Introduction

The association between teacher-related variables (i.e., teaching behavior) and adolescents’ mental health (i.e., depressive symptoms, positive & negative affect) is critical for understanding adolescents’ mental health and how professionals serving adolescents can positively influence students. However, the majority of previous research is cross-sectional which limits the conclusions that can be drawn about if and what teaching behaviors impact adolescents’ mental health and the impact of adolescents’ mental health on how teachers act. Thus, in the present 2-wave longitudinal study high school students will be asked to report on four types of teaching behavior of a teacher they identify as most similar to themselves and their own positive (e.g., cheerful, lively) and negative affect (e.g., ashamed, gloomy) twice, four months apart.

As many as 27% of adolescents develop depressive symptoms (Bertha & Balázs 2013, Kessler et al. 2012) with the highest prevalence of depression being reported in high school age adolescents (National Institute of Mental Health 2023). Depression and depressive symptoms in adolescence become even more concerning when considering the many academic implications (e.g., decreased grades, poorer attendance, reduced homework completion; Suldo et al. 2014) that are associated with them. Beyond the associations between depressive symptoms and academic outcomes, it is critical to consider the importance and context of school in adolescents’ lives when trying to understand adolescent depression. Notably, adolescents spend most of their waking hours in school and under the supervision of teachers (Hofferth & Sandberg 2001, Larson et al. 2001). As one might expect, a number of teacher-related variables, including teacher support (Pössel et al. 2018 & Pössel et al. 2013) and teaching behavior (Pittard et al. 2015, 2017), are related to adolescent mental health. Consistent with the difference in the prevalence of depression between middle and high school age adolescents (National Institute of Mental Health 2023), multiple differences have been found between how middle and high school students experience school. As Barber & Olson (2004) point out, the size of schools is negatively associated with opportunity for decision making and teacher support, and positively associated with anonymity and teacher control. With high school enrollment sizes being on average significantly larger than middle schools’, it is not surprising that high school freshmen report experiencing less support from teachers and more depressive symptoms than they reported the previous year while they were still in middle school. One possible explanation for this could be that as students transition from middle to high school, the size of the school, average class size, and number of teachers that students have per semester increases (Akos & Galassi 2004). This means that the time a student spends with an individual teacher and their connection to this teacher is likely to decrease, which may explain the reduction in perceived teacher support. Surprisingly, teacher support significantly predicted depressive symptoms in high school but not in middle school aged adolescents (Pössel et al. 2018). Given the relevance of depressive symptoms for adolescents and of the outlined relations, it is important to examine how teacher-related variables and high school aged adolescents’ depressive symptoms are related to each other.

One way to understand depressive symptoms is through the tripartite model (Clark & Watson 1991) n which depression is conceptualized as a combination of high negative affect and low positive affect. Negative affect refers to unpleasant experiences like being ashamed or sad while positive affect refers to excitement or pleasant experiences like being energetic or cheerful (Clark & Watson 1991). Research demonstrated that measuring affect was comparable to measuring depression in adolescence, giving support to the tripartite model’s application to this age group (Gaylord-Harden et al. 2011). Empirical findings demonstrate that children and adolescents with depression report less positive affect and more negative affect than youth without a depressive disorder (Forbes et al. 2004). Further, the use of the tripartite model is a preferred method for conceptualizing adolescent psychopathology because it provides more comprehensive information about the developmental course of disorders as compared to other depression-specific conceptualizations (De Bolle & De Fruyt 2010). These findings suggest that conceptualizing depression in adolescents as a combination of high negative affect and low positive affect is appropriate and advantageous. Given the support for the tripartite model, the National Institute of Mental Health uses positive and negative affect as part of the Research Domain Criteria (Woody & Gibb 2015). Based on this, the tripartite model was used in this study to conceptualize depression in high school students and positive and negative affect were measured.

