTR indicates the speaker used a wider variety of words; low TTR indicates the speaker used the same few words repeatedly. A second measure of lexical diversity, number of different words, also was computed in SALT.
Word frequency (familiar words; Dickinson & Porsche, 2011) was identified using an online program called VocabProfile (http://www.lextutor.ca) that compared samples with West’s (1953) list of the 2,000 most frequently used word families in English. VocabProfile also counted content (e.g., nouns, verbs, adjectives) and function words (e.g., auxiliaries, conjunctions) and calculated lexical density, or the proportion of content words to total words. A high-density sentence has a high ratio of lexical to grammatical terms and is more complex than a low-density sentence (Bradac et al., 1977). Finally, VocabProfile provided the number of academic words in each sample found on Coxhead’s (2000) academic word list.
Individual clarity codes are defined in Table 3. Mazes, errors, and abandoned utterances (Loban, 1963) were combined to form an index of disfluency. Similarly, mental state verbs (Spanoudis et al., 2007), figurative language (Lazar et al., 1989), and ambiguous language (Bugental et al., 1999; Crossan & Olson, 1969) were combined to form an index of abstract language. All six variables were summed and divided by the number of teacher utterances to determine proportion of utterances that contained vagueness markers.
After each RA achieved 85% reliability on transcribing and segmenting on three consecutive samples, 33% of the remaining samples were double transcribed and compared in Microsoft Word. Using the formula [agreements / (agreements + disagreements)] × 100, interobserver agreement was 91% for transcription, 89% for segmenting, 93% for the subordination index, and 88% for clarity codes.
In this nonexperimental observational study, our first analysis described naturally occurring rates of teacher talk behaviors known to impede comprehension in the sample as a whole. To compute sample means and standard deviations for each dependent variable (see Table 3), we pooled the three lessons into a single 30-minute sample for each teacher. We aggregated dependent variables according to their original metric: Count variables were summed (e.g., total number of words), and mean or proportion variables were averaged (e.g., WPM).
Next, we analyzed stability and variability of teacher talk within individuals and between groups of teachers using multilevel modeling to account for the nested structure of the data (i.e., lessons within teachers). We performed all statistical analyses using HLM 7 software (Raudenbush, Bryk, & Congdon, 2011). We fit four separate models for the outcomes of interest: total complete words, the subordination index, academic words, and vagueness, representing each of the four main constructs of quantity, complexity, content, and clarity, respectively.
Level 1 within-teacher variables were three (disaggregated) 10-minute language samples (lessons) per teacher. Graphic representations of stability or variability within individual teachers also are provided (see Fig. 1a–1d). Level 2 between-group variables were setting (n = 14, GE; n = 14, SE) and grade level (n = 14, K–2; n = 14, grades 3–4). First, to examine stability of teacher talk across observations, we ran unconditional or null models (random-effects ANOVAs with no predictor variables) for each of the dependent variables. We then calculated the intraclass correlation (ICC), or proportion of between-group to total variance, using the formula τ00/(τ00 + σ2). The ICC can be interpreted as correlations between pairs of lessons within teachers. Therefore, “high ICCs indicate stable behavior … [and] are analogous to test–retest reliability estimates” (Smolkowski & Gunn, 2012, p. 323). Finally, we entered uncentered dichotomous predictors at Level 2 (random-effects means as outcomes models) to examine differences in each of the four outcomes of general and special educators and upper and lower elementary classrooms.
As noted, GE and SE teachers were matched on gender and grade (χ2 = 0, p = 1). Chi-square tests confirmed there were no other demographic differences between GE and SE teachers or students (see Tables 1 and 2). In addition, teacher ratings of problem behavior were not significantly different between students with EBD and their at-risk counterparts: Mean scores for both groups were in the clinical range of severity (see Table 3). As expected, however, students with EBD served in SE classrooms had significantly lower oral language (t = 3.13, p < .001, d = 0.85) and reading scores (t = 4.08, p < .000, d = 1.08) than their at-risk peers served in GE classrooms.
Means and standard deviations are reported for each of the dependent variables, reported as aggregated 30-minute language samples (see Table 3). A brief summary of results follows.
