Abstract
To address instructional challenges and poor academic outcomes of youth in juvenile correctional facilities (JCFs), we must understand how and why some teachers are effective and why students are responsive to instruction in these settings. We observed and coded teacher–student instructional interactions from 733 fifteen-minute classroom reading sessions for seven teachers and 40 students in a secure U.S. JCF school. We then applied a series of time-window sequential analytic procedures to assess connections between instructional approaches and teacher behaviors, and contingent student engagement and response behaviors. We also compared contingent probabilities for students with disabilities and students without disabilities. Across all students, our observations were characterized by larger proportions of passive student engagement. We also found a relatively low use of teacher praise. When teachers provided either directives or opportunities to respond, conditional probabilities for appropriate student responses were higher across students, particularly when directives were provided to students with disabilities. We discuss additional results and implications for research and practice.
A majority of incarcerated youth have histories of academic and behavioral problems that impact their compliance with teacher demands and engagement in academic tasks (Gagnon et al., 2022). Although available information is rather dated, evidence suggests students commonly experience academic failure, as at least 75% of incarcerated youth have failed a course, 40% have been retained, and up to 40% enter the juvenile correctional facility (JCF) with no or few high school credits (Foley, 2001). The development of reading skills is particularly problematic for incarcerated youth who perform about 4 years behind publicly schooled peers on tests of reading achievement (Krezmien & Mulcahy, 2008; Wilkerson et al., 2012). The performance deficits of incarcerated youth exist across domains including decoding, word reading, and comprehension as exemplified by achievement scores that are at least a standard deviation below their non-incarcerated peers (Davis et al., 2014; Krezmien et al., 2013).
Incarcerated youths’ frustration and failure with reading can be further compounded by emotional and behavioral difficulties (Wilkerson et al., 2012). Juvenile correctional facility teachers have noted youth behavior is one of the most significant challenges they face (Houchins et al., 2009). In addition to inappropriate classroom behavior, about 61% of incarcerated youth have a diagnosed conduct disorder (CD; Beaudry et al., 2021). CD may manifest as rule violations, as well as more serious behaviors including physical or verbal aggression, property destruction, and theft (American Psychiatric Association, 2013).
The complicated academic and behavioral difficulties of youth in JCF are notable for their chronic and severe nature, leading many to be identified with a disability. Of youth in JCF, researchers have reported that 33% are classified with a disability (Quinn et al., 2005). However, due to poor child find procedures, other researchers have asserted this percentage could be as high as 85% (National Council on Disability, 2015). This contrasts with the 15% of students with disabilities found within public schools (U.S. Department of Education, 2021). Quinn and colleagues (2005) also reported that 48% of those with a disability in JCF are classified with an emotional disturbance (ED) and 39% with a learning disability (LD).
Educating youth in JCF is made more challenging by the highly transient nature of students, as well as the fact that they are less likely to benefit from education-related protective factors (Barber & Gagnon, 2020; Krezmien et al., 2008; Sedlak & Bruce, 2010). The disruption inherent when changing schools can have a negative effect (National Research Council, 2013), as youth who enter the juvenile justice system may experience disruption in access to and continuity of education both during and after their case proceedings. Lack of coordination between the juvenile justice and educational systems may also inhibit the provision and continuity of services as youth experience numerous transitions during justice system involvement (Suitts et al., 2014).
While some (Leone et al., 2005) have contended that education in JCF may serve as a protective factor for various negative outcomes, others (e.g., Gagnon & Barber, 2010; Gatti et al., 2009) point to potential adverse effects of confinement and the low quality of educational programming in JCF. In fact, the variation in student skill levels within a single class and employment of unqualified teachers (Mason-Williams & Gagnon, 2017) often results in teacher’s reliance on ineffective or unsupported approaches, including individual student work in textbooks or “work packets” in lieu of evidence-based instructional practices (e.g., explicit instruction; Gagnon & Ross Benedick, 2021).
Teacher–Student Interactions
Despite the inherent challenges of teaching students with significant academic and behavioral deficits and disabilities, “organizing environments [that are] supportive of appropriate behavior must be a primary concern to ensure the provision of a free appropriate public education for incarcerated youth” (Krezmien et al., 2015, p. 276). From a theoretical perspective, Sameroff and colleagues’ transactional model (Sameroff, 1975; Sameroff & Chandler, 1975; Sameroff & MacKenzie, 2003) suggests that reciprocal interchanges between behavior and the environment (e.g., interactions with teachers) can be helpful to understand youth’s adjustment problems over time.
Practically, the dynamics inherent to teacher–student interactions have been shown to predict students’ academic and behavioral outcomes (Reinke et al., 2016). For example, opportunities to respond (OTRs) are an antecedent strategy wherein a teacher provides a prompt, question, or task to elicit a student’s response (Conroy et al., 2008). Stimuli can be visual (e.g., response card), verbal (e.g., teacher asking a question), or a combination of the two (Rila et al., 2019). Opportunities to respond can evoke individual or unison response from student(s). The intent of an OTR is to provide an occasion for a student to correctly respond, as well as for the teacher to monitor student progress and provide corrective feedback.
