Autonomy support from support staff in higher education and students' academic engagement and psychological well-being

Abstract

The present research has examined the role of autonomy support provided by support staff in higher education, specifically teaching assistants (TAs), at a university in Japan in Study 1 and political instructors (PIs), at universities in China in Study 2. Self-determination theory was used to derive a model in which autonomy support related to the satisfaction of three core psychological needs, which in turn led to academic engagement and well-being. In Study 1, university students in Japan completed measures of autonomy support, need satisfactions, timely engagement, satisfaction with campus life, and depressive symptoms on two occasions. Results show that autonomy support provided by TAs was positively related to students’ timely engagement and satisfaction with campus life, and negatively related to students’ depressive symptoms. These relationships were mediated by need satisfactions, except for satisfaction with campus life. Results of Study 2 show that autonomy support from PIs was positively related to students' satisfaction with campus life but not mediated by need satisfaction. Implications of autonomy support from non-significant others in educational settings are discussed.

Keywords:

• Autonomy support
• need satisfaction
• support staff
• engagement
• well-being

Given recent decades’ rapid economic growth and dramatic increase in the number of students in tertiary education in Asia, higher education is shifting from massification to near-universal access (Huang, 2012). The number of students per lecturer in countries, such as Japan, Korea, China, Singapore, and Malaysia has been increasing constantly; as a consequence, the widespread provision of one-on-one personal support by teachers is nearly impossible to achieve, which leads to ineffective learning and low levels of student well-being (Brinton et al., 2015). This concern is particularly acute in Japan, where the problem has spread beyond higher education and has manifested itself in all educational settings, including phases of compulsory education. Teachers’ average working hours in Japan were, at the time of data collection, the highest among all 48 countries or regions examined across the world, and the average class size in Japan was the second-largest, including those in more depopulated areas (OECD, 2019).

Introducing a system that supports both students and teachers—such as via teaching assistants (TAs)—is one way to ease this problem. Almost every university has systems of support staff that are directly involved with students’ campus lives. Yet, research that focuses on the kind of influence such support staff has on students remains sparse. For instance, Philipp et al. (2016) found that compared to students without TAs, students with TAs in a chemistry class had significantly higher GPAs on their final exam scores.

The research above indicates that support staff (such as TAs) might play an important role in students’ campus lives, but the type of teaching style of the support staff that best benefits students remain unclear. One central tenet of self-determination theory (SDT; Deci & Ryan, 2000; Ryan & Deci, 2017) is that social contexts predict the quality of individuals’ motivation, performance, and the mental health of individuals who operate within them. The theory uses the concept of autonomy support vs. control to characterise the quality of social environments, hypothesising that autonomy-supportive social contexts tend to facilitate self-determined motivation, healthy development, and optimal functioning (Reeve et al., 2004). The present research aims to explore the extent to which autonomy support from support staff at universities relates to students’ academic engagement and psychological well-being, mainly via a focus on TAs in Japan.

TAs in Japan are graduate students who assist teachers with instructional responsibilities—including supporting students in class, tracking attendance, checking homework, and so forth—depending upon the class. Every university in Japan has a TA system to support students in their academics, yet no research has reported how they relate to students' learning outcomes. There is a unique type of support staff available at universities in China: political instructors (PIs). In contrast to TAs, PIs are not particularly involved with students in the academic domain; instead, PIs were originally involved in guiding the development of students’ political ideology. Recently, however, they have been assigned additional duties—such as providing information about campus life or offering counselling when students face interpersonal problems on campus (CCP Central Committee General Office and PRC State Council, 2015). Similar to the TA system in Japan, the PI system is extremely common in universities in China, but is less known in other countries and has not yet been studied.

Effects of autonomy support

Autonomy support in educational settings is a coherent cluster of teacher-provided instructional behaviours that collectively communicate an interpersonal tone of support and understanding to students (Reeve, 2009). Deci et al. (1994) first found that three interpersonal conditions are necessary for individuals to feel that their autonomy is supported: the provision of meaningful rationale (e.g. verbal explanations that help others understand why self-regulation of activity would have personal utility), acknowledgement of negative feelings (e.g. tension-alleviating acknowledgment that one’s request to others clashes with their personal inclinations and that their feelings of conflict are legitimate), and the use of non-controlling language (e.g. communication that minimises pressure; absence of the terms ‘should’, ‘must’, and ‘have to’, so as to convey a sense of choice and flexibility via phrasing). Autonomy support has been shown to contribute to positive outcomes with regard to learning activities, such as better academic performance and higher autonomous motivation to learn. Furthermore, such positive benefits of a teacher-created autonomy-supportive environment have been found across lines of race, ethnicity, and class by numerous empirical studies over the past decades (Jang et al., 2009; Nalipay et al., 2020).

SDT alleges that three basic psychological needs exist: the need for autonomy, competence, and relatedness. Autonomy refers to the experience of volition and willingness, whereby people experience a sense of integrity and satisfaction when their actions, thoughts, and feelings are self-endorsed; competence is the experience of effectiveness and mastery—it becomes satisfied as one capably engages in activities and experiences opportunities for using and extending skills and expertise; relatedness denotes the experience of warmth, bonding, and care, which is satisfied by connecting and feeling significant to others (Ryan & Deci, 2017, p. 86). These three needs are universal, and the satisfaction of these needs is essential for optimal development, integrity, and well-being. Conversely, the frustration of not attaining these needs leads to deficiencies and degradation in psychological integrity and social development, affecting both one’s wellness and vitality (Ryan & Deci, 2017, p. 86).

