Development of an Optimal Time for Learning Questionnaire in Chinese University Students


 Background
 Autonomous learning is crucial for students’ academic achievements, however, there lacks a structure-validated questionnaire to measure the optimal time for autonomous learning with different learning contents.
Methods
 Based on the previous investigations, we have designed a matrix of 49 items based on seven learning contents and seven time periods of a day to measure the optimal time for autonomous learning and invited 305 Chinese university students to answer the matrix.
Results
 Through both exploratory and confirmatory factor analyses, we have developed an Optimal Time for Learning Questionnaire with a satisfactory model structure of five factors (7, 6, 6, 6, 5 items per factor respectively) namely Noon, Late Night, Nightfall, Morning and Afternoon. The internal reliabilities of these factors were acceptable, and their inter-correlations were significant, albeit in low or medium levels.
Conclusions
 The Optimal time for Learning Questionnaire may help students find their optimal learning efficiency individually.


Introduction
Autonomous learning, also named self-managed learning (Lizzio & Wilson, 2005), self-regulated learning (Boekaerts & Niemivirta, 2000;Entwistle & McCune, 2004), or self-directed learning (Garrison, 1997), denotes a process where students possess su cient opportunity to organize and regulate their own learning. In autonomous learning, students' learning e ciency may vary with the learning contents (or concepts) and time epochs during a day. Previous studies have suggested that there exists a time-of-day effect in one's learning e ciency. For example, young participants showed superior memory accuracy in the afternoon compared to that in the morning (Hourihan & Benjamin, 2014). When identi ed as morning type or evening type, students reported differences in memory, interest, motivation, and achievement across time of day (Anderson et al., 1991;Itzek-Greulich et al., 2016). In the eld of logical reasoning, students' speed and performance improved markedly from morning to noon, and subsequently performance fell off rapidly from afternoon to night (Folkard, 1975).
Several other factors in uence cognitive performance and thus contribute to intra-individual variance in learning e ciency. Most importantly, the circadian rhythms in cognitive processes affect school related activities such as attention and memory (Valdez, 2019). Circadian rhythms have been found in three basic cognitive processes which are interconnected with each other: attention, working memory and executive functions. Attention denotes to the act of selectively concentrating on a discrete aspect of information while ignoring other perceivable information, which also interacts effectively with the environment (Nobre & Kastner, 2014); working memory comprises the storage, retrieval, and use of information (Baddeley & Hitch, 1974); and executive functions refer to the ability to program or regulate behavior, which is essential for problem solving, decision making as well as self-control (Lezak et al., 2012). These cognitive processes roughly reach their lowest levels during nighttime and early morning, improve around noon, and are at even higher levels during afternoon and in the evening (Valdez et al., 2005;Ramírez et al., 2006;García et al., 2012;Gallegos et al., 2019), demonstrating their coordination with the physiological rhythms of the human body (Valdez, 2019).
The learning e ciency and time course of cognitive performance are nonetheless also modulated by other conditions. The "chronotype" is a reproducible and stable trait of individuals, which encompasses morning type, evening type and intermediate type according to preferred time of learning or other activities (Kantermann & Eastman, 2018). The morning and evening types show an advance or a delay in cognitive rhythms respectively (Roenneberg et al., 2003). Hunger and sleep deprivation in uence the rhythms of cognitive performance as well (Benau et al., 2014;Killgore & Weber, 2014), leading to the impairment of attention, memory and executive functions (Benton & Parker, 1998;Doniger et al., 2006;Krause et al., 2017). The uctuations of attention interrelate with emotion (Taylor & Fragopanagos, 2005), and certain emotional states appear to facilitate memory control (Engen & Anderson, 2018) and decision making (Woodcock et al, 2020). Finally, exposure to environmental noise deteriorates cognitive performance with regard to comprehension, memory, attention, as well as executive functions (Clark & Paunovic, 2018).
A classical theory of learning skills is the "Building Blocks of Learning" (Goldstein & Mather, 1998). Learning skills are regarded as comprising ten building blocks that re ect foundational skills, symbolic or perceptual processing skills, and conceptual or thinking skills, including attention and self-regulation, emotion, behavior, self-esteem, phonological processing, orthographical processing, motor processing, thinking with language, thinking with images, and thinking with strategies (Abu-Hamour, 2014; Goldstein & Mather, 1998). In the current study, we would separate the learning contents into seven elements, i.e., language learning, problem analyzing, memorizing, comprehension, logic thinking, drawing and writing.
There are several inventories that have been developed to assess learning e ciency and learning style (Horne & Östberg, 1976;Fleming & Mills, 1992;Duff, 2004;Fleming, 2001;Roenneberg et al., 2003). To date however, there is no single study which illustrates a structure-validated questionnaire that investigates the optimal time for autonomous learning of In the current study, we adopted both exploratory and con rmatory factor analyses in order to validate the structure of a questionnaire. Furthermore, we roughly divided the day into the early morning, morning, noon, afternoon, nightfall, evening, and the late night. On the other hand, based on previous studies, the optimal time of learning refers to the time epoch when an individual is highly focused, does not easily fatigue, stable in mood, uent in thinking, and with the highest learning e ciency. Thus, we recruited a group of university students to answer a matrix of statements regarding the optimal time for autonomous learning. Considering that different cognitive performances show similar rhythms in a day (Valdez, 2019), we hypothesized that the e ciency of autonomous learning is primarily affected by the time of day instead of the learning contents.

