Edited by: Carl Senior,Aston University, United Kingdom
Reviewed by: Patrício Soares Costa, University of Minho, Portugal; Panagiota Dimitropoulou, University of Crete, Greece
This article was submitted to Educational Psychology, a section of the journal Frontiers in Psychology
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Popular descriptions of studying frequency show remarkable discrepancies: students complain about their workload, and alumni describe freedom and pleasure. Unfortunately, empirical evidence on student time use is sparse. To investigate time use and reveal contributing psychological factors, we conducted an e-diary study. One hundred fifty-four students reported their time use and valence hourly over 7 days, both at the start of the semester and during their examination period. Motivational problems, social support and self-control were assessed once via questionnaires. Whereas the mean academic time use was in the expected range, the between-subject differences were substantial. We used multilevel modeling to separately analyze the within- and between-subject associations of valence as within factor and time use and social support, self-control, and motivation as between factors and time use. The analyses revealed the importance of affective factors on a within-subject level. Before studying, valence was already low, and it deteriorated further during studying. As expected at the between-subject level, motivational problems were related to less time studying, whereas surprisingly, self-control had no effect. The findings at the start of the semester were replicated in the examination period.
There are remarkable discrepancies in people’s descriptions of studying. Whereas students complain about their workload and stress, alumni describe how much they enjoyed studying and were motivated by their freedom of choice in terms of topics and courses, personal time management, and learning with other students. The latter reports support psychological and educational models and theories that link self-regulated learning to greater motivation and positive affect (
The importance of assessing academic time use has been addressed frequently in research studies (
Fortunately, recent technological advances have facilitated investigating academic time use and the accompanying psychological processes prospectively in daily life. Different terms have been used to assess data in real time in daily life with these new technologies, including ambulatory assessment (
In the past, paper–pencil diary studies have been used to investigate academic time use. However, accuracy of paper diaries has been questioned, as electronic timestamps are not possible. Using an electronically manipulated paper diary,
Considerable interindividual differences in academic time use have also been reported by other studies (
At the theoretical level, several psychological theories include assumptions about the between-subject mechanisms that influence academic time use. Models of self-regulated learning attempt to explain how students acquire knowledge and skills that encompass cognitive, motivational, and affective strategies. In particular, the concept of self-control has received considerable attention. Self-control is typically defined as “the ability to suppress prepotent responses in the service of a higher goal” (
Models of self-regulated learning also encompass social support (
The importance of motivation for successful studying in relation to well-being, adjustment to university life, perceived stress (
The idea that affect as another component of self-regulated learning plays an important role in the academic learning context has increased over the last years (
Depending on leisure time characteristics, studies show different associations with well-being. Whereas passive leisure time (e.g., watching TV and computer-related activities done without social interaction) was negatively associated with well-being, active leisure time (e.g., social contact with friends, and physical activity) was positively associated with well-being in a traditional questionnaire study by
Unfortunately, the associations between time use and psychological variables are complex, as time use is compositional and limited to 24 h a day (
In sum, self-control, social support, motivational problems and valence, should explain interindividual differences in academic time use. Unfortunately, few studies that have assessed academic time use have separated within- and between-subject variability and used the experience sampling approach. To improve methodological quality and to generate reasonable estimates of students’ time use, we (i) assessed time use hourly using e-diaries to circumvent backfilling and retrospective distortions and (ii) assessed time use for 1 week at the start of the semester and for an additional week during the examination period to cover fluctuations over the semester.
First, we hypothesized that students’ academic time use would match the 40 h/week requirement from the Bologna Process, as accreditation boards in Europe evaluate study courses on matching this criterion and work on achieving this criterion. Second, we assumed that there would be a systematic shift in time use, with active and passive leisure time dominating the start of the semester and more time with learning-oriented activities during the examination period, as previous studies reported increased demands during the examination period (
To consider within-subject workload differences during the semester, we defined the following two measurement points: 1 week at the beginning of the semester (
Students from particular courses were asked to participate, which were selected for practical reasons (such as courses that primarily were attended by students in their first and third semesters). In groups of approximately 20, the students were informed about the study, asked to complete the first set of individual-level questionnaires, and started the e-diary. After the study week, the students returned the devices and completed another set of personal-level questionnaires, which was different from the first set. The set was divided in two to balance participant burden, because there was no reason to believe that any differences in assessment time would affect the reported data. Later in the semester, the students informed the research team about their examination dates. On that basis, appointments for the second measurement were made individually exactly 1 week before an examination. The e-diary assessment ended the evening before the examination. After the examination, the devices were returned. Again, personal-level questionnaires were completed before and after the second assessment week. The devices, e-diary questions, procedure, and timetable were the same in the examination period, except a few of the trait questionnaires that differed. To ensure compliance, the students received an individual report of their results with a personal coaching session on how to address their stressors. All students provided written informed consent. Ethical approval was not required for this study in accordance with the national and institutional guidelines.
