research paper about students' behavior

• Review article
• Open access
• Published: 30 January 2021

Understanding students’ behavior in online social networks: a systematic literature review

• Maslin Binti Masrom 1 ,
• Abdelsalam H. Busalim   ORCID: orcid.org/0000-0001-5826-8593 2 ,
• Hassan Abuhassna 3 &
• Nik Hasnaa Nik Mahmood 1  

International Journal of Educational Technology in Higher Education volume  18 , Article number:  6 ( 2021 ) Cite this article

181k Accesses

29 Citations

20 Altmetric

Metrics details

The use of online social networks (OSNs) has increasingly attracted attention from scholars’ in different disciplines. Recently, student behaviors in online social networks have been extensively examined. However, limited efforts have been made to evaluate and systematically review the current research status to provide insights into previous study findings. Accordingly, this study conducted a systematic literature review on student behavior and OSNs to explicate to what extent students behave on these platforms. This study reviewed 104 studies to discuss the research focus and examine trends along with the important theories and research methods utilized. Moreover, the Stimulus-Organism-Response (SOR) model was utilized to classify the factors that influence student behavior. This study’s results demonstrate that the number of studies that address student behaviors on OSNs have recently increased. Moreover, the identified studies focused on five research streams, including academic purpose, cyber victimization, addiction, personality issues, and knowledge sharing behaviors. Most of these studies focused on the use and effect of OSNs on student academic performance. Most importantly, the proposed study framework provides a theoretical basis for further research in this context.

Introduction

The rapid development of Web 2.0 technologies has caused increased usage of online social networking (OSN) sites among individuals. OSNs such as Facebook are used almost every day by millions of users (Brailovskaia et al. 2020 ). OSNs allow individuals to present themselves via virtual communities, interact with their social networks, and maintain connections with others (Brailovskaia et al. 2020 ). Therefore, the use of OSNs has continually attracted young adults, especially students (Kokkinos and Saripanidis 2017 ; Paul et al. 2012 ). Given the popularity of OSNs and the increased number of students of different ages, many education institutions (e.g., universities) have used them to market their educational programs and to communicate with students (Paul et al. 2012 ). The popularity and ubiquity of OSNs have radically changed education systems and motivated students to engage in the educational process (Lambić 2016 ). The children of the twenty-first century are technology-oriented, and thus their learning style differs from previous generations (Moghavvemi et al. 2017a , b ). Students in this era have alternatives to how and where they spend time to learn. OSNs enable students to share knowledge and seek help from other students. Lim and Richardson ( 2016 ) emphasized that one important advantage of OSNs as an educational tool is to increase connections between classmates, which increases information sharing. Furthermore, the use of OSNs has also opened new communication channels between students and teachers. Previous studies have shown that students strengthened connections with their teachers and instructors using OSNs (e.g., Facebook, and Twitter). Therefore, the characteristics and features of OSNs have caused many students to use them as an educational tool, due to the various facilities provided by OSN platforms, which makes learning more fun to experience (Moghavvemi et al. 2017a ). This has caused many educational institutions to consider Facebook as a medium and as a learning tool for students to acquire knowledge (Ainin et al. 2015 ).

OSNs including Facebook, YouTube, and Twitter have been the most utilized platforms for education purposes (Akçayır and Akçayır 2016 ). For instance, the number of daily active users on Facebook reached 1.73 billion in the first quarter of 2020, with an increase of 11% compared to the previous year (Facebook 2020 ). As of the second quarter of 2020, Facebook has over 2.7 billion active monthly users (Clement 2020 ). Lim and Richardson ( 2016 ) empirically showed that students have positive perceptions toward using OSNs as an educational tool. A review of the literature shows that many studies have investigated student behaviors on these sites, which indicates the significance of the current review in providing an in-depth understanding of student behavior on OSNs. To date, various studies have investigated why students use OSNs and explored different student behaviors on these sites. Although there is an increasing amount of literature on this emerging topic, little research has been devoted to consolidating the current knowledge on OSN student behaviors. Moreover, to utilize the power of OSNs in an education context, it is important to study and understand student behaviors in this setting. However, current research that investigates student behaviors in OSNs is rather fragmented. Thus, it is difficult to derive in-depth and meaningful implications from these studies. Therefore, a systematic review of previous studies is needed to synthesize previous findings, identify gaps that need more research, and provide opportunities for further research. To this end, the purpose of this study is to explore the current literature in order to understand student behaviors in online social networks. Accordingly, a systematic review was conducted in order to collect, analyze, and synthesize current studies on student behaviors in OSNs.

This study drew on the Stimulus-Organism-Response (SOR) model to classify factors and develop a framework for better understanding of student behaviors in the context of OSNs. The S-O-R model suggests that various aspects of the environment (S), incite individual cognitive and affective reactions (O), which in turn derives their behavioral responses (R) (Mehrabian and Russell 1974 ). In order to achieve effective results in a clear and understandable manner, five research questions were proposed as shown below.

What was the research regional context covered in previous studies?

What were the focus and trends of previous studies?

What were the research methods used in previous studies?

What were the major theories adopted in previous studies?

What important factors were studied to understand student usage behaviors in OSNs?

This paper is organized as follows. The second section discusses the concept of online social networks and their definition. The third section describes the review method used to extract, analyze, and synthesize studies on student behaviors. The fourth section provides the result of analyzing the 104 identified primary studies and summarizes their findings based on the research questions. The fifth section provides a discussion on the results based on each research question. The sixth section highlights the limitations associated with this study, and the final section provides a conclusion of the study.

• Online social networks

Since online social networks such as Facebook were introduced last decade, they have attracted millions of users and have become integrated into our daily routines. OSNs provide users with virtual spaces where they can find other people with similar interests to communicate with and share their social activities (Lambić et al. 2016 ). The concept of OSNs is a combination of technology, information, and human interfaces that enable users to create an online community and build a social network of friends (Borrero et al. 2014 ). Kum Tang and Koh ( 2017 ) defined OSNs as “web-based virtual communities where users interact with real-life friends and meet other people with shared interests” . A more detailed and well-cited definition of OSN was introduced by Boyd and Ellison ( 2008 ) who defined OSNs as “web-based services that allow individuals to (1) construct a public or semipublic profile within a bounded system, (2) articulate a list of other users with whom they share a connection, and (3) view and traverse their list of connections and those made by others within the system” . Due to its popularity, many researches have examined the effect of OSNs on different disciplines such as business (Kujur and Singh 2017 ), healthcare (Chung 2014 ; Lin et al. 2016 ; Mano 2014 ), psychology (Pantic 2014 ), and education (Hamid et al.  2016 , 2015 ; Roblyer et al. 2010 ).

The heavy use of OSNs by students has led many studies to examine both positive and negative effects of these sites on students, including the time spent on OSNs usage (Chang and Heo 2014 ; Wohn and Larose 2014 ), engagement in academic activities (Ha et al. 2018 ; Sheeran and Cummings 2018 ), as well as the effect of OSN on students’ academic performance. Lim and Richardson ( 2016 ) stated that the main reasons for students to use OSNs as an educational tool is to increase their interactions and establish connections with classmates. Tower et al. ( 2014 ) found that OSN platforms such as Facebook have the potential to improve student self-efficacy in learning and develop their learning skills to a higher level. Therefore, some education institutions have started to develop their own OSN learning platforms (Tally 2010 ). Mazman and Usluel ( 2010 ) highlighted that using OSNs for educational and instructional contexts is an idea worth developing because students spend a lot of time on these platforms. Yet, the educational activities conducted on OSNs are dependent on the nature of the OSNs used by the students (Benson et al. 2015 ). Moreover, for teaching and learning, instructors have begun using OSNs platforms for several other purposes such as increasing knowledge exchanges and effective learning (Romero-Hall 2017 ). On the other hand, previous studies have raised some challenges of using OSNs for educational purposes. For example, students tend to use OSNs as a social tool for entraining rather than an educational tool (Baran 2010 ; Gettman and Cortijo 2015 ). Moreover, the active use of OSNs on daily basis may develop students’ negative behavior such as addiction and distraction. In this context, Kitsantas et al. ( 2016 ) found that college students in the United States reported some concerns such as the OSNs usage can turn into addictive behavior, distraction, privacy threats, the negative impact on their emotional health, and the inability to complete the tasks on time. Another challenge of using OSNs as educational tools is gender differences. Kim and Yoo ( 2016 ) found some differences between male and female students concerning the negative impact of OSNs, for example, female students are more conserved about issues related to security, and the difficulty of task/work completion. Furthermore, innovation is a key aspect in the education process (Serdyukov 2017 ), however, using OSNs as an educational tool, students could lose creativity due to the easy access to everything using these platforms (Mirabolghasemi et al. 2016 ).

Review method

This study employed a Systematic Literature Review (SLR) approach in order to answer the research questions. The SLR approach creates a foundation that advances knowledge and facilitates theory development for a specific topic (Webster and Watson 2002 ). Kitchenham and Charters ( 2007 ) defined SLR as a process of identifying, evaluating, and synthesizing all available research that is related to research questions, area of research, or new phenomenon. This study follows Kitchenhand and Charters’ guidelines (Kitchenham 2004 ), which state that the SLR approach involves three main stages: planning the review, conducting the review, and reporting the review results. There are several motivations for carrying out this systematic review. First, to summarize existing knowledge and evidence on research related to OSNs such as the theories, methods, and factors that influence student behaviors on these platforms. Second, to discover the current research focus and trends in this setting. Third, to propose a framework that classifies the factors that influence student behaviors on OSNs using the S-O-R model. The reasons for using S-O-R model in this study are twofold. First, S-O-R is a crucial theoretical framework to understand individuals’ behavior, and it has been extensively used in previous studies on consumer behavior (Wang and Chang 2013 ; Zhang et al. 2014 ; Zhang and Benyoucef 2016 ), and online users’ behavior (Islam et al. 2018 ; Luqman et al. 2017 ). Second, using the S-O-R model can provide a structured manner to understand the effect of the technological features of OSNs as environmental stimuli on individuals’ behavior (Luqman et al. 2017 ). Therefore, the application of the S-O-R model can provide a guide in the OSNs literature to better understand the potential stimulus and organism factors that drive a student’s behavioral responses in the context of OSNs. The SLR was guided by five research questions (see “ Introduction ” section), which provide an in-depth understanding of the research topic. The rationale and motivation beyond considering these questions are stated in Table 1 .

Stage one: Planning

Before conducting any SLR, it is necessary to clarify the goal and the objectives of the review (Kitchenham and Charters 2007 ). After identifying the review objectives and the research questions, in the planning stage, it is important to design the review protocol that will be used to conduct the review (Kitchenham and Charters 2007 ). Using a clear review protocol will help define criteria for selecting the literature source, database, and search keywords. Review protocol reduce research bias and specifies the research method used to perform a systematic review (Kitchenham and Charters 2007 ). Figure  1 shows the review protocol used for this study.

Review protocol

Stage two: Conducting the review

In this stage relevant literature was collected using a two-stage approach, which was followed by the removal of duplicated articles using Mendeley software. Finally, the researchers applied selection criteria to identify the most relevant articles to the current review. The details of each step of this stage are discussed below:

Literature identification and collection

This study used a two-stage approach (Webster and Watson 2002 ) to identify and collect relevant articles for review. In the first stage, this study conducted a systematic search to identify studies that address student behaviors and the use of online social networks using selected academic databases, including the Web of Science, Wiley Online Library ScienceDirect, Scopus, Emerald, and Springer. The choice of these academic databases is consistent with previous SLR studies (Ahmadi et al. 2018 ; Balaid et al. 2016 ; Busalim and Hussin 2016 ). Derived from the structure of this review and the research questions, these online databases were searched by focusing on title, abstract, and keywords. The search in these databases started in May 2019 using the specific keywords of “students’ behavior”, “online social networking”, “social networking sites”, and “Facebook”. This study performed several searches in each database using Boolean logic operators (i.e., AND and OR) to obtain a large number of published studies related to the review topic.

The results from this stage were 164 studies published between 2010 and 2018. In the second stage, important peer-reviewed journals were checked to ensure that all relevant articles were collected. We used the same keywords to search on information systems and education journals such as Computers in Human Behavior, International Journal of Information Management, Computers and Education, and Education and Information Technologies. These journals among the top peer-reviewed journals that publish topics related to students' behavior, education technologies, and OSNs. The result from both stages was 188 studies related to student behaviors in OSN. Table 2 presents the journals with more than two articles published in these areas.

Study selection

Following the identification of these studies, and after deleting duplicated studies, this study examined title, abstract, or the content of each study using three selection criteria: (1) a focus on student behavior; (2) an examination of the context of online social networks; (3) and a qualification as an empirical study. After applying these criteria, a total of 96 studies remained as primary studies for review. We further conducted a forward manual search on a reference list for the identified primary studies, through which an additional 8 studies were identified. A total of 104 studies were collected. As depicted in Fig.  2 , the frequency of published articles related to student behaviors in online social networks has gradually increased since 2010. In this regard, the highest number of articles were published in 2017. We can see that from 2010 to 2012 the number of published articles was relatively low and significant growth in published articles was seen from 2013 to 2017. This increase reveals that studying the behavior of students on different OSN platforms is increasingly attractive to researchers.

Timeline of publication

For further analysis, this study summarized the key topics covered during the review timeline. Figure  3 visualizes the development of OSNs studies over the years. Studies in the first three years (2010–2012) revolved around the use of OSNs by students and the benefits of using these platforms for educational purposes. The studies conducted between 2013 and 2015 mostly focused on the effect of using OSNs on student academic performance and achievement. In addition, in the same period, several studies examined important psychological issues associated with the use of OSNs such as anxiety, stress, and depression. In the years 2016 to 2018, OSNs studies were expanded to include cyber victimization behavior, OSN addiction behavior such as Facebook addiction, and how OSNs provide a collaborative platform that enables students to share information with their colleagues.

Evolution of OSNs studies over the years

Review results

To analyze the identified studies, this study guided its review using four research questions. Using research questions allows the researcher to synthesize findings from previous studies (Chan et al. 2017 ). The following subsection provides a detailed discussion of each of these research questions.

RQ1: What was the research regional context covered in previous studies?

As shown in Fig.  3 , most primary studies were conducted in the United States (n = 37), followed by Asia (n = 21) and Europe (n = 15). Relatively few studies were conducted in Australia, Africa, and the Middle East (n = 6 each), and only five studies were conducted in more than one country. Most of these empirical studies used university or college students to examine and validate the research models. Furthermore, many of these studies examined student behavior by considering Facebook as an online social network (n = 58) and a few studies examined student behavior on Microblogging platforms like Twitter (n = 7). The rest of the studies used multiple online social networks such as Instagram, YouTube, and Moodle (n = 31).

As shown in Fig.  4 , most of the reviewed studies are conducted in the United States (US). Furthermore, these studies considered Facebook as the main OSN platform. However, the focus on examining the usage behavior of Facebook in Western countries, particularly the US, is one of the challenges of Facebook research, because Facebook is used in many countries with 80% of its users are outside of the US (Peters et al. 2015 ).

Distribution of published studies by region

RQ2: What were the focus and trends of previous studies?

The results indicate that the identified primary studies for student behaviors on online social networks covered a wide spectrum of different research contexts. Further examination shows that there are five research streams in the literature.

The first research stream focused on using OSNs for academic purposes. The educational usage of OSNs relies on their purpose of use. OSNs can improve student engagement in a course and provide them with a sense of connection to their colleagues (Lambić 2016 ). However, the use of OSNs by students can affect their education as students can easily shift from using OSNs for educational to entertainment purposes. Thus, many studies under this stream focus on the effect of OSNs use on student academic performance. For instance, Lambić ( 2016 ) examined the effect of frequent Facebook use on the academic performance of university students. The results showed that students using Facebook as an educational tool to facilitate knowledge sharing and discussion positively impacted academic performance. Consistent with this result, Ainin et al. ( 2015 ) found that data from 1165 university students revealed a positive relationship between Facebook use and student academic performance. On the other hand, Paul et al. ( 2012 ) found that time spent on OSNs negative impacted student academic behavior. Moreover, the results statistically highlight that increased student attention spans resulted in increased time spent on OSNs, which eventually results in a negatively effect on academic performance. The results from Karpinski et al. ( 2013 ) showed that the effect of OSNs usage on student academic performance could differ from one country to another.

In summary, previous studies on the relationship between OSN use and academic performance show mixed results. From the reviewed studies, there were disparate results due to a few reasons. For example, recent studies found that multitasking plays an important role in determining the relationship between OSN usage and student academic performance. Karpinski et al. ( 2013 ) found a negative relationship between using social network sites (SNSs) and Grade Point Average (GPA) that was moderated by multitasking. Moreover, results from Junco ( 2015 ), illustrated that besides multitasking, student class rank is another determinant of the relationship between OSN platforms like Facebook and academic performance. The results revealed that senior students spent significantly less time on Facebook while doing schoolwork than freshman and sophomore students.

The second research stream is related to cyber victimization. Studies in this stream focused on negative interactions on OSNs like Facebook, which is the main platform where cyber victimization occurs (Kokkinos and Saripanidis 2017 ). Moreover, most studies in this stream examined the cyberbullying concept on OSNs. Cyberbullying is defined as “any behavior performed through electronic media by individuals or groups of individuals that repeatedly communicates hostile or aggressive messages intended to inflict harm or discomfort on others” (Tokunaga 2010 , p. 278). For instance, Gahagan et al. ( 2016 ) investigated the experiences of college students with cyberbullying on SNSs, and the results showed that 46% of the tested sample witnessed someone who had been bullied through the use of SNSs. Walker et al. ( 2011 ) conducted an exploratory study among undergraduate students to investigate their cyberbullying experiences. The results of the study highlighted that the majority of respondents knew someone who had been bullied on SNSs (Benson et al. 2015 ).

The third research stream focused on student addiction to OSNs use. Recent research has shown that excessive OSN use can lead to addictive behavior among students (Shettar et al. 2017 ). In this stream, Facebook was the main addictive ONS platform that was investigated (Shettar et al. 2017 ; Hong and Chiu 2016 ; Koc and Gulyagci 2013 ). Facebook addiction is defined as an excessive attachment to Facebook that interferes with daily activities and interpersonal relationships (Elphinston and Noller 2011 ). According to Andreassen et al. ( 2012 ), Facebook addiction has six general characteristics including salience, tolerance, mood modification, withdrawal, conflict, and relapse. As university students frequently have high levels of stress due to various commitments, such as assignment deadlines, exams, and high pressure to perform, they tend to use Facebook for mood modification (Brailovskaia and Margraf 2017 ; Brailovskaia et al. 2018 ). On further analysis, it was noticed that Facebook addiction among students was associated with other factors such as loneliness (Shettar et al. 2017 ), personality traits (i.e., openness agreeableness, conscientiousness, emotional stability, and extraversion) (Błachnio et al. 2017 ; Tang et al. 2016 ), and physical activities (Brailovskaia et al. 2018 ). Studies have examined student addiction behavior on different OSNs platforms. For instance, Ndasauka et al. ( 2016 ), empirically examined excessive Twitter use among college students. Kum Tang and Koh ( 2017 ) investigated the prevalence of different addiction behaviors (i.e., food and shopping addiction) and effective disorders among college students. In addition, a study by Chae and Kim (Chae et al. 2017 ) examined psychosocial differences in ONS addiction between female and male students. The results of the study showed that female students had a higher tendency towards OSNs addiction than male students.

The fourth stream of research highlighted in this review focused on student personality issues such as self-disclosure, stress, depression, loneliness, and self-presentation. For instance, Chen ( 2017 ) investigated the antecedents that predict positive student self-disclosure on SNSs. Tandoc et al. ( 2015 ) used social rank theory and Facebook envy to test the depression scale between college students. Skues et al. ( 2012 ) examined the relationship between three traits in the Big Five Traits model (neuroticism, extraversion, and openness) and student Facebook usage. Chang and Heo ( 2014 ) investigated the factors that explain the disclosure of a student’s personal information on Facebook.

The fifth reviewed research stream focused on student knowledge sharing behavior. For instance, Kim et al. ( 2015 ) identified the personal factors (self-efficacy) and environmental factors (strength of social ties and size of social networks) that affect information sharing behavior amongst university students. Eid and Al-Jabri ( 2016 ) examined the effect of various SNS characteristics (file sharing, chatting and online discussion, content creation, and enjoyment and entertainment) on knowledge sharing and student learning performance. Moghavvemi et al. ( 2017a , b ) examined the relationship between enjoyment, perceived status, outcome expectations, perceived benefits, and knowledge sharing behavior between students on Facebook. Figure  5 provides a mind map that shows an overview of the research focus and trends found in previous studies.

Reviewed studies research focus and trends

RQ3: What were the research methods used in previous studies?

As presented in Fig.  6 , previous studies used several research methods to examine student behavior on online social networks. Surveys were the method used most frequently in primary studies to understand the different types of determinants that effect student behaviors on online social networks, followed by the experiment method. Studies used the experiment method to examine the effect of online social networks content and features on student behavior, For example, Corbitt-Hall et al. ( 2016 ) had randomly assigned students to interact with simulated Facebook content that reflected various suicide risk levels. Singh ( 2017 ) used data mining techniques to collect student interaction data from different social networking sites such as Facebook and Twitter to classify student academic activities on these platforms. Studies that investigated student intentions, perceptions, and attitudes towards OSNs used survey data. For instance, Doleck et al. ( 2017 ) distributed an online survey to college students who used Facebook and found that perceived usefulness, attitude, and self-expression were influential factors towards the use of online social networks. Moreover, Ndasauka et al. ( 2016 ) used the survey method to assess the excessive use of Twitter among college students.

Research method distribution

RQ4: What were the major theories adopted in previous studies?

The results from the SLR show that previous studies used several theories to understand student behavior in online social networks. Table 3 depicts the theories used in these studies, with Use and Gratification Theory (UGT) being the most popular theory use to understand students' behaviors (Asiedu and Badu 2018 ; Chang and Heo 2014 ; Cheung et al. 2011 ; Hossain and Veenstra 2013 ). Furthermore, the social influence theory and the Big Five Traits model were applied in at least five studies each. The theoretical insights into student behaviors on online social networks provided by these theories are listed below:

Motivation aspect: since the advent of online social networks, many studies have been conducted to understand what motivates students to use online social networks. Theories such as UGT have been widely used to understand this issue. For example, Hossain and Veenstra ( 2013 ) conducted an empirical study to investigate what drives university students in the United States of America to use Social Networking Sites (SNSs) using the theoretical foundation of UGT. The study found that the geographic or physical displacement of students affects the use and gratification of SNSs. Zheng Wang et al. ( 2012a , b ) explained that students are motivated to use social media by their cognitive, emotional, social, and habitual needs as well as that all four categories significantly drive students to use social media.

Social-related aspect: Social theories such as Social Influence Theory, Social Learning Theory, and Social Capital Theory have also been used in several previous studies. Social Influence Theory determines what individual behaviors or opinions are affected by others. Venkatesh, Morris, Davis, and Davis (2003) defined social influence as “the degree to which an individual perceives that important others believe he or she should use a new system” . Cheung et al. ( 2011 ) applied Social Influence Theory to examine the effect of social influence factors (subjective norms, group norms, and social identity) on intentions to use online social networks. The empirical results from 182 students revealed that only Group Norms had a significant effect on student intentions to use OSNs. Other studies attempted to empirically examine the effect of other social theories. For instance, Liu and Brown ( 2014 ) adapted Social Capital Theory to investigate whether college students' self-disclosure on SNSs directly affected their social capital. Park et al. ( 2014a , b ) investigated the effect of using SNSs on university student learning outcomes using social learning theory.

Behavioral aspect: This study have noticed that the Theory of Planned Behavior (TPB), Theory of Reasoned Action (TRA), Technology Acceptance Model (TAM), Unified Theory of Acceptance, and Use of Technology (UTAUT) were also utilized as a theoretical foundation in a number of primary studies. These theories have been widely applied in the information systems (IS) field to provide insights into information technology adoption among individuals (Zhang and Benyoucef 2016 ). In the context of online social networks, there were empirical studies that adapted these theories to understand student usage behaviors towards online social networks such as Facebook. For example, Doleck et al. ( 2017 ) applied TAM to investigate college student usage intentions towards SNSs. Chang and Chen ( 2014 ) applied TRA and TPB to investigate why college students share their location on Facebook. In addition, a recent study used UTAUT to examine student perceptions towards using Facebook as an e-learning platform (Moghavvemi et al. 2017a , b ).