Participants
Altogether 305 students (116 men, mean age: 21.39 years ± 1.94 S.D., range: 19-28 years; and 189 women, mean age: 21.31 ± 1.85, range:18-26) were recruited in the current study. The participants were enrolled in the Zhejiang University (Hangzhou, China), a typical comprehensive Chinese university with a large student population. They were all of Han ethnicity and majored in Modern Medicine, Science, Engineering, Management, Agronomy, History, Sociology, Literature, and Philosophy. There was no signi cant age difference between the male and female participants (t = .35, p = .728, 95% Con dence Interval = -.36 ~ .52). All participants were free from any somatic or psychiatric illnesses, had not experienced stressful life-events recently, and were medication or alcohol free at least 72h prior to testing. The study protocol was approved by the Ethic Committee of the School of Public Health, Zhejiang University and all participants had given their written informed consents.

Measures
Participants were asked to respond to a matrix of 49 items regarding the optimal time for autonomous learning in a quiet room. They were asked to use a ve-point Likert rating scale: 1-very unlike me, 2-moderately unlike me, 3-somewhat like and unlike me, 4-moderately like me, and 5-very like me. The matrix concerned following aspects (seven items each based on time periods below): 1) the optimal time for language learning, 2) the optimal time for analyzing a problem, 3) the optimal time for memorizing, 4) the optimal time for comprehension, 5) the optimal time for logical thinking, 6) the optimal time for drawing, and 7) the optimal time for writing. The participants were asked to choose between the following time periods: 1) early morning (between getting up and breakfast), 2) morning, 3) noon (between lunch and midday rest), 4) afternoon, 5) nightfall (around dinner), 6) evening, and 7) late night. The presentation of the items was randomized in the matrix.

Statistical Analyses
Responses to the 49 items were subjected to principal component analysis using the Predictive Analytics Software Statistics, Release Version 18.0 (SPSS Inc., 2009, Chicago, IL). The factor loadings were rotated orthogonally through the varimax normalized methods. Items which loaded less heavily (below .50) on a target factor, or cross-loaded heavily (above .25) on more than one factors were removed from subsequent analyses one-by-one. The procedure continued until no further item was needed to be removed. Then, the t of the remaining data (i.e., components extracted as latent indices were used to identify the model t: the χ 2 /df, the goodness of t index, the adjusted goodness of t index, the comparative t index, the Tucker-Lewis Index, and the root mean square error of approximation (Bentler & Bonett, 1980;Bentler, 1990;Browne & Cudeck, 1993).
When factors and their related items had been identi ed, the factor scores of each gender group and the internal reliabilities (the Coe cient H) were then calculated in all participants. The gender difference of each factor scores was evaluated by two-way ANOVA (i.e., gender × factor score) plus the post-hoc Student t test. A p value below .05 was considered signi cant. Moreover, the Pearson correlation test was used to search for possible relationships within the factors, with a p value below .01 was considered signi cant.