Teaching Behavior and Students’ Depressive Symptoms/Affect

As described above, authors of previous studies examined the association between students’ depressive symptoms and affect on the one hand and multiple teacher-related variables on the other hand (Pittard et al. 2015, 2017, Pössel et al. 2018), including teacher behaviors. Four types of teaching behaviors commonly discussed in the literature include instructional, organizational, socio-emotional, and negative teaching behaviors (Pianta & Hamre 2009, Pössel et al. 2013). Instructional teaching behavior refers to teachers' academically supportive actions, how they deliver instructions, provide feedback to students, and encourage students’ responsibility and autonomy. Organizational teaching behavior is defined as teachers' use of classroom management strategies, how they establish rules, provide structure, maximize the use of class time, and encourage productivity. Socio-emotional teaching behavior refers to teachers’ warmth and responsiveness in interactions with their students and how they promote feelings of belonging and acceptance in the classroom. Finally, negative teaching behavior is defined as counter-productive or unpleasant actions by the teacher that are perceived as threatening or punishing by students (Pianta & Hamre 2009, Pössel et al. 2013). These four types of teaching behavior have been found to be associated with high school students’ depressive symptoms (Pittard et al. 2015) and positive and negative affect (Pössel et al. 2013). Notably, previous research investigating the associations between teaching behavior and depressive symptoms or affect have primarily utilized cross-sectional designs and have not examined the temporal directionality of these associations. However, in order to better understand if and what teaching behaviors impact adolescents’ mental health and how adolescents’ mental health impacts teachers behaviors it is important to establish directionality. So far, in only one study was the temporal directionality of teaching behavior and affect was examined (Burton & Pössel 2022). However, the authors of this study did so in a middle school sample. Given that the prevalence of depression is the highest in high school age adolescents (National Institute of Mental Health 2023) and the above-described differences in associations between teacher support and depressive symptoms in middle and high school age adolescents (Barber & Olson 2004, Pössel et al. 2018), it is likely that the relation between teaching behavior and affect differs between high school and middle school students as well. Therefore, the transferability of Burton & Pössel’s (2022) findings from middle school to high school students is likely limited, and the current study aims to examine this expectation and fill this gap in the literature.

Instructional Teaching Behavior and Students’ Depressive Symptoms/Affect

The relation between instructional teaching behavior with depressive symptoms, and high negative affect and low positive affect, are variable across the school years. In a cross-sectional study the association between instructional teaching behavior and depressive symptoms was examined in both a middle and high school sample (Pittard et al. 2015). The results of this study showed that while there was no significant association for high school students, instructional teaching behavior was negatively associated with depressive symptoms in middle school students. Further, in a retrospective study, college freshmen reported on the teaching behavior of the one teacher whom they felt most similar to during their previous schooling. The results demonstrated a negative association between retrospective report of instructional teaching behavior and students’ current depressive symptoms (Pittard et al. 2017), similar to the above reported finding with middle school students (Pittard et al. 2015). With regard to affect, instructional teaching behavior seems to be negatively associated with negative affect in elementary (Barnard et al. 2017) and high school students (Pössel et al. 2013) and positively associated with positive affect in elementary school students but not in high school students. In middle school students, instructional teaching behavior was not associated with later negative affect, however negative affect was negatively associated with later instructional teaching behavior (Burton & Pössel 2022). In connecting this to the tripartite model (Clark & Watson 1991), the findings from the elementary sample about the temporal directionality of the relation between instructional teaching behavior and affect are consistent with Clark and Watson’s conceptualization (Barnard et al. 2017), whereas findings from the middle and high school sample were not (Burton & Pössel 2022, Pössel et al. 2013).

Organizational teaching behavior and students’ depressive symptoms/affect

The findings on the relation between organizational teaching behavior and affect are not as expected based on the tripartite model (Clark & Watson 1991). Research investigating the association between organizational teaching behavior and depressive symptoms in middle and high school students found a positive association in the middle school sample, but no significant association for high school students (Pittard et al. 2015). However, the retrospective study mentioned above investigating these associations in college freshmen found a third pattern of findings, such that organizational teaching behavior was negatively associated with current depressive symptoms (Pittard et al. 2017). Regarding affect, in high school students a negative association between organizational teaching behavior and negative affect, but no significant association with positive affect was found (Pössel et al. 2013). Further, in elementary (Barnard et al. 2017) and middle (Burton & Pössel 2022) school students, no significant association was found between organizational teaching behavior and either type of affect. Notably, these findings regarding the temporal directionality of the relation between affect and organizational teaching behavior are not consistent with the tripartite model (Clark & Watson 1991), as neither study demonstrated a combination of low positive affect and high negative affect.

