What constitutes a good life? Is this signaled by adequate material wealth? Nowadays, an increasing number of people, especially the young, find themselves trapped in an anxious state even though their wallets grow plumper. A survey conducted by the Anxiety and Depression Association of America reported that seven out of ten American adults claim to experience stress or anxiety at least at a moderate level on a daily basis [1
]. As a prevalent emotional disorder that interferes with psychosocial functioning [2
], anxiety has also become a serious public health problem in China and among university students, having an impact on their daily life [3
]. The gap between economic growth in China and anxiety reminds us to seek clues from other areas besides economics to answer questions in relation to what constitutes a good life.
Addressing this question, Csikszentmihalyi [4
] conducted extensive investigations of rock climbers, chess players, athletes, artists, and those well-known in their fields about why they often perform time-consuming, difficult, and even dangerous daily activities, even though these challenges do not repay any discernable and extrinsic reward. The answer points towards the concept of flow. In general, flow describes a state in which an individual is fully involved in the activity or task at hand without a sense of time, fatigue, or awareness of other irrelevant matters. One focuses on nothing else but merely the activity itself, feeling an intense sense of inner bliss likened to a stream flowing through one’s heart: giving rise to the name flow. There are three typical characteristic features of flow that are frequently reported by people experiencing it: the merging of action and awareness, a sense of control, and an altered sense of time [4
A good life, according to Seligman [7
], is considered to be a pleasant, meaningful, and engaged life that seeks to have more flow experience. Living without anxiety disorder is one of aspects of living a good life. It is beyond dispute that more support should be given to help anxiety disorder sufferers gain a sense of control of their inner mind. At the same time, flow which is characterized by self-consciousness and concentration can demonstrate a kind of self-control. Although flow has been investigated in association with many psychological constructs such as eudaimonic well-being [8
] and subjective well-being [10
], little research has explored its relationship to anxiety and the associated underlying mechanisms. So, can flow alleviate anxiety?
Anxiety, according to Craske and Stein, has become one of the most prevalent psychiatric disorders of modern times: it refers to a set of disorders characterized by excessive fear, anxiety, or avoidance of an array of external and internal stimuli [11
]. Anxiety has been experienced by a growing proportion of university students in recent years. Beiter and colleagues investigated what types of college students tend to experience the most serious anxiety symptoms and found that transfers, upperclassmen, and those living off-campus are the most anxious groups, experiencing anxiety in confronting three major concerns: academic performance, pressure to succeed, and post-graduation plans [1
]. While academic study can be perceived as a positive challenge for university students, potentially increasing learning capacity and competency, such stress can be detrimental to student mental health if viewed negatively [12
We therefore proposed the present study as a novel trial with a unique perspective, hypothesizing flow (optimal experience) as a potentially important way to reduce anxiety and therefore promote happiness. Moreover, we sought to provide preliminary empirical support for flow–anxiety relations by exploring the underlying mechanisms taking into consideration some possible influencing factors such as academic self-efficacy and self-esteem.
3. Materials and Methods
3.1. Participants and Procedures
A total of 630 participants who were registered at Southwest Jiaotong University were contacted, of which 627 responded. Excluding those individuals who finished the questionnaire in an unreasonably short time and for whom there were missing data on crucial study variables, a final sample of 590 participants remained. Of the participants, 46.78% were women and 53.22% were men, and they were aged between 18 and 25 years (M = 20, SD = 1.6). Of those in the sample, 23.05% were freshmen, 25.08% were sophomores, 28.64% were juniors, 13.90% were seniors, and 9.32% were postgraduates. For the majority of respondents (77.3%), their monthly disposable incomes were within the range between 1000 and 2000 RMB (about 130 and 260 Euro). Over one-third (36.95%) of respondents reported being in an indebted condition.