When there is support for autonomy, people are also more capable of seeking out and finding satisfaction relating to both competence and relatedness. That is, when teachers, parents, or significant others are autonomy-supportive, they are responsive to the perspectives and important issues faced by the individuals they guide, lead, or care for. This, in turn, will facilitate the satisfaction of multiple needs (Deci & Ryan, 2000). As such, SDT assumes that autonomy support is associated with adaptive outcomes in the educational domain, and these relationships are mediated by the satisfaction of basic psychological needs. For instance, Kaplan (2018) found that teachers' control negatively predicts students' satisfaction of their needs; furthermore, teachers' autonomy support positively predicts students' need satisfactions, which in turn positively affects students' autonomous motivation in learning. Autonomous motivation, in turn, predicts positive emotions and engagement in learning. A 5-week intervention study performed by Wang et al. (2016) found that, after the intervention, students in groups whose teachers were trained in the autonomy-support program displayed a significant improvement in their perceived autonomy support from teachers, need satisfactions, and grades when compared to pre-intervention measures. Importantly, they also found that students in the autonomy-supportive groups were more likely to use self-regulatory strategies that help students focus on planning, monitoring, and controlling their cognition than those students in the control condition. They argue that an autonomy-supportive environment enhances autonomous motivation, which leads students' autonomous self-regulation, that is, initiating, regulating, and organising their behaviours to achieve their academic goals.

Autonomy support from teachers not only affects students' learning processes and outcomes but has also been proven to have a positive influence on students' psychological well-being. Due to rapid changes in physical growth and psychological and social development, adolescence is considered to be a transitional period wherein individuals are highly susceptible to maladjustments, such as low life satisfaction, anxiety, and even depression (Hein et al., 2015). A longitudinal study by Yu et al. (2016) found that teachers’ autonomy support of seventh-graders in autumn increased the basic psychological need satisfaction of the learners in the spring, subsequently boosting school engagement and decreasing anxiety and depression of the students when they reached the eighth grade. More recently, Froiland (2021) observed that, when students set intrinsic learning goals in their own words, their basic psychological needs were being fulfilled, leading to various aspects of well-being—such as a greater intrinsic motivation to learn, fewer symptoms of depression, and greater happiness. To summarise, autonomy support has positive effects beyond learning and leads individuals to be better psychologically adapted in life.

Although the results of previous research on autonomy support have been consistent over time, most studies have a focus on the effects of autonomy support from sources of primary influence in educational settings, such as teachers and parents. However, based on the rationale of SDT, autonomy support should remain effective even when provided by individuals with lesser influence/contact. However, no research has, thus far, focussed on the influences of autonomy support from individuals who are not primary influencers—such as neighbours in communities or support staff in schools. Elucidating the effects of autonomy support provided by support staff could be a missing element in the effective structuring of a learning environment aimed at improving both students’ academic engagement and their psychological well-being. In the present research, we aim to reveal the relationship between autonomy support provided by support staff and students' learning and well-being outcomes by taking a close look at TAs in the educational context. In terms of mitigating ineffective learning and the poor well-being of students as caused by a lack of direct support from teachers, emphasising the autonomy support of these support staff in an education context should benefit students and be useful throughout higher education or any educational system with similar issues.

Present study

The main purpose of this research is to reveal the relationship between autonomy support from support staff and university students’ academic engagement and psychological well-being. First, we focus on students' engagement in learning, and specifically with respect to the behaviour of engaging in tasks in a timely manner with approach motivation—such as gaining an advantage in completing the task—referred to as a ‘timely engagement’ in the present study, which is the most desirable autonomously self-regulated behaviour for university students to be supported and nurtured (Strunk et al., 2013). We focus on this issue because researchers have consistently identified that between 40 and 60% of students procrastinate to a moderate or high degree (Onwuegbuzie, 2004). Furthermore, there is evidence that procrastination results in detrimental academic performance and life satisfaction (Balkis, 2013). Moreover, previous research has found that autonomy support from teachers has a positive impact on students' autonomous self-regulation in learning and preventing procrastination as a result (Won & Shirley, 2018). In the present research, we verified whether autonomy support from support staff could also positively relate to students' timely engagement in learning activities and whether this relationship was mediated by need satisfactions.

Secondly, we focus on psychological well-being specifically with respect to students' campus satisfaction and potential depressive symptoms. Campus satisfaction in the present research refers to students' satisfaction with the overall university environment and their interpersonal relationships. Students' satisfaction with their lives and schools are considered to be an important part of academic success as they form the basis of academic development and general psychosocial adjustment (Wach et al., 2016). Typical depressive symptoms—low self-esteem, loss of interest in normally enjoyable activities, low energy, and pain without a clear cause frequently (APA, 2013)—are mainly caused by stressful experiences within academic and interpersonal domains, and have been identified as major influences on a students’ academic performance in both Japan and China (Li & Chen, 2004; Steptoe et al., 2007). Tsukahara (2011) found that depressive symptoms among university students in Japan increased continuously over a decade. A meta-analysis by Lei et al. (2016) revealed an overall prevalence of depression among Chinese university students (23.8%), which is exceedingly high. These results suggest that it is imperative to pay more attention to the development of appropriate mental healthcare and support strategies for university students in both Japan and China. Autonomy support from significant others, such as teachers and parents have been shown to be positively related to campus life satisfaction and negatively related to depressive symptoms (Gutiérrez & Tomás, 2019; Yu et al., 2016). In the present research, we examine the relationships between autonomy support received from support staff and university students' campus satisfaction and general depressive symptoms, and whether these relationships were mediated by need satisfactions.

Study 1

Based on results from previous studies assessing the effects of autonomy support, we first hypothesised that autonomy support from TAs would positively relate to university students’ timely engagement in academic engagement, satisfaction with the university campus, and negatively related to depressive symptoms. Second, we predicted that these relationships would be mediated by the satisfaction of needs. Additionally, rather than merely focussing on the relationships between autonomy support and outcomes at a single point in time, we determined whether changes in the autonomy support provided by TAs related to changes in all dependent variables.

