Microsoft Word - Kärner et al_publication.docx


!
!
!
!
!

Frontline!Learning!Research!Vol.5!No.!1!(2017)!16!<!42!
ISSN!2295<3159!!

1 Contact information: Tobias Kärner, Chair of Business Education and Educational Management, University of 
Bamberg, Kärntenstraße 7, 96052 Bamberg, Germany. Email: tobias.kaerner@uni-bamberg.de DOI: 
http://dx.doi.org/10.14786/flr.v5i1.137 
!

Analysis of person-situation interactions in educational settings 
via cross-classified multilevel longitudinal modelling: illustrated 

with the example of students’ stress experience 

Tobias Kärnera1, Detlef Sembilla, Christian Aßmannb, Edgar Friederichsa, and  
Claus H. Carstensenc 

a Chair of Economics and Business Education, University of Bamberg, Germany 
b Chair of Statistics and Econometrics, University of Bamberg, Germany 

c Chair of Psychology and Methods of Educational Research, University of Bamberg, Germany 

 

Article received 9 December / revised 13 October / accepted 16 November / available online 10 February 

 

Abstract 

The investigation of learning processes by assessing students’ experience along with objective 
characteristics within a classroom context has a long tradition in empirical learning process research (e.g. 
Sembill, 1984 et passim; Wild & Krapp, 1996). However, most of the existing studies confine themselves to 
psychological variables that seem to be too narrowly considered, as there is theoretical and empirical 
evidence proving the involvement of somatic and psychological processes in learning and in stress 
reactions. Furthermore there is a lack of studies that investigate situation-related experience (states) as an 
outcome of interactions between relatively stable characteristics (traits) and continuously changing 
“objective” context conditions. Against this background, we will present an approach for cross-classified 
multilevel longitudinal modelling of person-situation interactions in naturalistic educational settings. We 
illustrated our model with the example of students’ stress experience referring to empirical data that we 
measured within a multidisciplinary research project (pedagogy, psychology, adolescent medicine). 53 
students at a public German vocational school were investigated during 9 lessons. There are up to 38 state 
measurements per person, resulting in 2,014 measurements in total. Taking into account that states are 
nested within persons and within situations we applied a cross-classified multilevel model to analyse effects 
on students’ stress experience. The analysis shows significant person-situation interactions between 
academic self-concept and classroom demands and between baseline cortisol concentration and classroom 
demands: the relation between classroom demands and stress experience depends on relatively stable 
person-related characteristics. A deeper knowledge about the complex interrelations between traits, states, 
and continuously changing context conditions seems to be essential for a more holistic understanding of 
learning at school and for the identification of crucial aspects for an evidence-based design and 
implementation of teaching and learning arrangements. 

Keywords: Learning process research; Experience sampling; Video analysis; Classroom demands; Stress; 
Cortisol; Heart rate variability; Academic self-concept 

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!



Kärner&et &al &
&

 
 
 

 | F L R !
!

17!

1.  Introduction 

School, as an essential part of adolescents’ learning environments, involves a number of 
potential stressors that increase individual chances of stress reactions. Difficulty of learning material, 
time pressure, and pressure to succeed have been noted as potential school-related stressors (e.g. 
Achtenhagen, 1978; Kouzma & Kennedy, 2004). Stress is not a one-, but a multi-dimensional 
phenomenon, as psychological and somatic processes constituting a complex system of relations are 
involved (Adam, 2006; Kemeny, 2003). Various existing studies pursue top-down approaches in order 
to translate psychological constructs into basic physiological mechanisms (e.g. Michaud et al., 2008; 
Roubinov et al., 2012; summarizing Ellenbogen, 2012); fewer studies consider a bottom-up approach 
assessing the effects of physiological variables on psychological variables (e.g. Buchanan & Lovallo, 
2001). The present study assumes that students’ stress experience is affected by physiological and 
psychological characteristics, situation-specific states, and “objective” characteristics of the classroom 
context. Furthermore, we have to consider the interactions between person and classroom context, 
which play an important role in students’ stress responses. In that regard, stress is seen as a specific 
relationship between person and environment, and more specifically, as an interrelationship between 
situational demands and personality characteristics (Lazarus, 1966 et passim). Here, internal or 
external demands that tax or exceed one’s adaptive resources may lead to stress reactions that involve 
psychological and somatic systems (Lazarus, 1999). !