The use of OTRs has been shown to increase appropriate behavior and compliance, thereby decreasing disruptive behavior (Adamson & Lewis, 2017; Gage et al., 2018; Haydon et al., 2010). For the current study, appropriate behavior and compliance are considered within the context of three defined student behaviors. First, Appropriate Response is a verbal, physical, or gestural behavior that is appropriate to an OTR or directive. Second, Disruption is defined as a verbal, physical, or gestural behavior that is excessive or inappropriate. Third, Noncompliance includes three sub-categories of behavior: Withdrawal (a student physically removes himself from an interaction with the teacher), Inaction (no student response), and Refusal (the student maintains a physical presence, but refuses verbally or does not engage in the appropriate behavior).
The use of OTRs is also associated with students’ instructional engagement (Adamson & Lewis, 2017; Cavanaugh, 2013; Pas et al., 2015). Engagement is a critical issue for teachers in JCF and can consequently influence student academic and behavioral performance (Harbour et al., 2015; Houchins et al., 2009). Student engagement is a malleable construct that teachers can directly influence and includes emotional, cognitive, and behavioral characteristics (Wang & Holcombe, 2010). In this study, we focus on behavioral engagement, or directly observable participation in academic tasks and learning activities (e.g., raising your hand, following school rules, writing the solution to math problems, typing on a computer). We specifically consider engagement to consist of two types: (a) Active Engagement, wherein the target student is actively attending to assigned work or activity (e.g., reading out loud, writing), and (b) Passive Engagement, wherein the target student is passively attending to assigned work or activity (e.g., listening to teacher’s instruction). These forms of engagement contrast with a student being Off-Task, or engaging in any activity that has not been assigned by the teacher.
Directives are another common teacher behavior distinct from OTRs (Kern & Clemens, 2007). Specifically, teacher directives are verbal commands requiring a specific and immediate physical response by target student(s) that can be either specific (e.g., “Sit at your desk.”) or general (e.g., “Stop that.”). While directives are a necessary aspect of the classroom context, concerns exist that teachers in JCF may focus too much of their time giving directives, as is the case in research on teachers of youth with and at-risk for ED (Downs et al., 2019).
Teacher praise is one of the most studied, easily implemented, and effective approaches to promote appropriate student responding. Praise can be defined as a verbal statement of approval regarding precise behaviors exhibited by target student(s). Praise may be behavior-specific (e.g., nice job raising your hand) and non-behavior-specific praise (e.g., well done). The effective use of praise has been shown to reliably improve student task engagement (Blaze et al., 2014; Hawkins & Heflin, 2011; Hollingshead et al., 2016) and decrease disruptive behavior and improve compliance (Blaze et al., 2014; Haydon & Musti-Rao, 2011; Parsonson, 2012; Pisacreta et al., 2011). Praise may be underutilized, however, particularly with youth who exhibit behavior problems. Although research conducted in JCF schools is not available, teachers of students with and at-risk for ED in other settings have been noted to rarely use praise and more frequently employ directives, telling students “what to do” or “what not to do” (Downs et al., 2019).
Despite existing knowledge regarding effective strategies (e.g., response opportunities, praise, and directives) to increase student engagement and appropriate responding, questions remain concerning their use by JCF teachers and whether potential implementation differences exist when applied to students with disabilities and students without disabilities. Moreover, given student’s histories of academic failure, poor reading skills, behavior problems, and difficulties associated with incarceration, it is important to understand the likelihood of their engagement and appropriate responding to effective instructional techniques such as OTRs, praise, and directives while receiving instruction in JCF schools.
Observational Research
For decades, researchers (e.g., Shores & Wehby, 1999; Sutherland et al., 2002) have acknowledged the importance of observational research that focuses on teacher–student interactions. Direct observation within the authentic learning environment is a valuable measurement strategy and technology has made the collection and analysis of observed and video-recorded teacher and student behaviors even more feasible (Lewis et al., 2014). To date, there has been only one (Pytash & Kosko, 2021) direct observational study in JCF schools that focuses on classroom instructional interactions between teachers and students. However, these interactions are fundamental social processes that promote student learning, social and emotional well-being, and appropriate behavior (Luckner & Pianta, 2011). Thus, there is a need for research that focuses on teachers’ use of instructional strategies such as OTRs and praise that may increase student engagement and appropriate responding (Lewis et al., 2004; Sutherland et al., 2002). In addition, teacher’s interactions with students during reading instruction are of particular interest because reading competence is consequential to youth behavior (Servoss, 2017), achievement (Wigfield et al., 2016), future employment and earnings (Gagnon & Barber, 2014; Hanuschek et al., 2015), and to avoid recidivism (Wexler et al., 2014). Finally, given that incarcerated students with disabilities receive more negative behavioral consequences than incarcerated youth without disabilities (Krezmien et al., 2015), it is critical to understand potentially differential treatment of youth during teacher–student interactions.