Methods

Participants and procedure

In Study 1, 290 undergraduate students (154 females and 136 males, mean years in university = 2.56) who were taking an optional psychology class at a private university in Japan participated in our research. Course credit was given as a reward for participation in the study. Informed consent was obtained from all participants, and the research was authorised by the faculty of psychology's ethics review committee of the university. Data were collected using a questionnaire survey at two-time points: the first week of the new semester in April, and during the mid-term exam week in June 2015.

The data of participants who did not complete both questionnaires were excluded, and 199 participants (113 females and 86 males, mean years in university = 2.57) were included in the final analysis. We had rather large attrition in Study 1 for three reasons: first, our participants were not obligated to participate in our study at both Time 1 and Time 2. They had the right to choose whether to complete the questionnaires on both occasions or just once. Second, at Time 2, there were other research groups recruiting participants from the same class. Therefore, we considered that students might not have the time or energy to participate in all studies. Third, in Study 1, we recruited our participants from an optional psychology class that did not take attendance, which is why students who participated in our study at Time 1 might have been absent from the class at Time 2. Regarding the large attrition, t-test results indicated no significant differences in the means of measures between the data we used in the final model and the data excluded from the original sample. Moreover, the result of Little's MCAR tests showed that the mechanisms of whole-wave missing data in Study 1 were missing completely at random (χ² = 2019.445, df = 1918, p = .053).

Measures

Perceived autonomy support, need satisfaction, timely engagement, satisfaction with campus life, and depressive symptoms were measured using self-report questionnaires. All variables were measured at both Time 1 and Time 2. The instruments used to measure perceived autonomy support and timely engagement were originally developed in English; therefore, an English-to-Japanese translation was performed by a graduate student who is fluent in both languages. The translations were back-translated by an undergraduate student who is a native English speaker and fluent in Japanese and were checked by a professor in the psychology faculty.

Perceived autonomy support

Students’ perception of TAs’ autonomy-supportive behaviours was assessed using the 15-item ‘Learning Climate Questionnaire’ (Williams & Deci, 1996) (e.g. ‘I feel that my TA gives me choices and options’) which is widely used in research of autonomy support (Black & Deci, 2000; Núñez et al., 2012). Responses to the questions are given using a Likert-type scale ranging from 1 (strongly disagree) to 7 (strongly agree). In this study, we instructed participants to limit their answers to the TA that they contacted the most in the (then) present semester.

Need satisfactions

Students’ basic need satisfactions were measured using the 12-item Japanese version of the ‘Basic Psychological Need satisfactions Scale—Academic Domain’ (Nishimura & Sakurai, 2013), with the scale adapted from the ‘Basic Psychological Needs Scale’ (Deci & Ryan, 2000). The scale consists of three subscales which measure need satisfactions in academic settings, and specifically: satisfaction of the need for autonomy measures concerning whether participants have feelings of control over their own learning activities (e.g. ‘I can decide what I need to learn without others reminding me’); satisfaction concerning the need for competence measures whereby participants feel confident in their learning outcomes (e.g. ‘I am confident that I can get a good grade for the test’); and satisfaction concerning the need for relatedness measures in which participants feel belongingness and connection to others at the university (e.g. ‘I have teachers or friends at school that I can to talk to if I have questions about studying’). Responses are provided using a Likert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree).

Timely engagement

Students’ timely engagement was assessed using a subscale of a 2 × 2 model of time-related academic behaviour (Strunk et al., 2013). This subscale contains six items that measure students’ timely engagement/behaviours towards academic tasks (e.g. ‘I work further ahead of the deadline, at a slower pace, because it helps me perform better’). The responses were codified using a Likert-type scale ranging from 1 (strongly disagree) to 7 (strongly agree).

Campus satisfaction

Students’ satisfaction with campus life was assessed using the nine-item ‘Campus Satisfaction Scale’ (Makino & Mori, 2002), which assesses students’ satisfaction with lectures (e.g. lecture content or curriculum), interpersonal relationships (e.g. their relationships with teachers or classmates), and the overall physical environment (e.g. classrooms and the canteen). Responses were codified via a Likert-type scale ranging from 1 (not satisfied at all) to 5 (very much satisfied).

Depressive symptoms

Students’ depressive symptoms were measured with the 20-item ‘Self-Rating Depression Scale’ (Zung, 1965), which has been used in several studies in Japan and China (e.g. Kera et al., 2015; Wang, 2008). Respondents rate each item in terms of how they felt during the previous week (e.g. ‘I feel down-hearted and blue’). Item responses are ranked from 1 (a little of the time) to 4 (most of the time).

Data analyses

We utilised the latent change model (LCM) (Allemand et al., 2007; Hertzog & Nesselroade, 2003; Shimizu & Miho, 2011) to assess the relationships between perceived autonomy support, need satisfactions, timely engagement, campus satisfaction, and depressive symptoms, with structural equation modelling, using IBM AMOS 24. Cross-lagged panel modelling is used more often in research that focuses on relationships between variables over time, which only focuses on how inter-individual differences at the baseline explain inter-individual differences at the time of the follow-up; however, such analyses cannot make inferences about intra-individual changes over time (Nesselroade & Ghisletta, 2000). Therefore, we used the same LCM as was used in the approach to capture intra-individual changes and change-change associations so as to further examine the dynamic features of the proposed relations within SDT (for an application, see Stenling et al., 2015). Measurement adequacy of measures is tested by testing the measurement model including all variables (Appendix Table A.1). We also tested the longitudinal measurement invariance of all the scales we used between Time 1 and Time 2 (Appendix Table A.2).

We controlled for the genders of the participants so as to ensure that the findings were not artefacts of such plausible variables. Results of previous research indicated that there were gender differences in academic engagement, academic life satisfaction, and depression (Balkis & Duru, 2017; Girgus & Yang, 2015). Specifically, females display lower levels of academic procrastination and higher levels of academic life satisfaction than males, while females also tend to be more likely to display depressive symptoms than males. Furthermore, we controlled for the university year of participants, considering that students in their first year of university life tend to engage in more maladaptive behaviours due to the significant changes in their overall environment and lifestyle (Villatte et al., 2017).