The referred paradigm is known as “interactionism” and focuses on the relationship between 
individuals, with their relatively stable characteristics, and the objective context-specific 
characteristics. This relationship is supposed to manifest itself in situation-related experience and in 
specific behaviours (Carver & Scheier, 2008; Nezlek, 2007). In an educational context, the 
interactional perspective is discussed within the scope of aptitude (trait) treatment interaction (ATI) 
theory. ATI theory assumes that a specific outcome (e.g. behaviour, experience) depends on the 
interaction between students’ traits and the characteristics of the classroom context they are exposed to 
(Cronbach & Snow, 1977; Snow, 1989; Yeh, 2012). Existing research on trait-treatment interactions is 
mostly based on cross-sectional methods within laboratory studies that do not account for longitudinal 
processes in naturalistic learning environments. Therefore, in the current paper we use a multilevel 
longitudinal approach to examine continuously changing conditions in the classroom (especially 
demanding conditions during the lessons) along with students’ experience. 

 

2.  Background 

Before stating our research aims and hypotheses, we will provide some information about the 
background of the referred approaches and constructs that are related to students’ stress response. 
Here, we first introduce the hypothalamic-pituitary-adrenal axis and the sympathetic and 
parasympathetic nervous system, afterwards presenting some information about the connection 
between traits (specifically academic self-concept as an important coping resource) and situational 
experience, finally looking at potential classroom demands. 

2.1  Hypothalamic-pituitary-adrenal axis activity 

Stress induction leads to an activation of the hypothalamic-pituitary-adrenal (HPA) axis (e.g. 
Michaud et al., 2008). This activation is initiated by a secretion of corticotropin-releasing hormone, 
which stimulates the anterior pituitary gland. In turn, this involves a secretion of adrenocorticotropic 
hormone that enters the blood circulation, triggering the adrenal cortex to release the glucocorticoid 
cortisol (Dickerson & Kemeny, 2004; Kemeny, 2003). A typical pattern of cortisol concentration 



Kärner&et &al &
&

 
 
 

 | F L R !
!

18!

exists during the day that corresponds to a circadian rhythm and ideal-typically accords to an inverted 
U-shaped curve. Various studies show that individual baseline cortisol concentration is an indicator 
for chronic stress (e.g. Caplan et al., 1979; Pruessner et al., 2003; Takahashi et al., 2005). Cortisol 
concentration is normally high in the morning and low in the evening; therefore, potential energy is 
expended during (things like) coping processes throughout the day (Hall, 2011). Wüst et al. (2000) 
found that the awakening cortisol response is significantly associated with perceived chronic stress. 
Takahashi et al. (2005) identified a significant positive correlation between individual trait anxiety and 
baseline cortisol concentration (r = .58) and a positive but non-significant correlation between state 
anxiety and baseline cortisol concentration (r = .25). The authors conclude that individuals “with 
chronic high levels of cortisol may have blunted acute neuroendocrine response to experimental social 
stress, possibly due to saturated HPA reactivity” (Takahashi et al., 2005, p. 353). In endocrinological 
studies, a frequently used method for assessing overall secretion of cortisol over a specific time period 
is the “area under the curve with respect to the ground” (AUCg) (Pruessner et al., 2003; for details see 
section “4.2.1 Baseline cortisol concentration”). There is theoretical and empirical evidence that 
AUCg is related to chronic stress, but current research shows that high chronic stress levels are not 
always associated with elevated baseline cortisol. Differences in found empirical relations between 
AUCg and chronic stress may result from methodological, theoretical, and analytical issues (for a 
further discussion see Saxbe, 2008). 