Socio-emotional teaching behavior and students’ depressive symptoms/affect

Given the definition of socio-emotional teaching behaviors, one might expect that they would be positively related to positive affect. However, research findings on this relation are not as straightforward as one might predict. An examination of the association between socio-emotional teaching behavior and depressive symptoms found no significant associations for either middle or high school students (Pittard et al. 2015). In another study with middle school students examining the impact of teacher support, which shared a similar definition to socio-emotional teaching behavior in the current study, perceived teacher support was responsible for 16% of the variance in adolescents self-reported well-being (Suldo et al. 2009). College freshmen’s retrospective reports of socio-emotional teaching behavior were positively associated with current depressive symptoms (Pittard et al. 2017). Considering affect, previous findings with elementary (Barnard et al. 2017) and high school students (Pössel et al. 2013) demonstrate a positive association between socio-emotional teaching behavior and both positive and negative affect. A study with middle school students found a positive bidirectional association between socio-emotional teaching behavior and negative affect, while no relation between socio-emotional teaching behavior and positive affect in either direction was revealed (Burton & Pössel 2022). Thus, the association between socio-emotional teaching behavior and affect is not as straightforward as one might expect. The pattern of these findings regarding affect and socio-emotional teaching behavior are not consistent with the tripartite model, as the directions of the associations are all positive, rather than an inverse combination as suggested by Clark & Watson (1991) (i.e., low positive affect and high negative affect). Instead, this is consistent with Pittard et al. (2015) non-significant findings in middle and high school students.

Negative teaching behavior and students’ depressive symptoms/affect

Findings on the temporal directionality of the association between negative teaching behavior and affect are also mixed. While one empirical study found no significant association for middle school students (Pittard et al. 2015), another found that negative teaching behavior was positively associated with later negative affect for middle school students (Burton & Pössel 2022). In samples of high school students (Pittard et al. 2015) and college freshmen (Pittard et al. 2017) negative teaching behavior was also positively associated with depressive symptoms. However, another study found that in high school students, negative teaching behavior had an inverse relation with positive affect and a positive association with negative affect (Pössel et al. 2013). In elementary students, while negative teaching behavior is positively associated with negative affect, there seems to be no significant association with positive affect (Barnard et al. 2017). Findings with high school students (Pössel et al. 2013) are consistent with the tripartite model (Clark & Watson 1991) and with associations between depressive symptoms and negative teaching behavior in a high school (Pittard et al. 2015) and college sample (Pittard et al. 2017). However, many of the aforementioned studies (Pittard et al. 2015, 2017, Pössel et al. 2013) utilized a cross-sectional design to examine the associations between teaching behavior and students’ affect or depressive symptoms. Consequently, these studies were not able to investigate the temporal directionality of the associations. In order to better understand the associations between teaching behavior and students’ affect it is important to establish directionality, which in turn can aid in better identifying the target of intervention.

Temporal Directionality of Teaching Behavior and Students’ Depressive Symptoms/Affect

The importance of investigating the associations between teaching behavior and students’ affect is clear given the significant amount of time students spend with (Hofferth & Sandberg 2001, Larson et al. 2001) and the well-established associations between teacher-related variables and students’ depressive symptoms (Burton & Pössel 2022, Pössel et al. 2013) and affect (Barnard et al. 2017, Pittard et al. 2015, 2017, Pössel et al. 2013). However, previous studies examining these associations have almost exclusively used cross-sectional designs. As a result, the design of these studies did not allow for an exploration of the temporal directionality or possible bidirectional nature of the associations, pointing to the need for more studies that utilize a longitudinal design in order to better understand the associations between these variables. Unfortunately, the few longitudinal studies that do exist have primarily explored the impact of teacher-related variables on student outcomes, such as depressive symptoms (Burton & Pössel 2022, Pössel et al. 2013), while the possible impact of students’ depressive symptoms or affect on their teachers’ behaviors has received little to no attention. Studies examining the temporal directionality of these associations may be useful in identifying where and how to intervene in order to promote positive affect and reduce negative affect in adolescents.

Current Study

Although the temporal directionality of the associations between teaching behaviors and students’ affect has been explored in middle school students, no studies to date have explored these associations in a high school sample. This relation is likely different among high school students given that the prevalence of depression is highest in high school age adolescents and the above-described differences in associations between teacher support and depressive symptoms in middle and high school age adolescents. Therefore, the current study aims to fill this gap in the literature by conducting a two-wave study with high school students to investigate whether and which teaching behaviors predict positive and negative affect, or vice versa. Based on the above-described previous research, it is expected that negative affect will be negatively associated with later instructional teaching behavior. Organizational teaching behavior and affect will not be significantly associated with each other in either direction. Socio-emotional teaching behavior and negative affect will be positively and bidirectionally associated. Negative teaching behavior will be positively associated with later negative affect.