Purpose of the Study
Our purpose in this study was to describe the teacher–student instructional interactions observed within the classrooms of one southeastern U.S. JCF school for adolescent males. To characterize the observed instructional interactions, we evaluated the frequencies of teachers’ use of instructional procedures, frequencies of student response behaviors, and duration of student engagement. We also investigated the associations between occurrences of instructional procedures and student response behaviors, as well as between occurrences of instructional procedures and student engagement. This study extends previous research on teacher–student interactions and behaviors in four ways. First, we observed the frequencies of teacher and student behaviors, as well as the duration of student engagement. Second, we observed the teacher–student interactions within a JCF school for adjudicated youth, a previously unstudied context for such observations. Third, we conducted specific analyses to compare interactions between students with disabilities and students without disabilities. Finally, we designed behavioral codes and coder training to improve observational accuracy and increase interobserver agreement (IOA), a concern noted with previous observational research (e.g., Shores et al., 1993).
We addressed five research questions in this study (see Table 1 for specific teacher and student behaviors):
Instructional Procedures, Student Engagement and Behavioral Responses, with Coding Definitions.
Note. OTR = opportunities to respond.
Method
Facility
We conducted this study in a privately run JCF commitment facility in a rural area of a southeastern state in the United States. The medium-level security facility utilized continuous staff supervision, secure walls, fences, and locked doors. The facility involved supervision by trained staff 24 hours per day, 7 days per week to male youth ages 12 to 18 years old with an average length of stay of 6 to 9 months. The maximum housing capacity was 150 youth, although the actual student population varied daily. The local education agency provided education at the facility.
Teachers
Throughout the study period, a total of seven teachers provided instruction to the 40 students of focus. Four of the teachers were female and three were male. Of those, five were considered highly qualified. Teachers held a bachelor’s degree (n = 4), master’s degree (n = 2), or doctoral degree (n = 1). Two teachers had been at the facility for 4 years, and one teacher had been at the facility for each of 7, 5, 2, 1.5, and 0.25 years, respectively. Five of the teachers had a general education certification in secondary English and three of those were also certified as a special education teacher. One teacher also held general education certification at the middle grade intermediate and one had temporary (undefined) certification. Each teacher provided instruction in 4, 31, 81, 414, 57, 21, and 125 of the video segments, respectively. We were not able to address differences due to teacher effectiveness because of the large imbalance among the number of video segments per teacher.
Reading Classes
Reading classes included both students with disabilities and students without disabilities. While the number of students in each class varied according to the student population at the facility, a typical class would include nine to fifteen students. The instructional approach within the facility reading classes varied, with some teachers using Read 180 (Scholastic Research, 2007) and other teachers using a combination of the New Century Learning System (New Century Education Foundation, 2010) software and teacher-led activities including individual silent reading, group oral reading, teacher-generated comprehension questions, and teacher-developed writing assignments. It is not uncommon in JCF for variation to exist regarding curriculum materials and teachers are often responsible for deciding their own curricular and instructional approaches (Gagnon, 2010; Gagnon et al., 2009). With this consideration and because our focus was on specific teacher and student behaviors, we aggregated data across the approaches taken by the teachers for reading instruction.
Students
Upon admission to the facility, we obtained informed assent and guardian consent for the study participation of each student. No students or guardians refused participation. Students participated in a reading class for 100 min per day because they did not pass the state reading assessment and had also not received a high school diploma or passed the General Educational Development (GED) test. We used a stratified random sampling procedure to select students for inclusion. Our procedure was based on the existence of a minimum of 500 min of videotaped instruction to ensure sufficient coverage for observational analyses. To mirror the 40% of youth at the facility with a disability, we randomly selected 16 students with disabilities (eight ED, two ED and other health impairment [OHI], five LD, and one LD and ED) and 24 students without disabilities. Of the 40 students sampled, 23 were African American, 14 were Caucasian, and three were Hispanic. Six students were 14 years of age, while 11 were 15 years, 12 were 16 years, and 11 were 17 years.
Procedures
Data Collection
We employed a direct observation design considered appropriate for collecting in-depth behavioral information (Atkinson & Hammersley, 1994). Over 31 months of daily reading instruction, we randomly selected and videotaped 20% of all instructional periods. Specifically, 1 day per week was randomly selected for each of the Read 180 (Scholastic Research, 2007) and New Century Learning System (New Century Education Foundation, 2010) classes to be video recorded. To acclimate students and teachers to the videotaping procedure, we incorporated cameras into the classrooms over a 2-week period in which we did not collect behavioral data.