2.2  Sympathetic and parasympathetic activity 

Another somatic system involved in human stress reactions is the autonomic nervous system, 
which ensures—compared to the reaction of the HPA axis—a much more rapid adaptation to stressful 
situations. It contains two essential components (Baert et al., 2012): on the one hand, the 
parasympathetic nervous system, controlling physiological resting functions; and on the other hand, 
the sympathetic nervous system, which is primarily activated under stress (Kemeny, 2003). As one 
indicator of sympathetic and parasympathetic activity, the activity of the cardiovascular system, 
particularly heart rate variability, is taken into account as an important marker. Within the frequency 
spectrum, the parasympathetic component of the autonomic nervous system is reflected by high 
frequency domain (HF); the activity of the sympathetic nervous system by low frequency domain (LF) 
(Agelink et al., 2001; Malik et al., 1996). The LF/HF ratio, as a frequency-based indicator of heart rate 
variability, reflects the sympathovagal balance (Sleight & Bernardi, 1998). Various studies report a 
significant increase of the LF/HF ratio in experimental stress situations that is caused by a decrease of 
the HF component (compared to the LF component), that in turn indicates a predominant state of the 
sympathetic nervous system under stress (Baert et al., 2012; Hjortskov et al., 2004; Isowa et al., 2006). 
HPA axis and autonomic nervous system are related (Kemeny, 2003); Stadler et al. (2011) as well as 
Looser et al. (2010) found significant correlations between heart rate variability and cortisol awaking 
response. Looser et al. (2010) suppose “a relative independence in the regulation of the HPA axis and 
the automatic nervous system in response to everyday stressors but synchrony of both systems in 
highly stressful situations” (p. 281). 

2.3  Coping resources and situational experience 

Lazarus (1966) describes traits as relatively stable personal characteristics. The term 
“relatively stable” means that traits can indeed change over longer periods of time (e.g. three years) 
but remain quite stable over a defined time period (e.g. three months) (Nezlek, 2007). On the other 
hand, psychological states—in terms of situational experience—are situation-specific and transient. In 
reference to the transactional model of stress and coping, it has to be taken into account that stress 
results from subjective appraisal of a stressful or demanding situation and the situational coping within 
the current situation. Here, subjective appraisal implies emotional processes involving cognitive-
affective states and physiological processes (Lazarus, 1984; Lazarus & Folkman, 1984).  



Kärner&et &al &
&

 
 
 

 | F L R !
!

19!

An important trait variable is the academic self-concept, which describes a person’s perception 
of his performance-related self-identity. As a construct that refers to mastery experience as a major 
information source, it is similar to the construct of self-efficacy beliefs (Bong & Skaalvik, 2003). 
Academic self-concept is not merely a cognitive construct. It also integrates motivational and 
emotional aspects (Shavelson et al., 1976) and mediates appraisal and coping processes, thus causing 
it to be seen as an important coping resource (Goetz et al., 2008). Within this context, coping is 
defined by Lazarus and Folkman (1984) as “constantly changing cognitive and behavioural efforts to 
manage specific external and/or internal demands that are appraised as taxing or exceeding the 
resources of a person” (p. 141). Characteristics like academic self-concept may affect emotional, 
cognitive, and motivational states (Sembill, 2004; Sembill et al., 2002). There are direct as well as 
indirect relationships between traits and states; coping processes mediate this relationship (Spangler et 
al., 2002). 

Regarding the different ontological levels involved in human stress reactions (cf. Sembill, 
Rausch, & Kögler, 2013), relationships also exist between the psychological level and adrenocortical 
responses (Adam, 2006). Spangler et al. (2002) for example report an interaction between trait anxiety 
and the development of cortisol concentration during an exam situation. A significant increase in 
cortisol was revealed amongst high anxiety students. On the other hand, a significant decrease was 
reported in low anxiety students. These results are congruent to the study of Takahashi et al. (2005). 
The authors summarise that “high trait anxiety [... is] associated with chronic elevation in the level of 
HPA activation” (p. 353). In addition there are significant negative correlations between emotion-
oriented coping and cortisol concentration (r = –.38 and r = –.44, see Spangler et al., 2002). 

2.4  Classroom demands 

Current research lacks examples of well-founded empirical evidence for the stress-triggering 
potential of various aspects of the classroom context. Here, it is necessary to consider the main effects 
of classroom context characteristics on students’ stress experience and to take into account trait-
treatment effects.  