Methods

Sample

Students (N = 188) from one public high school located in a medium-sized, suburban city in the Southern United States participated in the current study. Of these students, 131 (69.7%) identified as female and 57 (30.3%) identified as male. Their ages ranged from 14–19 years, with a mean age of 16.02 years (SD = 1.23). About one quarter of the students reported that they were in 9th grade (25.0% or n = 47), 23.9% in 10th grade (n = 45), 27.1% in 11th grade (n = 51), and 23.9% in 12th grade (n = 45). A majority of the students identified their race/ethnicity as White (88.8% or n = 167), followed by multiracial (6.4% or n = 12), Asian or Pacific Islander (2.1% or n = 4), Hispanic (1.1% or n = 2), another race/ethnicity (1.1% or n = 2), and Black (0.5% or n = 1). The 188 students were nested in 38 teachers, with an average of 5 students per teacher (SD = 5.27; range = 1–28). Students were nested within teachers based on the teacher that they responded about on the Teaching Behavior Questionnaire (i.e., the one teacher they perceived to be most similar to themselves). This practice is consistent with previous literature, which indicates that teachers who are perceived to be most similar are found to be the most influential in regard to students’ future depressive symptoms (Pössel and Smith 2019). There were no exclusion criteria and students did not receive any incentive for their participation. The data collection was completed before the COVID-19 epidemic.

Measures

Teaching Behavior Questionnaire (TBQ)

The TBQ (Pössel et al. 2013) is a 37-item instrument developed to measure student perceptions of teaching behavior across four types: Instructional teaching behavior (13 items; e.g., “My teacher uses examples I understand”); Organizational teaching behavior (5 items; e.g., “My teacher makes sure I understand the classroom rules”); Socio-Emotional teaching behavior (10 items; e.g., “My teacher talks with me about non-school related problems”); and Negative teaching behavior (9 items; e.g., “My teacher threatens to punish me when I misbehave”). At each wave, students were asked to indicate the frequency of each teaching behavior for the one teacher that they perceive to be the most similar to themselves using a four-point Likert type scale (from 1 = never, to 4 = always). The TBQ scale scores were obtained by calculating the mean of the item responses for each of the four scales, with a higher score representing a higher frequency of a particular teaching behavior. In a high school sample the internal consistency reliability estimates for the TBQ scales ranged from 0.78 (Organizational teaching behavior) to 0.97 (Instructional teaching behavior; Pössel et al. 2013). The internal consistency reliability estimates for the four teaching behavior scales at wave 1 and wave 2 are presented in Table 1.

Table 1 Correlations, internal consistencies, and descriptives of TBQ and PANAS-C scales
Full size table

Positive Affect and Negative Affect Scale for Children (PANAS-C)

The PANAS-C (Laurent et al. 1999) is a student-report instrument used to measure positive affect and negative affect in youth. While this instrument was developed for children ages 9–17, studies have used the PANAS-C in research conducted with high school students (Roxas & Glenwick 2014), which include students up to an age of 19 years, as in the current study. The PANAS-C includes 30 items that are evenly distributed across two subscales, Positive Affect (15 items, e.g., “cheerful,” “lively”) and Negative Affect (15 items; e.g., “ashamed,” “gloomy”). Students indicate on a five-point Likert type scale (1 = very slightly or not at all to 5 = extremely) the extent to which they felt each item during the past few weeks. Typically, in order to calculate the Positive Affect and Negative Affect subscale scores, item responses from each subscale are summed separately. However, because the current study used Available Item Analysis to address missing data, the Positive Affect and Negative Affect subscale scores were obtained by calculating the mean of the item responses for each of the scales. High scores indicate higher levels of affect. Based on the tripartite model, a combination of low levels of positive affect and high levels of negative affect are conceptualized as depressive symptoms (Clark and Watson 1991). The internal consistency reliability estimates from the scale’s development were 0.89 for Positive Affect and 0.94 for Negative Affect (Laurent et al. 1999). The internal consistency reliability estimates for positive and negative affect at wave 1 and wave 2 are presented in Table 1.

Procedure

After gaining approval from the Institutional Review Boards of the university and the public school district, teachers at the participating high school were consented. Subsequently, informational letters and parent consents were sent to the parents of all students enrolled in one of the participating teachers’ classes. On the date of the questionnaire administration, students with parent consent were assented to participate. The participating teachers administered the questionnaires via pencil and paper forms in their classroom during school hours. Wave 1 was administered in the fall semester and wave 2 was administered in the spring semester, four months apart. Because students’ schedules change at the semester, some students participated in only wave 1, only wave 2, or both waves of the study, depending upon their classroom teachers’ participation in the study. More specifically, 269 students completed the questionnaires at wave 1 and 274 students completed the questionnaires at wave 2. Given that the purpose of the current study is to identify the temporal direction of the associations between teaching behavior and affect, only the 192 students who participated in both wave 1 and wave 2 of data collection were included in the analyses.