The amount of talk varied considerably among teachers: The total number of words ranged from 2,067 to 5,435 (M = 3,061, SD = 706), showing that some teachers produced more than twice the number of words as others in the same amount of time. Rate in WPM ranged from 70.6 to 181.4 (M = 103, SD = 23). On average, each teacher turn consisted of four utterances of 5.81 (SD = 0.87) words, with a mean of 24.6 (SD = 8.13) words per turn.
The proportions of fragment, simple, and complex teacher utterances were 32%, 41%, and 27%, respectively. Fragments often consisted of a single word such as no, okay, a student’s name, or a phrase such as good job. Most complex utterances consisted of two to four clauses, many contained five to seven, and the longest contained 12 clauses. The mean subordination index was 1.15 (SD = 0.19) for the entire sample, which also indicates that teachers tended to use relatively little complex syntax. That is, the ratio of clauses per utterance was relatively low and more closely resembled conversational (typical range for speakers ages 8 to 44 is 1.17–1.39) than expository language (range: 1.42–1.49; Nippold, Hesketh, Duthie, & Mansfield, 2005).
On average, teachers used 257 (SD = 37) different words per 30-minute aggregated sample. In the current sample, 94% of teacher talk was composed of the 2,000 most frequently encountered words in English. Only 1% of spoken words were found on the academic word list (e.g., concentrate, specific, participate, job), whereas 5% were classified as off list (e.g., applesauce, sticker, announcement, okay).
On average, 44.2% (SD = 9.3) of teachers’ utterances contained abstract language, and 29.3% (SD = 14.2) contained disfluencies. The overall proportion of vagueness markers was 73.6%. The index for vagueness was composed primarily of maze words (31%), followed by mental state verbs (23%), figurative language (20%), and ambiguous terms (17%). Errors and abandoned utterances contributed relatively little to the total number of vagueness markers (6% and 3%, respectively).
Table 4. Within- and Between-Group Outcomes
| Fixed Effects | Random Effects | ICC | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| β | SE | t | df | p | τ00 | σ2 | df | χ2 | p | τ00/(τ00 + σ2) | ||
| Quantity: total complete words: | ||||||||||||
| Null model | γ00 | 1,020.32 | 43.67 | 23.36 | 27 | <.001 | 48,598.27 | 20,361.40 | 27 | 220.33 | <.001 | .71 |
| GE classes | γ00 | 998.57 | 43.58 | 22.91 | 26 | <.001 | ||||||
| SE classes | γ01 | 43.50 | 86.96 | .50 | 26 | .62 | 50,219.02 | 26 | 218.38 | <.001 | ||
| Grades 2–3 | γ00 | 1,236.71 | 131.20 | 9.41 | 26 | <.001 | ||||||
| Grades 3–4 | γ01 | −144.26 | 82.98 | −1.74 | 26 | .09 | 45,125.38 | 26 | 198.87 | <.001 | ||
| Complexity: SI: | ||||||||||||
| Null model | γ00 | 1.15 | 3.03 | 33.70 | 27 | <.001 | .019 | .04 | 27 | 65.27 | <.001 | .32 |
| GE classes | γ00 | 1.16 | .05 | 22.91 | 26 | <.001 | ||||||
| SE classes | γ01 | −.03 | .07 | −.48 | 26 | .64 | .021 | 26 | 64.73 | <.001 | ||
| Grades 2–3 | γ00 | 1.25 | .09 | 13.29 | 26 | <.001 | ||||||
| Grades 3–4 | γ01 | −.07 | .072 | −1.02 | 26 | .32 | .019 | 26 | 62.93 | <.001 | ||
| Content: academic words: | ||||||||||||
| Null model | γ00 | 1.25 | .10 | 12.07 | 27 | <.001 | .075 | .69 | 27 | 35.75 | .12 | .10 |
| GE classes | γ00 | 1.17 | .11 | 10.48 | 26 | <.001 | ||||||
| SE classes | γ01 | .17 | .20 | .83 | 26 | .42 | .078 | 26 | 34.89 | .11 | ||
| Grades 2–3 | γ00 | 1.03 | .29 | 3.58 | 26 | <.001 | ||||||
| Grades 3–4 | γ01 | .15 | .20 | .72 | 26 | .48 | .081 | 26 | 35.10 | .11 | ||
| Clarity: vagueness markers: | ||||||||||||
| Null model | γ00 | .88 | .04 | 21.99 | 27 | <.001 | .035 | .03 | 27 | 110.95 | <.001 | .51 |
| GE classes | γ00 | .84 | .05 | 16.57 | 26 | <.001 | ||||||
| SE classes | γ01 | .09 | .08 | 1.13 | 26 | .26 | .035 | 26 | 106.08 | <.001 | ||
| Grades 2–3 | γ00 | .83 | .11 | 7.42 | 26 | <.001 | ||||||
| Grades 3—4 | γ01 | .03 | .08 | .43 | 26 | .67 | .037 | 26 | 110.22 | <.001 | ||
Note. Intercept for the reference group = γ00; outcome for the comparison group = γ01; SE = standard error; ICC = intraclass correlation.