Main effects of classroom characteristics on students’ stress experience may result from 
aspects of the didactical design of the lessons. In regard to learning processes, empirical studies show 
that student-centred learning strongly supports higher levels of motivation and increased competence 
in complex problem-solving within students. Moreover, student-centred learning results in increased 
self-determination (Deci & Ryan, 1985; Schunk & Zimmerman, 1994; Seifried, 2004; Sembill et al., 
2002). However, in terms of “learning with risk” according to Sembill et al. (2002), in student-centred 
periods learners may also be confronted with higher degrees of complexity regarding domain-specific 
learning material, uncertainty, ambiguity, time pressure or public-speaking situations that may be 
stressful for them.  

Furthermore, main effects of classroom characteristics on students’ stress experience may 
result from the complexity of learning contents and from the associated cognitive challenge of 
handling a given learning material. This in turn may result in demands on the information system, 
specifically on the working memory capacity. In that regard, Sweller and colleagues differentiate 
between intrinsic cognitive load and extraneous cognitive load. Intrinsic cognitive load is associated 
with the structure of the information that is necessary for students to reach their learning goals. Task 
difficulty, informational complexity, and the perceived understanding of the learning content play an 
important role in considering to what extent the student perceives the learning material to be 
demanding and stressful (Sweller, 1988; 1994). The complexity of learning contents within the 
classroom context can be described via Bloom’s taxonomy of learning objectives providing a 
hierarchical classification of the cognitive challenge of handling a given learning material (such as 
“applying”, “analysing”, “synthesizing”, and “evaluating”; Bloom et al., 1956). The extraneous 
cognitive load also requires working memory resources and is associated, not with the intrinsic 



Kärner&et &al &
&

 
 
 

 | F L R !
!

20!

structure of the processed information, but with the learning activities in which students have to 
engage (Sweller, Ayres, & Kalyuga, 2011). Here, student-centred learning in particular is associated 
with a high degree of cognitive challenge during education that is characterised by complex tasks and 
active student engagement (Rieser et al., 2013). We assume that when taken together, the amount of 
student-centred learning and the level of cognitive challenge when handling a given learning material 
constitute demanding conditions for students. 

Besides main effects of classroom characteristics on students’ stress experience, one has to 
also take trait-treatment effects into account. Whether a task or an activity is perceived as demanding 
or stressful depends not only on the objective conditions in class, but also to a great extent on the 
individual’s disposition. In that regard, Krohne (1990, p. 27) states that the “process of cognitively 
transforming objective characteristics into subjective representations is clearly codetermined by 
personality dispositions” such as coping dispositions or control expectancies. The referred theoretical 
concept is known as the “interactional paradigm” and can be found in personnel psychology in terms 
of person-organisation-fit models (Kristof-Brown, Zimmerman, & Johnson, 2005). In clinical and 
developmental psychology, the aforementioned paradigm is discussed in terms of the diathesis-stress 
model. It describes (psychopathological) development as a result of an interaction between 
predispositional vulnerability and stressful context conditions: the higher the vulnerability, the greater 
the responsiveness to stressful conditions (Belsky & Pluess, 2009). In educational psychology, the 
corresponding approach is known as aptitude (trait) treatment interaction theory (Cronbach & Snow, 
1977; Snow, 1989) and has been extensively investigated in the context of anxiety research 
(Spielberger, 1989). The corresponding state-trait (interaction) model of anxiety postulates combined 
effects of individual differences in trait anxiety and objective situational conditions in terms of the 
type of threat on state anxiety response (Shedletsky & Endler, 1974): persons with high trait anxiety 
values show greater state anxiety reactions under stress than persons with low trait anxiety values 
(Kendall, 1981). 

 

3.  Research aims and hypotheses 

Existing research on students’ stress responses mostly refers to cross-sectional (laboratory) 
studies but show a lack of longitudinal studies specifically considering naturalistic learning processes 
at school. Ellenbogen (2012) points out that one of the most important challenges in this context is 
closing the gap between laboratory research and naturalistic studies. Further, few studies exist that 
analyse different levels of stress- and coping-related systems and processes simultaneously (e.g. 
Adam, 2006; Beck & Sczesny, 1993; Goetz et al., 2008). In that regard, there is a lack of studies 
assessing psychological and physiological states in naturalistic learning environments and taking 
special account of continuously changing context conditions assessed during class via objective 
observational methods. Based on the described considerations, we consider the following research 
question and hypotheses: 

To what extent is students’ stress experience associated with baseline cortisol concentration, 
academic self-concept, heart rate variability (LF/HF ratio), situational coping, and classroom 
demands?  