Analyses

Missing item-level data were examined and Available Item Analysis (AIA) was selected as a means to address missing data (Parent 2013). As results of AIA are only biased when missing data is extreme (50%; Schlomer et al. 2010), following a recommendation from Parent (2013), 75% was selected as a level of tolerance for missing data. In other words, 75% of items in each questionnaire must have been completed. In the current sample, no cases were excluded from the analyses as all participants responded to a sufficient number of items within each questionnaire. The relevant assumptions in hierarchical linear modeling HLM; i.e., assumptions of normality, the absence of outliers, and assumptions of homogeneity of variance (Garson 2013, Osborne 2013); were checked and the data were cleaned prior to conducting analyses. Doing so lead to the identification of four cases as outliers; consequently, these cases were removed prior to conducting analyses, bringing the total number of participants from 192 to 188.

In order to test for the hypothesized bidirectional associations between teaching behavior and high school students’ positive and negative affect, several two-level HLM analyses were calculated using HLM version 7.01 (Raudenbush et al. 2011). Students were nested within the teacher about whom they responded to on the TBQ. Full maximum likelihood estimation was used, for robustness (Garson 2013) and to test for homogeneity of variance (Osborne 2013). Prior to the primary analyses, intraclass correlations (ICC) were calculated to determine the proportion of variance in the dependent variables that exists between groups. The results demonstrated that the ICC was greater than 0 for each model and thus, the use of HLM was indicated (McCoach & Adelson 2010).

To examine the associations between teaching behavior at wave 1 and students’ positive and negative affect at wave 2, two separate HLMs were conducted with all four TBQ scale scores at wave 1 simultaneously entered as predictors of both PANAS-C Positive and Negative Affect scale scores at wave 2. To examine the associations between students’ positive and negative affect at wave 1 and teaching behavior at wave 2, four separate HLMs were conducted with PANAS-C Positive and Negative Affect scale scores at wave 1 simultaneously entered as predictors of each of the four TBQ scale scores at wave 2.

To test the null hypothesis for organizational teaching behavior and affect, it is important to demonstrate that the current study has enough statistical power to accurately detect an effect for this parameter of interest. Simulation studies demonstrated that at least 30 level-2 units (i.e., teachers) were needed in order to produce parameter estimates for the regression slopes and variance components at level-1 and level-2 with little bias in the samples (Maas & Hox 2005). Therefore, the current study fulfills this criterion with 38 level-2 units, reducing bias in the sample that may result in Type I or Type II error.

To determine if there were systematic differences between students who participated in only one wave of the study and those who participated in both wave 1 and wave 2, a MANOVA comparing both student groups in regard to the four teaching behaviors and positive and negative affect was calculated. Further, a χ2 test was conducted to determine whether these student groups differed on their self-reported race/ethnicity or sex. Last, linear regression were used to determine whether these student groups differed by age.

Results

Intraclass Correlations

The ICC was calculated for each of the six models in order to determine the proportion of variance in the dependent variables that exists between groups. The ICC from the unconditional model with the Instructional Teaching Behavior scale at wave 2 as the dependent variable demonstrated that 7.1% of the variability in instructional teaching behavior can be attributed to between-teacher differences, while the remainder (92.9%) can be attributed to within-teacher differences. Further, the ICC for the Organizational Teaching Behavior scale at wave 2 was 6.4%, the Socioemotional Teaching Behavior scale at wave 2 was 8.8%, and the Negative Teaching Behavior scale at wave 2 was 9.3%. Next, the ICC from the unconditional model with the Positive Affect scale at wave 2 was equal to 0.2%, suggesting that there is almost no variance between teachers for this variable. Last, the ICC from the unconditional model with the Negative Affect scale at wave 2 was equal to 12.5%. Overall, only a small portion of the variance in the outcome variables is between teachers and approximately 90% of the variance is accounted for within teachers. In other words, students’ clustered within the same teacher (e.g., students who responded about teacher A on the TBQ) shared more variance in their scores compared to students who rated different teachers (e.g., students responding about teacher A compared to students responding about teacher B). Notably, these estimates are similar to ICCs typically reported in school effects research, which range from 10–20% (McCoach & Adelson 2010).