One hundred fifty-four students gave their informed consent. Most of the participants were male (79%) and studying industrial engineering and management (85%). The mean age of all participants was 21.1 years (
To promote consistency throughout the paper, data from the e-diary will be labeled momentary data, whereas data from the questionnaires and the personal-level aggregated momentary data will be called personal-level data. For the statistical analysis, we refer to momentary data as within-subject data and personal-level data as between-subject data.
The e-diary emitted a signal every full hour (e.g., 9:00, 10:00, 11:00 a.m.) during the waking hours of each day during both assessment weeks. We chose such a non-random sampling scheme to improve the accuracy of the time use estimates. We assumed that it would be easier to report time use from full hour to full hour (e.g., from 9:00 to 10:00 a.m.), rather than for two random assessment points (e.g., 9:27 to 10:48 a.m.). We allowed for a 10-min maximum response delay. If the student did not answer within this time frame, the data were recorded as missing. Students put the e-diary in sleep mode before going to bed and started it again in the morning. The e-diary software MyExperience Movisens Edition (Movisens GmbH, Karlsruhe, Germany;
We used the following 10 different categories to classify time use: courses (e.g., lectures, workshops, tutorials), learning-oriented activities (e.g., reading relevant literature, thesis work, presentation preparation, literature research, explaining things to other students), other academic activities (e.g., borrowing books from the library, printing documents, organizing things at the study office), transport and idle time, household, eating and body care, job, active leisure time (e.g., sport, social contacts), passive leisure time (e.g., watching television, playing on the computer), sleeping and other activities. This allowed the students to split up the 60 (plus 10) minutes of their actual time use into the 10 categories. For example, a possible result would be 34 min of learning-oriented activities, 10 min of body care, 12 min of transport, and 7 min of idle time.
To assess momentary mood, we used the momentary Multidimensional Mood Questionnaire (
We assessed self-control via the German version of the Self-Control Schedule (SCS-D;
We used the Social Support Questionnaire (FsozU;
To assess a wide range of parameters that might be relevant to students’ time use, we used the student survey developed by
Missing data are unavoidable in e-diaries, as completing an entry while driving a car, swimming or showering, for example, is not possible. In addition, prompting signals may not be heard in noisy environments. Additionally, e-diary software can have technical problems, or participants can be unwilling to complete the e-diary on time (e.g., during visiting lectures or while at the opera). Fortunately, standard analyses of e-diary data such as multilevel regression models automatically handle missing data. However, this is only the case if the analyses focus on within-subject effects, which is the standard case in e-diary research such as the prediction of momentary mood by stressors or time. Unfortunately, these models are unhelpful in terms of our first hypothesis as we were interested in the cumulative value (the sum instead of the mean) of academic time use over the whole week to compare it to the standard of 40 h/week. Therefore, an additional imputation procedure was necessary to obtain ratings for every waking hour to achieve an estimation of total academic time use over the week.
To take into account both within- and between-subject sources of variance, we used the following linear equation model to estimate the missing data (see
Yij = Yi.+Y.j–Y.., where
Yij = the estimated value of a category (Y) of a person (i) at a given timeslot (j),
Yi. = the person’s mean score for this category over all timeslots,
Y.j = the mean of a timeslot for a given category over all persons,
Y.. = the grand mean for this category over all timeslots and persons.
In some rare cases, the estimation resulted in negative values. They were negligible in number and were set to zero. In addition, if participants forgot to activate the sleep mode at night, we set the sleeping time to a maximum of 10 h and defined the remaining hours as missing data.