(1) It will be hypothesised that the baseline cortisol concentration and the LF/HF ratio are 
positively associated with students’ stress experience (cf. Baert et al., 2012; Hjortskov et al., 
2004; Isowa et al., 2006; Takahashi et al., 2005) and that academic self-concept and 
situational coping are negatively related to students’ stress experience (cf. Goetz et al., 2008; 
Spangler et al., 2002).  



Kärner&et &al &
&

 
 
 

 | F L R !
!

21!

(2) With respect to the classroom context we hypothesise that classroom demands are positively 
associated with students’ stress experience (cf. Sembill et al., 2002; Sweller, 1988; 1994).  

(3) Furthermore, we hypothesise that there are person<situation interactions between baseline 
cortisol concentration and classroom demands, as well as between academic self-concept and 
classroom demands, as follows:  
a) Persons with high baseline cortisol values should show a greater stress-response under 

high degrees of classroom demands than persons with low baseline cortisol values, as high 
cortisol baseline values are associated with physiological vulnerability—as a possible 
result from accumulated chronic stress—that should affect the relationship between 
stressful context conditions and stress experience (cf. Belsky & Pluess, 2009; Takahashi et 
al., 2005; Wüst et al., 2000);  

b) Persons with high academic self-concept values should show a lower stress response 
under high degrees of classroom demands than persons with low academic self-concept 
values, as a person’s self-related perception of his/her coping abilities is seen as a 
substantial resource against external demands (cf. Lazarus & Folkman, 1984; Goetz et al., 
2008). 

 

4.  Method 

4.1  Participants and data collection 

Within a short-term longitudinal study, students from a public German vocational training 
school were investigated during 9 school lessons on the subject “economic business processes”. The 
sample was composed of 53 students (18 male, 35 female) within two classes and with a mean age of 
19.53 years (SD = 4.76). The study was approved by the Bavarian Ministry of Education and Cultural 
Affairs and was conducted in accordance with the Declaration of Helsinki. Participants were of full 
age, and if they were not, the parents of the underage participant signed declarations of consent prior 
to participating. 

Our study was designed to investigate effects of individual preconditions and characteristics 
within the classroom context on students’ states during learning processes (see Figure 1). Before 
investigating the lessons in school we measured academic self-concept and baseline cortisol 
concentration. Two months later, during the following learning processes we measured the students’ 
experience during the lessons along with their cardiovascular activity. We also recorded the lessons 
for subsequent video-based analysis of classroom characteristics (especially classroom demands). 

 

 

Figure 1. Data collection 

time

* Academic self-concept

* Baseline cortisol concentration
(8:00a.m./12:00noon/16:00p.m./20:00p.m.)

* Continuous sampling of students’ experience 
* Cardiovascular activity (LF/HF ratio)
* Video-based observation of classroom demands

Data collection before the  
learning processes

Data collection during the 
learning processes

two months 
interim 
period



Kärner&et &al &
&

 
 
 

 | F L R !
!

22!

4.2  Measures and operationalisation 

4.2.1 Baseline cortisol concentration 

Baseline cortisol concentration was measured using biochemical salivary analysis. The 
students received small plastic tubes and were instructed to provide saliva samples during a relaxed 
weekend day without school or work at the following times: 8:00 a.m., 12:00 noon, 16:00 p.m., and 
20:00 p.m. The saliva samples were immediately transported to a laboratory that performed the 
analysis (Immumed Ltd., Institute for applied Immunology, Munich, Germany), carried out via a 
laboratory machine by Roche Diagnostics®. An electrochemiluminescence immunoassay was applied 
to measure the concentration of salivary cortisol. With a correlation of r = .94, the general test-retest 
reliability of cortisol measures defined by repeated measurements carried out via Elecsys 2010® was 
found to be acceptable (Cobas, 2010). Testing for reliability we calculated Cronbach’s alpha 
coefficient for reliability over the four measures of cortisol and found a satisfactory interrelationship 
(α = .86).  