RQ5: What important factors were studied to understand student usage behaviors in OSNs?

Throughout the SLR, this study has been able to identify the potential factors that influence student behaviors in online social networks. Furthermore, to synthesize these factors and provide a comprehensive overview, this study proposed a framework based on the Stimulus-Organism-Response (S-O-R) model. The S-O-R model was developed in environmental psychology by Mehrabian and Russell ( 1974 ). According to Mehrabian and Russell ( 1974 ), environmental cues act as stimuli that can affect an individual’s internal cognitive and affective states, which subsequently influences their behavioral responses. To do so, this study extracted all the factors examined in 104 identified primary studies and classified them into three key concepts: stimulus, organism, and response. The details on the important factors of each component are presented below.

Online social networks stimulus

Stimulus factors are triggers that encourage or prompt students to use OSNs. Based on the SLR results, there are three stimulus dimensions: social stimulus, personal stimulus, and OSN characteristics. Social stimuli are cues embedded in the OSN that drive students to use these platforms. As shown in Fig.  7 , this study has identified six social stimulus factors including social support, social presence, social communication, social enhancement, social network size, and strength of social ties. Previous studies found that social aspects are a potential driver of student usage of OSNs. For instance, Kim et al. ( 2011 ) explored the motivation behind college student use of OSNs and found that seeking social support is one of the primary usage triggers. Lim and Richardson ( 2016 ) stated that using OSNs as educational tools will increase interactions and establish connections between students, which will enhance their social presence. Consistent with this, Cheung et al. ( 2011 ) found that social presence and social enhancement both have a positive effect on student use of OSNs. Other studies have tested the effect of other social factors such as social communication (Lee 2015 ), social network size, and strength of social ties (Chang and Heo 2014 ; Kim et al. 2015 ). Personal stimuli are student motivational factors associated with a specific state that affects their behavioral response. As depicted in Table 4 , researchers have tested different personal student needs that stimulate OSN usage. For instance, Zheng Wang et al. ( 2012a , b ) examined the emotional, social, and cognitive needs that drive students to use OSNs. Moghavvemi et al. ( 2017a , b ) empirically showed that students with a hedonic motivation were willing to use Facebook as an e-learning tool.

Classification framework for student behaviors in online social networks

OSN website characteristics are stimuli related to the cues implanted in an OSN website. In the reviewed studies, it was found that the most well studied OSN characteristics are usefulness and ease of use. Ease of use refers to student perceptions on the extent to which OSN are easy to use whereas usefulness refers to the degree that students believed that using OSN was helpful in enhancing their task performance (Arteaga Sánchez et al. 2014 ). Although student behaviors in OSNs have been widely studied, few studies have focused on OSN characteristics that stimulate student behaviors. For example, Eid and Al-Jabri ( 2016 ) examined the effect of OSN characteristics such as chatting, discussion, content creation, and file sharing. The results showed that file sharing, chatting, and discussion had a positive impact on student knowledge sharing behavior. In summary, Table 4 shows the stimulus factors identified in previous studies and their classification.

Online social networks organisms

Organism in this study’s framework refers to student internal evaluations towards using OSNs. There are four types of organism factors that have been highlighted in the literature. These types include personality traits, values, social, and cognitive reactions. Student personality traits influence the use of OSNs (Skues et al. 2012 ). As shown in Table 4 , self-esteem and self-disclosure were the most examined personality traits associated with student OSN behaviors. Self-esteem refers to an individual’s emotional evaluation of their own worth (Chen 2017 ). For example, Wang et al. ( 2012a , b ) examined the effect of the Big Five personality traits on student use of specific OSN features. The results found that students with high self-esteem were more likely to comment on other student profiles. Self-disclosure refers to the process by which individuals share their feelings, thoughts, information, and experiences with others (Dindia 1995 ). Previous studies have examined student self-disclosure in OSNs to explore information disclosure behavior (Chang and Heo 2014 ), location disclosure (Chang and Chen 2014 ), self-disclosure, and mental health (Zhang 2017 ). The second type of organism factors is value. It has been noticed that there are several value related factors that affect student internal organisms in OSNs. As shown in Table 4 , entertainment and enjoyment factors were the most common value examined in previous studies. Enjoyment is one of the potential drivers of student OSN use (Nawi et al. 2017 ). Eid and Al-Jabri ( 2016 ) found that YouTube is the most dominant OSN platform used by students for enjoyment and entertainment. Moreover, enjoyment and entertainment directly affected student learning performance.

Social organism refers to the internal social behavior of students that affect their use of OSNs. Students interact with OSN platforms when they experience positive social reactions. Previous studies have examined some social organism factors including relationship with faculty members, engagement, leisure activities, social skills, and chatting and discussion. The fourth type of organism factors is cognitive reactions. Parboteeah et al. ( 2009 ) defined cognitive reaction as “the mental process that occurs in an individual’s mind when he or she interacts with a stimulus” . The positive or negative cognitive reaction of students influences their responses towards OSNs. Table 5 presents the most common organism reactions that effect student use of OSNs.

Online social networks response

In this study’s framework, response refers to student reactions to OSNs stimuli and organisms. As shown in Table 5 , academic related behavior and negative behavior are the most common student responses towards OSNs. Studying the effect of OSN usage on student academic performance has been the most common research topic (Lambić 2016 ; Paul et al. 2012 ; Wohn and Larose 2014 ). On the other hand, other studies have examined the negative behavior of students during their usage of ONS, mostly towards ONS addiction (Hong and Chiu 2016 ; Shettar et al. 2017 ) or cyberbullying (Chen 2017 ; Gahagan et al. 2016 ). Table 6 summarizes student responses associated with OSNs use in previous studies.

Discussion and implications

The last two decades have witnessed a dramatic growth in the number of online social networks used among the youth generation. Examining student behaviors on OSN platforms has increasingly attracted scholars. However, there has been little effort to summarize and synthesize these findings. In this review study, a systematic literature review was conducted to synthesize previous research on student behaviors in OSNs to consolidate the factors that influence student behaviors into a classification framework using the S-O-R model. A total of 104 journal articles were identified through a rigorous and systematic search procedure. The collected studies from the literature show an increasing interest in the area ever since 2010. In line with the research questions, our analysis offers insightful results of the research landscape in terms of research regional context, studies focus trends, methodological trends, factors, and theories leveraged. Using the S-O-R model, we synthesized the reviewed studies highlighting the different stimuli, organism, and response factors. We synthesize and classify these factors into social stimuli, personal stimuli, and OSN characteristics, organism factors; personality traits, value, social, and cognitive reaction, and response; academic related behavior, negative behavior, and other responses.

Research regional perspective

The first research question focused on research regional context. The review revealed that most of the studies were conducted in the US followed by European countries, with the majority focusing on Facebook. The results show that the large majority of the studies were based on a single country. This indicates a sustainable research gap in examining the multi-cultural factors in multiple countries. As OSN is a common phenomenon across many counties, considering the culture and background differences can play an essential role in understanding students’ behavior on these platforms. For example, Ifinedo ( 2016 ) collected data from four countries in America (i.e., USA, Canada, Argentina, and Mexico) to understand students’ pervasive adoption of SNSs. The results from the study revealed that the individualism–collectivism culture factor has a positive impact on students' pervasive adoption behavior of SNSs, and the result reported high level of engagement from students who have more individualistic cultures. In the same manner, Kim et al. ( 2011 ) found some cultural differences in use of the SNSs platforms between Korean and US students. For example, considering the social nature of SNSs, the study found that Korean students rely more on online social relationships to obtain social support, where US students use SNSs to seek entertainment. Furthermore, Karpinski et al. ( 2013 ) empirically found significant differences between US and European students in terms of the moderating effect of multitasking on the relationship between SNS use and academic achievement of students. The confirms that culture issues may vary from one country to another, which consequently effect students’ behavior to use OSNs (Kim et al. 2011 ).

Studies focus and trends

The second research question of this review focused on undersigning the topics and trends that have been discussed in extant studies. The review revealed evidence of five categories of research streams based on the research focus and trend. As shown in Fig.  5 , most of the reviewed studies are in the first stream, which is using OSNs for academic purposes. Moreover, the trend of these studies in this stream focus on examining the effect of using OSNs on students’ academic performance and investigating the use of OSNs for educational purposes. However, a number of other trends are noteworthy. First, as cyber victimization is a relatively new concept, most of the studies provide rigorous effort in exporting the concept, and the reasons beyond its existence among students; however, we have noticed that no effort has been made to investigate the consequences of this negative behavior on students’ academic performance, social life, and communication. Second, we identified only two studies that examined the differences between undergraduate and postgraduate students in terms of cyber victimization. Therefore, there are many avenues for further research to untangle the demographic, education level, and cultural differences in this context. Third, our analysis revealed that Facebook was the most studied ONS platform in terms of addiction behavior, however, over the last ten years, the rapid growth of using image-based ONS such as Instagram and Pinterest has attracted many students (Alhabash and Ma 2017 ). For example, Instagram represents the fastest growing OSNs among young adult users age between 18 and 29 years old (Alhabash and Ma 2017 ). The overwhelming majority of the studies focus on Facebook users, and very few studies have examined excessive Instagram use (Kırcaburun and Griffiths 2018 ; Ponnusamy et al. 2020 ). Although OSNs have many similar features, each platform has unique features and a different structure. These differences in OSNs platforms urge further research to investigate and understand the factors related to excessive and addiction use by students (Kircaburun and Griffiths 2018 ). Therefore, based on the current research gaps, future research agenda including three topics/trend need to be considered. We have developed research questions for each topic as a direction for any further research as shown in Table 7 .

Theories and research methods

The third and fourth research questions focused on understanding the trends in terms of research methods and theories leveraged in existing studies. In relation to the third research question, the review highlighted evidence of the four research methods (i.e., survey, experiment, focus group/interview, and mix method) with a heavy focus on using a quantitative method with the majority of studies conducting survey. This may call for utilizing a variety of other research methods and research design to have more in-depth understanding of students’ behavior on OSN. For example, we noticed that few studies leveraged qualitative methods such as interviews and focus groups (n = 5). In addition, using mix method may derive more results and answer research questions that other methods cannot answer (Tashakkori and Teddlie 2003 ). Experimental methods have been used sparingly (n = 10), this may trigger an opportunity for more experimental research to test different strategies that can be used by education institutions to leverage the potential of OSN platforms in the education process. Moreover, considering that students’ attitude and behavior will change over time, applying longitudinal research method may offer opportunities to explore students’ attitude and behavior patterns over time.

The fourth research question focused on understanding the theoretical underpinnings of the reviewed studies. The analysis revealed two important insights; (1) a substantial number of the reviewed studies do not explicitly use an applied theory, and (2) out of the 34 studies that used theory, nine studies applied UGT to understand the motivation beyond using the OSN. Our findings categorized these theories under three aspects; motivational, social, and behavioral. While each aspect and theory offers useful lenses in this context, there is a lack of leveraging other theories in the extant literature. This motivates researchers to underpin their studies in theories that provide more insights into these three aspects. For example, majority of the studies have applied UGT to understand students’ motivate for using OSNs. However, using other motivational theories could uncover different factors that influence students' motivation for using OSNs. For example, self-determination theory (SDT) focuses on the extent to which individual’s behavior is self-motivated and determined. According to Ryan and Deci ( 2000 ), magnitude and types both shape individuals’ extrinsic motivation. The extrinsic motivation is a spectrum and depends on the level of self-determination (Wang et al. 2019 ). Therefore, the continuum aspect proposed by SDT can provide in-depth understanding of the extrinsic motivation. Wang et al. ( 2016 ) suggested that applying SDT can play a key role in understanding SNSs user satisfaction.

Another theoretical perspective that is worth further exploration relates to the psychological aspect. Our review results highlighted that a considerable number of studies focused on an important issue arising from the daily use of OSNs, such as excessive use/addiction (Koc and Gulyagci 2013 ; Shettar et al. 2017 ), Previous studies have investigated the behavior aspect beyond these issues, however, understanding the psychological aspect of Facebook addiction is worth further investigation. Ryan et al. ( 2014 ) reviewed Facebook addiction related studies and found that Facebook addiction is also linked to psychological factors such as depression and anxiety.

Factors that influence students behavior: S-O-R Framework

The fifth research question focused on determining the factors studies in the extant literature. The review analysis showed that stimuli factors included social, personal, and OSNs website stimuli. However, different types of stimuli have received less attention than other stimuli. Most studies leveraged the social and students’ personal stimuli. Furthermore, few studies conceptualized the OSNs websites characterises in terms of students beliefs about the effect of OSNs features and functions (e.g., perceived ease of use, user friendly) on students stimuli; it would be significant to develop a typology of the OSNs websites stimuli and systematically examine their effect on students’ attitude and behavior. We recommend applying different theories (as mentioned in Theories and research methods section) as an initial step to further identify stimuli factors. The results also highlight that cognitive reaction plays an essential role in the organism dimension. When students encounter stimuli, their internal evaluation is dominated by emotions. Therefore, the cognitive process takes place between students’ usage behavior and their responses (e.g., effort expectancy). In this review, we reported few studies that examined the effect of the cognitive reaction of students.

Response factors encompass students’ reaction to OSNs platforms stimuli and organism. Our review revealed an unsurprisingly dominant focus on the academic related behavior such as academic performance. While it is important to examine the effect of various stimuli and organism factors on academic related behavior and OSNs negative behavior, the psychological aspect beyond OSNs negative behavior is equallty important.

Limitations

Similar to other systematic review studies, this study has some limitations. The findings of our review are constrained by only empirical studies (journal articles) that meet the inclusion criteria. For instance, we only used the articles that explicitly examined students’ behavior in OSNs. Moreover, other different types of studies such as conference proceedings are not included in our primary studies. Further research efforts can gain additional knowledge and understanding from practitioner articles, books and, white papers. Our findings offer a comprehensive conceptual framework to understand students’ behavior in OSNs; future studies are recommended to perform a quantitative meta-analysis to this framework and test the relative effect of different stimuli factors.

Conclusions

The use of OSNs has become a daily habit among young adults and adolescents these days (Brailovskaia et al. 2020 ). In this review, we used a rigorous systematic review process and identified 104 studies related to students’ behavior in OSNs. We systematically reviewed these studies and provide an overview of the current state of this topic by uncovering the research context, research focus, theories, and research method. More importantly, we proposed a classification framework based on S-O-R model to consolidate the factors that influence students in online social networks. These factors were classified under different dimensions in each category of the S-O-R model; stimuli (Social Stimulus, Personal Stimulus, and OSN Characteristics), organism (Personality traits, value, social, Cognitive reaction), and students’ responses (academic-related behavior, negative behavior, and other responses). This framework provides the researchers with a classification of the factors that have been used in previous studies which can motivate further research on the factors that need more empirical examination (e.g., OSN characteristics).

Availability of data and materials

Not applicable.

Ahmadi, H., Arji, G., Shahmoradi, L., Safdari, R., Nilashi, M., & Alizadeh, M. (2018). The application of internet of things in healthcare: A systematic literature review and classification. Universal Access in the Information Society . https://doi.org/10.1007/s10209-018-0618-4 .

Article   Google Scholar  

Ainin, S., Naqshbandi, M. M., Moghavvemi, S., & Jaafar, N. I. (2015). Facebook usage, socialization and academic performance. Computers and Education, 83, 64–73. https://doi.org/10.1016/j.compedu.2014.12.018 .

Akbari, E., Naderi, A., Simons, R.-J., & Pilot, A. (2016). Student engagement and foreign language learning through online social networks. Asian-Pacific Journal of Second and Foreign Language Education, 1 (1), 4. https://doi.org/10.1186/s40862-016-0006-7 .

Akcaoglu, M., & Bowman, N. D. (2016). Using instructor-led Facebook groups to enhance students’ perceptions of course content. Computers in Human Behavior, 65, 582–590. https://doi.org/10.1016/j.chb.2016.05.029 .

Akçayır, G., & Akçayır, M. (2016). Research trends in social network sites’ educational use: A review of publications in all SSCI journals to 2015. Review of Education, 4 (3), 293–319.

Alhabash, S., & Ma, M. (2017). A tale of four platforms: Motivations and uses of Facebook, Twitter, Instagram, and Snapchat among college students. Social Media + Society, 3 (1), 205630511769154. https://doi.org/10.1177/2056305117691544 .

Amador, P., & Amador, J. (2014). Academic advising via Facebook: Examining student help seeking. Internet and Higher Education, 21, 9–16. https://doi.org/10.1016/j.iheduc.2013.10.003 .

Andreassen, C. S., Torsheim, T., Brunborg, G. S., & Pallesen, S. (2012). Development of a Facebook addiction scale. Psychological Reports, 110 (2), 501–517. https://doi.org/10.2466/02.09.18.PR0.110.2.501-517 .

Arteaga Sánchez, R., Cortijo, V., & Javed, U. (2014). Students’ perceptions of Facebook for academic purposes. Computers and Education, 70, 138–149. https://doi.org/10.1016/j.compedu.2013.08.012 .

Asiedu, N. K., & Badu, E. E. (2018). Motivating issues affecting students’ use of social media sites in Ghanaian tertiary institutions. Library Hi Tech, 36 (1), 167–179. https://doi.org/10.1108/LHT-10-2016-0108 .

Asterhan, C. S. C., & Bouton, E. (2017). Teenage peer-to-peer knowledge sharing through social network sites in secondary schools. Computers & Education, 110, 16–34. https://doi.org/10.1016/j.compedu.2017.03.007 .

Balaid, A., Abd Rozan, M. Z., Hikmi, S. N., & Memon, J. (2016). Knowledge maps: A systematic literature review and directions for future research. International Journal of Information Management, 36 (3), 451–475.

Baran, B. (2010). Facebook as a formal instructional environment. British Journal of Educational Technology, 41 (6), 146–149. https://doi.org/10.1111/j.1467-8535.2010.01115.x .

Benson, V., & Filippaios, F. (2015). Collaborative competencies in professional social networking: Are students short changed by curriculum in business education? Computers in Human Behavior, 51, 1331–1339. https://doi.org/10.1016/j.chb.2014.11.031 .

Benson, V., Saridakis, G., & Tennakoon, H. (2015). Purpose of social networking use and victimisation: Are there any differences between university students and those not in HE? Computers in Human Behavior, 51, 867–872. https://doi.org/10.1016/j.chb.2014.11.034 .

Błachnio, A., Przepiorka, A., Senol-Durak, E., Durak, M., & Sherstyuk, L. (2017). The role of personality traits in Facebook and Internet addictions: A study on Polish, Turkish, and Ukrainian samples. Computers in Human Behavior, 68, 269–275. https://doi.org/10.1016/j.chb.2016.11.037 .

Borrero, D. J., Yousafzai, Y. S., Javed, U., & Page, L. K. (2014). Perceived value of social networking sites (SNS) in students’ expressive participation in social movements. Journal of Research in Interactive Marketing, 8 (1), 56–78. https://doi.org/10.1108/JRIM-03-2013-0015 .

Borrero, J. D., Yousafzai, S. Y., Javed, U., & Page, K. L. (2014). Expressive participation in Internet social movements: Testing the moderating effect of technology readiness and sex on student SNS use. Computers in Human Behavior, 30, 39–49. https://doi.org/10.1016/j.chb.2013.07.032 .

Boyd, D. M., & Ellison, N. B. (2008). Social network sites: Definition, history, and scholarship. Journal of Computer-Mediated Communication, 13 (1), 210–230. https://doi.org/10.1111/j.1083-6101.2007.00393.x .

Brailovskaia, J., & Margraf, J. (2017). Facebook Addiction Disorder (FAD) among German students—A longitudinal approach. PLoS ONE, 12 (12), 1–15. https://doi.org/10.1371/journal.pone.0189719 .

Brailovskaia, J., Ströse, F., Schillack, H., & Margraf, J. (2020). Less Facebook use—More well-being and a healthier lifestyle? An experimental intervention study. Computers in Human Behavior, 108 (March), 106332. https://doi.org/10.1016/j.chb.2020.106332 .

Brailovskaia, J., Teismann, T., & Teismann, T. (2018). Physical activity mediates the association between daily stress and Facebook Addiction Disorder. Computers in Human Behavior, 86, 199–204. https://doi.org/10.1016/j.chb.2018.04.045 .

Busalim, A. H., & Hussin, A. R. C. (2016). Understanding social commerce: A systematic literature review and directions for further research. International Journal of Information Management, 36 (6), 1075–1088. https://doi.org/10.1016/j.ijinfomgt.2016.06.005 .

Cain, J., Scott, D. R., Tiemeier, A. M., Akers, P., & Metzger, A. H. (2013). Social media use by pharmacy faculty: Student friending, e-professionalism, and professional use. Currents in Pharmacy Teaching and Learning, 5 (1), 2–8. https://doi.org/10.1016/j.cptl.2012.09.002 .

Chae, D., Kim, H., & Kim, Y. A. (2017). Sex differences in the factors influencing Korean College Students’ addictive tendency toward social networking sites. International Journal of Mental Health and Addiction . https://doi.org/10.1007/s11469-017-9778-3 .

Chan, T. K. H., Cheung, C. M. K., & Lee, Z. W. Y. (2017). The state of online impulse-buying research: A literature analysis. Information & Management, 54 (2), 204–217. https://doi.org/10.1016/j.im.2016.06.001 .

Chang, C. W., & Chen, G. M. (2014). College students’ disclosure of location-related information on Facebook. Computers in Human Behavior, 35, 33–38. https://doi.org/10.1016/j.chb.2014.02.028 .

Chang, C. W., & Heo, J. (2014). Visiting theories that predict college students’ self-disclosure on Facebook. Computers in Human Behavior, 30, 79–86. https://doi.org/10.1016/j.chb.2013.07.059 .

Chen, B., & Marcus, J. (2012). Students’ self-presentation on Facebook: An examination of personality and self-construal factors. Computers in Human Behavior, 28 (6), 2091–2099. https://doi.org/10.1016/j.chb.2012.06.013 .

Chen, H. (2017). Antecedents of positive self-disclosure online: An empirical study of US college students Facebook usage. Psychology Research and Behavior Management, 10, 147–153. https://doi.org/10.2147/PRBM.S136049 .

Cheung, C. M. K., Chiu, P. Y., & Lee, M. K. O. (2011). Online social networks: Why do students use facebook? Computers in Human Behavior, 27 (4), 1337–1343.

Chung, J. E. (2014). Social networking in online support groups for health: How online social networking benefits patients. Journal of Health Communication, 19 (6), 639–659. https://doi.org/10.1080/10810730.2012.757396 .

Čičević, S., Samčović, A., & Nešić, M. (2016). Exploring college students’ generational differences in Facebook usage. Computers in Human Behavior, 56, 83–92. https://doi.org/10.1016/j.chb.2015.11.034 .

Clement, J. (2020). Facebook: Number of monthly active users worldwide 2008–2020 Published by J. Clement, Aug 10, 2020 How many users does Facebook have? With over 2.7 billion monthly active users as of the second quarter of 2020, Facebook is the biggest social network world . Retrieved from https://www.statista.com/statistics/264810/number-of-monthly-active-facebook-users-worldwide/

Corbitt-Hall, D. J., Gauthier, J. M., Davis, M. T., & Witte, T. K. (2016). College Students’ responses to suicidal content on social networking sites: An examination using a simulated facebook newsfeed. Suicide and Life-Threatening Behavior, 46 (5), 609–624. https://doi.org/10.1111/sltb.12241 .

Deng, L., & Tavares, N. J. (2013). From Moodle to Facebook: Exploring students’ motivation and experiences in online communities. Computers & Education, 68, 167–176. https://doi.org/10.1016/j.compedu.2013.04.028 .

Dindia, K. (1995). Self-disclosure: A sense of balance. Contemporary psychology: A journal of reviews . Vol. 40. New York: Sage Publications. https://doi.org/10.1037/003319 .