Descriptive Analysis

Analyses were conducted to determine whether there were systematic differences between students who participated in only one wave of the study and those who participated in both wave 1 and wave 2. A MANOVA demonstrated that instructional teaching behavior at wave 2 significantly differed for students who only participated at wave 2 compared to students who participated at both waves (M only participated at wave 2 = 3.28; M for students with both waves = 3.43; F(1, 271) = 4.66, p = 0.032); all other comparisons were non-significant. A χ2 test demonstrated that sex at wave 2 significantly differed for students who only participated at wave 2 compared to students who participated at both waves (χ2 (2) = 13.20; p = 0.001; males at both waves = 60; males only participated at wave 2 = 42; females at both waves = 132; females only participated at wave 2 = 38); all other comparisons were non-significant. A linear regression was used to determine whether these student groups differed by age; the results were not significant.

Primary Analyses

Means, standard deviations, internal consistencies, and correlations among all scales are presented in Table 1. Results of the HLMs investigating the bidirectional associations between the four teaching behaviors and positive and negative affect are presented in Tables 2 and 3. Consistent with the hypotheses, the TBQ Organizational teaching behavior scale at wave 1 was not significantly associated with the PANAS-C Positive Affect scale at wave 2 (p = 0.983) or the PANAS-C Negative Affect scale at wave 2 (p = 0.172). As hypothesized, neither the PANAS-C Positive Affect scale at wave 1 (p = 0.515) nor the PANAS-C Negative Affect scale at wave 1 (p = 0.432) were significantly associated with the TBQ Organizational teaching behavior scale at wave 2. Also consistent with hypotheses, the TBQ Instructional teaching behavior scale at wave 1 was not significantly associated with the PANAS-C Positive Affect scale (p = 0.122) or PANAS-C Negative Affect scale (p = 0.172) at wave 2. Further, neither the PANAS-C Positive Affect scale at wave 1 (p = 0.746) nor the PANAS-C Negative Affect scale at wave 1 (p = 0.410) were significantly associated with the TBQ Instructional teaching behavior scale at wave 2. Also as hypothesized, the TBQ Negative teaching behavior scale at wave 1 was significantly positively associated with the PANAS-C Negative Affect Scale at wave 2 (p = 0.001). The TBQ Negative teaching behavior scale at wave 1 was not significantly associated with the PANAS-C Positive Affect Scale at wave 2 (p = 0.923). Partially consistent with the hypothesis, the PANAS-C Negative Affect Scale at wave 1 was marginally significantly negatively associated with the TBQ Negative teaching behavior scale at wave 2 (p = 0.086). The PANAS-C Positive Affect Scale at wave 1 was not significantly associated with TBQ Negative teaching behavior scale at wave 2 (p = 0.426).

Table 2 Estimated fixed effects of the TBQ scales at wave 1 on the PANAS-C scale positive and negative affect at wave 2
Full size table
Table 3 Estimated fixed effects of the PANAS-C scale positive and negative affect at Wave 1 on the TBQ scales at wave 2
Full size table

Inconsistent with hypotheses, the TBQ Socio-Emotional teaching behavior scale at wave 1 was significantly positively associated with the PANAS-C Positive Affect Scale at wave 2 (p = 0.032). The TBQ Socio-Emotional teaching behavior scale at wave 1 was not significantly associated with the PANAS-C Negative Affect Scale at wave 2 (p = 0.376). Further, the PANAS-C Positive Affect Scale at wave 1 was significantly positively associated with the TBQ Socio-Emotional teaching behavior scale at wave 2 (p = 0.016). The PANAS-C Negative Affect Scale at wave 1 was not significantly associated with TBQ Socio-Emotional teaching behavior scale at wave 2 (p = 0.278).

Discussion

Although the temporal directionality of the associations between teaching behaviors and students’ affect has been explored in middle school students (Burton & Pössel 2022), no studies to date have explored these associations in a high school sample. Previously, in only one study was the temporal directionality of the four types of teaching behavior and adolescents’ affect examined; however, sample of this study was comprised of middle school students (Burton & Pössel 2022). Given that the prevalence of depression is highest in high school age adolescents (National Institute of Mental Health 2023) and the above-described differences in associations between teacher support and depressive symptoms in middle and high school age adolescents (Barber & Olson 2004, Pössel et al. 2018), it is likely that the previous findings with middle school students cannot be generalized to high school aged adolescents. Thus, the purpose of the current study was to fill a gap in the literature by conducting a two-wave study with high school age adolescents to investigate the temporal directionality of the associations between teaching behaviors and students’ depressive symptoms conceptualized in the tripartite model as low positive and high negative affect (Clark & Watson 1991). Summarized, the results of the current study indicate that there were significant positive associations between socio-emotional teaching behavior and later positive affect, positive affect and later socio-emotional teaching behavior, and negative teaching behavior and later negative affect. There was also a marginally significant positive association between negative affect and later negative teaching behavior. All other associations were not significant.