We collected discreet 15-minute video segments of the participants’ instructional periods. We chose this length of observation to avoid instances of observational drift. We omitted segments of poor audio quality and/or lost or interrupted footage, as well as segments in which the target student did not remain in the frame for at least 90% of the observation period. We chose the first 21 segments from the resulting randomized list, resulting in a final dataset of 733 observation segments across the 40 participants. Only one participant appeared in two segments; one appeared in eight segments; and one appeared in 12 segments. Ten participants appeared in 15 to 19 segments, 26 appeared in 20 segments, and one appeared in 21 segments. Moreover, 306 video segments (41.7%) involved students with identified disabilities; the remaining 427 video segments (58.2%) involved students without identified disabilities. Since this split closely aligned with the number of participants with disabilities (n = 16; 40%) and without disabilities (n = 24; 60%), we did not further correct for differences in the number of video segments for existence of a disability in the analyses.
Coding
We randomly distributed the observational segments to nine trained undergraduate and graduate student coders. Coders used the Multi-Option Observation System for Experimental Studies (MOOSES; Tapp et al., 1995) software and Pro-Coder DV (Tapp, 2003) to view and code data for the video segments. Using MOOSES and Pro-Coder DV, coders were able to stop the video and attach codes at specific reference points in the video. We used two types of codes (i.e., event and timed-event) to describe observed occurrences within the instructional segments (see Table 1). Differences in event and timed-event codes related to the nature of each coded unit; for events, the coded unit was the exact second in which the behavior of interest was observed. That is, we coded single seconds for events in a non-exhaustive manner. Thus, we obtained and used the frequency of event-coded behaviors for analyses. Event-coded teacher instructional procedures included opportunity to respond, directive, and praise. Event-coded student responses were appropriate response, disruption, and noncompliance. For timed-events, we exhaustively recorded the difference (in seconds) between the offset and onset of each behavior. Timed-event-coded student engagements were active engagement, passive engagement, and off-task.
We trained coders by providing explicit explanations, as well as examples and non-examples of each code, both verbally and contextually within video segments. After coders reviewed the codes, they practiced coding on a training video developed by project personnel. We selected training video segments for their inclusion of both event and timed-event codes. We compared each coder’s accuracy to a master video; agreement with the master video codes had to meet a minimum criterion of at least 90% for both event and timed-event codes prior to being allowed to code videos. We held weekly follow-up group meetings to discuss any concerns and make decisions on discrepant codes.
Interobserver Agreement
To ensure the accuracy of coded videos, we created groups of video segments that were coded by the same coder in a similar timeframe and randomly selected one of five videos to be recoded (coders recoded a total of 150 segments for reliability). We followed an occurrence IOA procedure (Yoder & Symons, 2010) wherein the IOA index for event-coded teacher and student behaviors was equal to the proportion of agreements out of the sum of agreements and disagreements. We defined agreement as both coders coding the same behavior within the 5-s time window (see, for example, Shores et al., 1993; Sutherland et al., 2002). For timed-event-coded behaviors, we defined the IOA index as the proportion of total duration agreed upon by both coders divided by the total duration that both coders agreed or disagreed with. Following our initial IOA analyses, we identified 15 groups of videos with the lowest IOA and recoded all 75 video segments in those groups. Overall, agreement was relatively high with a percent agreement of 0.85 for teacher event codes, 0.81 for student event codes, and 0.85 for student timed-event codes (see Table 2).
Interobserver Agreement for Instructional Procedure, Student Response, and Student Engagement.
Analytic Strategy
The student was the unit of analysis for each research question. As we described under the “Students” section, 37 of the 40 students appeared in 15 to 21 video segments and the ratio of segments involving students with disabilities and students without disabilities was close to the ratio of students with disabilities and students without disabilities at the JCF. Thus, we did not consider additional corrections to be necessary to account for video segment differences among students.
We employed a similar analytic strategy for Research Questions 1, 2, and 3. That is, we calculated the number of behavior occurrences separately for students with disabilities and those without disabilities, as well as the overall occurrence sums for event-coded behaviors in Research Questions 1 and 2. Calculating overall occurrence sums allowed us to report frequencies for each behavior. We also calculated standardized mean difference (i.e., Cohen’s d) to identify differences between students with disabilities and those without disabilities. In addition, we calculated the duration of each engagement behavior separately for students with disabilities and those without disabilities, and the overall occurrence duration for timed-event-coded student behaviors in Research Question 3. These overall sums allowed us to report the duration of each engagement behavior. We then calculated the standardized mean difference (i.e., Cohen’s d) to identify differences between students with disabilities and those without disabilities. To indicate small, moderate, and large differences, we used cutoff scores of 0.2, 0.5, and 0.8, respectively (Cohen, 1988). Although a t-test or Mann–Whitney U test may be considered more intuitive, the small sample size and distributional differences among groups precluded multiple tests.