As we wanted to assess the chronic elevation of HPA activation by considering multiple data 
points, we used the individual areas under the curves with respect to the ground (AUCg) for assessing 
baseline cortisol secretion, a common method used to derive information from multiple measures over 
time (Pruessner et al., 2003). AUCg indicates the total area under the curve of all cortisol measures 
over the day. Fekedulegn et al. (2007) point out that AUCg takes into account the sensitivity and the 
intensity of multiple measurements of cortisol. Furthermore, it acts as an indicator for the “total 
hormone output” (Pruessner et al., 2003, p. 928). For calculating the individual areas under the curves, 
we used the trapezoid formula for the area under the curve with respect to the ground following 
Pruessner et al. (2003), with identical periods of time in between taking the measurements (Eq. 1): 

 AUCg = ! (! !!! !!!)
!

!!!
!!!    (1),  

with mi denoting the individual measurement and n denoting the total amount of measures. 
Overall we found a mean AUCg of M = 10.21 (SD = 7.96, Min. = .75, Max. = 40.42) in the current 
sample that shows a marked variability. 

4.2.2 Heart rate variability (LF/HF ratio) 

Heart rate variability was measured via equipment from Medeia Ltd.® (Danev, 2010) and was 
operationalised by the LF/HF ratio (with LF from .04 Hz to .15 Hz and HF from .15 Hz to .4 Hz; cf. 
Agelink et al., 2001; Malik et al., 1996). The cardiovascular activity of the students was continuously 
measured during the investigated lessons. For further analysis, the LF/HF ratio measures were 
aggregated per person at 10-minute intervals via arithmetic means to synchronise the LF/HF ratio 
measures with psychological states and context conditions. The students received chest belts and 
storage devices that recorded the cardiovascular activity data wirelessly on an integrated memory chip 
(cf. Danev, 2010). LF/HF ratio was calculated via the software Qhrv Assessment (Medeia Ltd.®) 
resulting in an overall LF/HF ratio of M = 1.07 (SD = .32, Min. = .41, Max. = 3.22). 

4.2.3 Academic self-concept 

Academic self-concept was measured via six items that were adopted from Rost et al.’s (2007) 
Scale for assessing Self-concept of School related Achievement and Skills (SKSLF-8, e.g. “It's easy for 
me to resolve problems in school”). Learners rated items on a six-point Likert-type scale (1 = 
“Strongly disagree”, 6 = “Strongly agree”). The six self-concept items were averaged to a total scale, 
demonstrating satisfactory reliability (α = .81; M = 4.05, SD = .78, Min. = 2.17, Max. = 5.67). 

 

 



Kärner&et &al &
&

 
 
 

 | F L R !
!

23!

4.2.4 Students’ experience during class 

Students’ stress experience and situational coping were sampled at 10-minute intervals using 
mobile handheld-computers (Palm Tungsten E2®). The continuous-state-sampling method used 
(Sembill et al., 2002; Sembill, Seifried, & Dreyer, 2008) offers a high degree of ecological validity 
and differs from other experience-sampling methods (e.g. Csíkszentmihályi & Larson, 1987) due to 
the high frequency of its measurements and the equidistance of its measurement points, as the data are 
not sampled at randomly chosen times. There are a maximum number of 38 measurement points per 
person in the current study, adding up to a cumulative total of 2,014 measures. The number of 
measurement points equals the quantity of physiological states (LF/HF ratio) measures. For further 
analysis, physiological and psychological states were chronologically synchronised.  

Students had the opportunity to rate their actual experience on a continuous rating scale from 0 
(= “I fully disagree”) to 100 (= “I fully agree”). The scale referred to is comparable to the visual 
analogue mood scale from Luria (1975) and has been proved to be successful in previous process-
related video studies of the research group, for example in the field of self-organised learning (e.g. 
Sembill et al., 2002). Situational coping (SC) was measured using two items (“I can cope with actual 
demands”, M = 75.40, SD = 21.19, Min. = 0, Max. = 100; “I understand subject matter”, M = 79.80, 
SD = 20.06, Min. = 0, Max. = 100). Students’ stress experience (SSE), the dependent variable in our 
analysis, was measured via two items (“I’m under time pressure”, M = 17.44, SD = 25.01, Min. = 0, 
Max. = 100; “I’m under pressure to succeed”, M = 20.62, SD = 25.43, Min. = 0, Max. = 100). The 
items were formulated by the research group with regard to perceived time pressure and pressure to 
succeed as frequently mentioned terms for school related stressors (e.g. Kouzma & Kennedy, 2004) 
and to subjective coping competence and situational understanding as indicators for self-confidence in 
ability to follow the lesson (Kärner, 2015; Sembill et al., 2002). Assessing the factorial structure of the 
four items, we applied an exploratory factor analysis with varimax rotation and referred to the Kaiser 
criterion (eigenvalue > 1). The items form a two-factor solution that accounts for 85.29 % of the total 
variance (compared to the one-factor solution, accounting for only 43.17 %). Both items relating to 
students’ stress experience (r = .73) and both items of situational coping (r = .68) are correlated 
moderately to each other. For further analysis we used the factor scores as estimated values of the 
factors “situational coping” and “students’ stress experience”. 