If the time between the previous diary entry (e.g., 9:03 a.m.) and the next entry (e.g., 10:00 a.m.) was under 60 min and caused, for example, by the delayed response of the participant in the last diary entry (e.g., 9:03 a.m.) or technical problems, we increased all the ratings proportionally to full hours. For the compliance calculation, we used a conservative approach and added these “increased” ratings to the missing data. The overall compliance (calculated as the number of missing data entries divided by completely filled data entries with a 60-min time frame) was very good at the start of the semester (80%) and at the examination period (89%). We used the imputed data only to test our first and second hypotheses but not for the multilevel model that was developed to test the third hypothesis, as relationships and not sums were the main focus.
To calculate academic time use to test hypothesis 1, we used the following two different approaches: we first calculated academic time use by summing the time used for learning-oriented activities, course attendance, and other academic activities. Academic time use in the ECTS Bologna System is defined as the estimated time that a student typically uses for learning activities, such as attending classes, projects, practical work, or independent study to reach defined learning goals (
To compare the e-diary time use data from the start of the semester to that of the examination period, to test our second hypothesis, we used paired
To investigate which psychological processes were related to student academic time use, our third hypothesis, we chose a more conservative approach to estimate hourly workload. We simply summed the time used with learning-oriented activities and attending courses over the previous 60 min without including the two categories of other academic activities and transport and idle time. We did this because we assumed that the time use scores of learning-oriented activities and attending courses would be more homogenous in content than the combined score of all four categories. Due to the nested data structure, we used a multilevel regression model with a three-level structure. Occasions (level 1) were nested within days (level 2), which were nested within individuals (level 3). At the beginning of the semester, the maximum data points used were 17,107 at level 1, 1,106 at level 2, and 146 at level 3. In the examination period, the maximum data points used were 17,374 at level 1, 1,052 at level 2, and 145 at level 3. One advantage of such a multilevel model is that a different quantity of data points per person can be handled and, thus, used in the analysis. To make a clear and interpretable separation of the within- and between-subject effects, we centered our variables and included (mean individual) valence as an additional personal-level variable at the between-subject level (see
Weekend effects were controlled for using a dummy coded variable, with “weekend” coded as one and “working day” coded as zero. The control variable “hours of the day” was centered on 12 o’clock noon. We calculated six multilevel models as follows: (1) learning-oriented activities and courses at the start of the semester, (2) active leisure time at the start of the semester, (3) passive leisure time at the start of the semester, (4) learning-oriented activities and courses during the examination period, (5) active leisure time during the examination period, and (6) passive leisure time during the examination period. Even though these are different models, their outcomes are not totally independent. For example, if during a given hour, 60 min were spent on learning-oriented activities, then active and passive leisure time had to be zero. However, learning-oriented activities, courses, active and passive leisure time did not always add up to 60 min, as there were other possibilities such as housework, eating, body care, other academic activities, transport and idle time, and time invested in a part-time job. We included as level-three variables the personal-level variables of social support, self-control, and motivational problems. Full maximum likelihood estimation was used for all six multilevel models. All fixed effects from level 1 and level 2 were allowed to vary randomly.
Data management, especially the imputation procedure, was done using SAS©9.3 (SAS Institute, Inc., Cary, NC, United States). Statistical analyses for the first and second analyses were performed using SPSS©21 (SPSS, Inc., Chicago, IL, United States). For the multilevel models used to test the third hypothesis, we used HLM©7 (
Average for each time use category from the e-diary assessment at start of semester and in the examination period (
As predicted in our second hypothesis,
To examine between-subject differences more closely (hypothesis 3), we summed the time use scores for the learning-oriented activities and courses categories to form an academic time use score and plotted its distribution for both assessment points (see