Book   Google Scholar  

Doleck, T., Bazelais, P., & Lemay, D. J. (2017). Examining the antecedents of social networking sites use among CEGEP students. Education and Information Technologies, 22 (5), 2103–2123. https://doi.org/10.1007/s10639-016-9535-4 .

Eid, M. I. M., & Al-Jabri, I. M. (2016). Social networking, knowledge sharing, and student learning: The case of university students. Computers and Education, 99, 14–27. https://doi.org/10.1016/j.compedu.2016.04.007 .

Elphinston, R. A., & Noller, P. (2011). Time to Face It! Facebook intrusion and the implications for romantic jealousy and relationship satisfaction. Cyberpsychology, Behavior, and Social Networking, 14 (11), 631–635. https://doi.org/10.1089/cyber.2010.0318 .

Enskat, A., Hunt, S. K., & Hooker, J. F. (2017). A generational examination of instructional Facebook use and the effects on perceived instructor immediacy, credibility and student affective learning. Technology, Pedagogy and Education, 26 (5), 545–557. https://doi.org/10.1080/1475939X.2017.1354065 .

Facebook. (2020). Facebook Reports First Quarter 2020 Results . Retrieved from http://investor.fb.com/releasedetail.cfm?ReleaseID=842071 .

Fasae, J. K., & Adegbilero-Iwari, I. (2016). Use of social media by science students in public universities in Southwest Nigeria. The Electronic Library, 34 (2), 213–222. https://doi.org/10.1108/EL-11-2014-0205 .

Gahagan, K., Vaterlaus, J. M., & Frost, L. R. (2016). College student cyberbullying on social networking sites: Conceptualization, prevalence, and perceived bystander responsibility. Computers in Human Behavior, 55, 1097–1105. https://doi.org/10.1016/j.chb.2015.11.019 .

George, D. R., Dellasega, C., Whitehead, M. M., & Bordon, A. (2013). Facebook-based stress management resources for first-year medical students: A multi-method evaluation. Computers in Human Behavior, 29 (3), 559–562. https://doi.org/10.1016/j.chb.2012.12.008 .

Gettman, H. J., & Cortijo, V. (2015). “Leave Me and My Facebook Alone!” understanding college students’ relationship with Facebook and its use for academic purposes. International Journal for the Scholarship of Teaching and Learning Article, 9 (1), 1. https://doi.org/10.20429/ijsotl.2015.090108 .

Ha, L., Joa, C. Y., Gabay, I., & Kim, K. (2018). Does college students’ social media use affect school e-mail avoidance and campus involvement? Internet Research, 28 (1), 213–231. https://doi.org/10.1108/IntR-11-2016-0346 .

Hamid, S., Bukhari, S., Ravana, S. D., Norman, A. A., & Ijab, M. T. (2016). Role of social media in information-seeking behaviour of international students: A systematic literature review. Aslib Journal of Information Management, 68 (5), 643–666. https://doi.org/10.1108/AJIM-03-2016-0031 .

Hamid, S., Waycott, J., Kurnia, S., & Chang, S. (2015). Understanding students’ perceptions of the benefits of online social networking use for teaching and learning. Internet and Higher Education, 26, 1–9. https://doi.org/10.1016/j.iheduc.2015.02.004 .

Hong, F. Y., & Chiu, S. L. (2016). Factors influencing facebook usage and facebook addictive tendency in university students: The role of online psychological privacy and facebook usage motivation. Stress and Health, 32 (2), 117–127. https://doi.org/10.1002/smi.2585 .

Hossain, M. D., & Veenstra, A. S. (2013). Online maintenance of life domains: Uses of social network sites during graduate education among the US and international students. Computers in Human Behavior, 29 (6), 2697–2702. https://doi.org/10.1016/j.chb.2013.07.007 .

Ifinedo, P. (2016). Applying uses and gratifications theory and social influence processes to understand students’ pervasive adoption of social networking sites: Perspectives from the Americas. International Journal of Information Management, 36 (2), 192–206. https://doi.org/10.1016/j.ijinfomgt.2015.11.007 .

Islam, T., Sheikh, Z., Hameed, Z., Khan, I. U., & Azam, R. I. (2018). Social comparison, materialism, and compulsive buying based on stimulus-response- model: A comparative study among adolescents and young adults. Young Consumers, 19 (1), 19–37. https://doi.org/10.1108/MRR-09-2015-0216 .

Josefsson, P., Hrastinski, S., Pargman, D., & Pargman, T. C. (2016). The student, the private and the professional role: Students’ social media use. Education and Information Technologies, 21 (6), 1583–1594. https://doi.org/10.1007/s10639-015-9403-7 .

Junco, R. (2012). The relationship between frequency of Facebook use, participation in Facebook activities, and student engagement. Computers and Education, 58 (1), 162–171. https://doi.org/10.1016/j.compedu.2011.08.004 .

Junco, R. (2015). Student class standing, Facebook use, and academic performance. Journal of Applied Developmental Psychology, 36, 18–29. https://doi.org/10.1016/j.appdev.2014.11.001 .

Karpinski, A. C., Kirschner, P. A., Ozer, I., Mellott, J. A., & Ochwo, P. (2013). An exploration of social networking site use, multitasking, and academic performance among United States and European university students. Computers in Human Behavior, 29 (3), 1182–1192. https://doi.org/10.1016/j.chb.2012.10.011 .

Kim, J., Lee, C., & Elias, T. (2015). Factors affecting information sharing in social networking sites amongst university students. Online Information Review, 39 (3), 290–309. https://doi.org/10.1108/OIR-01-2015-0022 .

Kim, S., & Yoo, S. J. (2016). Age and gender differences in social networking: effects on south Korean students in higher education. In Social networking and education (pp. 69–82). Switzerland: Springer. https://doi.org/10.1007/978-3-319-17716-8_5 .

Kim, Y., Sohn, D., & Choi, S. M. (2011). Cultural difference in motivations for using social network sites: A comparative study of American and Korean college students. Computers in Human Behavior, 27 (1), 365–372. https://doi.org/10.1016/j.chb.2010.08.015 .

Kircaburun, K., & Griffiths, M. D. (2018). Instagram addiction and the Big Five of personality: The mediating role of self-liking. Journal of Behavioral Addictions, 7 (1), 158–170. https://doi.org/10.1556/2006.7.2018.15 .

Kırcaburun, K., & Griffiths, M. D. (2018). Problematic Instagram use: The role of perceived feeling of presence and escapism. International Journal of Mental Health and Addiction . https://doi.org/10.1007/s11469-018-9895-7 .

Kitchenham, B. (2004). Procedures for performing systematic reviews. Keele University, UK and National ICT Australia, 33, 28.

Google Scholar  

Kitchenham, B., & Charters, S. (2007). Guidelines for performing systematic literature reviews in software engineering. Keele University and University of Durham, 2, 1051.

Kitsantas, A., Dabbagh, N., Chirinos, D. S., & Fake, H. (2016). College students’ perceptions of positive and negative effects of social networking. In Social networking and education (pp. 225–238). Switzerland: Springer, Cham. https://doi.org/10.1007/978-3-319-17716-8_14 .

Koc, M., & Gulyagci, S. (2013). Facebook addiction among Turkish College students: The role of psychological health, demographic, and usage characteristics. Cyberpsychology, Behavior, and Social Networking, 16 (4), 279–284. https://doi.org/10.1089/cyber.2012.0249 .

Kokkinos, C. M., & Saripanidis, I. (2017). A lifestyle exposure perspective of victimization through Facebook among university students. Do individual differences matter? Computers in Human Behavior, 74, 235–245. https://doi.org/10.1016/j.chb.2017.04.036 .

Krasilnikov, A., & Smirnova, A. (2017). Online social adaptation of first-year students and their academic performance. Computers and Education, 113, 327–338. https://doi.org/10.1016/j.compedu.2017.05.012 .

Kujur, F., & Singh, S. (2017). Engaging customers through online participation in social networking sites. Asia Pacific Management Review, 22 (1), 16–24. https://doi.org/10.1016/j.apmrv.2016.10.006 .

Kumar Bhatt, R., & Kumar, A. (2014). Student opinion on the use of social networking tools by libraries. The Electronic Library, 32 (5), 594–602. https://doi.org/10.1108/EL-09-2012-0110 .

Kuo, T., & Tang, H. L. (2014). Relationships among personality traits, Facebook usages, and leisure activities—A case of Taiwanese college students. Computers in Human Behavior, 31 (1), 13–19. https://doi.org/10.1016/j.chb.2013.10.019 .

Lambić, D. (2016). Correlation between Facebook use for educational purposes and academic performance of students. Computers in Human Behavior, 61, 313–320. https://doi.org/10.1016/j.chb.2016.03.052 .

Lee, S. (2015). Analyzing negative SNS behaviors of elementary and middle school students in Korea. Computers in Human Behavior, 43, 15–27. https://doi.org/10.1016/j.chb.2014.10.014 .

Lim, J., & Richardson, J. C. (2016). Exploring the effects of students’ social networking experience on social presence and perceptions of using SNSs for educational purposes. The Internet and Higher Education, 29, 31–39. https://doi.org/10.1016/j.iheduc.2015.12.001 .

Lin, W.-Y., Zhang, X., Song, H., & Omori, K. (2016). Health information seeking in the Web 2.0 age: Trust in social media, uncertainty reduction, and self-disclosure. Computers in Human Behavior, 56, 289–294. https://doi.org/10.1016/j.chb.2015.11.055 .

Liu, C. C., Chen, Y. C., & Diana Tai, S. J. (2017). A social network analysis on elementary student engagement in the networked creation community. Computers and Education, 115 (300), 114–125. https://doi.org/10.1016/j.compedu.2017.08.002 .

Liu, D., & Brown, B. B. (2014). Self-disclosure on social networking sites, positive feedback, and social capital among Chinese college students. Computers in Human Behavior, 38, 213–219. https://doi.org/10.1016/j.chb.2014.06.003 .

Luqman, A., Cao, X., Ali, A., Masood, A., & Yu, L. (2017). Empirical investigation of Facebook discontinues usage intentions based on SOR paradigm. Computers in Human Behavior, 70, 544–555. https://doi.org/10.1016/j.chb.2017.01.020 .

Mano, R. S. (2014). Social media and online health services: A health empowerment perspective to online health information. Computers in Human Behavior, 39, 404–412. https://doi.org/10.1016/j.chb.2014.07.032 .

Mazman, S. G., & Usluel, Y. K. (2010). Modeling educational usage of Facebook. Computers and Education, 55 (2), 444–453. https://doi.org/10.1016/j.compedu.2010.02.008 .

Mehrabian, A., & Russell, J. A. (1974). An approach to environmental psychology . Cambridge: The MIT Press.

Meier, A., Reinecke, L., & Meltzer, C. E. (2016). Facebocrastination? Predictors of using Facebook for procrastination and its effects on students’ well-being. Computers in Human Behavior, 64, 65–76. https://doi.org/10.1016/j.chb.2016.06.011 .

Mirabolghasemi, M., Iahad, N. A., & Rahim, N. Z. A. (2016). Students’ perception towards the potential and barriers of social network sites in higher education. In Social networking and education (pp. 41–49). Switzerland: Springer. https://doi.org/10.1007/978-3-319-17716-8_3 .

Moghavvemi, S., Paramanathan, T., Rahin, N. M., & Sharabati, M. (2017). Student’s perceptions towards using e-learning via Facebook. Behaviour and Information Technology, 36 (10), 1081–1100. https://doi.org/10.1080/0144929X.2017.1347201 .

Moghavvemi, S., Sharabati, M., Paramanathan, T., & Rahin, N. M. (2017). The impact of perceived enjoyment, perceived reciprocal benefits and knowledge power on students’ knowledge sharing through Facebook. International Journal of Management Education, 15 (1), 1–12. https://doi.org/10.1016/j.ijme.2016.11.002 .

Mostafa, R. B. (2015). Engaging students via social media: Is it worth the effort? Journal of Marketing Education, 37 (3), 144–159. https://doi.org/10.1177/0273475315585825 .

Nawi, N. B. C., Al Mamun, A., Nasir, N. A. B. M., Shokery, N. B., Raston, N. B. A., & Fazal, S. A. (2017). Acceptance and usage of social media as a platform among student entrepreneurs. Journal of Small Business and Enterprise Development, 24 (2), 375–393. https://doi.org/10.1108/JSBED-09-2016-0136 .

Ndasauka, Y., Hou, J., Wang, Y., Yang, L., Yang, Z., Ye, Z., et al. (2016). Excessive use of Twitter among college students in the UK: Validation of the Microblog Excessive Use Scale and relationship to social interaction and loneliness. Computers in Human Behavior, 55, 963–971. https://doi.org/10.1016/j.chb.2015.10.020 .

Nwagwu, W. E. (2017). Social networking, identity and sexual behaviour of undergraduate students in Nigerian universities. The Electronic Library, 35 (3), 534–558. https://doi.org/10.1108/EL-01-2015-0014 .

Pantic, I. (2014). Online social networking and mental health. Cyberpsychology, Behavior, and Social Networking, 17 (10), 652–657. https://doi.org/10.1089/cyber.2014.0070 .

Parboteeah, D. V., Valacich, J. S., & Wells, J. D. (2009). The influence of website characteristics on a consumer’s urge to buy impulsively. Information Systems Research, 20 (1), 60–78. https://doi.org/10.1287/isre.1070.0157 .

Park, N., Song, H., & Lee, K. M. (2014). Social networking sites and other media use, acculturation stress, and psychological well-being among East Asian college students in the United States. Computers in Human Behavior, 36, 138–146. https://doi.org/10.1016/j.chb.2014.03.037 .

Park, S. Y., Cha, S.-B., Lim, K., & Jung, S.-H. (2014). The relationship between university student learning outcomes and participation in social network services, social acceptance and attitude towards school life. British Journal of Educational Technology, 45 (1), 97–111. https://doi.org/10.1111/bjet.12013 .

Paul, J. A., Baker, H. M., & Cochran, J. D. (2012). Effect of online social networking on student academic performance. Computers in Human Behavior, 28 (6), 2117–2127. https://doi.org/10.1016/j.chb.2012.06.016 .

Peters, A. N., Winschiers-Theophilus, H., & Mennecke, B. E. (2015). Cultural influences on Facebook practices: A comparative study of college students in Namibia and the United States. Computers in Human Behavior, 49, 259–271. https://doi.org/10.1016/j.chb.2015.02.065 .

Ponnusamy, S., Iranmanesh, M., Foroughi, B., & Hyun, S. S. (2020). Drivers and outcomes of Instagram addiction: Psychological well-being as moderator. Computers in Human Behavior, 107, 106294. https://doi.org/10.1016/j.chb.2020.106294 .

Rap, S., & Blonder, R. (2017). Thou shall not try to speak in the Facebook language: Students’ perspectives regarding using Facebook for chemistry learning. Computers and Education, 114, 69–78. https://doi.org/10.1016/j.compedu.2017.06.014 .

Raymond, J., & Wang, H. (2015). Social network sites and international students ’ cross-cultural adaptation. Computers in Human Behavior, 49, 400–411. https://doi.org/10.1016/j.chb.2015.03.041 .

Roblyer, M. D., McDaniel, M., Webb, M., Herman, J., & Witty, J. V. (2010). Findings on Facebook in higher education: A comparison of college faculty and student uses and perceptions of social networking sites. Internet and Higher Education, 13 (3), 134–140. https://doi.org/10.1016/j.iheduc.2010.03.002 .

Romero-Hall, E. (2017). Posting, sharing, networking, and connecting: Use of social media content by graduate students. TechTrends, 61 (6), 580–588. https://doi.org/10.1007/s11528-017-0173-5 .

Rui, J. R., & Wang, H. (2015). Social network sites and international students’ cross-cultural adaptation. Computers in Human Behavior, 49, 400–411. https://doi.org/10.1016/j.chb.2015.03.041 .

Ryan, R. M., & Deci, E. L. (2000). Self-determination theory and the facilitation of intrinsic motivation. Social Development, and Well-Being, 55 (1), 68–78.

Ryan, T., Chester, A., Reece, J., & Xenos, S. (2014). The uses and abuses of Facebook: A review of Facebook addiction. Journal of Behavioral Addictions, 3 (3), 133–148. https://doi.org/10.1556/JBA.3.2014.016 .

Serdyukov, P. (2017). Innovation in education: What works, what doesn’t, and what to do about it? Journal of Research in Innovative Teaching & Learning, 10 (1), 4–33. https://doi.org/10.1108/jrit-10-2016-0007 .

Sheeran, N., & Cummings, D. J. (2018). An examination of the relationship between Facebook groups attached to university courses and student engagement. Higher Education, 76, 937–955.

Shettar, M., Karkal, R., Kakunje, A., Mendonsa, R. D., & Chandran, V. V. M. (2017). Facebook addiction and loneliness in the post-graduate students of a university in southern India. International Journal of Social Psychiatry, 63 (4), 325–329. https://doi.org/10.1177/0020764017705895 .

Shim, M., Lee-Won, R. J., & Park, S. H. (2016). The self on the Net: The joint effect of self-construal and public self-consciousness on positive self-presentation in online social networking among South Korean college students. Computers in Human Behavior, 63, 530–539. https://doi.org/10.1016/j.chb.2016.05.054 .

Sin, S. C. J., & Kim, K. S. (2013). International students’ everyday life information seeking: The informational value of social networking sites. Library and Information Science Research, 35 (2), 107–116. https://doi.org/10.1016/j.lisr.2012.11.006 .

Singh, A. (2017). Mining of social media data of University students. Education and Information Technologies, 22 (4), 1515–1526. https://doi.org/10.1007/s10639-016-9501-1 .

Skues, J. L., Williams, B., & Wise, L. (2012). The effects of personality traits, self-esteem, loneliness, and narcissism on Facebook use among university students. Computers in Human Behavior, 28 (6), 2414–2419. https://doi.org/10.1016/j.chb.2012.07.012 .

Smith, R., Morgan, J., & Monks, C. (2017). Students’ perceptions of the effect of social media ostracism on wellbeing. Computers in Human Behavior, 68, 276–285. https://doi.org/10.1016/j.chb.2016.11.041 .

Special, W. P., & Li-Barber, K. T. (2012). Self-disclosure and student satisfaction with Facebook. Computers in Human Behavior, 28 (2), 624–630. https://doi.org/10.1016/j.chb.2011.11.008 .

Tally, S. (2010). Mixable blends Facebook with academics to improve student success . Purdue: Purdue University News.

Tandoc, E. C., Ferrucci, P., & Duffy, M. (2015). Facebook use, envy, and depression among college students: Is facebooking depressing? Computers in Human Behavior, 43, 139–146. https://doi.org/10.1016/j.chb.2014.10.053 .

Tang, C. S. K., & Koh, Y. Y. W. (2017). Online social networking addiction among college students in Singapore: Comorbidity with behavioral addiction and affective disorder. Asian Journal of Psychiatry, 25, 175–178. https://doi.org/10.1016/j.ajp.2016.10.027 .

Tang, J. H., Chen, M. C., Yang, C. Y., Chung, T. Y., & Lee, Y. A. (2016). Personality traits, interpersonal relationships, online social support, and Facebook addiction. Telematics and Informatics, 33 (1), 102–108. https://doi.org/10.1016/j.tele.2015.06.003 .

Tashakkori, A., & Teddlie, C. (2003). Issues and dilemmas in teaching research methods courses in social and behavioural sciences: US perspective. International Journal of Social Research Methodology: Theory and Practice, 6 (1), 61–77. https://doi.org/10.1080/13645570305055 .

Teo, T., Doleck, T., & Bazelais, P. (2017). The role of attachment in Facebook usage: a study of Canadian college students. Interactive Learning Environments, 4820 (April), 1–17. https://doi.org/10.1080/10494820.2017.1315602 .

Tokunaga, R. S. (2010). Following you home from school: A critical review and synthesis of research on cyberbullying victimization. Computers in Human Behavior . https://doi.org/10.1016/j.chb.2009.11.014 .

Tower, M., Latimer, S., & Hewitt, J. (2014). Social networking as a learning tool: Nursing students’ perception of efficacy. Nurse Education Today, 34 (6), 1012–1017. https://doi.org/10.1016/j.nedt.2013.11.006 .

Van Hoof, J. J., Bekkers, J., & Van Vuuren, M. (2014). Son, you’re smoking on Facebook! College students’ disclosures on social networking sites as indicators of real-life risk behaviors. Computers in Human Behavior, 34, 249–257. https://doi.org/10.1016/j.chb.2014.02.008 .

Venkatesh, V., Morris, M. G., Davis, G. B., & Davis, F. D. (2003). User acceptance of information technology: Toward a unified view. MIS Quarterly, 27 (3), 425–478. https://doi.org/10.2307/30036540 .

Walker, C. M., Sockman, B. R., & Koehn, S. (2011). An exploratory study of cyberbullying with undergraduate students. Tech Trends, 55 (2), 31–38. https://doi.org/10.1007/s11528-011-0481-0 .

Wang, J. C., & Chang, C. H. (2013). How online social ties and product-related risks influence purchase intentions: A Facebook experiment. Electronic Commerce Research and Applications, 12 (5), 337–346.

Wang, J. L., Jackson, L. A., Gaskin, J., & Wang, H. Z. (2014). The effects of Social Networking Site (SNS) use on college students’ friendship and well-being. Computers in Human Behavior, 37, 229–236. https://doi.org/10.1016/j.chb.2014.04.051 .

Wang, J. L., Jackson, L. A., Zhang, D. J., & Su, Z. Q. (2012a). The relationships among the Big Five Personality factors, self-esteem, narcissism, and sensation-seeking to Chinese University students’ uses of social networking sites (SNSs). Computers in Human Behavior, 28 (6), 2313–2319. https://doi.org/10.1016/j.chb.2012.07.001 .

Wang, X., Li, Y., & Wang, X. (2016). Users’ satisfaction with social network sites: A self-determination perspective. Journal of Computer Information Systems , 4417 (February). https://doi.org/10.1080/08874417.2015.11645800

Wang, X., Lin, X., & Spencer, M. K. (2019). Exploring the effects of extrinsic motivation on consumer behaviors in social commerce: Revealing consumers’ perceptions of social commerce benefits. International Journal of Information Management, 45 (March 2018), 163–175. https://doi.org/10.1016/j.ijinfomgt.2018.11.010 .

Wang, Z., Tchernev, J. M., & Solloway, T. (2012b). A dynamic longitudinal examination of social media use, needs, and gratifications among college students. Computers in Human Behavior, 28 (5), 1829–1839. https://doi.org/10.1016/j.chb.2012.05.001 .

Webster, J., & Watson, R. (2002). Analyzing the past to prepare for the future: Writing a literature review. MIS Quarterly, 26 (2), 13–23.

Wickramanayake, L., & Muhammad Jika, S. (2018). Social media use by undergraduate students of education in Nigeria: A survey. The Electronic Library, 36 (1), 21–37. https://doi.org/10.1108/EL-01-2017-0023 .

Wodzicki, K., Schwämmlein, E., & Moskaliuk, J. (2012). “Actually, I Wanted to Learn”: Study-related knowledge exchange on social networking sites. Internet and Higher Education, 15 (1), 9–14. https://doi.org/10.1016/j.iheduc.2011.05.008 .

Wohn, D. Y., & Larose, R. (2014). Effects of loneliness and differential usage of Facebook on college adjustment of first-year students. Computers and Education, 76, 158–167. https://doi.org/10.1016/j.compedu.2014.03.018 .

Yazdanparast, A., Joseph, M., & Qureshi, A. (2015). An investigation of Facebook boredom phenomenon among college students. Young Consumers, 16 (4), 468–480. https://doi.org/10.1108/YC-02-2015-00506 .

Zhang, H., Lu, Y., Gupta, S., & Zhao, L. (2014). What motivates customers to participate in social commerce? The impact of technological environments and virtual customer experiences. Information Management, 51, 1017–1030.

Zhang, K. Z., & Benyoucef, M. (2016). Consumer behavior in social commerce: A literature review. Decision Support Systems, 86, 95–108.

Zhang, R. (2017). The stress-buffering effect of self-disclosure on Facebook: An examination of stressful life events, social support, and mental health among college students. Computers in Human Behavior, 75, 527–537. https://doi.org/10.1016/j.chb.2017.05.043 .