Results
Responses to the 49 items measuring the optimal time for autonomous learning were rst submitted to a principal component analysis. Results of the pre-analysis check were acceptable (KMO = .84; the Bartlett test of sphericity = 7099.99; p < .001). The analysis disclosed 11 factors with eigenvalues greater than 1.0, and the scree plot indicated a level-off from the sixth factor on. The eigenvalues of the rst ve factors were 7.50, 7.27, 3.25, 2.85 and 2.42 respectively, which altogether accounted for 47.50% of the total variance (the rst four factors altogether accounted for 42.57%). After the varimax normalized rotation, in the six-factor model, there were no validly consistent items in the sixth factor. Therefore, a ve-factor model was chosen for the con rmatory factor analysis.
Three AMOS t-models of ve-factors with different items were constructed (Table 1) and the 30-item model (7, 6, 6, 6, 5 items respectively for the ve factors) was the best among the models. The standardized factor correlations for the 30item model structure were also acceptable ( Figure 1). Based on these 30 items, we developed an Optimal Time for Learning Questionnaire (OTL, Table 2), and used it in the subsequent analyses.  Factor 1 was called "Noon", which described the time between lunch and midday rest and was applicable to all the seven learning contents. Factor 2 was called "Late Night", which represented the dead of night and was applicable to all learning contents except for language learning and drawing. Factor 3 was called "Nightfall", which re ected the time around dinner. This factor was applicable to the other six learning contents excluding language learning. Factor 4 was called "Morning", which comprised the time between getting up and lunch. The learning contents in Factor 4 consisted of language learning, logical thinking, memorizing, and drawing. Factor 5 was called "Afternoon", which referred to the time between lunch and dinner and includes ve learning contents, other than drawing and writing.
There was no signi cant difference in the OTL factor scores between two gender groups (F [1, 303] = .47, p = .495, mean squared effect = 17.44). The internal reliabilities of OTL ve factors were all above .80, and their inter-correlations were signi cant but remained on a low or medium level (Table 3). Note: * p < .01 (for intercorrelations).

Discussion
Using both exploratory and con rmatory factor analysis on 49 items regarding optimal time for autonomous learning, we have constructed a satisfactory model structure of 30 items with ve scales, namely Noon, Late Night, Nightfall, Morning and Afternoon. These scales had acceptable internal reliabilities and low or medium inter-correlations, which con rmed our hypothesis that the e ciency of autonomous learning was mainly affected by time epoch of day. Although we have included various learning contents, the distribution of OTL items was primarily based on time of day, indicating that the time effect was most prominent in one's learning e ciency. Our results are consistent with the idea of circadian rhythms of cognitive performance, in that the different aspects such as memory, comprehension, logical thinking and language learning followed similar pattens of cognitive rhythm, i.e., with higher level during daytime hours and lower during nighttime and early in the morning. This maintains a phasic relationship to body temperature . Factor 4 (Morning) and Factor 5 (Afternoon) are correlated with diet as well, but the effect of sleep or rest also appears in these factors, as these periods take place immediately after a period of sleep. Previous studies have suggested that sleep plays a crucial role in memory stabilization and integration, and functions as a brain state optimizing memory consolidation ( 2018) reported that sleeping more than usual the night before testing was associated with better performance, suggesting even a single night's sleep can bene t in learning. Actually, the habit of siesta is prevalent in the Chinese population (Kang et al., 2017). The Factor 5 (Afternoon) of our study might also be explained by the effect of sleep or rest owing to the noon break in study schedule when students could take a nap or get some rest.
Factor 2 (Late Night) refers to the silence of night with a peaceful atmosphere, which is often connected with the calmful of the surrounding physical-environment. The optimal learning e ciency of different task types have been identi ed under environmental interactions. One study has shown that although the comfortable ambient-temperatures were varied, participants involved in perception, memory, problem-solving and attention tasks performed best under relatively/ fairly quiet and bright or moderately light surroundings (Xiong et al., 2018), suggesting that quiet environment helps to enhance the cognitive performance. Moreover, healthy environment might be related with emotional stability, which bene ts the ability of memory control (Engen & Anderson, 2018) and decision making (Woodcock et al, 2020). On the contrary, exposure to noise has a terrible impact on the ability of perception, comprehension, memory as well as writing (Klatte et al., 2013).
Our study also suffered from several design limitations. Firstly, although we recruited students from different academic majors, the participants were mainly undergraduates of clinical medicine, so whether the results can be generalized to other majors or grades needs further research. Secondly, we did not record the chronotypes of the participants.
Chronotypes have a nonnegligible impact on the circadian rhythms in cognitive performance (Goldstein et al., 2007). Thirdly, recall bias may have in uenced our measurements of optimal learning time for different contents. Fourthly, our study was exclusively conducted in China, where noon-time nap is common, which may negatively affect the generalizability of the present ndings. Fifthly, we did not record the nutritional habits of the participants, which may have in uenced their responses to the questionnaire, since comfortable diets are fuels for optimal level of cognition. Nevertheless, using both exploratory and con rmatory factor analyses in Chinese university students, we have developed a structure-validated scale of optimal learning time, which might help students of different specialties to optimize their everyday training e ciency.