We employed a time-window sequential analytic procedure (Bakeman & Quera, 1995) to determine whether a target behavior occurred after the antecedent behavior within a time window for Research Questions 4 and 5. For example, we identified if a student responded appropriately following an OTR. While common indices for results of sequential analyses include Yule’s Q, Allison Likert-type’s Z, Pearson’s R, and transformed kappa, these require exhaustive coding for both antecedent and target behaviors (Yoder & Symons, 2010). Since all behaviors in Research Question 4 and teacher instructional procedures in Research Question 5 were coded non-exhaustively, it was necessary to use conditional probabilities (i.e., probability of observing the target behavior within a specific time window after observing the antecedent behavior; Bakeman & Quera, 1995). We set a 5-s time window to assess sequential relationships. Estimating the conditional probability required two calculations (i.e., frequency of an instructional procedure (1) was and (2) was not followed by a student response within 5 s). We defined the resulting probability value for behavior pairs as followed occurrences divided by the sum of followed and not followed occurrences.
We estimated conditional probabilities for Question 5 using a procedure similar to that used to answer Question 4. Insofar as we exhaustively coded rates of student engagement (i.e., duration of behavior occurrence), we were able to count the seconds an instructional procedure was accompanied or not accompanied by a student engagement behavior. Thus, we defined the probability value for a behavior pair as the result of occurrences divided by the sum of the occurrences and non-occurrences.
In these analyses, we did not focus only on the onset of engagement behaviors because it is possible for a student behavior to start before a teacher behavior. As such, we cannot strictly conclude the existence of cause–effect relationships. If a student engagement behavior was continued upon the introduction of an antecedent behavior, however, we considered it possible that a potential (but not sequential) association between behaviors existed.
We acknowledge that the low occurrence of some coded behaviors may raise a validity concern. When the occurrence of a behavior was extremely low, we concluded it was not possible to determine whether the occurrence was due to chance alone. To avoid issues related to chance occurrence, we adopted a sufficiency criterion set forth by Bakeman and Gottman (1997) in which both antecedent and target behaviors must be observed at least five times for event-coded behaviors, and at least 25 s for timed-event-coded behaviors. Although this may seem counterintuitive insofar as the unit of analysis was the student and not the video segment, this criterion is quite liberal due to the high number of video segments coded per student. Upon confirming sufficient data for the 15-min video segment, we pooled behaviors among individuals (a common practice to deal with insufficient data). This approach assumes the relationship between antecedent and target behaviors is not significantly different across video segments for a given participant (i.e., “stationary assumption”; see Yoder & Symons, 2010). Finally, if we did not observe any antecedent or target behavior at least five times (or 25 s) for a given student, we considered the conditional probability as missing and did not calculate a probability estimate.
We estimated nine total conditional probabilities (i.e., three antecedents by three response pairs) for each of the 40 students for Research Questions 4 and 5. After calculating the conditional probabilities, we ran Wilcoxon rank sum tests (Wilcoxon, 1945) to determine whether the probability of observing the target behavior, given the antecedent behavior, was statistically different among students with disabilities and students without disabilities. We also calculated the Wilcoxon effect size r to identify the magnitude of group differences. A Wilcoxon effect size r between 0.10 and 0.30 is considered a small effect, between 0.30 and 0.50 is a moderate effect, and greater than 0.50 is considered a large effect (Tomczak & Tomczak, 2014). We performed Wilcoxon rank sum tests using the “wilcox.test” function within the “stats” package in the statistical program R, version 4.0.1 (R Core Team, 2020). Wilcoxon effect size r is obtained through the “wilcox_effsize” function in the “rstatix” package (Kassambara, 2020) within the same program. In addition to Wilcoxon rank sum tests, we also reported the averaged conditional probabilities for each behavior pair for students with disabilities and students without disabilities, as well as for the overall sample.
Results
Several patterns emerged from the data. Regarding teacher instructional procedures addressed by Research Question 1, the most frequently used procedure was OTR with a total frequency of 2,620, followed by directives (f = 1,125) and praise statements (f = 255; see Table 3). Compared to students with disabilities, students without disabilities received more teacher directives (Cohen’s d = −0.31) and praise statements (Cohen’s d = −0.40), although these differences are considered small. We found no other notable differences for other instructional procedures used for students with versus without disabilities. In terms of the student response behaviors addressed in Research Question 2, we identified disruption as the most frequently observed response behavior (total frequency of 3,484), followed by appropriate responding (f = 1,955) and noncompliance (f = 437; see Table 4). Compared to students with disabilities, students without disabilities responded with noncompliance more often (Cohen’s d = −0.32), although this difference is considered small. We found no other notable differences in response behaviors for students with versus without disabilities. For Research Question 3, students exhibited passive engagement for the longest total duration (369,944 seconds), followed by off-task (197,200 seconds) and active engagement (57,455 seconds; see Table 5). Compared to students with disabilities, students without disabilities were more actively engaged (Cohen’s d = −0.30). Conversely, compared to students with disabilities, students without disabilities were less frequently passively engaged (Cohen’s d = 0.43), although these differences are considered small. We found no other notable differences in engagement behaviors for students with versus without disabilities.