Consistent with previous studies and as expected, in the current study organizational teaching behavior and affect were not significantly associated in either direction. Thus, organizational teaching behavior does not seem to impact students’ later affect and students’ affect does not impact later organizational teaching behavior. As predicted based on previous cross-sectional (Pössel et al. 2013) and longitudinal (Burton & Pössel 2022) studies, the association between negative teaching behavior and later negative affect was positively significant. Contrary to the hypotheses and Burton & Pössel’s (2022) longitudinal study with middle school students but consistent the findings of a cross-sectional study with high school students (Pössel et al. 2013), there was a marginally significant positive relation between negative affect and later negative teaching behaviors in the current study. Findings from the current study suggest that perhaps increases in students’ negative affect could be a greater source of frustration for teachers, possibly evoking more negative teaching behaviors. As adolescents transition from middle to high school, the average class size and the number of teachers that students have per semester increase (Akos & Galassi 2004). Therefore, high school teachers may have more difficulty managing the negative affect of their students because of their increased class sizes resulting in more negative teaching behaviors.

Unexpectedly and contrary to some previous studies (Pittard et al. 2015, 2017), a positive bidirectional relation between socio-emotional teaching behavior and positive affect was revealed. Previous research indicated that supportive teaching behaviors, similar to those measured as socio-emotional teaching behavior, were related to depressive symptoms (i.e., high negative affect, low positive affect) for adolescents with low levels of stressful life events (Pössel et al. 2013). However, in adolescents with average to high levels of stressful life events, socio-emotional teaching behavior was associated with fewer reported depressive symptoms over time (Pössel et al. 2013). While stressful life events were not measured in the current study, it is possible that adolescents in this study experienced average to high levels of stress meaning that teacher’s socio-emotional teaching behavior could have been needed and was beneficial, leading to more reported positive affect.

Also different from expectations, there were no significant associations between negative affect and later instructional teaching behavior or socio-emotional teaching behavior in either direction. One possible explanation for this unexpected finding may be related to the internal nature of constructs such as affect and depressive symptoms. Previous findings indicate that teachers tend to be good informants for externalizing behaviors, such as attention and hyperactivity, but may not be as good of informants for internalizing behaviors such as depressive symptoms (Barry et al. 2013). In turn, it may be that teachers are not as impacted by adolescents’ affect or depressive symptoms because they are not as easily noticeable by teachers compared to externalizing behaviors. Specifically, it may be that teachers are more aware of adolescents externalizing behaviors, as these behaviors are more likely to require the teacher to redirect a student and take time away from instruction. Further, externalizing problems may be a greater source of frustration for teachers, possibly evoking more negative teaching behaviors and making it more difficult to form a positive relationship and consistently respond to adolescents with warmth (i.e., socio-emotional teaching behavior). Although this is one possible explanation as to why this study did not find the predicted associations for affect predicting later teaching behavior, a previous study (Burton & Pössel 2022) investigating the temporal direction of these associations found that adolescents’ affect or depressive symptoms are associated with later teaching behavior making this explanation unlikely. Nevertheless, researchers should consider examining the associations between teaching behavior and adolescents’ internalizing and externalizing problems to determine the relative percentages of variance explained in teaching behavior by each construct.

Another possible explanation for why proposed associations between negative affect and later instructional teaching behavior or socio-emotional teaching behavior and negative affect in either direction were not found may be related to differences in sample characteristics. Two of five participating middle schools in Burton and Pössel’s (2022) study were private Catholic/parochial schools, whereas adolescents in the current study were recruited from one public high school. Oftentimes in private parochial schools, students have the same teacher for more than one subject during the same semester, and sometimes even have the same teacher across grades 6, 7 and 8. In contrast, public high school students typically have a particular teacher for just one subject, and even switch to a new teacher for that subject every semester (Akos & Galassi 2004). Therefore, the student-teacher relationship and experiences middle school students in private parochial schools have with their teacher may be quite different from students in public high schools in terms of duration and frequency. Based on the above, it is possible that adolescents in Burton & Pössel’s (2022) study may have been under the supervision of the teacher they rated for the complete duration of the study (both wave 1 and wave 2 of data collection) while the high school students in the current study may have only encountered the teacher they rated for one semester. Further, high school students in the current study were asked to rate the one teacher that they perceived to be the most similar to themselves and as a result, it is not known whether adolescents in the current study rated a teacher that they currently have or a teacher from a previous school year. It is feasible to assume that the impact of teaching behavior is greater while adolescents are under a teacher’s supervision, findings which are well-established by cross-sectional studies (Barnard et al. 2017, Pittard et al. 2015, 21), but not long after the adolescent has been removed from the teacher’s supervision. However, the previous longitudinal study with middle-school students investigating the temporal direction of these associations found enduring effects (Burton & Pössel 2022). In order to test the aforementioned proposals as to why the predicted associations were not found, researchers should replicate the current study using longitudinal designs with three or more time points over a longer time period. This would allow researchers to determine how long lasting the impact of teaching behavior is on students’ mental health, and vice versa. Researchers may also consider examining whether the length of time a student spent under their teacher’s supervision moderates the examined relations.