Download references

Acknowledgements

This paper is supported by Fundamental Research Grant Scheme (FRGS) (Vote No. R.K130000.7840.4F245), and UTM Razak School of Engineering and Advanced Technology research grant or DPUTMRAZAK (Vote No. R.K13000.7740.4J313).

Author information

Authors and affiliations.

Razak Faculty of Technology and Informatics, Universiti Teknologi Malaysia, 54100, Kuala Lumpur, Malaysia

Maslin Binti Masrom & Nik Hasnaa Nik Mahmood

Irish Institute of Digital Business, DCU Business School, Dublin City University, Dublin 9, Ireland

Abdelsalam H. Busalim

Faculty of Social Sciences & Humanities, School of Education, Universiti Teknologi Malaysia, UTM, 81310, Skudai, Johor, Malaysia

Hassan Abuhassna

You can also search for this author in PubMed   Google Scholar

Contributions

The corresponding author worked in writing the paper, all authors worked collaboratively to write the literature review and discussion. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Abdelsalam H. Busalim .

Ethics declarations

Competing interests.

This paper is an original work, this paper conducted a systematic literature review of students’ behavior and OSNs studies to explicate to what extent students behave on these platforms.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Masrom, M.B., Busalim, A.H., Abuhassna, H. et al. Understanding students’ behavior in online social networks: a systematic literature review. Int J Educ Technol High Educ 18 , 6 (2021). https://doi.org/10.1186/s41239-021-00240-7

Download citation

Received : 16 July 2020

Accepted : 13 January 2021

Published : 30 January 2021

DOI : https://doi.org/10.1186/s41239-021-00240-7

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Students’ behavior
• Social media
• Systematic review
• Stimulus-organism-response model

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• My Account Login
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 04 February 2022

Fluctuations in behavior and affect in college students measured using deep phenotyping

• Constanza M. Vidal Bustamante 1 , 2   na1 ,
• Garth Coombs 3rd 1 , 2   na1 ,
• Habiballah Rahimi-Eichi 1 , 2 , 3 , 4 ,
• Patrick Mair 1 ,
• Jukka-Pekka Onnela 5 ,
• Justin T. Baker 3 , 4 &
• Randy L. Buckner 1 , 2 , 3 , 6  

Scientific Reports volume  12 , Article number:  1932 ( 2022 ) Cite this article

5088 Accesses

6 Citations

1 Altmetric

Metrics details

• Human behaviour
• Risk factors

A Publisher Correction to this article was published on 25 July 2022

An Author Correction to this article was published on 29 March 2022

This article has been updated

College students commonly experience psychological distress when faced with intensified academic demands and changes in the social environment. Examining the nature and dynamics of students’ affective and behavioral experiences can help us better characterize the correlates of psychological distress. Here, we leveraged wearables and smartphones to study 49 first-year college students continuously throughout the academic year. Affect and sleep, academic, and social behavior showed substantial changes from school semesters to school breaks and from weekdays to weekends. Three student clusters were identified with behavioral and affective dissociations and varying levels of distress throughout the year. While academics were a common stressor for all, the cluster with highest distress stood out by frequent report of social stress. Moreover, the frequency of reporting social, but not academic, stress predicted subsequent clinical symptoms. Two years later, during the COVID-19 pandemic, the first-year cluster with highest distress again stood out by frequent social stress and elevated clinical symptoms. Focus on sustained interpersonal stress, relative to academic stress, might be especially helpful to identify students at heightened risk for psychopathology.

Similar content being viewed by others

Monitoring sleep using smartphone data in a population of college students

Personalized mood prediction from patterns of behavior collected with smartphones

Smartphone keyboard dynamics predict affect in suicidal ideation

Introduction.

First-year college students must navigate multifaceted change in their daily lives. In addition to completing demanding classes, assignments and exams, students must adapt to a different social environment, make new friends while also managing previous social relationships, and deal with roommates, finances, and health matters with increased independence from their childhood caregivers. These various academic, social, and personal demands, in addition to students’ oftentimes poor sleep and physical activity, can all contribute to psychological distress and increased vulnerability to mental illness 1 , 2 , 3 , 4 , 5 .

About one in three first-year college students present with at least one mental health disorder, most commonly depression and anxiety 6 , and an even larger proportion of students report experiencing psychological distress that affects their academic performance and day-to-day functioning 7 , 8 . Persistent psychological distress increases the likelihood of dropping out of college and also of engaging in self-injurious behavior, with potentially long-term consequences 1 , 9 , 10 , 11 . Especially as the share of individuals accessing higher education continues to increase 12 , understanding and addressing student mental health remains a pressing task.

Examining the nature and dynamics of students’ multidimensional affective and behavioral experiences can help us better characterize the correlates of psychological distress. The various demands associated with the transition to college life (academics, social relationships, and otherwise) are likely to have varying degrees of impact on students’ distress 5 , 13 . Moreover, the intensity of these demands is not uniform over time 14 , 15 , and their dynamics might help identify periods of heightened vulnerability to distress.

The increasing ubiquity of smartphones and wearables afford new opportunities to examine individuals at frequent sampling rates, for extended periods of time, and with relatively low participant burden 16 , 17 . Recent studies with students across college years have used these tools to collect a wide variety of passive sensing and survey-based measures over several weeks, including sleep, mobility, and studying and socializing behaviors. Researchers have adopted this deep phenotyping approach to describe behavioral patterns and predict students’ academic outcomes, stress levels, and depression symptoms 14 , 18 , 19 , 20 , 21 , and most recently to compare behavior and clinical symptoms before and after the start of the COVID-19 global pandemic 22 .

Here, we build on this emergent line of deep phenotyping research with two main goals. First, we sought to capture a full “year in the life” of a first-year college student by simultaneously examining multiple affective and behavioral experiences that have been previously associated with mental health outcomes. These include sleep, physical activity, academic and social behavior, and perceived stress levels and sources of stress (e.g., academics, social relationships, status, health, etc.). We were especially interested in assessing how these experiences fluctuate in relation to the academic calendar (e.g., during exams periods, extended school breaks, weekdays and weekends, etc.), and whether some experiences are more persistent than others across changing contextual demands. Our second goal was to explore individual differences in these experiences and their relation to mental health outcomes. Specifically, we sought to explore the presence of student subgroups with distinct affective and behavioral phenotypes, and how their distinguishing features relate to their respective levels of psychological distress.

To achieve these goals, we leveraged continuous wristband actigraphy data and daily smartphone-based self-report surveys in a sample of 49 first-year students for the full academic year, for a total of close to 10,000 daily observations. While emerging work combines passive data from several sensors to infer a wide range of complex behaviors and emotional experiences, here we limited our passive-sensing work to estimates with well-validated data processing and analysis pipelines, namely sleep detection and relative physical activity 23 . The rest of our metrics reflect participants’ daily self-reports of sleep quality and physical activity, stress levels and sources, positive and negative affect, and academic and social behavior. When the COVID-19 nationwide shut-down occurred two years later, we followed the same students again, seizing on the unique opportunity to prospectively assess their affective and behavioral patterns as they underwent the first three months of this unprecedented life transition.

Our results show substantial fluctuations in affect and behavior over the course of the year. For the average first-year student, these fluctuations followed the structure of the academic calendar, including higher stress at the beginning of the year and during exams periods, and marked changes from school semesters to school breaks and from weekdays to weekends. Clustering analyses revealed three student subgroups with multiple behavioral and affective dissociations and varying levels of distress throughout the year. A critical dissociation of the clusters was revealed by how frequently they endorsed stress sources related to social relationships (e.g., friends, family, etc.) as compared to sources related to academics (e.g., homework, grades). While academics were the most common source of stress for all, the cluster with highest distress stood out by their frequent social stress. These dissociations were observed again in the follow-up COVID-19 dataset. We identify frequent reports of social stress, in contrast to academic stress, as a helpful marker of current and subsequent psychological distress and clinical symptoms.

Table 1 presents summary statistics for all actigraphy- and survey-based metrics (for more details on missing data, see Supplementary Information and Supplementary Fig. 1 ). Below, we present the results of our two main research goals: (1) a description of the temporal dynamics of the average student’s daily stress, sleep, physical activity, and academic and social behavior, and (2) an exploration of student subgroups with distinct affective-behavioral phenotypes and patterns of psychological distress. We then extend these results with a 3-month follow-up study with the same students during the COVID-19 pandemic, allowing for the prospective replication of the student subgroup patterns.

Sleep patterns show school break and weekly fluctuations

Broadly, there is a pattern of increased sleep and lower activity when students are released from structured academic demands. Group-averaged time series for daily actigraphy-derived sleep and wake-time activity metrics are shown in Fig.  1 . The academic year has a Fall Semester and a Spring Semester (~ 16 weeks each) ending with Reading and Exams periods. During Reading period students work on assignments and prepare for final examinations. The two semesters are separated by a five-week class-free Winter Break. Additionally, there is a five-day Thanksgiving Break in Fall and a week-long Break in Spring.

Activity and sleep patterns show school break and weekly fluctuations. Time series show daily actigraphy-derived observations averaged across the full sample. Vertical lines indicate landmark events in the academic calendar, labeled at the top. Black circles indicate Monday through Friday and white circles indicate Saturday and Sunday (sleep observations correspond to the day when the participant woke up). Dashed gray lines indicate the mean. The top panel shows the probability that survey-based observation is missing on any given day (i.e., the proportion of missing individual observations out of the total sample), and the gray shading across panels indicates days with more than 50% missing observations. H-Y Harvard-Yale football game, Festival music festival hosted on campus.

Over Winter Break the average student had lower Wake-time Activity and longer Sleep Duration relative to the school semesters. Similar patterns were observed during Thanksgiving and Spring Breaks. Within the school semesters, Sleep Duration was longer during weekends relative to weekdays. Wake Duration and Wake-time Activity tended to peak on Fridays, the only day in the week when students are likely to both attend classes during the day and social events at night. In contrast, Wake Duration and Activity were the lowest on Sundays.

Social events also affected activity and sleep patterns. Wake-time Activity peaked at the start of the Harvard-Yale football game weekend, the day of the spring music Festival, and on Housing Day, when students received their housing assignments. Meanwhile, Sleep Duration showed peaks the night following the Harvard-Yale football game and the night following the spring music Festival, and it dropped sharply the night before Housing Day, when students are known to stay up late with friends prior to receiving their housing assignment. These dynamics reinforce that we captured real-world behavioral fluctuations relevant to daily student life.

Behavior and affect also reveal marked fluctuations

Students experienced a period of markedly reduced Stress during Winter Break (Fig.  2 a, Supplementary Fig. 2 a). During this period they also reported slightly better Sleep Quality, spending less time on Schoolwork, more time Interacting with others, and fewer daily stressors. Decreases in Stress and increases in Positive Affect and Social Interaction extended to Thanksgiving and Spring Breaks, to days with university-wide social events, and to weekend days within the school semesters (Fig.  2 a, Supplementary Supplementary Fig. 2 b). Thus, at the group level, structured academic demands and releases from them are associated with large fluctuations in behavior and experienced distress.

School-related fluctuations extended to various behavioral and affective domains. ( a ) Time series show survey-derived daily observations averaged across the full sample. Vertical lines indicate landmark events in the academic calendar, labeled at the top. Black circles indicate Monday through Friday and white circles indicate Saturday and Sunday. The top panel shows the probability that survey-based observation is missing on any given day (i.e., the proportion of missing individual observations out of the total sample), and the gray shading across panels indicates days with more than 50% missing observations. ( b ) Some of the individual stressor items that made up the composite academic, social, and status stress categories (displayed in the last panel of ( a )) are shown in the first, second, and third panels, respectively. TB thanksgiving break, R reading period, E exams period, WB winter break, HD housing day, SB spring break, H-Y  Harvard-Yale football game, Festival music festival hosted on campus.

Of note, not all items showed clear school-related changes: Energy, feeling Connected to others, Negative Affect, and Social Stress showed similar levels from school semesters to Winter Break, and from weekdays to weekends. As will be illustrated later, there were individual differences tied to clinical symptoms of distress that are distinct from those tied closely to academic demands.

Academics is the most common stressor but social stress is relentless

Stress was commonplace and had multiple sources (Fig.  2 b). Academic stress sources were the most frequently reported during the school semesters. Homework-related stress, in particular, was two to eight times more frequent than stressors related to social relationships or status. However, Academic Stress was situational, with a near complete attenuation during school breaks.

Social Stress was less common but displayed a distinct sustained pattern. The probability of experiencing Social Stress from Friends and Family was present throughout the year, including during school breaks. Friends were the most frequent source of Social Stress during the school semesters. Family stress increased during the school breaks (when students usually go back home), reaching a similar and at times higher probability than Friend-related stress.

Latent states of distress reflect the structure of the academic calendar in the group

The shared fluctuations in behavior and affect described above were summarized as latent distress states using a multivariate hidden Markov model (HMM) over the group-averaged time series. The three-state model had the lowest Bayesian Information Criterion (BIC) (Supplementary Table 1 ). The three states identified by the model were intuitive and reflected a progression from lowest to highest distress with overall increasing mean values for Stress, Negative Affect and Energy, and decreasing mean values for Sleep Quality and feeling Connected to others. We refer to these states as Lowest, Medium, and Highest Distress states (these labels are relative; e.g., the “Highest” distress state had the highest mean value for feeling Stress, but this value corresponds to “moderate” stress on the survey scale).

The estimated sequence of Latent Distress States over the year revealed a clear, structured pattern of distress that reflects academic demands (Fig.  3 ). The semesters were characterized by the Medium and the Highest Distress levels, with a weekly pattern: Medium Distress on Fridays and Saturdays, and Highest Distress the remainder of the week. The Lowest Distress state was present during the extended breaks including Thanksgiving, Winter, and Spring Breaks.

Latent states of distress reflect the structure of the academic calendar. Latent states were estimated via multivariate hidden Markov modeling over the group mean time series of stress, Negative affect, Energy, Sleep quality, and feeling Connected to others. TB thanksgiving break, R reading period, E exams period, WB winter break, HD housing day, SB spring break.

Clustering analysis identifies three distinct student profiles

Not all students experience their first year the same way. Descriptive analysis revealed substantial between-person variability in most affective and behavioral measures (Fig.  4 a). Moreover, moderate to large correlations between measures illustrated that there might be significant latent structure (Fig.  4 b). For example, Negative Affect, Stress, Academic Stress and Social Stress all showed strong positive correlations.

Between-person variability was substantial and showed latent structure. ( a ) Violin plots show distribution of participant-averaged metrics. ( b ) Correlation matrix shows between-person Pearson r correlation coefficients; stars represent statistical significance (*p < 0.05, **p < 0.01, ***p < 0.001).

Clustering analysis was employed to explore individual differences. A latent profile clustering analysis indicated that the three-cluster mixture was the best solution. The clustering structure is visualized in Fig.  5 a (for more details on this biplot, see Supplementary Information). The first dimension (62% of the eigenvalues) separated Cluster A (comprising 12 participants) from the other two clusters, along measures including Stress, Academic Stress, and number of daily Stressors, with Cluster A showing the lowest values on these variables. The second dimension (38% of the eigenvalues) separated Cluster B (17 participants) from Cluster C (20 participants), with the former showing lower values on measures including Procrastination and higher values for Sleep Duration and Sleep Quality. This second dimension also separated Cluster B from the other two clusters along Physical Activity, with Cluster B showing the lowest values.

Clustering analysis reveals three distinct student profiles. ( a ) Biplot shows both the maximal separation among the clusters, displayed as dots colored by cluster, and the original variables used in the analysis, shown as standardized basis vectors on the two dimensions. The length of the vectors represents standardized regression coefficients. The arrows for Interaction and Status Stress are almost fully overlapping, and the vector for Sleep Regularity is almost exactly at (0, 0). ( b ) Bar plots show between-person means of each clustering variable separately by cluster. Error bars represent standard errors of the mean. ( c ) Boxplots show participants’ cumulative grade point average (GPA) their first year of university (top) and mean Global Clinical symptom severity (middle). Asterisks represent statistically significant differences between cluster pairs (two-tailed t -tests; **p < 0.01, ***p < 0.001). Bottom bar plot shows the self-reported sex breakdown of the three clusters.

The between-subject mean values of each clustering variable by subgroup are illustrated in Fig.  5 b. On average, participants in Cluster A reported the lowest scores for Negative Affect, Stress, Social Stress, Academic Stress, and time spent on Schoolwork, and the highest levels of Physical Activity. On the other end of the spectrum, participants in Cluster C reported feeling the most Negative Affect and Stress, the highest number of daily Stressors, medium to high probability to experience Academic, Social, and Status Stress, the lowest Sleep Quality, lowest Energy, and highest Procrastination.

Cluster B was intermediate in terms of Stress but additionally stood out on other measures. Participants in this cluster had the highest Sleep Duration and the lowest Physical Activity. Like Cluster A, they reported high levels of time spent on social Interaction and feeling Connected to others, and low Negative Affect. But in contrast to Cluster A, Cluster B reported elevated Academic Stress. Crucially, this stress was selective to Academics, with low Social and Status Stress. In this sense, Cluster B is much like Cluster A in terms of general distress but has elevated distress focused around academic demands. By contrast, Cluster C has a pattern of distress that is pervasive across academic and social domains.

Student profiles associate with academic performance and clinical symptoms

To test whether separation of the participants into three subgroups was predictive of additional data, we turned to independently-acquired assessments of academic performance (first-year cumulative grade point average, GPA) and psychological distress as reflected by the Global Clinical Score (Fig.  5 c). Cluster B, which had a high probability to report Academic stressors (but not other stress sources), had the highest cumulative grades (GPA; ANOVA p  < 0.001; post hoc t- tests of Cluster B vs. Cluster A and Cluster B vs. Cluster C, ps < 0.01). Most critically, Cluster C, which had the highest scores for Negative Affect, Stress, and both Academic and Social Stress, also had the highest Global Clinical Scores (ANOVA p  < 0.001; post hoc t- tests of Cluster C vs. Cluster A and Cluster C vs. Cluster B, p s < 0.01). Women were more likely to be in Cluster C relative to the other two groups, while men showed an even distribution across the clusters. These sex differences are in line with the well-documented higher rates of anxiety and mood disorders among females compared to males 24 , 25 , 26 .

The three student profiles show distinct distress dynamics

To further understand how the year was experienced by the distinct subgroups of participants, the measures were graphed separately for each of the major academic year epochs: Fall Semester, Winter Break, and Spring Semester. All three of the subgroups showed clear differences between the School semester versus the Break (Fig.  6 a) (as well as weekday versus weekend patterns; Supplementary Fig. 3 ). However, while even Cluster C showed decreased Negative Affect and Stress scores during Winter Break and weekends, between-cluster differences were robust and present throughout the year. Notably, Cluster C’s Procrastination, Negative Affect, and Social Stress scores during Winter Break were higher than Clusters A and B’s levels of the same variables during the School terms.

The three student profiles show distinct distress dynamics. ( a ) Bar plots show between-person means of each variable separately by student profile and by term. Error bars represent standard errors of the mean. ( b ) Sequence of most likely latent states of Distress for the three student profiles. TB thanksgiving break, R reading period, E exams period, WB winter break, HD housing day, SB spring break. ( c ) Frequency of Social stressors, but not Academic stressors, prospectively predicts subsequent global clinical symptoms. **Statistical significance at p < 0.01, n.s. not statistically significant (p > 0.05).

We used a three-state multilevel HMM to examine if the sequence of latent Distress states identified at the group level (shown in Fig.  3 ) differed by cluster (Supplementary Table 2 ).

What emerged were distinct patterns of distress across the three subgroups (Fig.  6 b). Cluster A was in the Lowest or Medium Distress states almost all the time including during the academic semesters and exam periods. Cluster A’s marginal probability to be in the Highest Distress state remained near zero throughout the year (Supplementary Fig. 4 ). Cluster B was in the Medium Distress state for most of the academic semester and reached the Highest Distress state on a handful of days around periods of midterm examinations, consistent with their distress being associated with academic demands. Finally, Cluster C was in the Highest Distress state during most of the school semesters, dropping to the Medium distress state during the school breaks. This group did not reach the Lowest Distress state at all, with its marginal probability to be in the Lowest Distress state remaining at zero throughout the year (Supplementary Fig. 4 ).

Social stress, but not academic stress, predicts clinical symptoms

An interesting feature of the above descriptive analyses is the contrast between Academic and Social Stress sources in relation to their temporal dynamics and correlation to clinical symptoms. While Academic Stress is the most robust stressor experienced by the participants, the cluster with the highest Global Clinical Score stands out by their higher frequency of Social Stress. Social Stress was less common than Academic Stress but was also less situational. For example, as seen in Fig.  6 a, while Cluster C’s probability to report Academic Stress reduces during Winter Break, their probability to report Social Stress remained stable throughout the break. These observations suggested that frequent report of interpersonal stressors might be a more sensitive marker of clinical symptoms than academic stressors.

A post hoc mixed effects linear model including Social Stress and Academic Stress as predictors of Global Clinical Scores sought to formally test this observation across all participants. The frequency of Social Stress during the semester was a statistically significant predictor of subsequent Global Clinical Scores at the end of the semester ( B  = 9.56 (3.27), t  = 2.92, p  < 0.01), above and beyond baseline symptoms and Academic Stress (Fig.  6 c). In contrast, Academic Stress was not a statistically significant predictor of Global Clinical Scores ( B  =  − 4.24 (3.98), t  =  − 1.06, p  = 0.29) when controlling for Social Stress. As revealed in the next section, this pattern of dissociation continued in independent, prospectively acquired data when the same participants were followed years later during the COVID-19 pandemic.

Student profiles prospectively predict experiences during the COVID-19 pandemic

The outbreak of the COVID-19 pandemic posed a unique opportunity to assess the stability of our observations under a new stressful life transition. In March 2020, now in their third year of college, 43 out of the original 49 students (88% retention), enrolled in a fully remote three-month follow-up study. Surprisingly, there was no statistically significant difference in participants’ Global Clinical Scores between the first-year study and the follow-up study (paired t -test, p  = 0.07). Analyses focused on the student subgroups determined in the first-year data. What emerged is that the three subgroup clusters continued to have distinct patterns of distress, with Social Stress, but not Academic Stress, indicative of those individuals with severe distress (Fig.  7 ).

Student profiles prospectively predict experiences during the COVID-19 pandemic. ( a ) Boxplots on the left represent participant-means throughout the follow-up study period, grouped by the cluster labels assigned in the first-year study. Asterisks represent statistically significant differences between cluster pairs (two-tailed t -tests; *p < 0.05, **p < 0.01, ***p < 0.001). Time series on the right show cluster averages per each biweekly questionnaire timepoint. Error bars represent standard error of the mean. ( b ) Time series show daily observations averaged by first-year cluster labels. Diamonds emphasize the different trajectories of Academic versus Social Stress for Cluster C into Summer Break. R reading period, E exams period.

Cluster C, which had originally shown the highest overall distress, again stood out with elevated Global Clinical Scores during the COVID-19 pandemic (Fig. 7 a; ANOVA p  < 0.05; post hoc t- test of Cluster C vs. Cluster A, p  < 0.05; post hoc t- test of Cluster C vs. Cluster B, p  = 0.06) and higher Perceived Stress and Anxiety and Depression symptoms (ANOVAs ps < 0.01; post hoc t- tests of Cluster C vs. Cluster A and Cluster C vs. Cluster B, p s < 0.05). By contrast, Clusters A and B both reported low Perceived Stress and clinical symptoms (post hoc t- tests of Cluster A vs. Cluster B, p s > 0.11). Moreover, while at the beginning of the pandemic all three clusters showed similar levels of Global Clinical Scores, Clusters A and B’s overall trajectories showed decreases over time, but Cluster C’s remained consistent throughout.