Total Frequency of Teacher Instructional Procedure.
Note. SWD = students with disabilities; SND = students with no disabilities; CI = confidence interval of the mean difference among students with and without disabilities groups in teacher instructional procedures.
0.20–0.49 = Small standardized mean difference.
Total Frequency of Student Responses.
Note. SWD = students with disabilities; SND = students with no disabilities; CI = confidence interval of the mean difference among students with and without disabilities groups in student responses.
0.20–0.49 = Small standardized mean difference.
Total Duration of Student Engagement in Seconds.
Note. SWD = students with disabilities; SND = students with no disabilities; CI = confidence interval of the mean difference among students with and without disabilities groups in student engagement in seconds.
0.20–0.49 = Small standardized mean difference.
Research Question 4 addressed the associations between teacher instructional procedures and student response behaviors (see Table 6). Given an OTR, students were more likely to respond appropriately, with an overall averaged conditional probability of .32. This was followed in likelihood by disruption and noncompliance with conditional probabilities of .03 and .03, respectively. There was neither a significant group difference nor a moderate or greater effect size difference between students with disabilities and students without disabilities (see Table 7). Teacher directives were more likely to be associated with appropriate student responses, with an overall averaged conditional probability of .57, followed by noncompliance and disruption with conditional probabilities of .25 and .03, respectively (see Table 6). We observed neither a significant difference nor a moderate or greater effect size difference for students with disabilities and those without disabilities (see Table 7). Students’ appropriate responses were most highly associated with teacher praise, with an averaged conditional probability of .06, followed by disruption and noncompliance with conditional probabilities of .03 and .02, respectively (see Table 6). On average, the probability of observing noncompliance when given praise was .03 for students without disabilities versus .00 for students with disabilities, with a moderate Wilcoxon effect size (r = 0.41). We observed neither a significant difference nor a moderate or greater effect size difference for the remaining behavior pairs when comparing students with disabilities and students without disabilities.
Average Conditional Probabilities of Observing Student Response Given Teacher Instructional Procedure Among Students With and Without Disabilities.
Note. Number of students who have sufficient data for a particular behavior pair is reported in parentheses. SWD = students with disabilities; SND = students with no disabilities; CI = confidence interval of the mean conditional probability of observing student response given teacher instructional procedure differences among students with and without disabilities.
Summary Wilcoxon Rank Sum Tests for Conditional Probabilities of Observing Student Response Given Teacher Instructional Procedure.
Note. W = Wilcoxon rank sum test statistic; r = Wilcoxon effect size.
0.10 < r < 0.30: Small effect. b0.30 < r < 0.50: Moderate effect.
Regarding the association between teacher instructional procedures and student engagement explored in Research Question 5 (see Table 8), OTRs were most heavily associated with passive student engagement with an averaged conditional probability of .58, followed by active engagement and off-task with conditional probabilities of .21 and .20, respectively. We observed neither a group difference nor a moderate or greater effect size difference when comparing students with disabilities and students without disabilities (see Table 9). Given a teacher’s use of a directive, the largest associated student response was passive engagement, with an overall averaged conditional probability of .51, followed by off-task and active engagement with conditional probabilities of .31 and .14, respectively (see Table 8). We observed neither a significant difference nor a moderate or greater effect size difference for students with disabilities and students without disabilities (see Table 9). Given a teacher’s praise statement, passive engagement was the largest associated response with an average conditional probability of .62, followed by off-task and active engagement with conditional probabilities of .19 and .18, respectively (see Table 8). We observed neither a significant group difference nor a moderate or greater effect size difference for students with disabilities and students without disabilities (see Table 9).
Average Conditional Probabilities of Observing Student Engagement Given Teacher Instructional Procedure Among Students With and Without Disabilities.
Note. SWD = students with disabilities; SND = students with no disabilities; CI = confidence interval of the mean conditional probability of observing student response given teacher instructional procedure differences among students with and without disabilities.
Number of students who have sufficient data for a particular behavior pair is reported in parentheses.
Summary Wilcoxon Rank Sum Tests for Conditional Probabilities of Observing Student Engagement Given Teacher Instructional Procedure Among Students.
Note. W = Wilcoxon rank sum test statistic; r = Wilcoxon effect size.
0.10 < r < 0.30: Small effect.