Results and discussion

Table 1 shows the means, standard errors, and Omega coefficients for all variables (perceived autonomy support from TAs, need satisfactions, timely engagement, campus satisfaction, and depressive symptoms) at both Time 1 and Time 2. Table 2 shows the correlations between all variables at both Time 1 and Time 2.

Table 1. Means and standard errors (SE) of all variables in Study 1 (Japan) (N = 199).

Variables	Mean	SE	ω
Autonomy support
Time 1	3.916	.067	.931
Time 2	3.453	.091	.958
Need satisfactions
Time 1	3.308	.042
Time 2	3.248	.043
Need satisfaction for autonomy
Time 1	3.584	.052	.732
Time 2	3.553	.057	.789
Need satisfaction for competence
Time 1	2.998	.054	.831
Time 2	2.993	.057	.823
Need satisfaction for relatedness
Time 1	3.344	.062	.810
Time 2	3.201	.063	.811
Timely engagement
Time 1	3.528	.088	.868
Time 2	3.344	.088	.896
Campus satisfaction
Time 1	3.146	.048	.834
Time 2	3.111	.051	.836
Depressive symptoms
Time 1	2.201	.027	.777
Time 2	2.197	.028	.753

Table 2. Correlations between all variables in Study 1 (Japan) (N = 199).

Variables	1	2	3	4	5	6	7
1. Autonomy support		.297***	.196**	.363***	−.199**	−.018	−.095
2. Need satisfaction	.226**		.497***	.427***	−.418***	−.007	−.113
3. Timely engagement	.213**	.382***		.190**	−.252***	−.013	.058
4. Campus satisfaction	.392***	.266***	.137		−.472***	.157*	.113
5. Depressive symptoms	−.159*	−.321***	−.170*	−.329***		.102	−.115
6. Gender	−.072	−.012	−.044	−.130	.141*		.020
7. University year	−.047	−.030	−.087	−.037	−.137	.020

Note. Dummy code of gender: male = 0, female = 1. Correlations in Time 1 are showed at lower left, correlations in Time 2 are showed at upper right.

*p < .05, **p < .01, ***p < .001.

The study sought to examine the mediation effect of need satisfaction on the relationship between perceived autonomy support, academic engagement, and psychological well-being using the LCM. It has been suggested that the parameter estimation of structural equation models has an inherent problem whereby the form of the structural part affects the estimates of the measurement part (Burt, 1976; Mitsunaga et al., 2005); in response, Mitsunaga et al. (2005) have proposed an estimation method using factor scores. Thus, we applied the estimation method of factor scores for the measurement model as modified by Shimizu (2010), and used the scores for the main analyses. The final models were constructed with levels and latent differences of each variable as well as controlling variables. The dependent variables were modelled separately for the proper estimation of the factor scores. In Table 3, we do not present the result of the relationship with the level in the LCM but only focus on the relationship between the changes in the variables.

Table 3. Results of mediation analyses between autonomy support, need satisfactions, timely engagement, campus satisfaction, and depressive symptoms in LCM of Study 1 (Japan).

	Direct effect		95% CI		Indirect effect		95% CI		Direct effect		95% CI
	estimate	SE	LB	UB	estimate	SE	LB	UB	estimate	SE	LB	UB
	Timely engagement								Need satisfactions
Autonomy support	.082	.069	−.051	.225	.047*	.020	.016	.093	.194*	.054	.077	.292
Need satisfactions	.243*	.067	.092	.361
	χ²	df	CFI	RMSEA	SRMR
	18.942**	6	.934	.064	.056
	Campus satisfaction								Need satisfactions
Autonomy support	.258*	.067	.115	.383	.015	.023	−.025	.065	.219*	.063	.082	.333
Need satisfactions	.071	.095	−.144	.238
	χ²	df	CFI	RMSEA	SRMR
	10.845^†	6	.953	.064	.051
	Depressive symptoms								Need satisfactions
Autonomy support	−.067	.067	−.196	.082	−.040*	.018	−.082	−.011	.203*	.056	.082	.303
Need satisfactions	−.195*	.069	−.320	−.057
	χ²	df	CFI	RMSEA	SRMR
	14.963*	6	.940	.087	.055

LCM: latent change model; CI: confidence interval; CFI: comparative fit index; RMSEA: root mean square error of approximation; SRMR: standardised root mean residual; SE: standard error; LB: lower bound; UB: upper bound.

Note. Mediation analyses were performed using AMOS 24; 95% CIs calculated using bootstrapping (1000 repetitions). Estimates of gender and university year on dependent variables were not significant and were omitted for brevity.

^†p < .10, *p < .05, **p < .01.

As shown in Table 3, latent differences in perceived autonomy support are related to latent differences in need satisfactions (β = .194, 95% CI = [.077, .292]); differences in need satisfactions also related to differences in timely engagement (β = .243, 95% CI = [.092, .361]), indicating that changes in perceived autonomy support are positively correlated with changes in need satisfactions; changes in need satisfactions also have a positive correlation with the changes observed regarding timely engagement. Although the estimates were statistically significant, the effect size was small. To verify our hypothesis that autonomy support from TAs positively relates to students’ timely engagement (which is mediated by need satisfactions), a mediation analysis was performed via AMOS 24, making use of the bootstrapping method (1000 repetitions) to calculate 95% confidence intervals (CIs). As predicted, the indirect effect of autonomy support on timely engagement, which was mediated by need satisfactions, and statistically significant (β = .047, 95% CI = [.016, .093]). The direct effect of latent differences in autonomy support on latent differences in timely engagement was not significant (β = .082, 95% CI = [.092, .361]).