Prior to data collection, the required application forms to seek ethical approval for the research were prepared and submitted, and the project was approved at Southwest Jiaotong University by routine exemption, due to proper survey design, anonymity, and lack of harm to participants. The study followed by ethical principles based on guidelines from international scientific communities in psychology. Informed consent for participation was obtained after participants were presented with documents that outlined any risks, their choice to participate, and their ability to leave. Participants then proceeded to complete the survey. A total of 630 paper and pencil-based questionnaires were printed and then administered in the library and classrooms inside the university campus. Respondents received a small gift for completing the questionnaire survey. The data-gathering phase ran from 28 to 29 April in the year 2019.
Standard and specific instruments were used to measure flow, anxiety, self-esteem, and academic self-efficacy, based on the past month of an individual’s psychological state. The instruments were delivered by means of a self-report questionnaire consisting of three sections. The first section included 42 multiple-choice questionnaire items. The majority of the items were positively worded, while the five negatively worded items were reversely coded. Responses to all of the multiple-choice questions were registered on a 7-point Likert-type scale with answers ranging from 1 (“not at all characteristic of me”) to 7 (“completely characteristic of me”). The second section regarded participants’ financial status. The third section was on sociodemographic background.
By adapting a similar administration technique to that of Waterman [53
] and our previous work [32
], a question was developed to define the activities within which the subject subsequently had to provide responses regarding flow experience: “Think about one of your past month’s activities that was challenging but that you regularly engaged in with utmost of enjoyment (i.e., reading, sports, listening to music, etc.), please indicate your feelings based on the following eight statements____”. Then, the selected eight-item flow scale, corresponding to flow characteristics identified by Csikszentmihalyi was administered [5
]. These items were phrased as completions of a common stem anchored by not at all characteristic of me and completely characteristic of me, with moderately in the middle. The common stem was: “When I engage in this activity ____”, and the item completions were the following: (1) I feel I have clear goals; (2) I feel self-conscious (reverse scored), (3) I feel in control; (4) I lose track of time; (5) I feel I know how well I am doing; (6) I have a high level of concentration; (7) I forget personal problems; and (8) I feel fully involved. This scale has been widely used [54
], and its structure was tested with a reported alpha coefficient ranging from 0.80 to 0.83 in a multinational sample testing the associations between flow and personal identity [32
]. Cronbach’s α for flow in the present sample was 0.798.
Participants were asked to rate their overall psychophysiological state based on the past month using the anxiety part of the DASS 21 (Depression, Anxiety, and Stress Scale 21) [58
]. The scale contains nine items to access physical symptoms and mental state in relation to anxiety disorders. Mental state measurements are constituted of apprehension, panic, and worry about academic performance (i.e., “I felt apprehensive during this month”), while physical indicators included trembling, a shaky condition, dryness of the mouth, breathing difficulties, heart pounding, and palm sweatiness (i.e., “I felt sweatiness of the palms during this month”). The item which is described as “I am aware of dryness of the mouth” was eliminated after the pilot test, as it partially overlaps the item assessing sweatiness of the palms. Cronbach’s α for this scale for the present sample was 0.807.
Self-esteem was measured based on the renowned Rosenberg Self-Esteem Scale (RSES) [43
]. With feedback from the pilot test, moderate elimination of items was performed to optimize our proposed model. The item described as “I take a positive attitude toward myself” was removed, as it overlapped another item (“on the whole, I am satisfied with myself”) in pilot test. Another item, “I certainly feel useless at times”, was also excluded, due to its similarity to “I wish I could have more respect for myself” in our pilot test. Furthermore, we discarded the item “I wish I could have more respect for myself”, because it is difficult to reverse modify this while retaining understanding in a specific Chinese context. The remaining original items, which were reverse rated, were adjusted to adapt to our model. Cronbach α for this scale in the current study was 0.899.
3.2.4. Academic Self-Efficacy
The Academic Self-Efficacy Scale from Stagg [59
] was utilized to measure university students’ self-reported capability of academic performance. The scale includes eight items with three principal facets: learning efficiency, examination, and learning processes. Learning efficiency was assessed by finishing work on time to a good standard and effective management of one’s time. The examination facet was composed of passing an exam after revising hard and achieving expected grades. Other items including comprehending academic literature, effectively seeking background materials, taking notes in lectures, and answering questions in class, all of which can be regarded as specific aspects of learning processes. Cronbach’s α for academic self-efficacy in the present sample was 0.837.