Examination of the dynamic patterns for Academic and Social Stress again presented dissociations in this follow-up period (Fig.  7 b). Just like in their first year of college, the probabilities to report Academic Stress and Social Stress were both high for Cluster C and both low for Cluster A, while Cluster B’s probability to report Academic Stress was high relative to their low probability to report Social Stress. Furthermore, a critical observation comes from the trajectories of Academic and Social Stress: while the probability to report Academic Stress decreases substantially for all clusters toward the end of the final Exams period, Cluster C’s report of Social Stress remained elevated even into the beginning of summer break. The relentlessness of Cluster C’s Social Stress mirrors the stability of this cluster’s clinical scores, thus further pointing to Social Stress as the critical marker.

The current study used a year-long deep-phenotyping approach to measure actigraphy-derived sleep behavior and daily self-reported metrics of affect and sleep, academic, and social behavior in a group of 49 first-year college students. Our results provide new insights into the variability of students’ affective and behavioral experiences, both over time and across individuals, and their relationship to psychological distress.

Students’ behavior and affect fluctuated with academic demands, in line with results from previous deep-phenotyping studies with college students 15 , 18 , 22 . Stress levels were highest during the first few weeks of the academic year as well as during midterm and final exams periods. On school breaks, as well as on Fridays and Saturdays, the average student slept more, spent less time on schoolwork and more time interacting with others, felt less stress, and reported fewer academic- and status-related stressors compared to the school semesters and weekdays.

Not all participants experienced their first year of college the same way. Our clustering analysis identified three student subgroups with different affective and behavioral features and varying levels of global psychological distress on an external clinical inventory. A critical dissociation of the clusters was revealed by how frequently they endorsed social sources of stress (friends, family, roommate, and/or partner) compared to academic sources of stress (schoolwork and/or grades). These patterns replicated in the follow-up COVID-19 study. While academic stress was more frequently reported than social stress for all groups, the cluster with elevated clinical scores stood out by frequent social stress. The other two clusters had varying degrees of academic stress, but both had infrequent social stress and low clinical scores. In a post hoc mixed effects model further probing this dissociation, the frequency of social stress, but not academic stress, was a significant predictor of subsequent clinical scores at the end of the semester. Although our analyses did not systematically test all possible predictors of psychological distress, these observations suggest that, when it comes to sources of stress, academics are a robust, normative stressor for most students, while a focus on persistent interpersonal stress might be especially helpful to identify those at risk for psychopathology.

These results are in line with previous literature showing academics as a frequently reported source of stress among students 2 , 3 , 27 , and pointing to interpersonal stressors as more strongly associated than non-interpersonal ones with mental health outcomes (e.g., depression symptoms 27 , 28 , 29 ). Social relationships are at the core of individuals’ development and daily experience and play an important role in shaping psychological and emotional wellbeing. Social relationships can represent a critical source of support, but they can also represent a source of stress, either due to direct conflict (e.g., having an argument with friends or family) or to more indirect or subjective pressures and aggravations (e.g., worrying about others’ expectations, lack of personal space due to roommates, etc.). While occasional social conflict and hassles are inevitable and important for the development of social skills, our results suggest that sustained interpersonal stress, at least as perceived and measured through subjective report, is associated with greater psychological distress and clinical symptoms.

Our findings do not imply that academic stress is irrelevant to students’ psychological wellbeing and mental health. For example, the cognitive demands of academics might exacerbate interpersonal stress by making students more sensitive to social friction. However, frequent academic stress, by itself, does not associate with heightened clinical symptoms of distress. Cluster B had frequent reports of academic stress, but low daily negative affect and low psychological distress, as well as the highest grades among the three clusters. One possibility is that these students perceived academic stressors as a motivating challenge under their control rather than as an overwhelming pressure. Cluster C had frequent academic stress as well as the highest negative affect and psychological distress scores, and it is possible that academic stress contributed to their distress, but it was their level of social stress that set them apart.

While our group-level results suggest that an early focus on identifying, resolving, and preventing interpersonal stress might be especially important to help reduce students’ risk for psychological distress, assessing each participant’s data separately might provide more tailored recommendations. For example, future analyses could leverage our intensive longitudinal design to provide individualized models of psychological distress, using lead-lag analyses to assess the directionality of associations among stress and negative affect over days or weeks, and assess whether certain behaviors (e.g., sleep or physical activity) influence the relationship between a daily stressor (social, academic, or otherwise) and negative affect.

The meaningful structure evidenced extensively in the data validate the use of digital phenotyping tools to capture a wide range of behavioral metrics relevant to student life over extended periods of time. However, some limitations should be noted. Most enrolled participants were highly compliant, but there were non-trivial amounts of missing data, especially among the daily phone-based surveys toward the last few months of the study. Although missingness in the daily assessments were not associated with participants’ global clinical scores, it is difficult to know how different the results would look with perfect completion rates. Payment structures and phone notifications were designed for participant compliance and retention, but future work should explore additional strategies.

Our analyses relied heavily on self-report. While we were interested in participants’ subjective perception of stress and affect, participants’ report of their behavior (e.g., how much time they spent on schoolwork or interacting with others) might be inaccurate. Participants’ answers to these questions showed meaningful patterns (e.g., spending less time on schoolwork and more time interacting during weekends and school breaks), but objective metrics derived from validated passive sensing pipelines might provide greater accuracy 18 , 19 .

Finally, our sample was a small group of students living on the campus of an elite United States university, and might not be representative of all first-year college students. For example, if applied to larger or more diverse samples, our analyses would likely identify different and/or more student phenotypes. Our finding that frequent interpersonal stress is more closely associated with psychological distress than academic stress also warrants further examination in larger samples.

The transition to college involves a wide range of academic, social, and physical demands, all of which can contribute to students’ psychological distress and increased vulnerability to mental illness. Our year-long deep-phenotyping investigation provides new insights into the correlates of psychological distress by examining the nature and dynamics of these multidimensional affective and behavioral experiences. Our results revealed substantial variability, both over time and between individuals, in students’ stress levels and sources, sleep, academic and social behavior. Academic stress was common for most students. Those with highest psychological distress stood out by their frequent report of stressors related to social relationships. Although further research is needed in larger and more diverse samples, interpersonal stressors might represent a useful marker of distress and potential target for interventions seeking to support students’ college adjustment and overall wellbeing.

Participants

First-year students living on campus at an elite university in the United States were recruited within the first two weeks of their Fall semester for a year-long study via flyers and e-mail. Participants were required to be enrolled full-time in classes and own an Apple iPhone or Android smartphone compatible with the smartphone application used to collect daily surveys, Beiwe, which is part of the open-source Beiwe platform for digital phenotyping 30 . Participants were not excluded for psychiatric disorders or medication use. Interested participants scheduled an in-person consent session where study procedures were explained. Informed consent was obtained from all study participants. All study methods were approved by the Institutional Review Board of Harvard University and were performed in accordance with the relevant guidelines and regulations.

From an initial recruitment pool of 68 individuals, 19 were excluded from analysis based on issues with data acquisition including early withdrawal from the study (n = 7), technical failure of the actigraphy data (n = 1), poor quality actigraphy data (n = 2), and completion of < 100 daily surveys across the data collection period (n = 9). The final cohort was N = 49 (ages 18–19, mean age = 18.06; 25 female). Of this final sample 63% identified as White, 14% Black, 10% Asian, 4% American Indian, and 4% Mixed-Race. Twelve percent of the sample reported prior diagnosis of a psychiatric disorder (including anxiety, depression, and/or ADHD), of which 66% were active diagnoses. First-year college students have not yet declared their area of study, but participant-reported intended future occupation was 31% medicine, 14% business or finance, 12% academia or other research, 10% engineering, 10% policy or government, 8% law, and 6% other or undecided. Finally, 94% of participants were iPhone users and 6% were Android users.

Study design

This intensive longitudinal observational study collected data over the full academic year and a few days into the summer break. Participation involved completing a battery of online questionnaires at the beginning, middle, and end of the study, completing smartphone-based daily surveys and a voice-recorded diary, wearing an actigraphy wristband for continuous activity and sleep monitoring for the duration of the study (GENEActiv Original, Activinsights Ltd., Kimbolton, UK), and brief in-person check-ins every 3–4 weeks. Participants were compensated per hour for online surveys, $1 per each daily survey they submitted, and $1 per day for continuously wearing the actigraphy wristband. They were also given a milestone bonus for completing half of the study and a larger milestone bonus for completing the full study.

Measures and quality control

Objective activity and sleep measures.

Sleep duration, sleep timing regularity, and wake-time activity were derived from the accelerometer data. Participants were instructed to wear the wristband on their nondominant wrist continuously, including during sleep and when bathing. The wristband collected tri-axial acceleration with precise timestamps at a rate of 30 Hz while participants were on campus, and 10 Hz while participants were away for winter break (in order to extend battery life and memory while participants were not on campus). Participants were instructed to press the wristband’s button when they began trying to sleep at night and immediately after they awoke in the morning. Following the initial consent and receipt of the wristband, individuals exchanged their wristband for a fully charged one with reset memory at the in-person check-ins.

The DPSleep processing pipeline 23 was applied to the raw actigraphy data to automatically estimate minute-based activity and to detect the major sleep episode for each day. The pipeline first removes detected minutes when the individual was not wearing the device using the tri-axial acceleration variance, and then proceeds with its estimation. Days where one of the boundaries of the sleep episode (i.e., rises in relative activity both before and after a period of lower activity) could not be detected due to missing data were labeled as unusable. Two trained independent raters examined the automatically detected start and end times and usability label of each sleep episode against the minute-based activity levels and the participant button presses. If necessary, they adjusted the automatic times and labels. A full description of the processing pipelines applied to the actigraphy data, including quality control steps, can be found in Ref. 23 .

All data that passed quality control was included in analysis, including days with no detected sleep episode (i.e., with no extended periods of lower relative activity). The following metrics were derived and used in analyses:

Sleep duration

Daily sleep duration reflects the number of hours between the timestamps for the start and end of the day’s longest detected sleep episode.

Sleep timing regularity

Daily sleep regularity reflects the proportion (from 0 to 1) of overlap in sleep timing between each study day and the participant’s average sleep timing. See Ref. 23 for details. An individual who sleeps and wakes at the same time on both study day j and on their average sleep day would get a score of 1 for study day j . Conversely, if the sleep episode is completely non-overlapping with their average sleep day, the daily score would be 0.

Wake-time activity

Daily wake-time activity is the average of the minute-based, person-normalized activity percentiles (estimated by DPSleep) across the wake period (i.e., the period between two detected sleep episodes in consecutive days). If daily sleep episode data was missing, wake-time physical activity was also marked missing.

Daily phone-based surveys

Smartphone surveys were administered via the Beiwe application 30 . Each night, before they went to sleep, participants completed a 46-item self-report survey related to their daily lives. Questions were designed to assess a broad range of behaviors and internal states over the past 24 h, including general physical health, daily consumption habits, positive and negative affect, studying behaviors, stress levels and sources, and sociability and support 31 . Most questions were answered using a 5-point Likert scale. We limited our analyses to a subset of questions selected a priori that probed the range of affective and behavioral experiences stated in our research goals. Specifically, these questions probed subjective sleep quality (from “Terrible” to “Exceptional”), energy levels and physical activity levels (from “Very little or not at all” to “Extremely”), how much of their awake time they spent on schoolwork and socializing (from “0–20% of my time” to “80–100% of my time”), how much they felt stress (from “Very little or not at all” to “Extremely”), positive affect (individual items for happy, outgoing, excited, and relaxed) and negative affect (individual items for sad, upset, hostile, irritable, lonely, anxious), and participants’ sources of stress (selected from a checklist spanning academics, social relationships, status, health, and financial situation categories). A full list of the daily survey questions used in the present analysis and other survey details are included in the Supplementary Information.

Surveys submitted between 5PM (local time) the day the survey opened and 6AM the following day were considered on time. Surveys submitted past 6AM the day after the survey went live were discarded and marked as missing. A participant was included in analysis if they were compliant with at least 100 daily surveys across the entire data collection period, and only on-time surveys from those participants were included.

In-person and periodic web-based surveys

All participants completed basic demographic and physical and mental health questionnaires at their baseline in-person visit. The health questionnaire asked participants to report whether they had past or current diagnoses of a series of conditions listed on a checklist (Supplementary Information). Additionally, participants used REDCap, a secure online platform 32 to complete a clinical questionnaire (Symptoms Checklist 90 Revised, see below) 33 , 34 , other surveys (not analyzed here), and report their cumulative grade point averages (GPA). This REDCap-based survey battery was collected at three timepoints in the year: within the first month of enrollment (baseline), at the midpoint of the study period (during Winter Break after Fall semester final grades had been returned) and at the end of the study period (after Spring semester grades had been returned).

Symptoms Checklist 90 Revised

The SCL-90-R 33 , 34 is a 90-item self-report questionnaire that assesses the severity of a broad range of psychological problems and clinical symptoms, including somatization, internalizing, psychoticism, and other domains. Each question asked participants to indicate how much they were bothered by that problem during the past two weeks using a 5-point Likert scale, from “Not at all” (0) to “Extremely” (4). The Global Severity Index (GSI) is a subscale of the SCL-90-R that reflects the overall current level or depth of distress in terms of both number of symptoms endorsed and intensity of distress 33 . The GSI, which we also refer to as Global Clinical Score in this paper, was computed for each participant at each of the three timepoints, transformed to adolescent- and gender-normed T-scores, and then averaged to obtain a single, person-level summary score of global psychological distress (possible range following T-score transformation = 25–81). Symptoms are considered to be at clinical levels if the GSI T-score is greater than 63 33 . The GSI subscale has good sensitivity and reliability in psychiatric and non-psychiatric populations 35 , 36 , and had good internal consistency in all three timepoints in the first-year study (Cronbach’s α = 0.87).

COVID-19 prospective data collection during the COVID-19 pandemic

In March 2020, the COVID-19 global pandemic prompted the closure of the university’s on-campus activities. The following week, the 49 original participants were recontacted (now in their third college year) and invited to participate in a fully remote, 13-week follow-up study. Participation involved completing computer-based survey batteries via REDCap every two weeks and smartphone-based daily surveys via Beiwe. Wristband-based actigraphy data was not collected. Forty-three of the 49 year-long study participants enrolled in the new study (i.e., 88% retention rate, with > 82% retention within each of the three student clusters. Three of the 43 participants provided only the biweekly surveys). The other six original participants (one from Cluster A, three from Cluster B, and two from Cluster C) declined to participate or did not respond to our contacts. Data were collected starting the week after Spring Break for a total of 94 days, capturing the second half of the Spring semester (seven weeks) and six weeks into the summer break. Thirty-two percent of the sample (up from 12% in first-year year study) reported a lifetime psychiatric disorder diagnosis. Participants were compensated $10 for each completed biweekly survey and $1 for each daily survey, as well as given a milestone bonus for completing the study, which was scaled to reward few missed surveys. All participants were re-consented. All study procedures were approved by the Institutional Review Board of Harvard University and were performed in accordance with the relevant guidelines and regulations.

Biweekly survey batteries included the SCL-90-R (same as described in the year-long study; GSI subscale had good internal consistency in all timepoints in the follow-up study, Cronbach’s α = 0.85–0.91) as well as the following assessments:

Perceived stress

Perceived stress was measured with the Perceived Stress 14-item Scale (PSS-14 37 ). Participants indicated how often they felt the way described in the items using a 5-point scale (from 1 = “Never” to 5 = “Very often”), with higher scores indicating higher perceived stress. Answers to all items are summed to provide a total perceived stress score. The PSS-14 has been found to have good validity and reliability 37 , 38 , and had good internal consistency in all timepoints in the follow-up study (Cronbach’s α = 0.85–0.91).

Anxiety symptoms were measured with the Generalized Anxiety Disorder 7-item scale (GAD-7), which queries symptoms occurring in the last two weeks. Items are scored on a Likert scale ranging from 0 to 3 and summed to create a symptom severity score. The GAD-7 has good reliability and validity 39 , and had good internal consistency in all timepoints in the follow-up study (Cronbach’s α = 0.88–0.91).

Depression symptoms were measured with the Patient Health Questionnaire 9-item scale (PHQ-9), which assesses depression symptoms occurring in the last two weeks. The item assessing suicidal ideation was not included in the surveys. Items are scored on a Likert scale ranging from 0 to 3 and summed to create a symptom severity score. The PHQ-9 has good reliability and validity 40 , and had good internal consistency in all timepoints in the follow-up study (Cronbach’s α = 0.83–0.87).

Analytical approach

Daily survey-derived composite scores.

Daily-level composite scores were derived for Positive Affect, Negative Affect, and the stress source categories using items from the daily phone-based surveys. Individual items for Happy, Excited, Relaxed, and Outgoing were averaged into a composite Positive Affect score, and individual items Sad, Upset, Anxious, Irritable, Angry, Lonely, and Self-Dissatisfied were averaged into a composite Negative Affect score. Individual items within each composite score had positive same-day correlations (Supplementary Fig. 5 ). Additionally, we computed binary scores at the daily level to indicate whether the participant reported any of the stressors under three categories: Academic Stress (individual items for Homework, Grades), Social Stress (individual items for Friends, Family, Roommate, Partner), and Status Stress (individual items for Academic Standing, Social Status). Finally, we computed a daily Stressors variable that represents the sum of all individual stressors endorsed by the participant.

Group-averaged year-long time series and structure of the academic year

To compute group-level time series of each of the metrics of interest, data were averaged at the daily level (from the first day of the Fall semester to the last day of the Spring semester, i.e., 256 days) across all available participant observations in the final sample. Days with more than 50% missing observations across the sample are indicated with gray shading in the relevant figures.

Hidden Markov modeling to identify latent behavioral-affective states over time

A hidden Markov model (HMM) 41 , 42 was used to identify latent behavioral-affective states over the year of the average student. The R depmixS4 package (v1.4.2 43 ) fit a multivariate HMM over the group-averaged time series of five variables (Stress, Negative Affect, Energy, Sleep Quality, and feeling Connected to others) to keep it manageable and interpretable. These five variables were selected a priori to represent a range of sleep, social, academic, and affective components of participants’ daily experiences. Models with one to three state solutions were fitted to test whether the HMM would reflect the three types of time-related affective and behavioral fluctuations (breaks, weekdays, and weekends) we had observed in the previous step. Model selection was performed based on lowest Bayesian Information Criterion (BIC). The identified latent states were interpreted based on the means of the five dependent variables under each state. Additionally, the year-long sequence of estimated hidden states was examined to identify meaningful structure in the presentation of the states over time.

Clustering analysis to identify subgroups with distinct profiles

To explore the possibility of individual differences, a latent profile analysis (LPA) assessed the presence of subgroups with distinct student profiles (see Supplementary Information for LPA technical details). LPA used the R mclust package (v5.4.7 44 ) and included person-level means (excluding Winter Break data to focus on school-time student behavior) of the same 15 actigraphy- and daily survey-derived variables that had been selected for longitudinal examination: Sleep Duration, Sleep Regularity, Sleep Quality, Physical Activity, Energy, time spent on Schoolwork, time spent on social Interaction, Procrastination, feeling Connected to others, Negative Affect, Stress, number of Stressors, Academic Stress, Social Stress, and Status Stress. Given our relatively small sample size and to keep results interpretable, models with two to five cluster solutions were tested, and model selection was performed using BIC.

The package mclust offers a dimensionality reduction method similar to principal components analysis that identifies a set of linear combinations, ordered by importance as quantified by the associated eigenvalues, of the original features which capture most of the clustering structure contained in the data 45 . We used this dimensionality reduction method to generate a biplot (Fig.  5 a; see more details in Supplementary Information). This approach and the means of the fitted variables for each of the identified clusters were used to describe and interpret the meaning of each cluster.

As an external validation of the LPA clustering results, we turned to independently-acquired assessments of academic performance (first-year cumulative grade point average—GPA) and psychological distress as reflected by the Global Clinical Score. The R stats package fitted two separate one-way ANOVAs to assess differences in person-level GPA and differences in Global Clinical Scores by the identified clusters. Following statistically significant differences, post hoc two-tailed t -tests were used to assess differences between cluster pairs. Statistical significance was assessed at p  < 0.05.

Hidden Markov modeling across distinct student profiles

To better understand the subgroups identified by the clustering analysis, a multilevel HMM 46 was used to examine their respective sequences of latent distress states over time, expanding the full-sample HMM approach described above. The R mHMMbayes package 47 was used to fit a five-variable, three-state HMM similar to the model outlined above, but in a multilevel framework. Group-level parameters were estimated based on the cluster-averaged time series, and cluster-level parameters were subsequently sampled from the group-level distributions. Thus, while each cluster was allowed to have its own unique parameter values, these were all computed within the same HMM, with all clusters fitted by the same number and similar composition of the hidden states. The Bayesian estimation required initial values for the transition probabilities for each state and for the distribution (mean and variance) of each dependent variable in each state, and also the specification of hyperprior distributions for the dependent variables. To facilitate comparison, these values were set informed by the results from the previous, full-sample HMM. The model was run with 5000 iterations and a 500 burn-in period. The marginal probabilities of each of the three states were extracted from the model and used to compute each cluster’s most likely state on each day in the study.

Academic and social stressors as predictors of subsequent clinical symptoms

A final analysis performed on the first-year data used a mixed effects model to explore whether Social and Academic Stress sources predicted change in Global Clinical Scores over the year. The outcome variable was the SCL-90’s Global Severity Index (GSI) at two timepoints: end of Fall and at end of Spring, while the baseline GSI was entered as a fixed covariate. The predictors of interest were frequency of reported Social Stress sources, and frequency of reporting Academic Stress sources, entered separately per semester (Fall and Spring). Semester (Fall, Spring) and the interaction between semester and baseline symptom levels were also included as fixed effects. Random intercepts for each participant were specified in the model.

First-year subgroup clusters as predictors of COVID-19 clinical symptoms

To further probe the stability of the student subgroups identified in the first-year study, we assessed whether these cluster labels prospectively predicted differences in clinical symptoms under the COVID-19 follow-up study. Four separate one-way ANOVAs were used to assess differences in person-mean Global Clinical Scores, Perceived Stress, Anxiety symptoms, and Depression symptoms by the clusters identified in the first-year study, with the expectation that each of these measures should demonstrate significant effects, allowing for both prospective prediction and convergence across multiple measures. Following statistically significant differences, post hoc two-tailed t -tests were used to assess differences between cluster pairs. Statistical significance was assessed at p  < 0.05 and patterns that illustrated convergence, meaning significance across multiple variables, were interpreted. These analyses were run with the R stats package.

Change history

29 march 2022.

A Correction to this paper has been published: https://doi.org/10.1038/s41598-022-09563-5

25 July 2022

A Correction to this paper has been published: https://doi.org/10.1038/s41598-022-16757-4

Thorley, C. Not by Degrees: Improving Student Mental Health in the UK’s Universities (2017).

Beiter, R. et al. The prevalence and correlates of depression, anxiety, and stress in a sample of college students. J. Affect. Disord. 173 , 90–96 (2015).

Article   CAS   Google Scholar  

Ross, S. E., Niebling, B. C. & Heckert, T. M. Sources of stress among college students. Coll. Stud. J. 33 , 312 (1999).

Google Scholar  

Dyson, R. & Renk, K. Freshmen adaptation to university life: Depressive symptoms, stress, and coping. J. Clin. Psychol. 62 , 1231–1244 (2006).

Article   Google Scholar  

Gerdes, H. & Mallinckrodt, B. Emotional, social, and academic adjustment of college students: A longitudinal study of retention. J. Couns. Dev. 72 , 281 (1994).

Auerbach, R. P. et al. WHO World Mental Health surveys International College Student Project: Prevalence and distribution of mental disorders. J. Abnorm. Psychol. 127 , 623–638 (2018).

YouGov. One in Four Students Suffer from Mental Health Problems (YouGov, 2016).

American College Health Association. American College Health Association-National College Health Assessment III: Undergraduate Student Reference Group Data Report Fall 2019 (American College Health Association, 2019).

Auerbach, R. P. et al. Mental disorders among college students in the World Health Organization World Mental Health Surveys. Psychol. Med. 46 , 2955–2970 (2016).

Liu, C. H., Stevens, C., Wong, S. H. M., Yasui, M. & Chen, J. A. The prevalence and predictors of mental health diagnoses and suicide among US college students: Implications for addressing disparities in service use. Depress. Anxiety 36 , 8–17 (2019).