4.2.5 Classroom demands 

Classroom demands—as “objective” characteristics of the classroom context—were 
operationalised by the amount of student-centred learning and by the quality of cognitive challenge 
during education. Student-centring is defined by phases of individual or group work where learners 
work independently from the teacher on complex problems. It was assessed via video-based time-
sampling analysis using a defined category-system we adopted from Seidel et al. (2001). Time 
intervals of 15 seconds each were coded. Afterwards, the single coded 15-second intervals were 
aggregated to 10-minute intervals via sum scores, synchronising the context conditions—in terms of 
“micro-segments” of classroom context—and the person-related data. To assess the reliability of the 
codings, one third of the videos were coded by two independent coders, finding a satisfactory Cohen’s 
kappa of .73. Overall we found a mean of M = 3.16 minutes (SD = 3.63, Min. = 0, Max. = 9.5) student-
centring per 10 minutes of education.  

Cognitive challenge during class was coded by ratings of two independent coders: they coded 
one third of the videos (between-coder-correlation r = .82) on the basis of the observed classroom 
discussion and the learning material the students worked on. In order to assess content-related 
difficulty, we referred to the curriculum of the corresponding subject “economic business processes” 
and to Bloom’s taxonomy of learning objectives. Time intervals of 1 minute each were coded, and we 
used a four point Likert-type scale based on Bloom’s taxonomy to assess the complexity of learning 
contents (0 = “applying”, 1 = “analysing”, 2 = “synthesizing”, 3 = “evaluating”; cf. Bloom et al., 



Kärner&et &al &
&

 
 
 

 | F L R !
!

24!

1956). Afterwards, the single coded 1-minute intervals were arithmetically aggregated to 10-minute 
intervals. An overall mean of M = 1.50 (SD = .82, Min. = .44, Max. = 3.00) of cognitive challenge in 
the observed lessons was found. Assessing the factorial structure of the amount of student-centred 
learning and the cognitive challenge during education, we applied an exploratory factor analysis with 
varimax rotation and referred to the Kaiser criterion (eigenvalue > 1). The items form a one-factor 
solution that accounts for 87.55 % of the total variance, with both variables correlated moderately to 
each other (r = .75). Afterwards, we used the factor scores as estimated values of the factor “classroom 
demands”. 

4.3  Statistical analysis 

4.3.1 Previous analysis 

Pearson product-moment correlations were calculated in order to identify multicollinearity and 
to check characteristics of the independent variables.  