Furthermore, latent differences in perceived autonomy support were statistically significantly related to latent differences in campus satisfaction with a small effect size (β = .258, 95% CI = [.115, .383]), but differences in need satisfactions were not significantly relate to differences in campus satisfaction (β = .071, 95% CI = [−.144, .238]). The results of the mediation analysis indicate that the relationship between the differences in perceived autonomy support and campus satisfaction was not significantly mediated by the differences in need satisfactions (β = .015, 95% CI = [−.025, .065]).

Although latent differences in perceived autonomy support showed no statistically significant direct effects on differences in depressive symptoms (β = −.067, 95% CI = [−.196, .082]), the differences in need satisfactions were significantly related to differences observed regarding depressive symptoms and the effect size was small (β = −.195, 95% CI = [−.320, −.057]). The results of the mediation test show that the indirect effect of perceived autonomy support on depressive symptoms, which was significantly mediated by need satisfactions, is significant (β = −.040, 95% CI = [−.082, −.011]).

The above results support our first hypothesis, which suggests that autonomy support from TAs was not only positively related to students’ timely engagement in academic settings and satisfaction with campus life but also negatively related to students’ depressive symptoms. Although only the direct effect of autonomy support on campus satisfaction was statistically significant, the relationship between autonomy support and students’ timely engagement and depressive symptoms were significantly mediated by need satisfactions, as predicted. Therefore, our second hypothesis is partially supported. These results indicate that autonomy support from support staff, such as TAs, is also important and necessary to further improve students’ positive learning outcomes and mental health on campus.

Study 2

In Study 2, we focussed on the relationship between autonomy support from PIs in China and students’ academic engagement and well-being. As stated in the introduction, PIs in China mainly focus on providing information about campus life, as well as offering counselling when students face interpersonal problems on campus, rather than being directly involved with students’ learning activities. Therefore, we predicted that autonomy support from PIs would have a stronger correlation with students’ psychological well-being than their academic engagement. As in Study 1, we also predicted that the relationships between autonomy support from PIs and students' academic engagement and psychological well-being would be mediated by need satisfactions.

Methods

Participants and procedure

One hundred undergraduate students (79 females and 21 males, mean years in university = 3.14) from two public universities in China who were taking an optional class in politics participated in our research. A Starbucks point card worth 50 CNY (about 6 USD) was given as a reward for participation in this study. Informed consent was obtained from all participants, and research was authorised by the faculty of psychology's ethics review committee of the university. Data were collected using a questionnaire survey at two-time points: the first week of the new semester in March, and during the mid-term exam week in May 2015.

The data of participants who did not complete both questionnaires were excluded, and 87 participants (74 females and 13 males, mean years in university = 3.14) were included in the final analysis. t-Tests indicated no significant differences in the means of the measures between the data we used in the final model and the data excluded from the original sample. The result of Little's MCAR tests showed that the mechanisms of whole-wave missing data in Study 1 were missing completely at random (χ² = 97.830, df = 97, p = .457).

Measures

As in Study 1, perceived autonomy support, need satisfactions, timely engagement, campus satisfaction, and depressive symptoms were measured using self-report questionnaires. All variables were measured at both Time 1 and Time 2. English-to-Chinese and Japanese-to-Chinese translations were performed by a native Chinese-speaking graduate student who is fluent in all three languages. The translations were back-translated by another native Chinese-speaking graduate student who is also fluent in all three languages and was checked by a professor in the psychology faculty. All the measures used in Study 2 were the same as those in Study 1. Measurement adequacy of measures is tested by testing the measurement model including all variables (Appendix Table A.1). We also tested the longitudinal measurement invariance of all the scales we used between Time 1 and Time 2 (Appendix Table A.3).

Results and discussion

Table 4 shows the means and standard errors of all variables at both Time 1 and Time 2 (perceived autonomy support from PIs, need satisfactions, timely engagement, campus satisfaction, and depressive symptoms). Table 5 shows the correlations between all variables at both Time 1 and Time 2. As in Study 1, final models were constructed with levels and latent differences of each variable, and every model contained only one dependent variable (Table 6). Estimates of gender and university year on dependent variables, and the results of levels were omitted for brevity.

Table 4. Means and standard errors (SE) of all variables in Study 2 (China) (N = 87).

Variables	Mean	SE	ω
Autonomy support
Time 1	4.615	.145	.932
Time 2	4.636	.131	.936
Need satisfaction
Time 1	3.563	.054
Time 2	3.631	.050
Need satisfaction for autonomy
Time 1	3.845	.065	.727
Time 2	3.945	.061	.661
Need satisfaction for competence
Time 1	3.347	.076	.663
Time 2	3.434	.078	.828
Need satisfaction for relatedness
Time 1	3.497	.081	.680
Time 2	3.514	.076	.699
Timely engagement
Time 1	4.245	.117	.774
Time 2	4.362	.121	.843
Campus satisfaction
Time 1	2.659	.049	.737
Time 2	2.609	.048	.702
Depressive symptoms
Time 1	1.855	.041	.759
Time 2	1.915	.043	.822

Table 5. Correlations between all variables in Study 2 (China) (N = 87).

Variables	1	2	3	4	5	6	7
1. Autonomy support		.257*	.072	.554***	−.165	−.176	−.162
2. Need satisfaction	.243*		.222*	.259*	−.154	−.177	−.129
3. Timely engagement	.080	.270*		.139	−.288**	−.056	−.250*
4. Campus satisfaction	.446***	.001	.287**		−.406***	−.166	−.185
5. Depressive symptoms	−.164	−.160	−.302**	−.387***		.125	.260*
6. Gender	−.395***	−.113	−.034	−.211*	.138		.244*
7. University year	−. 210	−.108	−.119	−.035	.325**	.244*

Note. Dummy code of gender: male = 0, female = 1. Correlations in Time 1 are showed at lower left, correlations in Time 2 are showed at upper right.

*p < .05, **p < .01, ***p < .001.

Table 6. Results of mediation analyses between autonomy support, need satisfactions, timely engagement, campus satisfaction and depressive symptoms in LCM of Study 2 (China).