3.3. Data Analysis
Data were analyzed using structural equation modeling (SEM) with AMOS 21.0 software, in which both factorial analysis and path analysis were performed. At first, assessment of normality was conducted following the estimation methods of maximum likelihood (ML), generalized least squares (GLS), and asymptotic distribution-free (ADF), which could lead us to a picture that is closer to reality [60
]. Following this, the general goodness-of-fit and internal quality of the model were examined. On the basis of a model with excellent indices of goodness-of-fit, tests of mediating effects and the magnitude of indirect effects were conducted by bootstrapping procedures [61
], and multigroup analyses were performed.
4.1. Assessment of Normality
Results for the assessment of normality are given first, since the ML and GLS estimation methods are grounded on this. The maximum of absolute value of skewness was 1.131, while that of kurtosis was 1.430. So, we deemed that distribution of samples was within an acceptable range, and ML and GLS could be used to conduct the following analysis.
4.2. Convergent Validity
Due to the fact that validity is composed of convergent validity and discriminant validity, we chose composite reliability (CR) and average variance extracted (AVE) to examine convergent validity, and we performed six multimodel analyses to check discriminant validity. Test results for convergent validity are shown in Table 1
The model was deemed to have good convergent validity when observed variables of one construct were highly correlated and when they effectively indicated the corresponding latent variable. Composite reliability (CR) assessed the consistency of indicators of each latent construct, and average variance extracted (AVE) assessed the level of error variance which latent constructs could account for. As shown in Table 1
, values of CR were all above 0.6, and those of AVE were greater than 0.5, which suggested that the adjusted model of the current dataset had excellent convergent validity.
4.3. Discriminant Validity
Discriminant validity refers to when there are significant differences between the measurements of the constructs. We carried out six multimodel analyses to confirm the discriminant validity of the adjusted model. Results of multimodel analyses are shown in Table 2
. It appeared that all values of chi-squared given by the three estimation methods were statistically significant, which suggested that there were substantial differences between the latent constructs, i.e., that the constructs in the adjusted model had acceptable discriminant validity.
4.4. Estimated Structural Equation Model
The complete structural equation model estimated by ML is shown in Figure 3
. The causal path of flow on anxiety and that of academic self-efficacy on anxiety were eliminated, for they were shown to have no statistical significance (α = 0.01). As will be discussed subsequently, the general goodness-of-fit and other internal quality indices of the current model were confirmed to be acceptable. Therefore, we continued to perform tests of mediation and multigroup analyses based on the diagram presented in Figure 3
displays standardized estimates, with results of three estimation methods presented. The first part shows regression weights of the structural model, which indicates regression weights between latent variables. The latter three sections show regression weights of the measurement model, which refers to the factor loadings of each construct upon their observed variables.
From this table, we can identify that, except for certain items such as the path between flow and self-esteem, all of the other estimates by ML, GLS, and ADF appeared to be close to each other. This indicates that the samples for the current study are representative and that parameter estimates have robustness, to some degree. All the estimates had excellent statistical significance. Flow positively predicted self-esteem and academic self-efficacy. Academic self-efficacy exerted a positive influence on self-esteem, while self-esteem negatively influenced anxiety. These results confirmed our previously stated hypotheses.
shows values for overall fit indices as estimated by ML, GLS, and ADF. Results reveal that the model had favorable overall fit.
4.5. Mediating Effects and Total Effects
Despite the fact that all causal paths of the model in Figure 2
were confirmed to be statistically significant, further tests of mediating effects and total effects were essential in order to identify the effects that flow and other psychological constructs exerted on anxiety. Bootstrapping SEM in AMOS 21.0 was utilized to calculate the relevant p
Results based on standardized estimates of indirect effects and total effects by ML, GLS, and ADF are shown in Table 5
. It is clear that all indirect and total effects are significant, with p
values less than 0.01. Table 5
reports three mediating-effect tests and three total-effect tests. In one respect, self-esteem played a mediating role on the path between flow and anxiety, while academic self-efficacy mediated the path between flow and self-esteem. Self-esteem mediated the path between self-efficacy and anxiety. It can be concluded that self-esteem played a fully mediating role in two related causalities, while academic self-efficacy played a partial mediating role on the path between flow and self-esteem. These results confirmed our hypotheses H2, H5, and H6. From this table we also note that of all the effects for anxiety, self-esteem had the strongest total effect, followed by flow (optimal experience), while academic self-efficacy was the weakest predictor for anxiety.