Results (see Table 4) revealed no statistically significant between-groups effects for any of the outcome variables. That is, teachers in GE and SE classrooms were not statistically significantly different in quantity (p = .62), complexity (p = .64), content (p = .42), or clarity (p = .26) of teacher talk. Similarly, we found no statistically significant differences on any of these structural elements of teacher talk that we measured across early or late elementary grade classrooms. Visual analysis of Figures 1a–1d shows that some individuals could be considered outliers; however, removing extreme values did not change the outcomes. Therefore, all individuals are included and reported in Table 4.
Teacher talk has been shown to affect learners’ comprehension of academic content and is thought to influence behavioral outcomes as well. To determine whether teachers frequently used features of language that have been shown to interfere with comprehension, we analyzed clarity of teacher talk in the sample as a whole. Findings indicated that 74% of all teacher utterances included at least one marker of vague, disfluent, or ambiguous language. Additionally, teachers tended to be consistent in terms of quantity, clarity, and complexity. That is, participants tended to talk the same way regardless of what they were talking about. Finally, our samples provided no evidence that teachers encoded messages differently for children according to ability level: Despite the fact that children in the SE classrooms had significantly lower reading and oral language scores, there were no significant differences in the structural language of general and special educators. Similarly, there were no differences between teachers of upper (grades 3–4) or lower elementary (K–2) students. Next, we compare our findings with those from prior research.
In 1997, Sturm and Nelson observed 15 GE classrooms and generated a list of 10 rules of teacher talk in elementary school. The first rule, that “teachers mostly talk and students mostly listen” (Sturm and Nelson, 1997, p. 259), was supported in the current study: Teacher turns accounted for approximately 80% of the recorded utterances. We could not transcribe student responses; however, anecdotal evidence indicated that student turns rarely consisted of more than a single utterance consisting of a brief, often one-word response to a teacher’s question. That is, teacher-student discourse generally conformed to the traditional IRE format reported by earlier researchers (Cazden, 2001; Mehan, 1979; Sturm & Nelson, 1997).
Contrary to Sturm and Nelson’s (1997) findings, neither the amount nor the complexity of teacher talk increased with grade level in the current sample. An early study by Kean (1967) also failed to find differences between second- and fifth-grade teachers’ grammatical complexity. Kean concluded, “It appears that the teachers in both grades are using normal adult speech patterns that are not related specifically to any differences that might separate them from their students” (p. 1). As with quantity of talk, our findings also revealed that complexity was stable within individuals. That is, teachers who tended to use short, simple sentences did so consistently, regardless of lesson content (see Fig. 1b).