Distribution of the student’s time used for learning-oriented activities and courses (academic time use) in h/week at the start of the semester and in the examination period (
To explain between- and within-subject differences in students’ time use (hypothesis 3), we included basic psychological processes as predictor variables in six different multilevel models (see
Results from the
Start of semester |
Exam period |
|||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Model I |
Model II |
Model III |
Model IV |
Model V |
Model VI |
|||||||
Learning and coursesa |
Active leisure timea |
Passive leisure timea |
Learning and coursesa |
Active leisure timea |
Passive leisure timea | |||||||
Predictors | β(SE) | β(SE) | β(SE) | β(SE) | β(SE) | β(SE) | ||||||
Intercept | 23.66 (0.66)** | 35.93 | 3.04 (0.32)** | 9.38 | 4.54 (0.33)** | 13.68 | 29.62 (0.98)** | 30.10 | 3.37 (0.44)** | 7.71 | 5.07 (0.37)** | 13.63 |
Social support | 0.66 (0.88) | 0.75 | -0.83 (0.67) | -1.23 | 0.53 (0.72) | 0.74 | 3.73 (1.40)** | 2.66 | 1.49 (0.87) | -1.72 | -1.10 (0.69) | -1.59 |
Self-control | -0.01 (0.02) | -0.35 | -0.03 (0.02) | -1.79 | 0.01 (0.02) | 0.68 | 0.04 (0.04) | 1.07 | -0.03 (0.02) | -1.42 | 0.00 (0.02) | 0.05 |
Motivationd | -1.31 (0.28)** | -4.68 | 0.42 (0.16)** | 2.65 | 0.65 (0.19)** | 3.47 | -1.53 (0.48)** | -3.22 | 0.46 (0.22)* | 2.04 | 0.45 (0.19)* | 2.38 |
(weekly) Valence | -2.15 (0.59)** | -3.68 | 0.84 (0.45) | 1.87 | 0.72 (0.38) | 1.89 | -1.89 (0.84)* | -2.26 | 0.93 (0.56) | 1.65 | 0.48 (0.37) | 1.32 |
Valence (t-1) | -3.29 (0.41)** | -8.04 | 0.82 (0.26)** | 3.21 | 3.35 (0.27)** | 12.55 | -4.00 (0.41)** | -9.81 | 0.79 (0.26)** | 3.10 | 2.54 (0.25)** | 10.06 |
Valence (t-1–t0)b | 3.10 (0.29)** | 10.89 | -0.45 (0.18)* | -2.50 | -2.23 (0.21)** | -10.73 | 2.87 (0.29)** | 9.85 | -0.13 (0.17) | -0.77 | -1.54 (0.18)** | -8.75 |
Hours | 0.37 (0.05)** | 7.55 | 0.07 (0.05) | 1.49 | 0.23 (0.06)** | 3.74 | 1.37 (0.11)** | 12.84 | 0.16 (0.05)** | 3.09 | 0.50 (0.06)** | 8.39 |
Hours2 | -0.16 (0.01)** | -19.58 | 0.09 (0.01)** | 10.35 | 0.13 (0.01)** | 13.61 | -0.29 (0.02)** | -13.65 | 0.13 (0.01)** | 12.33 | 0.05 (0.01)** | 4.28 |
Weekendc | -8.13 (0.60)** | -13.65 | 3.44 (0.54)** | 6.38 | 8.85 (0.67)** | 13.17 | -1.93 (0.80)* | -2.41 | 0.41 (0.43) | 0.96 | 1.79 (0.50)** | 3.60 |
We calculated three different multilevel models to predict time spent on learning-oriented activities and courses (model I), active leisure time (model II), and passive leisure time (model III) at the start of the semester (
Momentary valence (t-1) did significantly predict subsequent time use (which is the actual time use from t-1–t0) in all three models (I, II, III). The association with time spent studying was negative (
Changes in valence (t-1–t0) were also significantly associated with time use. More specifically, there was a significant positive effect on time spent on learning-oriented activities and in courses (
Several psychological processes that were included in our multilevel model as between-subject predictors also revealed significance. Motivational problems had a significant negative effect on time spent studying (
In addition, the weekend versus weekday differentiation did have a significant effect on all three outcomes (all
Momentary valence (t-1) and changes in valence (t-1–t0) significantly predicted subsequent time use (time use from t-1–t0) in all three models (see
Again, among all between-subject predictors, motivational problems significantly influenced time spent studying (
Controlling for time variables, the weekend effect was significant for time spent studying (
To sum up, all findings from the beginning of the semester were replicated except the effect of perceived social support. There was a significant effect of social support, meaning, the more social support, the more time they spent studying hourly during the week.
As hypothesized, on average, the academic time use of students did not differ from the Bologna criterion of 40 h/week. Even if we considered 50% of transport and idle time, the numbers were only slightly above the target value of 40 h/week. This result is in contrast to the findings of other studies that showed lower academic time use (
The hypothesized systematic shift in the reported time use from the start of the semester to the examination period was clearly evident in our data. Across the semester, learning-oriented activities increased by nearly 20 h/week, whereas passive leisure time decreased. Surprisingly, active leisure time was higher during the examination period. Additionally, passive leisure time was still high during this period, with a mean of 18.6 h/week. Between-subject differences were huge. Almost one-quarter of the sample had an estimated academic time use for the whole semester of more than 50 h/week (using the mean of both measurement points as proxy for the whole semester). In addition, almost one-quarter of the students studied for fewer than 30 h/week. In the examination period, between-subject differences were even more pronounced.