	Direct effect		95% CI		Indirect effect		95% CI		Direct effect		95% CI
	Estimate	SE	LB	UB	Estimate	SE	LB	UB	Estimate	SE	LB	UB
	Timely engagement								Need satisfactions
Autonomy support	.053	.107	−.152	.266	−.007	.028	−.069	.032	.096	.090	−.077	.282
Need satisfactions	−.071	.212	−.408	.360
	χ²	df	CFI	RMSEA	SRMR
	1.229	6	1.000	.000	.024
	Campus satisfaction								Need satisfactions
Autonomy support	.322*	.109	.104	.535	−.005	.011	−.037	.008	.077	.060	−.031	.207
Need satisfactions	−.069	.113	−.278	.171
	χ²	df	CFI	RMSEA	SRMR
	8.864	6	.985	.074	.073
	Depressive symptoms								Need satisfactions
Autonomy support	−.063	.125	−.359	.159	−.011	.014	−.067	.002	.060	.051	−.032	.173
Need satisfactions	−.189	.103	−.384	.016
	χ²	df	CFI	RMSEA	SRMR
	4.729	6	1.000	.000	.054

LCM: latent change model; CI: confidence interval; CFI: comparative fit index; RMSEA: root mean square error of approximation; SRMR: standardised root mean residual; SE: standard error; LB: lower bound; UB: upper bound.

Note. Mediation analyses were performed using AMOS 24; 95% CIs calculated using bootstrapping (1000 repetitions). Estimates of gender and university year on dependent variables were not significant and were omitted for brevity.

*p < .05.

Contrary to our hypothesis, latent differences in perceived autonomy support from PIs were not statistically significantly related to latent differences in students’ need satisfactions (β = .096, 95% CI = [−.077, .282]). Moreover, differences in timely engagement were not significantly related to differences in perceived autonomy support or differences in need satisfactions (β = .053, 95% CI = [−.152, .266], and β = −.071, 95% CI = [−.408, .360]), respectively).

Next, latent differences in perceived autonomy support were statistically significantly related to latent differences in campus satisfaction with a small effect size that is similar to Study 1 (β = .322, 95% CI = [.104, .535]); however, differences in need satisfactions were not significantly relate to differences in campus satisfaction (β = −.069, 95% CI = [−.278, .171]). Therefore, the relationship between the differences in perceived autonomy support and campus satisfaction was not significantly mediated by differences in need satisfactions (β = −.005, 95% CI = [−.037, .008]). Similar to Study 1, these results indicate that changes in autonomy support are positively related to changes in campus satisfaction.

Finally, latent differences in perceived autonomy support show no statistically significant direct effect on corresponding values for depressive symptoms (β = −.063, 95% CI = [−.359, .159]), and the direct effect of latent differences in need satisfactions on latent differences in depressive symptoms was also not significant (β = −.189, 95% CI = [−.037, .008]). Latent differences in need satisfactions do not significantly mediate the relationship between differences in perceived autonomy support and depressive symptoms (β = −.011, 95% CI = [−.067, .002]).

The results above partially support our first prediction, which held that autonomy support from PIs in China would have a positive correlation with students’ well-being, specifically campus satisfaction, but it is not statistically significantly related to timely engagement or depressive symptoms. Additionally, contrary to our hypothesis, the correlation between autonomy support from PIs and students’ campus satisfaction was not significantly mediated by need satisfactions. These results indicate that autonomy from PIs is essential to how students perceive their general satisfaction with the university. The limitations of Study 2 will be discussed in the next section.

General discussion

The results of the present research used LCMs to reveal that receiving autonomy support from support staff relates to university students’ academic engagement and psychological well-being. In Study 1, the change in autonomy support from TAs in Japan not only positively related to the change in students’ timely engagement and the change in satisfaction with campus life but also negatively related to their change in depressive symptoms. In Study 2, the change in autonomy support from PIs in China has a positive relationship with the change in students’ campus satisfaction, but the change in autonomy support from PIs was not related to the change in students’ timely engagement towards academic activities or depressive symptoms.

More specifically, in Study 1, the relationships between autonomy support and students’ timely engagement and depressive symptoms were mediated by need satisfactions, which is consistent with previous research (e.g. Froiland, 2021; Pesch et al., 2016). These results indicate that autonomy support has a positive correlation, even if it is from individuals who are not primary influencers, as the same mechanism of autonomy support is used by primary influencers, which may indicate the importance of autonomy support from support staff like TA. Although TAs only support students’ learning activities, autonomy support from TAs was not only related to students’ academic engagement, but also their psychological well-being, given that satisfaction has both top-down and bottom-up effects (Milyavskaya et al., 2013). That is, not only could need satisfactions in general influence overall outcomes but need satisfactions at the domain level can also exert an ‘upward’ influence on the general level of wellness in certain situations. For example, in the present research, participants were university students, whose stressors are often learning-related (Pluut et al., 2015); therefore, autonomy support from TAs who focus on students’ learning could also have a negative relationship with students’ depressive symptoms.

In Study 2, changes in autonomy support from PIs do not relate to changes in the need satisfactions of Chinese students. These findings contradict those of previous studies, and this might be caused by the very small sample size used in Study 2, which failed to reach the required power, which is considered one of the limitations of the present research. However, it should also be noted that the need satisfactions scale used in the present study (Nishimura & Sakurai, 2013) measures only need satisfactions within the academic domain. As PIs are not directly involved in university students’ learning activities, this scale may not entirely capture the relationships of autonomy support from PIs with students’ need satisfactions. The correlation between autonomy support from PIs and university students’ campus satisfaction could be mediated by overall need satisfactions (that can be measured using the ‘Basic Psychological Needs Scale’; Deci & Ryan, 2000). Considering the potential positive correlation between autonomy support of PIs and adapted outcomes of students' learning activities and psychological well-being, more well-designed research on the relationship should be carried out in the future.