National Alliance on Mental Illness. College Students Speak: A Survey Report on Mental Health . (2012).

UNESCO. Institute for Statistics (UIS) Database (Accessed 15 June 2021); http://data.uis.unesco.org/ (2020).

O’Donnell, M. B. et al. The college adjustment questionnaire: A measure of students’ educational, relational, and psychological adjustment to the college environment. J. Coll. Stud. Dev. 59 , 116–121 (2018).

Wang, R. et al. Tracking depression dynamics in college students using mobile phone and wearable sensing. In Proc. ACM Interactive, Mobile, Wearable Ubiquitous Technol., Vol. 2, 1–26 (2018).

Harari, G. M. et al. Patterns of behavior change in students over an academic term: A preliminary study of activity and sociability behaviors using smartphone sensing methods. Comput. Hum. Behav. 67 , 129–138 (2017).

Harari, G. M. et al. Using smartphones to collect behavioral data in psychological science: Opportunities, practical considerations, and challenges. Perspect. Psychol. Sci. 11 , 838–854 (2016).

Onnela, J. P. Opportunities and challenges in the collection and analysis of digital phenotyping data. Neuropsychopharmacology 46 , 45–54 (2021).

Wang, R. et al. StudentLife: Assessing mental health, academic performance and behavioral trends of college students using smartphones. In UbiComp 2014—Proceedings of the 2014 ACM International Joint Conference on Pervasive and Ubiquitous Computing 3–14 https://doi.org/10.1145/2632048.2632054 (ACM Press, 2014).

Sano, A. et al. Identifying objective physiological markers and modifiable behaviors for self-reported stress and mental health status using wearable sensors and mobile phones: Observational study. J. Med. Internet Res. 20 , e210 (2018).

DaSilva, A. W. et al. Correlates of stress in the college environment uncovered by the application of penalized generalized estimating equations to mobile sensing data. JMIR Mhealth Uhealth 7 (3), e12084 (2019).

Melcher, J., Hays, R. & Torous, J. Digital phenotyping for mental health of college students: A clinical review. Evid. Based Ment. Health 23 , 161–166 (2020).

Huckins, J. F. et al. Mental health and behavior of college students during the early phases of the COVID-19 pandemic: Longitudinal smartphone and ecological momentary assessment study. J. Med. Internet Res. 22 , e20185 (2020).

Rahimi-Eichi, H. et al. Open-source longitudinal sleep analysis from accelerometer data (DPSleep): Algorithm development and validation. JMIR MHealth UHealth 9 , e29849 (2021).

Merikangas, K. R. et al. Lifetime prevalence of mental disorders in US adolescents: Results from the National Comorbidity Survey replication-adolescent supplement (NCS-A). J. Am. Acad. Child Adolesc. Psychiatry 49 , 980–989 (2010).

Kessler, R. C. et al. Lifetime prevalence and age-of-onset distributions of DSM-IV disorders in the national comorbidity survey replication. Arch. Gen. Psychiatry 62 , 593–602 (2005).

Rosenfield, S. Gender and dimensions of the self: Implications for internalizing and externalizing behavior. In American Psychopathological Association Series (ed. Frank, E.) 23–36 (American Psychiatric Publishing, 2000).

O’Neill, S. C., Cohen, L. H., Tolpin, L. H. & Cimbolic Gunthert, K. Affective reactivity to daily interpersonal stressors as a prospective predictor of depressive symptoms. J. Soc. Clin. Psychol. 23 , 172-194 (2004).

Rudolph, K. D., Lansford, J. E. & Rodkin, P. C. Interpersonal theories of developmental psychopathology. Dev. Psychopathol. https://doi.org/10.1002/9781119125556.devpsy307 (2016).

Article   PubMed   PubMed Central   Google Scholar  

Hammen, C. The emergence of an interpersonal approach to depression. In The Interactional Nature of Depression: Advances in Interpersonal Approaches (eds Joiner, T. & Coyne, J. C.) (American Psychological Association, 1999).

Torous, J., Kiang, M. V., Lorme, J. & Onnela, J.-P. New tools for new research in psychiatry: A scalable and customizable platform to empower data driven smartphone research. JMIR Ment. Health 3 , e16 (2016).

Coombs, G. Using Single-Subject Designs to Probe Dynamics Associated With Stress and Transitions to College Life (Harvard University, Cambridge, 2020).

Harris, P. A. et al. Research electronic data capture (REDCap)-A metadata-driven methodology and workflow process for providing translational research informatics support. J. Biomed. Inform. 42 , 377–381 (2009).

Derogatis, L. R. SCL-90-R: Symptom Checklist-90-R: Administration, Scoring, and Procedures Manual (NCS Pearson, 1975).

Derogatis, L. R. & Unger, R. Symptom checklist-90-revised. In The Corsini Encyclopedia of Psychology (eds Weiner, I. B. & Craighead, W. E.) 1–2 (Wiley, 2010).

Rytilä-Manninen, M. et al. Psychometric properties of the Symptom Checklist-90 in adolescent psychiatric inpatients and age- and gender-matched community youth. Child Adolesc. Psychiatry Ment. Health 10 , 23 (2016).

Holi, M. Assessment of Psychiatric Symptoms Using the SCL-90 (University of Helsinki, Helsinki, 2003).

Cohen, S., Kamarck, T. & Mermelstein, R. A global measure of perceived stress. J. Health Soc. Behav. 24 , 385–396 (1983).

Lee, E. H. Review of the psychometric evidence of the perceived stress scale. Asian Nurs. Res. 6 , 121–127 (2012).

Spitzer, R. L., Kroenke, K., Williams, J. B. W. & Löwe, B. A brief measure for assessing generalized anxiety disorder: The GAD-7. Arch. Intern. Med. 166 , 1092–1097 (2006).

Kroenke, K., Spitzer, R. L. & Williams, J. B. The PHQ-9: Validity of a brief depression severity measure. J. Gen. Intern. Med. 16 , 606–613 (2001).

Zucchini, W., MacDonald, I. L. & Langrock, R. Hidden Markov Models for Time Series: An Introduction Using R (Chapman and Hall CRC Press, 2016).

MATH   Google Scholar  

Visser, I. Seven things to remember about hidden Markov models: A tutorial on Markovian models for time series. J. Math. Psychol. 55 , 403–415 (2011).

Article   MathSciNet   Google Scholar  

Visser, I. & Speekenbrink, M. depmixS4: An R package for hidden Markov models. J. Stat. Softw. 36 , 1–21 (2010).

Scrucca, L., Fop, M., Murphy, T. B. & Raftery, A. E. Mclust 5: Clustering, classification and density estimation using Gaussian finite mixture models. R J. 8 , 289–317 (2016).

Scrucca, L. Dimension reduction for model-based clustering. Stat. Comput. 20 , 471–484 (2010).

Altman, R. M. Mixed hidden Markov models: An extension of the Hidden Markov model to the longitudinal data setting. J. Am. Stat. Assoc. 102 , 201–210 (2007).

Article   MathSciNet   CAS   Google Scholar  

Aarts, E. mHMMbayes: Multilevel Hidden Markov Models Using Bayesian Estimation. R Package, Developer Version (2020).

Download references

Acknowledgements

We thank Laura Farfel, Marisa Marotta, Erin Phlegar, Lauren DiNicola, Arpi Youssoufian for their help collecting data. Timothy O'Keefe, Harris Hoke, Lily Jeong provided valuable assistance in neuroinformatics support. Kenzie W. Carlson provided support with Beiwe. Katherine Miclau, Amira Song, Emily Iannazzi and Hannah Becker helped with the actigraphy quality control procedure. Stephanie Kaiser and Francesca Davy-Falconi helped proofread the manuscript. This work was supported by a generous gift from Kent and Liz Dauten, NIMH Grants U01MH116925 and DP2MH103909, NIH Grant T90DA022759, the Sackler Scholar Programme in Psychobiology, and by the Harvard Foundations of Human Behavior Initiative.

Author information

These authors contributed equally: Constanza M. Vidal Bustamante and Garth Coombs 3rd.

Authors and Affiliations

Department of Psychology, Harvard University, Northwest Science Building 280.05, 52 Oxford Street, Cambridge, MA, 02138, USA

Constanza M. Vidal Bustamante, Garth Coombs 3rd, Habiballah Rahimi-Eichi, Patrick Mair & Randy L. Buckner

Center for Brain Science, Harvard University, Cambridge, MA, 02138, USA

Constanza M. Vidal Bustamante, Garth Coombs 3rd, Habiballah Rahimi-Eichi & Randy L. Buckner

Department of Psychiatry, Harvard Medical School, Boston, MA, 02114, USA

Habiballah Rahimi-Eichi, Justin T. Baker & Randy L. Buckner

Institute for Technology in Psychiatry, McLean Hospital, Belmont, MA, 02478, USA

Habiballah Rahimi-Eichi & Justin T. Baker

Department of Biostatistics, Harvard T.H. Chan School of Public Health, Harvard University, Boston, MA, 02115, USA

Jukka-Pekka Onnela

Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Boston, Charlestown, MA, 02129, USA

Randy L. Buckner

You can also search for this author in PubMed   Google Scholar

Contributions

G.C., R.L.B., J.T.B., J.P.O. designed the research. G.C. performed the research. C.M.V., G.C., H.R., R.L.B. analyzed the data. P.M. provided statistical support. C.M.V. and R.L.B. wrote the paper, and all other authors provided comments.

Corresponding author

Correspondence to Constanza M. Vidal Bustamante .

Ethics declarations

Competing interests.

J.P.O. is a cofounder and board member of a commercial entity, Beiwe, established in 2020, that operates in digital phenotyping. JTB has received consulting fees from Verily Life Sciences as well as consulting fees and equity from Mindstrong Health Inc. for work unrelated to the present work. All other authors declare no competing interests.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

The original online version of this Article was revised: The original version of this Article contained errors in Table 1, where the Metric ‘Daily survey: Available observations per subject’ was incorrect for the Mean (column ‘M’) and the Median (column ‘Med’). Full information regarding the corrections made can be found in the correction for this Article.

The original online version of this Article was revised: The original version of this Article contained an error in the Competing interests section.

Supplementary Information

Supplementary information., rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Vidal Bustamante, C.M., Coombs, G., Rahimi-Eichi, H. et al. Fluctuations in behavior and affect in college students measured using deep phenotyping. Sci Rep 12 , 1932 (2022). https://doi.org/10.1038/s41598-022-05331-7

Download citation

Received : 14 September 2021

Accepted : 05 January 2022

Published : 04 February 2022

DOI : https://doi.org/10.1038/s41598-022-05331-7

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

This article is cited by

Dynamic computational phenotyping of human cognition.

• Roey Schurr
• Daniel Reznik
• Samuel J. Gershman

Nature Human Behaviour (2024)

By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Class size, student behaviors and educational outcomes

Organization Management Journal

ISSN : 2753-8567

Article publication date: 17 February 2022

Issue publication date: 2 August 2022

While many business schools use large classes for the sake of efficiency, faculty and students tend to perceive large classes as an impediment to learning. Although class size is a contested issue, research on its impact is inconclusive, mainly focusing on academic performance outcomes such as test scores and does not address classroom dynamics. This study aims to expand the focus of class size research to include classroom dynamics and subjective educational outcomes (e.g. student learning outcomes and satisfaction).

Design/methodology/approach

Using Finn et al.’s (2003) theoretical framework and research conducted in introductory business classes, this study investigates how student academic and social engagement influence educational outcomes in different class sizes.

Results highlight the critical role that student involvement and teacher interaction play on student success and student satisfaction regardless of class sizes. In addition, the results indicate that students perceive lower levels of teacher interaction and satisfaction in larger classes.

Originality/value

This study applies Finn’s framework of student engagement in the classroom to understand the dynamics of class size in business education. The results reveal the influential roles of academic and social engagements on educational outcomes. Practical strategies are offered to improve learning outcomes and student satisfaction in large classes.

• Large class teaching
• Student satisfaction
• Assessment of learning
• Business education outcomes
• Student learning outcomes

Wang, L. and Calvano, L. (2022), "Class size, student behaviors and educational outcomes", Organization Management Journal , Vol. 19 No. 4, pp. 126-142. https://doi.org/10.1108/OMJ-01-2021-1139

Emerald Publishing Limited

Copyright © 2022, Liz Wang and Lisa Calvano.

Published in Organization Management Journal . Published by Emerald Publishing Limited. This article is published under the Creative Commons Attribution (CC BY 4.0) licence. Anyone may reproduce, distribute, translate and create derivative works of this article (for both commercial and non-commercial purposes), subject to full attribution to the original publication and authors. The full terms of this licence maybe seen at http://creativecommons.org/licences/by/4.0/legalcode

Introduction

According to the 2017 Inside Higher Ed Survey, 71% of 409 chief business officers agreed that higher education institutions were facing significant financial difficulties ( Jaschik, 2017 ). Many business schools use large classes to address the challenges of shrinking resources. Large classes may enable institutions to deploy faculty more efficiently and accommodate more students, especially when it is not feasible to expand facilities or increase hiring ( Guseman, 1985 ). Nevertheless, large class size is a contested issue for students and instructors because it is thought to affect student success ( Blatchford et al., 2009 ; Maringe & Sing, 2014 ).

Most studies on class size focus on student academic performance, but the results are inconclusive. At the elementary and secondary levels, some studies suggest that smaller classes positively impact academic performance ( Glass & Smith, 1978 ; Robinson, 1990). Others indicate that class size has limited or no impact on performance ( Hanushek, 1986 ; Hoxby, 2000 ). Similarly, in higher education, some studies indicate no difference on course grades between large and small classes ( Guseman, 1985 ; Raimondo et al., 1990 ; Karakaya et al., 2001 ) and others report negative effects on academic performance ( Paola et al., 2013 ; Maringe & Sing, 2014 ). Moreover, there is a dearth of research explaining how and why class size influences student behaviors and educational outcomes. Additional research is needed to better understand classroom dynamics related to class size ( Anderson, 2000 ; Finn et al., 2003 ); Blatchford et al., 2009 ).

Another issue in the class size literature is that most studies focus on grades or standardized test scores as the primary measure of student success. Thus, research that examines the relationship between class size and educational outcomes beyond academic performance should be included in learning assessment. For example, the Association to Advance Collegiate Schools of Business ( AACSB, 2021 ) endorses the use of “well-documented assurance of learning (AoL) processes that include direct and indirect measures for ensuring the quality of all degree programs that are deemed in scope for accreditation purposes.” In addition, the shift from teacher-directed to student-centered pedagogy means that student perception of learning has become an important educational outcome ( Maher, 2004 ; Adam, 2004 ). Today, student satisfaction is recognized as critical factor in attracting and retaining students ( Santini et al., 2017 ).

This study aims to fill the aforementioned gaps in the literature by applying Finn et al.’s (2003) theoretical framework of student engagement in the classroom. They suggest that student academic and social engagement with peers and teachers may influence academic achievement. This study uses Finn’s framework to investigate how student learning and social behaviors influence relevant educational outcomes in different class sizes. The purpose of the study is twofold: to better understand the dynamics of class size in business education and to provide practical strategies to improve educational outcomes and student satisfaction in large and small classes.

Theoretical background

Although business schools typically consider class size a factor in determining teaching loads, there is no accepted definition of a large class. Mateo and Fernandez (1996) propose a numerical taxonomy. For example, a large class contains between 60 and 149 students. Maringe and Sing (2014) define large class size qualitatively as “any class where the number of students poses both perceived and real challenges in the delivery of quality and equal learning opportunities to all students in the classroom” (p. 763). In practice, class size norms vary greatly across institutions and disciplines, with some business schools considering sections of 25–35 students to be small and between 200 and 350 to be large ( Raimondo et al., 1990 ).

Conceptual model

Finn et al. (2003) suggest student academic achievement is influenced by a combination of academic and social engagement in the learning process. Academic engagement refers to student learning behaviors related directly to the learning process, such as class participation. Social engagement is student social interactions with classmates and the instructor. Using group theory, Finn et al. (2003) argue that students in small classes are more visible and more likely to engage in learning and social behaviors during class. Conversely, large classes permit students to reduce their visibility. Also, smaller classes encourage participation or interaction as students may receive more support from classmates. Because social and academic interactions are the focal point of the higher education, these classroom dynamics are critical to positive learning outcomes ( Demaris and Kritsonis, 2008 ).

Blatchford et al. (2009) suggest that a negative relationship exists between class size and classroom processes. Class size differences may impact classroom processes, which in turn influence student attentiveness and active involvement with teachers and peers. Teachers in small classes are more likely to give individual attention to students, effectively control and manage the classroom and build better relationships with students. Similarly, students in small classes may be more engaged in classroom and more likely to interact with teachers and peers ( Blatchford et al., 2009 ).

Consistent with Finn et al. (2003) and Blatchford et al. (2009) , this study proposes a research model in Figure 1 suggesting how class size affects student learning and social behaviors, as well as learning outcomes. It includes perceived learning outcomes and satisfaction as additional educational outcomes.

Research hypotheses

Student attentiveness level will be higher in small classes than in large classes.

Student involvement level with the course will be higher in small classes than in large classes.

Student class participation level will be higher in small classes than in large classes.

Student interaction with classmates will be higher in small classes than in large classes.

Student perception of teacher encouragement will be higher in small classes than in large classes.

Student perception of teacher supportiveness will be higher in small classes than in large classes.

The factors affecting student academic performance may include attentiveness, involvement, class participation, interaction with classmates, perception of teacher encouragement, perception of teacher supportiveness and class size.

The factors affecting the student perceived learning outcome of business knowledge may include attentiveness, involvement, class participation, interaction with classmates, perception of teacher encouragement, perception of teacher supportiveness and class size.

The factors affecting the student perceived learning outcome of communication skills may include attentiveness, involvement, class participation, interaction with classmates, perception of teacher encouragement, perception of teacher supportiveness and class size

The factors affecting student satisfaction may include attentiveness, involvement, class participation, interaction with classmates, perception of teacher encouragement, perception of teacher supportiveness and class size.

This study was conducted in the business school of a medium-sized public university in the northeastern USA. Students were recruited from three introductory business courses (management, marketing and business law) that were offered in both small and large sections. The typical size of most classes in this business school is 25–40 students and a class with more than 50 students is considered large. In general, students are able to select their own classes and all majors were represented in all classes.

Survey participants were invited by their instructors to take an online survey voluntarily at the end of the semester. Participating students granted permission to the researchers to access to their academic records, including course grade. A total of 280 student participated with 52 respondents from small classes and 228 from large classes. The overall response rate was 43% (37% for small classes vs 45% for large classes). Approximately 47.9% of respondents were male and 51.4% were female. Nearly all (98.6%) respondents attended school full-time. The average number of credits completed was 65.

Student academic performance was measured using a student’s course grade (A = 4.0, A− = 3.670, B+ = 3.330, B = 3.000, B− = 2.670, C+ = 2.330, C = 2.000, C− = 1.670, D+= 1.330, D = 1.000, D− = 0.670, F = 0.00). The mean and the standard deviation for course grades were 3.223 and 0.497. Most measures of this study were adapted from previous research. Student attentiveness was assessed by items from Leufer’s (2007) research on the factors affecting the learning environment. Student involvement with the course was measured by using a personal involvement inventory and the questions were adapted to fit the context of the survey ( Zaichkowsky, 1994 ). Bai and Chang’s (2016) measures were adapted to assess class participation, student interaction with classmates, student perception of teacher encouragement and student perception of teacher supportiveness. Measures of the SLOs of business knowledge and communication skills were developed by the researchers. These two SLOs were selected because they are assessed as part of the institution’s AACSB AOL process. Student satisfaction was adapted from Eastman et al. (2017) . All questions used seven-point scales.

SPSS was used to run Cronbach’s coefficient alpha of each measure. The results ranged from 0.929 to 0.756, higher than the minimum requirement of 0.7 ( Nunnally, 1978 ). The scores of composite reliabilities for all constructs were higher than 0.6 and demonstrated the reliability of the measures. Harman’s single factor test was used to evaluate common method variance (CMV). The CFA results indicate 41% total variance explained and then no CMV issues ( Podsakoff et al., 2003 ). Table 1 displays the means, standard deviations, and correlation matrix of all constructs.

A one-way ANOVA was run to test H1 with class size as the independent variable and student attentiveness as the dependent variable. On Table 2 , there was no significant difference in student attentiveness between small and large class sizes (F = 1.00, p < 0.32).

Following the same logic, a series of ANOVA were run to test H2 – H6 . On Table 2 , the results indicate no significant differences between class sizes on student involvement ( H2 ), class participation ( H3 ), interaction with classmates ( H4 ) and student perceived teacher encouragement ( H5 ). But, the results support H6 (F = 6.25, p < 0.013), suggesting that class size differences impact perceived teacher supportiveness. Students in small classes perceived higher level of teacher supportiveness than those in large class (Mean = 6.14 > Mean = 5.68).

Multiple regression analysis was conducted to test H7 – H10 . A dummy variable was created to represent two categories of class size: small class (coded as 0); and large class (coded as 1). As discussed below, each educational outcome was a dependent variable and all factors involving student learning and social behaviors and the class-size dummy variable were independent variables. Table 3 displays the statistical results for H7 – H10 .

The regression model for H7 was not statistically significant (F (7, 272) = 1.424, p < 0.195) and the results did not support H7 . The regression analysis for H8 was statistically significant (F (7, 272) = 101.912, P < 0.000). The results indicate that student involvement ( β 2 = 0.321, p < 0.000), student interaction with classmates ( β 4 = 0.085, p < 0.048), student perception of teacher encouragement ( β 5 = 0.276, p < 0.000) and teacher supportiveness ( β 6 = 0.349, p < 0.000) impact student perceived learning outcome for business knowledge.

The regression model for H9 was statistically significant (F (7, 272) = 56.75, P < 0.000). The results reveal that student involvement ( β 2 = 0.176, p < 0.000), student class participation ( β 3 = 0.254, p < 0.000), student interaction with classmates ( β 4 = 0.262, p < 0.000), student perception of teacher encouragement ( β 5 = 0.173, p < 0.013) and teacher supportiveness ( β 6 = 0.131, p < 0.042) also influenced perceived learning outcome of communication skills.

The regression model for H10 was statistically significant (F (7, 272) = 67.174, P < 0.000). Student attentiveness ( β 1 = 0.11, p < 0.004), student involvement ( β 2 = 0.144, p < 0.000), student class participation ( β 3 = 0.099, p < 0.042), student perception of teacher encouragement ( β 5 = 0.239, p < 0.000), teacher supportiveness ( β 6 = 0.347, p < 0.000) and class size ( β 7 = −0.127, p < 0.001) were determinants of student satisfaction. Importantly, student satisfaction level was significantly lower in large classes than those in small classes by 12.7%.

Structural equation modeling (SEM) was conducted to understand how student learning behaviors, peer interaction, and teacher interactions influence the three significant educational outcomes. The partial least square SEM analysis (PLS-SEM) works efficiently with small sample sizes, multi-item measures and complex structural models. It makes no distributional assumptions for data ( Hair et al., 2017 ). We use SmartPLS 3 software to run PLS-SEM analysis, as it is an appropriate approach for our data set to assess the key drivers for educational outcomes.

Given the concern of discriminant validity for teacher encouragement and teacher supportiveness, these two factors are combined into one construct to represent teacher interaction. Figure 2 displays the structural model. The sample size for small classes in this study exceeds the minimum sample size for PLS SEM analysis based on the 10 times rules ( Hair et al., 2017 ).

The measurement model on convergent validity, reliability and the discriminant validity were assessed. On Table 4 , the values of Cronbach’s alpha and composite reliability of all the constructs exceeding the standard level of 0.70. The average variance extracted (AVE) for all the constructs exceeds the lower acceptable limit of 0.50 ( Hair et al., 2017 ) except for attentiveness. Discriminant validity is assessed with Fornell and Larcker criterion and the heterotrait–monotrait ratio of correlations (HTMT) ( Hair et al., 2017 ). Table 5 results support discriminant validity as the variance shared between constructs is lower than the variance shared by a construct with its indicators ( Fornell & Larcker, 1981 ). The bootstrapping results also support discriminant validity as all the HTMT values are significantly from 1 ( Hair et al., 2017 ). Except for attentiveness with a slightly lower AVE of 0.46 (< 0.5), the other constructs demonstrate good reliability, convergent and discriminant validity for the measurement model ( Hair et al., 2017 ).