The substantial within- and between-subject differences in academic workload allowed us to test the psychological processes that might explain these differences. Our multilevel models revealed three main findings. First, both measurement points (the start of the semester and examination period) provided highly comparable effects in association with the predictors, which increased our confidence in the findings.
Second, valence before time use was negatively associated with academic time use and positively associated with leisure time consistently across all six models. In other words, students were in a good mood before leisure time and a bad mood before learning-oriented activities. Given these findings, it seems appropriate to assume that students look forward to leisure time but approach studying uneasily. Another plausible explanation for students’ being in a bad mood before studying might be that they had studied before already. To control for that possibility statistically, we ran additional multilevel models, controlling for time use the hour before, which did not change the association between valence and studying. Moreover, valence diminished during learning-oriented activities, whereas valence improved during leisure time. Generally, these findings are in accordance with those of
Third, at the between-subject level, we investigated the contribution of motivational problems, self-control, and social support. Motivational problems showed a coherent pattern across all models. Specifically, fewer motivational problems were associated with more time spent on learning-oriented activities and courses. Motivation seems to be an important psychological process that drives academic time use.
Social support was only related to academic time use during the examination period. Perhaps social support helps in the exchange of information about the study subject. It might also be that students test their level of acquired knowledge in learning groups. It might also be that social support is needed more in the form of emotional support to relieve the high level of stress in this period, which leads to more studying.
Somewhat surprisingly, self-control as a cognitive factor did not have any significant between subject associations. Given the impressive literature demonstrating the positive relationship between self-control and academic functioning (see de
Even though we used a cutting-edge methodological assessment strategy that included real-time assessment with electronic devices that prevented backfilling and repeated assessment that enabled us to separate within- and between-variance components, we want to address some of the limitations of the current study. First, to ensure high compliance and low reactivity in e-diary studies, a fair balance between the number of assessment points and number of items is necessary to reduce participant burden. Given the hourly measurement points (up to 200 per individual) and the additional blood pressure and cortisol measurements (which were not reported), we had to manage participant burden by restricting the number of items and measurement weeks. Even though it would be tempting to have multiple questionnaires for each construct, we chose the student survey developed by
Second, we were not able to provide empirical evidence regarding the association between workload and grades, as we could not assess the latter due to data protection and privacy issues. However, investigations of the association between academic time use and grades have generated mixed results (
In our sample of students in engineering sciences, the students’ average academic time use seemed to conform to the specifications and guidelines administered by the Bologna reform. Our design enabled us to reveal large between- and within-subject differences in students’ academic time use and explain these differences with psychological processes. Valence appeared to be a strong predictor of time use, highlighting the important role of affective factors in self-regulation and motivational processes of academic time use. Future research is needed to investigate the role of the consequences of negative valence on learning processes in daily life. Studies should detangle affective states that accompany the motivational, self-regulatory components of study time and the learning process. They also should investigate what emotion regulatory strategies in which context in daily life help students address mistakes in a successful manner so that positive valence can be reestablished and learning goals can be met. This could be helpful for professionals in designing appropriate interventions regarding emotion regulation in the learning process. Future studies should also enhance the understanding of within and between person variables and processes that influence academic time use further, especially regarding individual differences in within-subject relations, which may be a next step in helping to discover problematic trajectories and facilitate specific interventions.
All students provided written informed consent. Ethical approval was not required for this study in accordance with the national and institutional guidelines.
SK-H, UE-P, and PS contributed to the design of the study. SK-H and PS performed the data collection, which was overseen by UE-P. SK-H performed and UE-P supervised data management. SK-H carried out the statistical analysis. SK-H and UE-P wrote the first draft of the manuscript. UE-P, PS, and AG wrote sections of the manuscript. All authors contributed to the manuscript revision, read and approved the submitted version.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
We thank S. Haas, M. Lindner, J. Kazmaier, C. Reinke, M. Scheurer, and D. Schell for their support in conducting the study.