Discussion
There remains an ongoing need for observational research to understand the instructional interactions in the JCF educational context. Observational research can provide meaningful insight into the instructional practices most effective for incarcerated youth (Pytash & Kosko, 2021), and prior observational research (see McKenna et al., 2015) has indicated meaningful patterns in the way that teachers interact with students with disabilities to promote subject area learning. In accordance with the need for observational research, it is necessary to also involve methodological improvements that build the capacity for understanding the nature of observed instructional interactions. In this study, we explored the complexities of teacher–student interactions during literacy instruction in a JCF for adjudicated youth. Our initial research question focused on identifying the frequency of instructional procedures used by teachers and the comparison of those frequencies for students with disabilities and students without disabilities. Teachers provided OTRs more frequently than any other instructional procedure, although the procedure was used at relatively low levels. The use of OTRs is encouraging given the strategies’ associations with increased student engagement and appropriate behavior (Adamson & Lewis, 2017; Gage et al., 2018; Haydon et al., 2010). The concern that teachers in JCF may focus primarily on giving students directives, as is the case in research on teachers of youth with and at-risk for ED (Downs et al., 2019), was not upheld. However, there was a relative overall absence of praise statements. Praise is a powerful instructional procedure (Blaze et al., 2014; Hawkins & Heflin, 2011; Hollingshead et al., 2016) that we observed to be underutilized by teachers in this study. While other researchers have noted students with disabilities, particularly those with and at-risk for ED, may receive fewer praise statements (Downs et al., 2019), we witnessed no difference for students with disabilities, as compared to students without disabilities in this study. However, it is important to note that limitations in the data did not allow for the disaggregation of students by disability classification.
Concerning Research Question 2, students exhibited instances of disruption (n = 3,484) far more than any other observed response. Given few OTRs and praise statements overall, we are not surprised students exhibited greater amounts of disruption (see Sutherland et al., 2000 for similar results). In addition, it appears that simply providing a greater number of OTRs does not guarantee appropriate student response. We did not evaluate the extent to which student responses were accurate in this study. However, an important aspect of providing OTRs is the appropriateness of the opportunity and ability of students to answer correctly (Sutherland et al., 2003). Contradictory to the high levels of disruption we observed, the frequency of noncompliance was somewhat low. This finding requires further exploration considering common student reading deficits and the need for research on accuracy of student responses (Krezmien et al., 2008). Finally, while the difference was small, it is interesting to note that students without disabilities responded more often with noncompliance. An additional inquiry could potentially identify the plausible explanations for this pattern.
Our third research question focused on the duration of student engagement behaviors. Overwhelmingly, the total duration of passive engagement was the longest, with students exhibiting this behavior almost two times more than behaviors of the next longest duration (i.e., off-task). There were small differences in effect size across disability status, as students without disabilities were generally observed to be more engaged. It is possible that this result indicates the need for additional reinforcement systems or other strategies to promote active engagement. Certainly, “finding ways to motivate and engage students in reading is an essential feature of adolescent literacy instruction” (Roberts et al., 2008, p. 67). However, specific teacher instructional approaches could also affect the type of student engagement. Additional research is needed to identify links between teacher strategies and student motivation.
Our fourth research question focused on the potential associations between teacher instructional behaviors and student responses, as well as differences in responding for students with disabilities versus students without disabilities. The overall conditional probabilities of appropriate responding following an OTR or a directive indicated a high likelihood that appropriate responding followed either procedure. Interestingly, it was also highly likely that when provided a directive, students would exhibit noncompliance. The associations between OTRs and appropriate responding in this study are similar to seminal observational research on student–teacher interactions (Shores et al., 1993; Van Acker et al., 1996) and confirm the importance of providing OTRs in JCF. It was unclear, however, whether the benefit of OTRs was maximized because our data did not indicate the extent of correct versus incorrect responses following the OTR.
Conclusions related to teacher directives are rather complicated. Due to the high probability of both appropriate responding and noncompliance, it is possible that some, but not all teachers, provided directives effectively. Specifically, Landrum et al. (2003) noted two methods for providing directives that increase appropriate responding: (a) Providing precision directives that are predictable, incorporate consequences, and give students time to comply; and (b) incorporating behavioral momentum, wherein the teacher provides a directive with a high probability of student compliance followed by a low-probability directive. We did not assess the quality or nature of the directives recorded in this study. As such, it was not possible to determine whether variations in the procedure led to appropriate responses versus noncompliance in a systematic fashion.
Our fifth research question examined associations between instructional procedures and rates of engagement and response behaviors for students with disabilities versus students without disabilities. Here again, conclusions about the conditional probabilities for praise statements are not prudent, given the relative rarity with which teachers provided them. However, we were able to note important associations between OTRs and teacher directives with student engagement. We believe a greater depth of information is needed on the qualities of actual teacher statements, as well as other instructional variables that may promote or inhibit student interaction.