Results of this study support Corrigan’s (2011) observation that “teacher-talk in many primary and elementary school classrooms is restricted to commonly known vocabulary items” (p. 752). In our sample, 94% of teacher talk consisted of the 2,000 most common words in English, with very few instances of academic vocabulary use. However, our results contradicted Corrigan’s statement that teachers tend to talk about “the here and now” (p. 752) in elementary classrooms, as 44% of teacher utterances in our sample contained abstract language. On average, 14.6% of teachers’ utterances contained figurative language. This proportion mirrors results reported by Lazar et al. (1989), in which 11.5% of teacher utterances contained at least one idiom. In both studies, teachers’ use of idioms varied among individuals but not across grade level. Use of opaque language increases the need for students to infer the meaning of teachers’ intended messages. However, inferencing and learning from context are known areas of difficulty for students with SLI (Bishop, 1992) and EBD (Mack & Warr-Leeper, 1992). Thus, nonsignificant differences between GE and SE classrooms that the current study yields may provide implications for instructional practice. That is, as our data suggest, SE teachers did not use language that was clearer than that of GE teachers. Because of the known benefit that clear, explicit instruction affords students who struggle, the lack of meaningful differences comparing educational settings is noteworthy.
Comprehension difficulties can only be compounded by use of disfluent language, which accounted for an additional 29% of teacher talk in this sample. Disfluencies consisted primarily of mazes, which occurred at a mean rate of 2.9 per minute. However, the number of mazes per minute ranged from 41 to 277 per 30-minute sample, or a rate of 1.3 to 9.2 per minute. Smith and Land (1980) reported that college students’ test performance was negatively affected by hearing only 5.1 mazes per minute. Therefore, it is highly likely that even the typically developing elementary students in our sample had difficulty understanding some teachers, even in the absence of excessive talk, complex syntax, or sophisticated semantic content.
Teachers must use language effectively to communicate academic content and facilitate interpersonal interactions. To achieve this goal, teachers also must adapt messages to meet different learners’ needs (e.g., DeThorne & Channell, 2007; Girolametto et al., 2000; Owen, 1996). However, recommended practices for modifying teacher talk may differ according to students’ linguistic strengths and needs. For example, simplifying teacher talk is recommended to maximize comprehension (Gruenewald & Pollack, 1990), but this approach is unlikely to promote language development (Dickinson, 2011). One way to address both language development and comprehension is to repeat or restate key information using multiple linguistic forms. Although it may increase the overall amount of talk, redundancy has been shown to facilitate comprehension for second-language learners (Park, 2002), GE students (Brophy, 1988; Crossan & Olson, 1969), and SE students (Lapadat, 2002). Possible benefits or drawbacks of these approaches are undetermined for students with LI and EBD.
Another way to modify adult talk is to take shorter turns: “The longer the speech during a clinician speaking turn, the denser the informational chunk, and the greater the oral literacy demand” (Roter, Erby, Larson, & Ellington, 2007, p. 1445). Reducing the amount of connected talk reduces demands on working memory and increases students’ ability to follow verbal directions (Engle, Carullo, & Collins, 1991). An effective way to reduce the length of teacher turns is to increase students’ opportunities to respond, an evidence-based strategy to increase academic engaged time and decrease problem behavior for children with EBD (Sutherland & Wehby, 2001).
Increasing students’ verbal participation also is important for language and literacy development. For example, Dickinson and Porsche (2011) reported that lower teacher-to-child speech ratios in preschool classrooms predicted higher reading comprehension, receptive vocabulary, and word recognition skills in fourth grade. Increasing wait time after instructional prompts also decreases the amount of teacher talk and increases both the quality and quantity of student responses (Tobin, 1986). Reducing the rate of speech also has been shown to benefit students with low language skills (Montgomery, 2004). Perhaps slowing down allows teachers time to encode messages thoughtfully. Interestingly, Carlsen (1993) noted that new teachers talked more when they understood relatively less about lesson content. Greenwood and Soar (1973) also found that teachers with lower ratios of student talk reported being less satisfied with teaching than those who allowed students more turns at talk. These observations support prior reflections that concise teacher talk reflects confidence and understanding (Loban, 1963).
Results of this study must be interpreted with consideration of some important limitations. First, data were collected as part of a larger study. These extant recordings were included in the current study as a convenience sample of instructional language in which teachers self-selected which lessons to record and submit. Therefore, teachers may have elected to submit only lessons they considered to be exemplary or that reflected frequent use of praise and opportunities to respond (the target behaviors of the self-monitoring intervention). Either of these scenarios could influence outcomes, such as academic vocabulary or ratio of teacher turns, and limit the extent to which findings can be generalized.