On the model results, the value of SRMR is 0.063 (< 0.08) which is considered a good fit. No VIF issue was found as each predictor construct’s VIF value was between 0.2 and 5 ( Hair et al., 2017 ). Then, we ran bootstrapping to assess the significance of path efficient. Table 6 displays each path coefficient, t -values and p -value in the structural model. Most of the paths were statistically significant. However, attentiveness has no significant path coefficients with knowledge and communication outcomes. Student class participation and peer interactions have no significant coefficients with knowledge outcome and satisfaction. The coefficients of determination ( R 2 value) for communication outcome, knowledge outcome and satisfaction are 0.60, 0.74 and 0.68 respectively, which indicate moderate predictive power ( Hair et al., 2017 ). Among all the paths, teacher interaction has large effects on knowledge outcome (coefficient = 0.601) and satisfaction (coefficient = 0.638). Involvement has a medium effect on knowledge outcome (coefficient = 0.315).

Multiple group analysis was tested for differences on path coefficients between small and large classes in the same structural model. The results only indicate a significant difference on the path coefficient from student peer interaction to communication outcome between the two groups. While the path coefficient for large class is 0.22, the value for small class is 0.54. It suggests that peer interaction in a small class exerts a stronger positive effect on student perceived communication outcome than that in a large class.

Discussion and implications

Large classes are unlikely to disappear given the financial pressures that most institutions face in the USA ( Maringe & Sing, 2014 ). However, large classes are associated with challenges in delivering high-quality and equitable learning opportunities ( Bligh, 2002 ). For example, students in large classes do not have the same opportunities to interact with the teacher compared to students in small classes ( Maringe & Sing, 2014 ). To offer large classes without sacrificing quality of education, educators must understand how and why class sizes influence student engagement behaviors and educational outcomes.

For all class sizes, this study found student involvement as the most influential academic engagement behavior and teacher interaction as the most influential social engagement behavior for positive educational outcomes. In addition, students only perceived teacher supportiveness more positively in small classes with no differences for other engagement behaviors. In terms of educational outcomes, this research found negative effects on student satisfaction in large classes, but no differences on course grade, knowledge and communication learning outcomes in large and small classes. Despite of the mixed effects of class sizes, the study reveals that in the large classes, students may perceive a lower level of teacher interactions and satisfaction. It raises critical concerns as teacher interaction was found as the most influential driver for all subjective educational outcomes. Given that student satisfaction is a key component of student and institutional success ( Santini et al., 2017 ), educators must develop strategies to enhance teacher interactions and satisfaction in large classes to maintain and enhance the quality of education and student success.

Practical implications

Because teacher interaction is the most influential factor for student satisfaction, business schools may consider allocating technology resources to support a more interactive learning environment in large classes. For example, the use of a student response systems (SRS), such as clickers or Poll Everywhere, has become more popular. Heaslip et al. (2014) found that students in large classes became more engaged and involved when clickers were in use. In addition, SRS enable students to have more equal opportunities to interact with the teacher easily and efficiently.

Choosing educational outcomes that accurately measures learning objectives is critical to monitor and improve education quality. Educational outcomes should reflect what the program wants the students to know and be able to do. For example, this study included a communication student learning outcome because it was assessed as part of AACSB AOL process. The results found that the communication learning outcome is associated with student participation and peer interaction. Thus, when a course focuses on communication goals, faculty should create more opportunities for student participation and peer interaction in course design. For example, in large classes, SRS allow students to participate in class discussion and also to see other students’ responses. Additionally, as peer interaction in a small class is stronger on communication outcome than that in a large class, a small class is a better choice for a course focusing on communication skills.

Different educational outcomes may involve different student learning and social behaviors in the classroom, whereas class size may not influence all these behaviors. What educational outcomes do schools expect for students? If, for example, student satisfaction is the key educational outcome, then our study suggests that large class sizes should not be used. On the other hand, if course grade is the key educational outcome, both large and small classes will work as grade differences are not related to class size.

Research implications and future research

While the extant literature focuses on academic performance as the primary learning outcome, this study shifted to a “student-centered” perspective and added three measures of subjective educational outcomes – student satisfaction and the perceived SLOs for knowledge and communication skills – to transcend the usual academic performance outcomes. As schools use AOL results for continuous improvement, educators should consider a broader set of educational outcomes.

There have been tremendous changes in the modality of course delivery in higher education since the covid pandemic. Drea (2021) notes that higher education is unlikely to fully return to pre-COVID-19 course delivery models, as students have now experienced the intensive integration of technology into their courses, and this has likely reset their expectations for the future. There are a variety of course delivery formats emerging since the pandemic. Educators must understand whether a change of modality of content delivery has an impact on quality. The current study recognized the importance of both academic and social engagements in student learning. It may offer a groundwork to investigate how student engagements influent their learning outcomes in different delivery formats, such as online, hybrid, synchronous online or asynchronous online.

Limitations

Because this study was conducted in a face-to-face classroom setting, the results cannot be generalized to different learning environments, such as online or hybrid. Similarly, the results are not generalizable across all types of higher education institutions because it was conducted at one university. In addition, the results are limited to introductory business courses and do not include advanced courses that require higher-order thinking and analytical skills. Finally, the sample included only undergraduate students and does not consider age as a salient factor when considering effects on classroom processes ( Blatchford et al., 2009 ).

Conclusions

Many schools use large classes to respond to shrinking resources. This study contributes to the existing literature by showing how and why class sizes influence student engagement behaviors and educational outcomes other than academic performance. Student involvement and teacher interaction are found as influential factors on student learning outcomes and satisfaction regardless of class sizes. However, the study results indicate students perceive lower levels of teacher interaction and satisfaction in larger classes. In conclusion, to offer large classes without sacrificing quality of education, we suggest faculty creating more opportunities to encourage more student–teacher interactions, such as using SRS technologies.

Research model

PLS-SEM structural model

Descriptive statistics and correlation matrix

Fornell–Lacker criterion

Structural model path coefficients

AACSB ( 2021 ) The 2020 AACSB business standards (July 2021) . Retrieved from www.aacsb.edu/accreditation/standards/business ( accessed 7 September 2021 ).

Adam , S. ( 2004 ). Using learning outcomes: A consideration of the nature, role, application and implications for European education of employing ‘learning outcomes’ at the local, national and international levels . 1-2 July , Heriot-Watt University (Edinburgh Conference Centre) Edinburgh . Scotland . Retrieved from www.aic.lv/bolona/Bologna/Bol_semin/Edinburgh/S_Adam_Bacgrerep_presentation.pdf ( accessed 9 February 2021 ).

Alderman , M. K. ( 2008 ). Motivation for Achievement. Possibilities for Teaching and Learning , (3rd ed.) , Routledge .

Anderson , L. W. ( 2000 ). Why should reduced class size lead to increased student achievement? . In M. C. , Wang & J. D. , Finn (Eds.), How Small Classes Help Teachers do Their Best , 3 – 24 .

Bai , Y. & Chang , T. ( 2016 ). Effects of class size and attendance policy on university classroom interaction in Taiwan . Innovations in Education and Teaching International , 53 ( 3 ), 316 – 328 . doi: 10.1080/14703297.2014.997776 .

Blatchford , P. , Russell , A. , & Brown , P. ( 2009 ). Teaching in large and small classes . in L. J. Saha & A. G. Dworkin , (Eds.), International handbook of research on teachers and teaching , pp. 779 – 790 . New York, NY : Routledge .

Bligh , J. ( 2002 ). The first year of doctoring: Still a survival exercise . Medical Education , 36 ( 1 ), 2 – 3 . doi: 10.1046/j.1365-2923.2002.01129.x .

Demaris , M. C. & Kritsonis , W. A. ( 2008 ). The classroom: Exploring Its effects on student persistence and satisfaction . Online Submission , 2 ( 1 ), 1 – 9 .

Drea , J. ( 2021 ). Online? In person? The power of letting students choose . Retrieved from https://hbsp.harvard.edu/inspiring-minds/online-in-person-the-power-of-letting-students-choose ( accessed 7 September 2021 )

Eastman , J. K. , Aviles , M. , & Hanna , M. D. ( 2017 ). Determinants of perceived learning and satisfaction in online business courses: An extension to evaluate differences between qualitative and quantitative courses . Marketing Education Review , 27 ( 1 ), 51 – 62 . doi: 10.1080/10528008.2016.1259578 .

Ehrenberg , R. G. , Brewer , D. J. , Gamoran , A. , & Willims , J. D. ( 2001 ). Class size and student achievement . Psychological Science in the Public Interest , 2 ( 1 ), 1 – 30 . doi: 10.1111/1529-1006.003 .

Fassinger , P. ( 1995 ). Understanding classroom interaction: Students' and professors' contributions to students' silence . The Journal of Higher Education , 66 ( 1 ), 82 – 96 . doi: 10.1080/00221546.1995.11774758 .

Finn , J. D. , Pannozzo , G. M. , & Achilles , C. M. ( 2003 ). The “why's” of class size: Student behavior in small classes . Review of Educational Research , 73 ( 3 ), 321 – 368 . doi: 10.3102/00346543073003321 .

Fornell , C. & Larcker , D. F. ( 1981 ). Evaluating structural equation models with unobservable variables and measurement error . Journal of Marketing Research , 18 ( 1 ), 39 – 50 . doi: 10.1177/002224378101800104 .

Glass , G. V. & Smith , M. L. ( 1978 ). Meta-Analysis of Research on the Relationship of Class Size and Achievement , San Francisco : Far West Laboratory for Educational Research and Development .

Guseman , D. ( 1985 ). Class size impact upon student learning and attitudes in the introductory marketing class . Journal of Marketing Education , 7 ( 1 ), 2 – 7 . doi: 10.1177/027347538500700102 .

Hair , J. F. , Hult , G. , Ringle , C. , & Sarstede , M. ( 2017 ). A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM) , Thousand Oaks, CA : Sage Publication .

Hanushek , E. ( 1986 ). The economics of schooling: Production and efficiency in public schools . Journal of Economic Literature , 24 , 1141 – 1177 .

Heaslip , G. , Donovan , P. , & Cullen , J. G. ( 2014 ). Student response systems and learner engagement in large classes . Active Learning in Higher Education , 15 ( 1 ), 11 – 24 . doi: 10.1177/1469787413514648 .

Hoxby , C. ( 2000 ). The effects of class size on student achievement: New evidence from population variation . Quarterly Journal of Economics , 115 , 1239 – 1285 .

Jaschik , S. ( 2017 ). The 2017 survey of admissions directors: Pressure all Around . Retrieved from www.insidehighered.com/news/survey/2017-survey-admissions-directors-pressure-all-around ( accessed 7 September 2021 ).

Karakaya , F. , Ainscough , T. L. , & Chopoorian , J. ( 2001 ). The effects of class size and learning style on student performance in a multimedia-based marketing course . Journal of Marketing Education , 23 ( 2 ), 84 – 90 . doi: 10.1177/0273475301232002 .

Leufer , T. ( 2007 ). Students’ perceptions of the learning experience in a large class environment . Nursing Education Perspectives , 28 ( 6 ), 322 – 326 .

Maher , A. ( 2004 ). Learning outcomes in higher education: Implications for curriculum design and student learning . The Journal of Hospitality Leisure Sport and Tourism , 3 ( 2 ), 46 – 54 . doi: 10.3794/johlste.32.78 .

Maringe , F. & Sing , N. ( 2014 ). Teaching large classes in an increasingly internationalising higher education environment: Pedagogical, quality and equity issues . Higher Education , 67 ( 6 ), 761 – 782 . doi: 10.1007/s10734-013-9710-0 .

Martirosyan , N. ( 2015 ). An examination of factors contributing to student satisfaction in Armenian higher education . International Journal of Educational Management , 29 ( 2 ), 177 – 191 .

Mateo , M. A. & Fernandez , J. ( 1996 ). Incidence of class size on the evaluation of university teaching quality . Educational and Psychological Measurement , 56 ( 5 ), 771 – 778 . doi: 10.1177/0013164496056005004 .

Nunnally , J. C. ( 1978 ). Psychometric theory , 2nd ed ., New York, NY : McGraw-Hill .

Paola , M. D. , Ponzo , M. , & Scoppa , V. ( 2013 ). Class size effects on students achievement: Heterogeneity across abilities and fields . Education Economics , 21 ( 2 ), 135 – 153 . doi: 10.1080/09645292.2010.511811 .

Podsakoff , P. M. , MacKenzie , S. B. , Lee , J.-Y. , & Podsakoff , N. P. ( 2003 ). Common method biases in behavioral research: A critical review of the literature and recommended remedies . Journal of Applied Psychology , 88 ( 5 ), 879 – 903 . doi: 10.1037/0021-9010.88.5.879 .

Raimondo , H. J. , Esposito , L. , & Gershenberg , I. ( 1990 ). Introductory class size and student performance in intermediate theory courses . Research in Economic Education , 1 , 369 – 381 .

Santini , F. D. , Ladeira , W. J. , Sampaio , C. H. , & da Silva Costa , G. ( 2017 ). Student satisfaction in higher education: A meta-analytic study . Journal of Marketing for Higher Education , 27 ( 1 ), 1 – 18 . doi: 10.1080/08841241.2017.1311980 .

Weaver , R. R. & Qi , J. ( 2005 ). Classroom organization and participation: College students’ perceptions . The Journal of Higher Education , 76 ( 5 ), 570 – 601 . doi: 10.1353/jhe.2005.0038 .

Wulff , D. H. , Nyquist , J. D. , & Abbott , R. D. ( 1987 ). Students’ perceptions of large classes . New Directions for Teaching and Learning , 32 , 17 – 30 .

Zaichkowsky , J. L. ( 1994 ). Research notes: The personal involvement inventory: Reduction, revision, and application to advertising . Journal of Advertising , 23 ( 4 ), 59 – 70 . doi: 10.1080/00913367.1943.10673459 .

Further reading

Statham , A. , Richardson , L. , & Cook , J. A. ( 1991 ). Gender and University Teaching: A Negotiated Difference , Albany : State University of New York, NY Press .

Corresponding author

Related articles, we’re listening — tell us what you think, something didn’t work….

Report bugs here

All feedback is valuable

Please share your general feedback

Join us on our journey

Platform update page.

Visit emeraldpublishing.com/platformupdate to discover the latest news and updates

Questions & More Information

Answers to the most commonly asked questions here

• Open access
• Published: 14 July 2022

The impact of virtual learning on students’ educational behavior and pervasiveness of depression among university students due to the COVID-19 pandemic

• Fatima M. Azmi   ORCID: orcid.org/0000-0001-9275-0965 1 ,
• Habib Nawaz Khan   ORCID: orcid.org/0000-0003-3519-264X 2 &
• Aqil M. Azmi   ORCID: orcid.org/0000-0002-0983-2861 3  

Globalization and Health volume  18 , Article number:  70 ( 2022 ) Cite this article

50k Accesses

15 Citations

2 Altmetric

Metrics details

One of the worst pandemics of recent memory, COVID-19, severely impacted the public. In particular, students were physically and mentally affected by the lockdown and the shift from physical person-to-person classrooms to virtual learning (online classes). This increased the prevalence of psychological stress, anxiety, and depression among university students. In this study, we investigated the depression levels in Saudi Arabian university students who were learning virtually because of the COVID-19 pandemic and examined its impact on their educational proficiency.

The study focused on two points: first, examining the depression levels among undergraduate students in Saudi Arabia, by adapting the Zung (Self-Rating Depression Scale) questionnaire. Second, whether there is an association between the levels of depression and various distress factors associated with virtual (online) learning resulting from the COVID-19 pandemic and its impact on students’ educational behaviors. The questionnaire was prepared using a monkey survey and shared online, via email, and on WhatsApp groups, with participants in two universities, a public and private university in the largest city of Saudi Arabia. A total of 157 complete responses were received. Data were analyzed using SPSS-24, the chi-square test, descriptive statistics, and multilinear regression.

The results indicated that three-fourths of the university students suffered from different depressive symptoms, half of which had moderate to extreme levels of depression. Our study confirmed that a boring virtual (online) learning method, stress, fear of examinations, and decreased productivity were significantly associated with increased depression. In addition, 75% and 79% of the students suffered from stress and fear of examinations, respectively. About half of the students were associated with increased depression. The outcome also indicated that female students experienced extreme depression, stress, and fear of examinations more than males.

These findings can inform government agencies and representatives of the importance of making swift, effective decisions to address students’ depression levels. It is essential to provide training for students to change their educational experience mindset, which might help decrease "depression and stress-related growth." There is also a need to search for a better virtual teaching delivery method to lessen students' stress and fear of examinations.

On March 11, 2020, the World Health Organization (WHO) declared the highly contagious coronavirus (COVID-19) a global pandemic [ 1 ]. As the cases of COVID-19 increased, China, and many other countries practiced partial or complete lockdowns. It is estimated that this drastic measure helped save 3 million lives across 11 European nations [ 2 ]. Toward the end of January 2022, the total number of confirmed COVID-19 cases worldwide was 360,578,392, and 5,620,865 confirmed deaths. The number of people who received vaccination doses globally was 9,679,721,754 [ 3 ].

To contain the virus, the lockdown caused academic disruptions. This resulted in the indeterminate closure of schools, universities, various institutes, shopping malls, and centers of economic activities [ 4 , 5 ]. Repetitive activities, transfer of educational mode to distance (virtual) learning, and change in social life amplified the prevalence of psychological stress, anxiety, depression, and acute stress reactions among university students [ 6 ]. Sociodemographic factors associated with low mental health include financial constraints, old age, infection risk, and fear of losing a relative or friend. In addition, COVID-19 pandemic-related educational stress may be attributed to (in no particular order): (a) transformed teaching and assessment methods; (b) skepticism about university education; (c) technological worries about online courses [ 7 , 8 ]; (d) uncertainty about the future because of academic disruptions; (e) fear of failing examinations; (f) inability to concentrate during lectures, and many more factors. All these factors have been detected in universities across the world [ 9 , 10 ]. A global study that inspected students’ experiences in about 62 different countries, including a university in the United States, found that students expressed worries about their academic achievements and professional careers and feelings of dullness, anxiety, and frustration [ 9 ]. Students in China also reported increased sadness, anger, anxiety, and fear [ 11 ]. The occurrence of depression, psychological distress, and anxiety from pandemics differed from country to country. A study in Italy reported that 15.4% of Italians suffered from extremely high levels of depression, 12.6% were highly stressed, and 11.5% were highly anxious [ 12 ]. In Malaysia, it was reported that severe to extremely severe levels of depression and anxiety were found in 9.2% and 13.2% of the subjects, respectively. Moderate stress was found in 9.5% of subjects, and severe to highly severe stress was found in 6.6% of subjects [ 13 ]. Furthermore, students in Switzerland manifested a decrease in social interface and higher levels of stress, anxiety, and loneliness [ 14 ]. Adults have also reported declining physical activity, while food eating increased during pandemic quarantine periods compared to previous times [ 15 ].

The first COVID-19 case appeared in the Kingdom of Saudi Arabia (KSA) on March 2, 2020 [ 16 ], while the lockdown was imposed on March 8, 2020. To keep students on track due to the pandemic, the education delivery mode was switched to virtual learning. It has been over one year since teaching was transferred online, and many countries worldwide have tried to revert to the standard path of education by opening schools and universities. Although the COVID-19 vaccine is available worldwide, some countries are still practicing lockdown because of the appearance of several more contiguous variants of the coronavirus, such as Delta, a SARS-CoV-2 strain that was first spotted in India [ 17 ]. The spread of COVID-19 presents a serious risk; in mid-April 2022, the confirmed cases in KSA were 751,717, out of which 736,910 had recovered, and 9,055 deaths were recorded [ 18 ].

The psychological consequences of COVID-19 have been observed and described in KSA. Al-Hanawi et al. [ 19 ] reported different levels of distress in 40% of the general Saudi population because of COVID-19. Moreover, Alkhamees et al. [ 20 ] reported moderate to severe psychological effects in 23.6% of the general Saudi population. In another study of the influence of the COVID-19 pandemic on Saudi Arabian residents, Alyami et al. [ 21 ] stated that the percentages of mild, moderate, moderately severe, and severe levels of depression were 41%, 20%, 6.2%, and 3.2%, respectively. Furthermore, Khoshaim et al. [ 22 ] reported that about 35% of students experienced moderate to extreme anxiety levels. Azmi et al. [ 23 ] observed that 75% of students suffered from various levels of depression, while 41% suffered from low levels of self-esteem.

Likewise, another study found that 35% of students in the western and northern regions of KSA had high rates of distress [ 24 ]. Following the observed rise of psychological disorders, the authorities posted health messages and distributed procedures to the public. For example, during the pandemic, the Saudi Center for Disease Control and Prevention (CDC) [ 25 ] supplied a precautionary manual for mental and social health focused on prevention, pressure, and fear control. From the foregoing, the COVID-19 pandemic has had a severe impact on the physical and mental health of the public in general and students in particular, as university students are among those most severely affected by the COVID-19 pandemic.

In this study, we investigated the depression levels of university students in Riyadh, the capital and largest city in KSA, who were learning virtually because of the COVID-19 pandemic. We also assessed the impact of virtual learning on their educational behaviors. The following questions were explored during the investigation:

What are the levels of depression among university students?

What is the impact of virtual learning on students’ educational behaviors and what are the relationship between depressive symptoms they exhibited and virtual learning?

To answer the second question, we explored the relationship between the levels of depression and various distress factors associated with virtual learning because of the COVID-19 pandemic and its impact on students’ educational behaviors. These factors were divided into two main categories: Category 1 dealt with factors relating to how virtual learning has affected students’ feelings from an educational perspective. Category 2: dealt with factors relating to how virtual learning affected students’ understanding of subjects/learning materials.

Once we ascertain the current levels of depression and their impact on students’ educational behavior, we may embark on helping them cope with the extraordinary situation. Hopefully, this will help lower their elevated depression levels. Furthermore, we hope our study will guide policymakers in searching for innovative ways of online teaching to make learning less stressful and more productive.

Design and sampling procedure

This study examines depression levels and investigates virtual learning-related distress factors, which might predict the increased level of depressive symptoms among university students in Riyadh City during the COVID-19 pandemic.

Research design

We conducted a descriptive survey-based study to obtain responses from students at large universities in Riyadh, the capital of KSA. The total size of the target population of the city of Riyadh is about 7 million [ 26 ]. The sampled population of both universities’ undergraduate students was approximately 0.027 million (27,000). The male-to-female ratio of undergraduate students at King Saud University (KSU) is about 67%: 33%; the male-to-female ratio of undergraduate students at Prince Sultan University (PSU) is about 28%: 72%, as this is a female-dominated university. Since the sampled population was largely heterogeneous, we minimized the heterogeneity by dividing the given population into sub-populations to obtain sampling units that are homogeneous internally and heterogeneous externally. Hence, we used a stratified random sampling technique, which is more appropriate than other sampling techniques for obtaining better estimates of the parameters of interest. To ensure the efficiency of the estimates, we used the proportional allocation technique to determine the sample size.

A Monkey survey was used to prepare the questionnaire, following the approval of PSU’s Institutional Review Board. The questionnaire included demographic questions, such as gender, age, and college. Zung’s Self-Rating Depression Scale (ZSDS), with 20 items on a 4-point Likert scale, was used to measure depression. The questionnaire also had questions to address distress factors associated with virtual learning because of the COVID-19 pandemic. The students were asked to read all the questions carefully and answer them.

The survey was written in English and Arabic side by side. A subject expert translated the questionnaire from Arabic to English. Thereafter, five more experts checked the same questions for more corrections and authenticity. The actual online survey took place from March to April 2021. The survey was voluntary, and the informed consent of the students was sought. We received reasonable responses from the students; however, we also received some incomplete responses. The missing/incomplete responses were discarded from the study so that the estimated results were not compromised. The valid responses received from males and females were 49.7% and 50.3%, respectively.

Measuring instruments

Demographic data and personal characteristics, such as age, gender, and area of study, were recorded.