The results of the current study should be interpreted with a consideration of the study’s strengths and limitations. Notably, the current study addresses a gap in the literature by examining the temporal direction of the associations between the four types of teaching behavior and affect in high school age adolescents. Previous studies examining similar associations in high school students have predominantly used cross-sectional designs to investigate which teaching behaviors predict adolescents’ affect (Pittard et al. 2015, 2017, Pössel et al. 2013) or depressive symptoms (Pössel et al. 2013). Thus, on one hand, the design of the current study can be considered a strength in that it allows for an examination of the bidirectional associations between teaching behaviors and adolescents affect, filling a gap in the literature. However, a longitudinal analysis with three or more time points would allow for an investigation into how enduring the associations are after an adolescent has been removed from their teacher’s supervision and would allow for the examination of possible non-linear trajectories. Thus, it is recommended that a longitudinal design is utilized in future studies to investigate whether the associations between teaching behavior and affect remain significant after the adolescent is no longer under the supervision of a particular teacher.

Another limitation of the current study is the generalizability of the findings given the composition of the sample. More specifically, participants included adolescents from one public high school located in a small, suburban city in the Southern United States with almost 90% of the students in this sample identifying as White and approximately 70% identifying as female. As such, it is unclear whether findings from this study are generalizable to adolescents of other racial/ethnic groups, male adolescents, those in other geographic locations, and adolescents in different school settings (e.g., private or parochial schools, elementary and middle schools). Therefore, authors of future studies may wish to build on the results of the current study by including samples that are diverse in both adolescent characteristics and school settings.

In addition to sample characteristics, another limitation related to the current study’s sample is that about 45% of students only participated in one wave of the data collection. In large part, this was because student participation was dependent upon whether the student’s teacher agreed to participate in the study, as teachers who agreed to participate administered the questionnaires to students in their classroom. Consequently, because students change teachers and classes from one semester to the next, some students participated in only wave 1, only wave 2, or both waves of the study, depending upon their classroom teachers’ participation in the study, which resulted in this loss of data. In order to address missing data, the use of multiple imputation and full information maximum likelihood (FIML) were considered. The authors of a simulation study to investigate best practice for managing missing data examined multiple imputation and FIML when data are missing at 10, 20, and 50% (Schlomer et al. 2010). The researchers found that when the amount of data missing is severe (i.e., 50%), these estimation methods introduce enough bias to be of concern. Given that about 45% of participants are missing all item-level data for one time point, it was determined that the use of multiple imputation or FIML may result in biased estimates and so these methods were not used for the current study. Further, when multiple imputation or FIML are used, the HLM software is unable to conduct the test of homogeneity of level-1 variance, and therefore it would not be possible to address the assumption of homogeneity of variance.

The final limitation of the present study is the exclusive use of student-report measures. When the same person provides information for all measured variables common method variance can result (Podsakoff et al. 2003) which might inflate the existing associations. In addition, students’ rating of their teachers’ behavior might be impacted by their affect which could cause the associations to be inflated as well. To avoid both reasons for a possible inflation of the examined associations, authors of future studies might find it beneficial to include teacher-report to measure teaching behaviors. However, it should be noted that student-report is the best predictor for student outcomes and teacher-reports of their own behavior has been shown to demonstrate bias (Douglas 2009). Thus, the inclusion of teacher reports of teaching behavior might be less beneficial than it might appear.

Conclusion

The understanding of the temporal directionality of the associations between teaching behaviors and students’ affect in high school students was limited due to a lack of studies. Thus, the purpose of the current two-wave longitudinal study was to fill this research gap. The current study’s findings suggest that socio-emotional teaching behavior and high school students’ positive affect influence each other, and that negative teaching behavior is associated with later negative affect while the association between negative affect and later negative teaching behavior is less clear. These findings clarify the impact that teachers have on high school students’ affect and following the tripartite model, depressive symptoms (Clark & Watson 1991). Further, it demonstrates the impact that students’ mental health has on teaching behavior, an area of study that has received limited attention.