In addition, sampling procedures were not standardized: Each recording represented the first 10 minutes of large-group instruction, but the number of recordings submitted (M = 7), duration of lessons (5–30 min), and content areas (science, math, language arts) were allowed to vary. Although teachers’ language use (Gamez & Lesaux, 2015) and instructional practices (Smolkowski & Gunn, 2012) tend to be consistent across the school year, it should be noted that that the time point at which samples were obtained across the 2-year study also varied.
A final and important limitation is that sampling procedures did not permit transcription of student utterances. This omission limits our ability to draw conclusions about rate and ratio variables (WPM and percentage teacher talk, respectively), as well as the ability to analyze teacher-student interactions. In addition, the study would have been strengthened if student outcomes had included proximal measures of student outcomes related to lesson content (e.g., comprehension; academic engaged time) in addition to norm-referenced assessments of academic and behavioral performance. Proximal measures are more directly theoretically linked to the types of interactions and environments for which clear communicative messages are most important.
Despite these limitations, this study provides a platform and direction for future research. As federal funding initiatives have recently recognized, research is needed to better understand the relationship between language deficits and behavior problems (Institute of Education Sciences, 2016). The foundational work this study presents will inform this important effort as part of a programmatic line of research assisting teachers supporting students with problem behavior and low language skills.
Contrary to expectations, this study revealed that teachers’ language use was not systematically different between types of elementary classrooms. The null results reported suggest that this sample of special educators did not modify structural elements of instructional language to accommodate their students’ lower language skills. It is therefore possible that teachers were unaware of these deficits, as has been noted throughout the literature. Of course, evidence for this statement is circumstantial, and other explanations are certainly possible (e.g., the data collection or coding systems were not sensitive to functional or individualized adaptations). Still, regardless of classroom setting, teachers in this sample frequently used features of language that impede comprehension. These preliminary results suggest that intervention efforts should include both general and special educators, particularly in schools and classrooms with high populations of struggling students. If future research supports these initial findings, improving teacher-student communication could be an important factor in developing universal, targeted, and intensive interventions as stand-alone techniques or within school-wide tiered systems of support (Hollo & Chow, 2015).
To realize these goals, future studies must include standardized procedures, for example, examining differences in teacher talk during vocabulary instruction or interactive book reading and determining effects on student academic and behavioral outcomes. In addition, it will be important to analyze teacher-student verbal interactions to identify features of adult language that support adaptive student behaviors and determine whether a mismatch in adult and child language is associated with increased conflict in classroom or therapeutic settings. Future experimental studies are needed to determine whether teacher talk is amenable to change, and, most importantly, what are the effects of modifying different features of teacher talk on student academic and social behaviors. Determining whether decreasing rates of speech, increasing length of pauses, or increasing opportunities to respond is most beneficial to student performance is an important empirical question. Equally important is to understand how to increase comprehensible input without sacrificing opportunities for language development, for example, by simplifying syntactic complexity while adding repetitions, recasts, and extensions (Park, 2002). Another fruitful area of research will be to examine the use of measures other than transcription to assess teacher talk variables. For example, direct observations, coding directly from recordings, or the use of rating scales to assess quality of communication (in lieu of or in addition to frequency and duration of discrete behaviors) may prove more efficient than employing the resource-intensive, word-by-word transcription procedures used in this study.
Clear communication is critical to ensure successful teacher-student interactions for all students: Without it, even established evidence-based practices cannot be implemented as intended and therefore may not achieve expected effects. Although results of this study must be interpreted with caution, preliminary findings suggest that elementary teachers may profit from learning to improve linguistic clarity. For example, simply changing from a conversational to a clear manner of speaking will likely benefit all students, but particularly those who struggle with language and learning (Bradlow et al., 2003). Whether such changes also affect student behavior remains unknown; however, this underdeveloped area of research is a potentially strong foundation on which to build supports for students with and at risk for EBD.
Alexandra Hollo is assistant professor in the Department of Special Education at West Virginia University; Joseph H. Wehby is associate professor and chair of the Department of Special Education at Vanderbilt University. Correspondence may be sent to Alexandra Hollo at alexandra.hollo@mail.wvu.edu.