4.6. Multigroup Analysis
To confirm cross-validation of adjusted model, multigroup analyses were performed, with gender and debt condition set as moderating variables. Each of them has two possible values. Multigroup analysis in AMOS was to compare whether statistically significant difference exists between the models of different groups. Put another way, cross-validation was confirmed if the models of different gender or debt groups classified by moderating variable fit each other well.
Results of multigroup analyses are exhibited in Table 6
. Due to a relatively small sample size, the index chi-squared/df can be regarded as credible and reliable to some extent. Values of RMSEA and GFI were extraordinarily prominent, which indicated remarkable fit. Hence, it was apparent that the models of different groups moderated by gender had no significant difference. The same result was also seen for models moderated by debt condition.
4.7. Summary of Results
As previously outlined in the introduction section, our main aim was to identify the role of optimal experience in reducing anxiety of university students. Results revealed by the present dataset may suggest interventions through enhancing flow as a potentially effective way to relieve anxiety. Firstly, the predicting effects of flow on anxiety, as well as the mediation effect of self-esteem and academic self-efficacy on the path of flow to anxiety, were tested and confirmed. Secondly, the significance of the mediating effects and total effects were obtained by bootstrapping methods based on the confirmed statistical model. Thirdly, factor loadings of observed variables on flow, self-esteem, and academic self-efficacy were shown to weight the predicting effects of different dimensions on anxiety. Overall, the results for the tested hypotheses are indicated in Table 7
It was found that flow affected anxiety by the fully mediating function of self-esteem and academic self-efficacy. Of all the direct effects on anxiety, only self-esteem on anxiety was shown to be statistically significant, which indicated that self-esteem may play a crucial mediating role on the path of flow to anxiety.
Accordingly, the proposed structural model in Figure 2
was slightly modified. The diagram of the adjusted structural model is displayed in Figure 3
and Figure 4
, where the direct effects of flow and academic self-efficacy on anxiety have been eliminated.
5. Discussion and Conclusions
The current study set out to examine the relationship between flow and anxiety and the roles of self-esteem and academic self-efficacy in the path between flow and anxiety. By using absolute instruments we found that the experience of flow negatively predicted anxiety among university students, which confirmed the previous results on negative flow–anxiety correlations [62
] while providing concrete support for the notion of flow serving as a protective factor for those who experience anxiety; such a finding is also consistent with previous work on antithetical flow–anxiety relations in Midwest American University students [40
]. The three loaded factors from the present investigation’s sample confirm Csikszentmihalyi’s [63
] theory that flow is experienced based on typical characteristics such as control, concentration, and time distortion. Such a finding is also consistent with our prior work on flow [15
]. Multiple studies have implied that flow experience is positively associated with positive affect and negatively linked to negative affect, and our results yielded from the present dataset also confirmed this finding. As anxiety may be considered to be a case of negative affect [40
], the present work, as a novel trial, provides concrete evidence and practical suggestions for facilitating university students’ flow experience in order to alleviate their academic anxiety. However, these results may be generalizable for a broad range of professionals who are always faced with high expectations and have high potential for achievement: e.g., performing artists, professional sportsmen and dancers, commanding officers, leaders and managers.
Our research also provides strong evidence on the relationship between flow and two important self-concepts: self-efficacy and self-esteem. Our research supports previous findings and extends them with a university student sample. It should also be remembered that some previous research contributions have already shown correlations among carrying out self-defining activities, the experience of flow, and the strengthening of one’s own identity by considering personal [15
], social [32
], and place [19
] identity. Thus, the present study reports results which coherently fit within the existing literature showing flow’s importance for the psychological self.