Depression measure

The ZSDS was used to measure the levels of depression. The tool is a 20-item self-reporting assessment device used for measuring depression levels [ 27 , 28 ]. This is divided into 10 positively worded and 10 negatively worded items. The latter items were reversely scored. Each item was scored on a Likert-type scale as follows: 1 =  Never , 2 =  Sometimes , 3 =  Often/most of the time , and 4 =  Always . The total raw scores ranged from 20–80, and when converted into the depression index (termed "ZSDS index"), the range becomes 25–100. To determine the level of depression, we classified the ZSDS index into four classes (levels). Therefore, ZSDS index scores were considered "normal" from 25–49, "Mildly Depressed,” from 50–59, “Moderately Depressed” from 60–69, and “Severely Depressed” from 70 and above [ 27 ]. In [ 29 ], the author translated the ZSDS measure into Arabic and further validated it. Question 6, “I still enjoy sex,” was deemed offensive religiously and culturally. Therefore, it was rephrased to “I enjoy looking at, talking to, and being with attractive women/men,” which is culturally more appropriate. The accuracy of the new version was verified in [ 29 ]. The Arabic and English languages were used side by side to prepare the questionnaire. The Cronbach’s alpha coefficient of this study was 0.87, showing high internal consistency.

Data on distress factors associated with virtual learning

Data on distress factors associated with virtual learning due to the COVID-19 pandemic were divided into two categories. The first category dealt with questions on how virtual learning due to the pandemic affected students’ feelings from an educational perspective and caused a) lack of motivation/boredom, b) stress, c) worry and fear of exams, and d) decreased productivity. The second category dealt with questions on virtual learning and its effect on students’ understanding of subjects/materials, such as a) It needs more self-effort to understand, b) It made learning and understanding harder for them, c) They need more time to understand the subject, i.e., the understanding pace became slower, d) Virtual learning is boring, and e) they had difficulty solving problems in academic subjects and writing down the solutions correctly. The answer to each question was either “Yes” or “No.”

Finally, the questionnaire had an open-ended question that offered students a chance to express in their own words how the lockdown and virtual teaching had affected their educational advancement.

Data analysis

Data were analyzed using IBM SPSS version 24 software. The categorical variable demographic data were analyzed descriptively to determine the essential characteristics of the sample and were presented as counts and percentages. The level of depression index among university students in Riyadh, and its association with gender, age, and their field of education, was analyzed using the chi-square test and descriptive statistics. Multilinear regression analysis was performed to investigate the connection between depression levels and various factors associated with virtual learning due to the COVID-19 pandemic. The statistically significant level was set at \(p \le 0.05.\)

Demographic characteristics

The total number of participants was 157 university students. Table 1 shows the demographic characteristics of the participants.

Students’ levels of depression and demographic variables

In the univariate analysis, chi-square tests were used to determine the associations between students’ demographic variables and the ZSDS level. Table 2 displays the association between depression levels with gender, age, and college. Among the demographic variables, only the association with gender was statistically significant at \({\chi }^{2}\) = 20.229, and p  < 0.001, while the association with age and college was not significant. A total of 74.4% of the students had various levels of depression. Of these, 37%, 21.7%, and 16% had mild, moderate, and severe depression levels, respectively. In addition, females (28%) had more depressive symptoms than males (4%).

Educational distress factors associated with virtual learning and descriptive statistics

The factors related to virtual learning sequel to the COVID-19 pandemic, and its impact on students’ educational behaviors were divided into two categories. Questions on virtual learning's effect on students' feelings from an educational perspective (Category 1) had four items, each with a "Yes" or "No" answer. Likewise, questions on virtual learning and its effect on students’ understanding of the subjects/materials (Category 2) had five items, each with a “Yes” or “No” answer. Table 3 demonstrates the descriptive statistics. In the first category, the highest percentage was feeling worried and having a fear of exams (79%), followed by stress (75.2%), lack of motivation, and decreased productivity (70%, each). In the second category, the highest percentage was 78%, who felt they had to put extra self-effort into understanding and studying.

Furthermore, 74.5% felt that virtual learning was more challenging for them to understand than physical learning. In addition, 73% said virtual learning was slow and extra time was needed to understand and learn the concepts, while 64% found it boring. Finally, 58.6% had difficulty solving problems and submitting properly written answers (for math and computer science subjects).

Distress factors related to virtual learning and depressive symptoms

Multilinear regression analysis was used to study whether various distress factors related to virtual learning can influence depressive symptoms among students.

The first category, which dealt with students’ feelings from the educational point of view, hypothesized that lack of motivation, stress, worry/fear of examinations, and decreased productivity would significantly impact the development of depressive symptoms among students.

Multi-regression analysis was used to test the hypotheses, with the Zung depression index as a dependent variable. The results show that 24.6% of the variance in Zung’s depression index can be accounted for by four predictors, collectively \(, F(4, 152) = 12.414, p < 0.001\) . Looking at the unique individual contribution of the predictors, the result shows that worry and fear of exams ( \(\beta =0.290, t=3.589, p<0.001)\) , stress ( \(\beta =0.202, t=2.566, p=0.011<0.05)\) , and decreased learning amount and not being productive ( \(\beta =0.211, t=2.783, p=0.006<0.05)\) , statistically significantly contributed to worsening depressive symptoms. The predictor, feeling lack of motivation, did not significantly impact developing depressive symptoms.

The second category dealt with virtual learning and its effect on students’ understanding of the subjects/materials. It was hypothesized that the need for extra self-effort to understand the subject, learning became harder, learning became slower, learning was boring, and difficulty in solving problems and writing answers properly would have a statistically significant impact on developing depressive symptoms among students.

Multi-regression analysis was used to test the hypotheses, with Zung’s depression index as a dependent variable. The test showed that 13% of the variance in Zung's depression index can be accounted for by the five predictors, collectively \(, F(5, 151) = 4.505, p < 0.001\) . Looking at the unique individual contribution of the predictors, the result shows that learning is not much fun or exciting ( \(\beta =0.250, t=3.060, p=0.003<0.05)\) , and facing difficulty in solving questions and writing answers properly ( \(\beta =0.176, t=2.067, p=0.05<0.05)\) , were statistically significantly associated with worsening depressive symptoms. While the other three predictors, learning became harder, learning became slower, and the need to put extra self-effort did not contribute significantly to depressive symptoms, as shown in Table 4 .

Furthermore, we explored two distress factors, stress, and worry/fear of exams, which contributed statistically significantly to worsening depressive symptoms. Using the chi-square test, we examined the association of the distress factors with depression levels; that is, what extent does stress or worry/fear of exams contribute to moderate or severe depression. The results showed a statistically significant association between stress and moderate to severe levels of depression ( \({\chi }^{2}\) = 17.179, and p  < 0.001). Likewise, there was a statistically significant association between worry/fear of exams and moderate to severe levels of depression ( \({\chi }^{2}\) = 30.236, and p  < 0.001), Table 5 .

The association between stress or worry/fear of exams and gender was examined using the chi-square test. There was a statistically significant association between these two factors and gender, with more females having higher stress levels (54%) than males (41%). Also, worry/fear of exams manifested in 60% of females and 40% of males during virtual learning, sequel to the COVID-19 pandemic. The results are presented in Table 6 .

Open-ended questions

The questionnaire ended with an open-ended question, in which students were asked to write in their own words how the lockdown has affected their educational advancement. Excerpts of the negative comments from students are outlined below:

“Virtual teaching and exam resulted in increased cheating." “Virtual teaching caused difficulty in understanding the subject, which resulted in lowering my grades.” "I have to sit in the same room with my siblings while learning online, as my home is small. So, I cannot concentrate at all; it just makes me very frustrated.”

From their comments, it is clear that a virtual learning environment is entirely different from a physical classroom teaching environment where exams are conducted with invigilators proctoring.

Significantly few students provided positive comments.

"Virtual teaching made me understand better and increased productivity and my grades."

In this study, we investigated the severity of depressive symptoms among university students while learning virtually because of the COVID-19 pandemic and its impact on educational behaviors in KSA We collected samples from different universities in Riyadh. The total number of complete responses was 157. The Zung Self-Rating Depression measure was used to measure depression levels. Our results indicate that 75% of the students suffer from different levels of depression (37%, 22%, and 16% of the students reported mild, moderate, and extremely severe levels of depression, respectively). This result is consistent with an American study, which reported that 44% of students in the USA experienced an augmented level of depressive thoughts [ 30 ].

The association between the levels of depression and various distress factors associated with virtual learning due to the pandemic and its impact on students’ educational behaviors was explored using multilinear regression. These factors are divided into two main categories: Category 1: These factors relate to how virtual learning has affected students’ feelings from an educational perspective. This consists of four items: lack of motivation, stress, worry/fear of exam, and decreased productivity. Category 2 factors relate to how virtual learning has affected students’ understanding of the subjects/materials. This category has five items, including need of extra self-effort, need to study harder, learning is slower, virtually learning is boring, difficulty in solving problems, and writing properly.

Consistent with our hypotheses, we confirmed that stress, worry/fear of examinations, and decreased productivity were significantly associated with an increased level of depression. Another recognized factor that contributes significantly to a higher risk of developing depressive symptoms among university students is that virtual teaching and learning becomes boring. Furthermore, students faced difficulty in solving mathematics and science problems and writing the answers properly due to online teaching. A few other factors, such as lack of motivation, learning became more complex and slower, and the need to put extra self-effort contributed to developing depressive symptoms.

Our results indicate that 75% of the students suffer from stress, and about half (47%) have high levels of depression. This is consistent with the results in [ 13 ]. Our findings also indicate that 79% of the students suffer from fear of exams, and about half of them (47%) experience moderate to severe levels of depression. It is usual for some students to have worries and fear for exams; however, it is highly unusual for more than three-fourths of the students to experience fear and worry. This is a clear indication that the changed mode of lecture delivery and exam administration because of COVID-19 has a significant role in raising depression levels among university students. Our findings indicate that a higher percentage of females experience extreme levels of depression than males (28% of females compared to only 4% of males), stress (59% females, vs. 41% males), and worry/fear of exams (60% females, vs. 40% males). This finding is consistent with many studies concerning college students, in which females were at a higher risk of suffering psychologically during virtual learning because of the COVID-19 pandemic [ 9 , 31 , 32 , 33 ]. Another study showed that Vietnamese female students had a higher percentage of depression compared to male students [ 34 ]. Furthermore, Huange et al. [ 35 ] reasoned that Chinese females experienced more anxiety than males during the COVID-19 pandemic. Thus, we assert that feamles are more commonly inclined toward depression and anxiety disorders than males [ 36 ].

The results of the open-ended responses demonstrated the students’ frustration and stress relating to online learning. In contrast, very few students positively indicated that online learning and studying from home felt relaxing.

COVID-19 has been a catastrophic experience. Although it has largely subsided, new variants are causing apprehension among health officials. Our research found that 75% of university students in Saudi Arabia suffer from some degree of depression. Half of these students showed moderate to extreme levels of depression. This is greater than the expected depression level in the overall population. Our study confirms that stress, worry, and fear of examinations, decreased productivity, and the fact that virtual learning is boring are significantly associated with increased depression. Our findings also indicate that 75% (79%) of the students suffer from stress (fear of exams), and that about half of them have increased levels of depression. It should be noted that the students are 18–24 year olds. This is consistent with the study [ 22 ], which found that psychological distress, stress, and anxiety were higher in the younger age group during the COVID-19 pandemic.

Remarkably, more female students experienced extreme depression, stress, and fear of exams than male students. This result supports previous reports that females were at higher risk of psychological distress during the COVID-19 pandemic [ 9 , 31 , 32 , 33 ].

Our observation calls for instant attention and sustenance for students. There is a requirement to explore potential coping policies that have been shown to be effective during pandemics [ 37 ]. The results of our research might direct policymakers to develop distress management protocols as part of their policy for dealing with future pandemics [ 38 ]. It is essential to provide training for students to redirect their educational experience mindset to focus on the “bright side” and expand instances that may guide "depression and stress-related growth.” A flexible mindset can also help students adapt to new ways of learning and developing tremendous gratitude for life. In addition, there is a need to explore better online teaching delivery methods to lower students’ stress and fear of exams.

Study strengths and limitations

The strength of this study is that it was conducted after students had received virtual teaching for more than one year because of the Pandemic. Therefore, this study accurately reflects students’ depression levels and how these impact their educational behaviors in KSA.

Furthermore, the study was conducted in Riyadh, the capital of KSA, hence our study sample is more reflective of the Saudi student population. Moreover, the depression assessment tool for the study, the Zung Self-Rating Depression Scale, is a reliable, universally acceptable scale.

The limitation of our study is that the sample was not randomly selected from all university students, as a convenience sampling method was used.

Availability of data and materials

The raw data supporting the results of this study will be made available by the corresponding author without undue reservation.

WHO. Rolling updates on corona virus 2019 (covid-19). World Health Organization; 2020. https://www.who.int/emergencies/diseases/novel-coronavirus-2019/events-as-they-happen Retrieved on 14th Apr 2020.

Flaxman S, Mishra S, Gandy A, et al. Estimating the effects of non-pharmaceutical interventions on COVID-19 in Europe. Nature. 2020;584(7820):257–61. https://doi.org/10.1038/s41586-020-2405-7 .

Article   CAS   PubMed   Google Scholar  

World Health Organization. Coronavirus disease (COVID-19) (who.int) 2019. https://www.who.int/emergencies/diseases/novel-coronavirus-2019 Accessed 25 Jan 2022.

Viner RM, Russell SJ, Croker H, Packer J, Ward J, Stansfield C, et al. School closure and management practices during coronavirus outbreaks including COVID-19: a rapid systematic review. Lancet Child Adol Health. 2020;4:397–404. https://doi.org/10.1016/S2352-4642(20)30095-X .

Article   CAS   Google Scholar  

Anderson RM, Heesterbeek H, Klinkenberg D, Hollingsworth TD. How will country-based mitigation measures influence the course of the COVID-19 epidemic? Lancet. 2020;395:931–4. https://doi.org/10.1016/S0140-6736(20)30567-5 .

Article   CAS   PubMed   PubMed Central   Google Scholar  

Sankhi S, Marasine NR. Impact of COVID-19 pandemic on mental health of the general population, students, and health care workers: a review. 2020. https://doi.org/10.20944/preprints202007.0616.v1  (preprints.org.)

Book   Google Scholar  

Azmi F, Khoshaim H. The COVID-19 pandemic and the challenges of e-assessment of calculus courses in higher education: a case study in Saudi Arabia. Ijlter. 2021;20(3):265–81. https://doi.org/10.26803/ijlter.20.3.16 .

Article   Google Scholar  

Azmi F. Undergraduate students' preference regarding the type of online homework questions in math and science subjects and instructors' analysis. In proceedings of Conference IC4E 2021: 12th International Conference on E-Education, E-Business, E-Management, and E-Learning. Tokyo, Japan, January 10–13, 2021. https://doi.org/10.1145/3450148.3450209

Aristovnik A, Kerzˇič D, Ravsˇelj D, Tomazˇevič N, Umek L. Impacts of the COVID-19 pandemic on life of higher education students: a global perspective. preprints.org. 2020. https://doi.org/10.20944/preprints202008.0246.v2

Oerther DB, Peters CA. Educating heads, hands, and hearts in the COVID-19 classroom. Environ Eng Sci. 2020;37(5):303–303.

Cao W, Fang Z, Hou G, Han M, Xu X, Dong J, et al. The psychological impact of the COVID-19 epidemic on college students in China. Psychiatr Res. 2020;287:112934. https://doi.org/10.1016/j.psychres.2020.112934 (PMID: 32229390).

Mazza C, Ricci E, Biondi S, Colasanti M, Ferracuti S, Napoli C, et al. A nationwide survey of psychological distress among Italian people during the COVID-19 pandemic: immediate psychological responses and associated factors. Int J Environ Res Public Health. 2020;17(9):3165.

Article   CAS   PubMed Central   Google Scholar  

Woon LS-C, Leong Bin Abdullah MFI, Sidi H, Mansor NS, Nik Jaafar NR. Depression, anxiety, and the COVID-19 pandemic: Severity of symptoms and associated factors among university students after the end of the movement lockdown. PLoS One. 2021;16(5):e0252481. https://doi.org/10.1371/journal.pone.0252481 .

Elmer T, Mepham K, Stadtfeld C. Students under lockdown: comparisons of students’ social networks and mental health before and during the COVID-19 crisis in Switzerland. PLoS ONE. 2020;15:e0236337-e236422. https://doi.org/10.1371/journal.pone.0236337 (PMID: 32702065).

Ammar A, Brach M, Trabelsi K, Chtourou H, Boukhris O, Masmoudi L, et al. Effects of COVID-19 home confinement on physical activity and eating behaviour preliminary results of the ECLBCOVID19 international online-survey. medRxiv.org. 2020:1–24. https://doi.org/10.1101/2020.05.04.20072447

Ministry of Health. MOH reports 3,121 new COVID-19 cases, reiterates the importance of social distancing. 2020. https://www.moh.gov.sa/en/Ministry/MediaCenter/News/Pages/News-2020-06-06-001.aspx Accessed July.

Katella K. 5 things to know about the delta variant. Yale Medicine. 2022. https://www.yalemedicine.org/news/5-things-to-know-delta-variant-covid .

Worldometer. Total Coronavirus Cases in Saudi Arabia. 2022. https://www.worldometers.info/coronavirus/country/saudi-arabia/ .

Google Scholar  

Al-Hanawi MK, Mwale ML, Alshareef N, Qattan AM, Angawi K, Almubark R, et al. Psychological distress amongst health workers and the general public during the COVID-19 pandemic in Saudi Arabia. Risk Manag Healthc Policy. 2020;13:733–42.

Article   PubMed   PubMed Central   Google Scholar  

Alkhamees AA, Alrashed SA, Alzunaydi AA, Almohimeed AS, Aljohani MS. The psychological impact of COVID-19 pandemic on the general population of Saudi Arabia. Compr Psychiatry. 2020;102:152192. https://doi.org/10.1016/j.comppsych.2020.152192 .

Alyami H, Naser A, Dahmash E, Alyami M, Alyami MS. Depression and anxiety during the COVID-19 pandemic in Saudi Arabia: a cross-sectional study. Int J Clin Pract. 2021;75:e14244. https://doi.org/10.1111/ijcp.14244 .

Khoshaim HB, Al-Sukayt A, Chinna K, Nurunnabi M, Sundarasen S, et al. Anxiety level of university students during COVID-19 in Saudi Arabia. Front Psychiatry. 2020;11:579750. https://doi.org/10.3389/fpsyt.2020.579750 .

Azmi FM, Khan HN, Azmi AM, Yaswi A, Jakovljevic M. Prevalence of COVID-19 pandemic, self-esteem and its effect on depression among university students in Saudi Arabia. Front Public Health. 2022;10:836688. https://doi.org/10.3389/fpubh.2022.836688 .

Elhessewi G, Almoayad F, et al. Psychological distress and its risk factors during COVID-19 pandemic in Saudi Arabia: a cross-sectional study. Middle East Curr Psychiatry. 2021;28:7. https://doi.org/10.1186/s43045-021-00089-6 .

Center for Disease prevention and control, Ministry of Health, Riyadh. Novel corona virus (2019-nCoV) infection guidelines V1.2. 2020.

Riyadh Population. Demographics, Maps, Graphs. 2020. https://worldpopulationreview.com/world-cities/riyadh-population .

Zung WW. A self-rating depression scale. Arch Gen Psychiatry. 1965;12:63–70.

Zung WWK. Factors influencing the Self-Rating Depression Scale. Arch Gen Psychiatry. 1967;16(5):543–7.

Kirkby R, AlSaif A, el-din Mohamed G. Validation of an Arabic translation of the Zung Self-Rating Depression Scale. Ann Saudi Med. 2005;25(3):205–8.

Son C, Hegde S, Smith A, Wang X, Sasangohar F. Effects of COVID-19 on college students’ mental health in the United States: interview survey study. J Med Internet Res. 2020;22(9):e21279.

da Silva ML, Rocha RSB, Buheji M, Jahrami H, Cunha KDC. A systematic review of the prevalence of anxiety symptoms during coronavirus epidemics. J Health Psychol. 2020;71:135910532095162–11. https://doi.org/10.1177/1359105320951620 (PMID: 32830577).

Xiong J, Lipsitz O, Nasri F, Lui LMW, Gill H, Phan L, et al. Impact of COVID-19 pandemic on mental health in the general population: a systematic review. J Affect Disord. 2020;277:55–64. https://doi.org/10.1016/j.jad.2020.08.001 (PMID: 32799105).

Sundarasen S, Chinna K, Kamaludin K, Nurunnabi M, Baloch GM, Khoshaim HB, et al. Psychological impact of COVID-19 and lockdown among university students in Malaysia: Implications and policy Recommendations. IJERPH. 2020;17:6206–13. https://doi.org/10.3390/ijerph17176206 (PMID:32867024).

Nguyen DT, Wright EP, Dedding C, Pham TT, Bunders J. Low self-esteem and its association with anxiety, depression, and suicidal ideation in Vietnamese secondary school students: a cross-sectional study. Front Psychiatry. 2019;10:698. https://doi.org/10.3389/fpsyt.2019.00698 .

Huang L, Xu F, Liu H. Emotional responses and coping strategies of nurses and nursing college students during COVID-19 outbreak. 2020. https://doi.org/10.1101/2020.03.05.20031898 .

Hunt J, Eisenberg D. Mental health problems and help-seeking behavior among college students. J Adolesc Health. 2010;46:3–10. https://doi.org/10.1016/j.jadohealth.2009.08.008 (PMID: 20123251).

Article   PubMed   Google Scholar  

Wang H, Xia Q, Xiong Z, Li Z, Xiang W, Yuan Y, et al. The psychological distress and coping styles in the early stages of the 2019 Coronavirus disease (COVID-19) epidemic in the general mainland Chinese population: A web-based survey. PLoS ONE. 2020;15:e0233410. https://doi.org/10.1371/journal.pone.0233410 .

Ross AGP, Crowe SM, Tyndall MW. Planning for the next global Pandemic. Int J Infect Dis. 2015;38:89–94. https://doi.org/10.1016/j.ijid.2015.07.016 .

Download references

Acknowledgements

Fatima Azmi would like to thank Prince Sultan University for funding the project and covering the publication fees.

This work was supported by a research project grant [Grant number: COVID-19-DES-2020–43] from Prince Sultan University, KSA.

Author information

Authors and affiliations.

Department of Mathematics and Sciences, College of Humanities and Sciences, Prince Sultan University, Riyadh, Saudi Arabia

Fatima M. Azmi

Department of Statistics, University of Science and Technology, Bannu, KPK, Pakistan

Habib Nawaz Khan

Department of Computer Science, College of Computer and Information Sciences, King Saud University, Riyadh, Saudi Arabia

Aqil M. Azmi

You can also search for this author in PubMed   Google Scholar

Contributions

Conceptualization: FMA; AMA. Data Curation: FMA; AMA. Formal Analysis: FMA. Methodology: FMA; HNK. Writing-Original Draft: FMA; AMA. Writing-Review and Editing: FMA; HNK; AMA. The author(s) read and approved the final manuscript.

Corresponding author

Correspondence to Fatima M. Azmi .

Ethics declarations

Ethics approval and consent to participate.

This study was approved by Prince Sultan University’s Institutional Review Board (IRB) (Reference Number: PSU IRB-2021–03-0077). Furthermore, informed consent was obtained from the participants.

Competing interests

The authors declare no conflict of interest.

Additional information

Publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Azmi, F.M., Khan, H.N. & Azmi, A.M. The impact of virtual learning on students’ educational behavior and pervasiveness of depression among university students due to the COVID-19 pandemic. Global Health 18 , 70 (2022). https://doi.org/10.1186/s12992-022-00863-z

Download citation

Received : 11 April 2022

Accepted : 29 June 2022

Published : 14 July 2022

DOI : https://doi.org/10.1186/s12992-022-00863-z

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Saudi Arabia
• Higher education
• Quality education

Globalization and Health

ISSN: 1744-8603

• Submission enquiries: [email protected]
• General enquiries: [email protected]