Some researchers have posited that the need for autonomy only applies to Western cultures because many Asian cultures have distinct concepts regarding individuality that insist on the fundamental relatedness among individuals (Markus & Kitayama, 1991; Miller, 1997). However, the current results are aligned with those of previous studies, which showed that SDT is also applicable in cultures, such as those of Japan and China, where the climate is considered overtly oriented towards collectivism (Nie et al., 2015; Yamauchi & Tanaka, 1998). More importantly, autonomy is considered a basic human need that differs from individualism and independence across cultures, races, and nations (Chirkov et al., 2003).

Theoretical and practical implications

By understanding the connections between the autonomy support delivered by those who are not primary influencers in an educational setting—specifically, support staff in higher education—and the educational and mental health-related outcomes in the two studies in different countries, the present study is the first to provide evidence for the merits and importance of receiving autonomy support from such individuals, adding a new aspect to the current literature on SDT. Based on the results from the present research, we consider that autonomy support delivered by those who are not primary influencers is as important as autonomy support provided by teachers or parents. This highlights how to improve the environment in higher education so that it becomes more supportive for students in multiple ways. Importantly, utilising autonomy support from support staff (such as TAs and PIs) in educational contexts could mitigate the ineffective learning and lower well-being of students caused by a lack of direct support from teachers, which many Asian countries are struggling with (Huang, 2012). In addition, this could be considered a key component in the efficient construction of environments that foster students’ autonomous motivation.

Our results indicate that the basic mechanism of how autonomy support provided by TAs and PIs is associated with students' engagement and well-being is the same as that of teachers, which is mediated by need satisfactions. However, we did not further explore whether the autonomy support delivered by those who are not primary influencers (e.g. TAs and PIs) and primary influencers (e.g. teachers and parents) relate to students' learning and well-being outcomes differently. Future studies should look into these two processes more closely, especially on how they might fulfill the three aspects of basic psychological needs differently, which could lead to the differential contributions of both support staff and teachers to students' learning and well-being outcomes.

Moreover, given the positive correlation between autonomy support from support staff and students' adapted outcomes on learning and well-being, and considering the effects of autonomy support intervention programs (Cheon et al., 2012), we propose establishing autonomy support training programs for support staff whose future roles entail assisting students. As we mentioned earlier in the discussion, a lack of emphasis on the importance of autonomy support, and the correct provision thereof, would lead to its decrease as time progresses. In addition, previous research has shown that, without detailed guidelines, the nature of PIs’ work (i.e. instructing students in the development of their political ideology) allows for the adoption of a controlling motivational style (Xu, 2003). Thus, to maximise the utility of autonomy support among support staff in educational settings, adequate knowledge regarding this concept is required for future professional training programs.

Limitations

Overall, the present results demonstrate the importance of TAs and PIs in university students’ lives as well as the usefulness of autonomy support from support staff. However, our conclusions are somewhat limited. First, the present research contains methodological limitations, such as the rather small sample size (especially in Study 2) although the overall effect seems to be small, and the fact that we only used self-reported data as our index of interesting variables. Moreover, the proportion of the gender in the present study was not even. Although we controlled for the effect of gender in the model, and the results show that the effect was not significant, this issue should be avoided in future studies to give more reliable generalisations. Second, the duration of students’ interactions with TAs or PIs was not measured. Because the duration of contact between TAs or PIs and students is typically longer for students requiring assistance—relative to those who do not require it—it is possible that students’ perceptions of TAs or PIs as being supportive reflect the duration of their interactions, rather than actual supportiveness. Third, we did not control for the types of classes in which TAs were involved in Japan; therefore, the interaction between TAs and students could differ, depending on class content. Fourth, as we mentioned earlier, we did not measure the autonomy support from primary educational influencers, so we could not control the effect of autonomy support from teachers or parents while estimating the effect of autonomy support from support staff. Future studies should put more effort into elaborating the relative importance of TA-initiated autonomy support compared to professor-initiated autonomy support, so we can benefit from simultaneously exploring the differential contributions of both teaching professionals to students' adapted outcomes on learning and well-being. Finally, we used a self-report scale to measure students' perception of autonomy support from TAs and PIs, which is a common method utilised in previous research; however, the distinction between the satisfaction of the need for autonomy (as satisfaction of a basic psychological need) and the perceived autonomy support from others, does not appear clearly. Even though autonomy support itself is conceptually different from the satisfaction of need for autonomy, we can argue that when students perceive their teacher or TA as being autonomy-supportive, their need for autonomy is fulfilled; at that point, we might be actually measuring the same psychological status. Observing the autonomy-supportive actions in the classroom or measuring the autonomy support by asking the person who performed it is one way to avoid this limitation (Reeve et al., 2004). In future studies, the issues above should be considered as covariates so as to enable further understanding of the effects of autonomy support from support staff.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

This work was supported by JSPS KAKENHI (Grant Number 16K11978, 16H06406).

Appendix

The model fit of the full measurement model that included single measurement models of each scale in confirmatory factor analyses in both Study 1 and Study 2 was evaluated using a likelihood-ratio chi-square test and three goodness-of-fit indices: Bentler’s (1990) comparative fit index; standardised root mean residual (Hu & Bentler, 1999); and Steiger’s (1990) root mean square error of approximation. A comparative fit index > 0.95, a root mean square error of approximation < 0.05, and a standardised root mean residual < 0.05 suggest an excellent model fit (Hu & Bentler, 1999). As shown in Table A.1, the full measurement model in both Study 1 and Study 2 resulted in acceptable fits.

Table A.1. Model fit of confirmatory factor analyses of full measurement models in Study 1 (Japan) and Study 2 (China).