Secondly, the mediation effects among these latent variables were verified, and the probable mechanism for how flow experience eliminates anxiety was revealed based on the present dataset. (1) The mediating role of self-esteem on academic self-efficacy and anxiety was confirmed. As we suggested, self-esteem featured as a mediator in different functioning processes [64
]. As with previous studies, the current study verified this mediation effect. However, in contrast to our proposition, results revealed that a full mediation effect exists between these three variables. In other words, higher academic self-efficacy will help students eliminate anxiety, entirely by promoting their self-esteem. (2) The most important finding of this study is that flow can alleviate anxiety through promoting academic self-efficacy and then self-esteem, and self-esteem has a direct influence on anxiety. This interesting finding highlights the importance of self-esteem in the relationship between flow, academic self-efficacy, and anxiety. The effect of academic self-efficacy on anxiety was fully mediated by self-esteem: this means that when university students are engaged in a flow state, their increased academic confidence and beliefs can reduce anxiety through the subsequently promoted self-evaluation of their inner worth.
The study also offers implications for educational practice. As we know, flow experience and anxiety affect occur with different levels of challenges and skills: instructors should help students in balancing academic challenges with their skills. Specifically, the important role of flow in anxiety provides strong support for the use of flow-promoting technology such as polling devices [66
] to eliminate anxiety and enhance learning experiences. Secondly, as we have shown that academic self-efficacy and self-esteem play essential roles in the process of how flow influences anxiety, instructors should enable students to perceive their self-efficacy, especially enhancing their self-esteem using a variety of approaches. In this way, facilitating flow (i.e., via setting appropriate and clear learning goals, providing prompt feedback on learning process, assigning the challenging task that can be overcome with a stretch of capabilities) may foster the individual’s high performance, high achievement and competence [67
], promote academic self-efficacy and self-esteem, and thus reduce anxiety.
In conclusion, the innovative contribution of this study is that we introduced flow theory to study anxiety, taking self-esteem and academic self-efficacy into consideration as mediating variables, and have proposed, tested, and confirmed a model that highlights the potential benefits of flow theory for effectively informing psychological therapies for anxiety relief in university students. This may support subsequent studies which aim to examine how and to what extent optimal experience may work to reduce anxiety amongst university students in clinical practice. However, the genuine predicting effect of flow on anxiety needs to be further verified in these efforts: the present contribution in fact provides correlation evidence, and experimental evidence is needed to prove cause–effect relations. The topics or areas of the flow-inducing activity need to be clarified: specifically, to see if any relation exists among the domain of the flow-generating activity, the domains of both self-efficacy and self-esteem, and finally the anxiety domain. Further studies need to clarify if these mediation effects happen across domains or if they are domain-constrained.
There are several limitations that warrant discussion. Firstly, the model is an approximation of the reality, and there are a number of other possible important variables which may play a role. The results remind us that there may be certain other crucial exogenous variables which need to be taken into account and which might improve the performance of the model. It seems that mere psychological constructs are not enough. This does not challenge our finding that optimal experience may be a potential way to reduce anxiety, but it does require us to reconsider this issue and the research framework with the extension of more factors. Our results implied that anxiety is an extremely sophisticated phenomenon. Despite psychological therapies which may be effective to some extent, factors from other disciplines such as economics and sociology should be integrated together in a proposed model to gain deeper insights into mechanisms for anxiety reduction. Finally, experimental evidence will be welcome in the future in order to ascertain the relations among flow, self-efficacy, and self-esteem, to finally target anxiety reduction.
A final note may stress how the elucidation of such psychological processes may help in designing strategies to build psychological sustainability, within or without clinical population and context. As anxiety is a common human experience, over and above its clinical relevance, more and more strategies and techniques are requested to manage it and to cope with it. Flow theory represents an option which can be implemented in terms of everyday ordinary activities, with very simple and clear ways of knowing how. Creating simple and available strategies for improving human psychological sustainability of commonly and optimally experienced activities and contexts (such as academic study) therefore represents a way to foster human resilience in the face of challenging, demanding, and stressing requests.