Limitations and Research Implications
This study represents one of the first observational studies of teacher and student interactions in a JCF for adjudicated youth, and the first to include specific analyses of differences observed for students with disabilities and students without disabilities. While the results can inform future research and teacher practice, certain limitations should be acknowledged. Regarding teachers’ use of instructional procedures, it is encouraging to note that teachers somewhat frequently provided OTRs during instruction. Researchers (Rila et al., 2019) recommend providing four to six OTRs per minute during instruction involving new material, and students should be able to respond appropriately 80% of the time. However, variations in the use of opportunities to respond may have impacted student responses and the likelihood that they exhibited appropriate responding. We were not able to disaggregate data by instructional task or approach. Future research should consider the coding of distinct variations in the learning context, as obtaining this additional information would provide greater specificity for evaluating the procedure’s effectiveness in eliciting accurate responding and aiding student understanding.
Future studies are also needed that employ sequential analyses and identify the type and nature of response opportunities provided by teachers. Researchers (MacSuga-Gage & Simonsen, 2015) have noted, for example, that OTRs can be either teacher-directed, delivered by peers (i.e., peer tutoring), or can incorporate technology such as computer games. Moreover, the nature of OTRs may yield individual, choral, or mixed opportunities that have differential effects on responding, particularly for students with ED (Haydon et al., 2010). Finally, OTRs may uniquely affect youth given their disability symptoms and classification. Limitations in the current data set did not allow us to pursue analyses by disability classification. In addition, issues with infrequently observing certain target behaviors have been voiced by researchers (Lewis et al., 2014) and are a concern in the current study related to teacher praise statements. Future studies will require additional observations to ensure an adequate number of target behaviors for analysis and improve power for detecting conditional probabilities in the aggregate and at the individual student level.
While teachers in this study frequently provided OTRs, they rarely offered students praise. Again, limits in our data set did not allow us to investigate the use of contingent and behavior-specific praise, two practices that are highly recommended to support learning for students with disabilities (Sutherland, 2010). Future research would likely require additional observation periods to obtain sufficient data to evaluate the usage of and response to praise procedures. Other classroom practices and contexts require further consideration (or even control within future studies). For example, it is important to consider the potential effects of student length of stay at the JCF and the impact of a transient student population. In addition, teachers’ classroom management systems (or the lack thereof), as well as delineation and use of rules and routines are also likely to impact instructional interactions (Shores et al., 1993; Sutherland et al., 2009). The potential differential effects of public versus private praise (see Blaze et al., 2014) may also be a valuable area of exploration.
Given the value of observational research, it is important that future research apply this methodological approach. Moreover, future studies should extend and improve specific aspects of observation employed in this study. For example, studies are needed that explore other instructional interactions within JCF and expand evaluation to include different subject matter areas and the content’s “bidirectional influences” on interactions (Sutherland & Oswald, 2005, p. 2). Research of this type could evaluate, for example, the probability that an OTR or directive would be followed by appropriate responding, possibly initiating a cycle of positive interaction that leads to praise and additional appropriate responding. In addition, increasing the number of observations would allow for data to be disaggregated by specific disability and not just by the existence of a disability, as in the current study.
Implications for Practice
Important implications for teacher practice in JCF are apparent from the current study findings. First, it is important for teachers in JCF to increase rates of praise for students, particularly for students with disabilities. As others have noted (Downs et al., 2019), low rates of praise statements may be insufficient for students with disabilities (particularly those with and at-risk for ED) to affect positive behavioral responses. Given the high levels of student disruption, coupled with the low levels of teacher praise noted in this study, it is imperative that JCF teachers focus on maintaining high levels of praise as a mechanism to prevent disruption.
Although we were not able to assess the quality of directives in this study, the similar probabilities for appropriate responding and noncompliance indicate the potential for some directives to be more effective than others. To ensure appropriate responding, researchers have suggested precision directives that are predictable, incorporate consequences, and give students time to comply, or using behavioral momentum to establish low rates of response failure (Landrum et al., 2003).
Finally, the high likelihood that an appropriate response followed the teacher’s use of OTRs in this study suggests this procedure yields favorable results when used with youth with and without disabilities in JCF. Yet, there is considerable potential for students to respond disruptively if the OTRs are not sufficiently frequent and do not provide the student with the opportunity to respond correctly. That is, the rate of opportunity and the possibility of “getting it right” should also be considered by teachers.
Conclusion
Adjudicated youth have complex academic and behavioral needs that significantly impact their behavior and can be a challenge for JCF teachers. Observational methods for assessing teacher–student interactions, such as those used in the current study, can provide meaningful information about the nature of teacher and student instructional behaviors within this context. This foundational research is the first step to understanding these interactions, and in documenting the lack of teacher praise, as well as the complicated associations between teacher OTRs, directives, and praise, and student response behaviors and engagement. In building upon this initial knowledge, future research is sorely needed to evaluate causal relationships between the behaviors inherent to student–teacher interactions in JCF schools.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
This study was conducted as part of Project LIBERATE, funded by the Institute of Education Sciences (IES), Award #R324A080006. The opinions expressed here are the authors’ and do not necessarily represent those of IES or the U.S. Department of Education.
Ethical Approval
All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.
Informed Consent/Assent
Informed assent was received from all youth and informed consent from the guardian of each youth.