	χ²	df	CFI	RMSEA	SRMR	Range of standard regression weights
Japan Time 1	68.338***	35	.967	.069	.069	.364–.952
Japan Time 2	71.938***	35	.970	.073	.057	.466–.985
China Time 1	44.437	35	.975	.056	.076	.326–.971
China Time 2	66.451***	37	.928	.096	.099	.417–.975

Note. ***p < .001.

Table A.2. Model fit indices of the models with configural (A), metric (B), and scalar (C) measurement invariance across informants in Study 1 (Japan).

		Model fit indices						Model comparisons
	Model	χ²	df	CFI	RMSEA	SRMR		Δχ^2	df	ΔCFI	ΔRMSEA	ΔSRMR
Autonomy support	A	594.353***	180	.899	.076	.052
	B	595.803***	194	.903	.072	.051	B vs. A	1.45	6	0.004	−0.004	−0.001
	C	610.131***	210	.908	.069	.050	C vs. B	14.328	16	0.005	−0.003	−0.001
Need satisfactions	A	486.178***	102	.912	.074	.061
	B	499.254***	111	.923	.068	.054	B vs. A	13.076	9	0.011	−0.006	−0.007
	C	516.492***	129	.915	.067	.048	C vs. B	17.238	18	−0.008	−0.001	−0.006
Timely engagement	A	58.917***	18	.955	.081	.046
	B	60.977***	23	.967	.073	.041	B vs. A	2.060	5	0.012	−0.008	−0.005
	C	69.052***	30	.966	.057	.040	C vs. B	8.075	7	−0.001	−0.016	−0.001
Campus satisfaction	A	355.621***	70	.897	.080	.058
	B	361.881***	79	.898	.078	.056	B vs. A	6.260	9	0.001	−0.002	−0.002
	C	367.371***	90	.922	.068	.051	C vs. B	5.490	11	0.024	−0.010	−0.005
Depressive symptoms	A	1323.046***	340	.867	.088	.059
	B	1338.734***	359	.877	.085	.058	B vs. A	15.688	19	0.010	−0.003	−0.001
	C	1371.687***	380	.925	.081	.053	C vs. B	32.953*	21	0.048	−0.004	−0.005

Note. Configural model = factor structural are equal; Metric model = loadings are equal; Scalar model = loadings and intercepts are equal.

*p < .05, ***p < .001.

Moreover, univariate outliers in both Study 1 and Study 2 were checked via boxplots and histograms, which indicated that the data were normal in this regard. Multivariate outliers were checked using the Mahalanobis distance statistic. The largest Mahalanobis d-squared value—or the observation of the furthest distance from the centroid—is 33.35 with a probability value <0.01, or a small probability of an unusual observation (i.e. an outlier) (Walker, 2010). Furthermore, the data were screened for instances of multicollinearity via an analysis of a variance inflation factor (VIF). Multicollinearity was not present, as all VIF measures were <3, meeting the noted cut-off point of <10 (Belsley et al., 1980).

We tested the longitudinal measurement invariance of all the scales we used between Time 1 and Time 2 by creating three models: (A) a configural model, in which the factor structure is the same across time points but factor loadings, intercepts, and residual variances are allowed to differ between time points; (B) a metric model, whereby the factor loadings are equal across time points but the intercepts are allowed to differ between time points; and (C) a scalar model, whereby loadings and intercepts are constrained to be equal across time points. If imposing invariance constraints resulted in a significant increase in the chi-square value and, additionally, in ΔCFI ≤ −0.010 supplemented by ΔRMSEA ≥ 0.015, or ΔSRMR ≥ 0.03, the respective constraint is not tenable (Chen, 2007). Table A.2 shows the results of tests of longitudinal measurement invariance of all the scales in Study 1, which indicated that the scalar model held for all scales, except for depressive symptoms; however, the indices in the metric model were acceptable, and the differences in fit from the freely estimated model were not sufficient to conclude that the equality constraints on the loadings and intercepts of the observed indicators (across the two samples) substantially deteriorate the approximation of the data. Table A.3 shows the results of tests of longitudinal measurement invariance of all the scales in Study 2, which indicated that the scalar model held for all scales.

Table A.3. Model fit indices of the models with configural (A), metric (B), and scalar (C) measurement invariance across informants in Study 2 (China).

		Model fit indices						Model comparisons
	Model	χ²	df	CFI	RMSEA	SRMR		Δχ^2	df	ΔCFI	ΔRMSEA	ΔSRMR
Autonomy support	A	390.230***	188	.878	.075	.054
	B	400.689***	194	.880	.079	.051	B vs. A	10.459	6	0.002	0.004	−0.003
	C	422.978***	210	.896	.077	.050	C vs. B	22.289	16	0.014	−0.002	−0.001
Need satisfactions	A	226.849***	102	.912	.068	.053
	B	238.298***	111	.895	.082	.057	B vs. A	11.449	9	−0.017	0.014	−0.004
	C	252.401***	129	.902	.075	.051	C vs. B	14.103	18	0.007	−0.007	−0.006
Timely engagement	A	48.917***	18	.917	.073	.052
	B	50.930***	23	.925	.061	.049	B vs. A	2.013	5	0.008	−0.012	−0.003
	C	57.680***	30	.926	.060	.047	C vs. B	6.750	7	0.001	−0.001	−0.002
Campus satisfaction	A	338.014***	70	.878	.076	.055
	B	345.928***	79	.882	.078	.060	B vs. A	7.914	9	0.005	0.002	−0.005
	C	354.188***	90	.896	.081	.058	C vs. B	8.260	11	0.014	0.003	−0.002
Depressive symptoms	A	751.700***	340	.890	.072	.060
	B	780.712***	359	.887	.083	.059	B vs. A	29.012	19	−0.003	0.009	−0.001
	C	802.890***	380	.924	.080	.055	C vs. B	22.178	21	0.037	−0.003	−0.004

Note. Configural model = factor structural are equal; Metric model = loadings are equal; Scalar model = loadings and intercepts are equal.

***p < .001.