4.3.2 Multilevel analysis 

Against our theoretical background, it seems to be crucial to measure time-varying states and 
objective context conditions in a synchronic way in addition to relatively enduring characteristics. 
Appropriate interrelationships can be investigated using multilevel analytic methods, as they provide 
the opportunity to simultaneously analyse different hierarchical data levels. In this context, 
longitudinal data can be seen as hierarchical data, with repeated measurements nested within persons 
(Bryk & Raudenbush, 1992; Goldstein et al., 1994; Heck & Thomas, 2009; Hox, 2002; Nezlek, 2007). 
Scollon et al. (2003) point out that multilevel modelling is useful in analysing continuously sampled 
data because multiple data points are nested in a single individual. In our case, students’ states are not 
only nested within persons but also within situations that are defined as “micro-segments” of the 
classroom context prevailing at the time of measurement. Therefore we applied a cross-classified 
multilevel model (cf. Heck et al., 2010). Cross-classification considers that the multiple state-measures 
are not only nested within persons but that they also belong to observation units of the classroom 
context at Level 2 corresponding to the time of measurement (cf. Goldstein, 1994; Hill & Goldstein, 
1998). Most of the existing approaches in researching person-situation interactions analyse how 
variations in situations affect individuals depending on their dispositions via analysis-of-variance or 
multivariate regression analysis (Carver & Scheier, 2008; Cronbach & Snow, 1977). But these 
approaches do not usually consider the hierarchical data structure implied by multiple measures, that is 
important because investigating trait-treatment interactions should focus on analysing how persons 
show changes in an outcome variable over time or across situations (Yeh, 2012). Multilevel methods 
have the advantage of combining the different approaches within a simultaneous analysis, therefore 
providing a potential method for investigating person<situation interactions in the classroom context 
(cf. Nezlek, 2007). In our analysis, the hierarchical data structure is reflected in the variables as 
follows. 

a) Level 3 represents the CLASS level: While the teachers of the two classes in our sample 
dealt with the same subject (“economic business processes”) within the same time frame, 
an ANOVA shows that the two classes differ in general levels of classroom demands: 
class a (M = –.52, SD = .48) demonstrates a lower degree of classroom demands than class 
b (M = .59, SD = 1.12) [F(1, 74) = 31.48, p < .001, η2 = .298]. Therefore, we consider the 
class-related means of classroom demands as a Level 3 variable. 

b) On the one hand, Level 2 represents the PERSON level: every student is characterised by 
individual baseline cortisol concentration and academic self-concept. We additionally 
consider person-related control variables. 



Kärner&et &al &
&

 
 
 

 | F L R !
!

25!

c) With regard to cross-classification, an additional Level 2 represents the level of 
CLASSROOM CONTEXT units where every 10-minute interval represents a “micro-
segment”: every single segment of the classroom context is characterised by its individual 
degree of classroom demands. Within each single segment, the state measures of all 
persons present at the same place at a fixed point in time are nested. 

d) Person<situation interactions are modelled via product terms of academic self-concept and 
classroom demands and of baseline cortisol concentration and classroom demands (cf. 
Baron & Kenny, 1986; Hayes & Matthes, 2009).  

e) At Level 1, within-person differences are modelled by time-varying states (students’ stress 
experience, situational coping, LF/HF ratio). 

It is assumed that students’ stress experience is nested as the dependent variable within 
PERSONS and within every single observation unit of the CLASSROOM CONTEXT at Level 2. 
Figure 2 illustrates the cross-classified data structure. 

To reflect the cross-classified structure in the unobserved model component, we calculated 
two separate models, comparing them afterwards. Model #1 considers random effects for intercept 
variance for the PERSON level and for the CLASSROOM CONTEXT level, changing at each interval 
of measurement. This allows for an estimation of the proportions of variance for the two nesting 
levels. Model #1 also considers random effects for slope variance, and we assume that slopes vary 
randomly across classroom context segments, depending on academic self-concept and on baseline 
cortisol concentration. As an alternative, model #2 focuses on the time dependence of measurements 
and considers an autoregressive covariance structure that “indicates that each person’s distance 
measurement at one time is correlated with their distance measurement at the previous time period” 
(Leech et al., 2008, p. 201). Therefore, we modelled a first-order autoregressive covariance structure 
with homogenous variances (cf. Hox, 2002; Littell, Pendergast, & Natarajan, 2000; Peugh & Enders, 
2005; Raudenbush & Bryk, 2002; Twisk, 2006).  

 

 
 

Figure 2. Cross-classified data structure with states nested within persons and within context segments 

 

1 … i … n 1 … i … n
1 sa,1,1 … sa,i,1 … sa,n,1

PERSONs of class b

CO
N

TE
X

T 
se

gm
en

ts
 o

f c
la

ss
 

a

PERSONs of class a

… … … … … …

CO
N

TE
X

T 
se

gm
en

ts
 o

f c
la

ss
 

a j sa,1,j … sa,i,j … sa,n,j

CO
N

TE
X

T 
se

gm
en

ts
 o

f c
la

ss
 

a
… … … … … …CO

N
TE

X
T 

se
gm

en
ts

 o
f c

la
ss

 
a

m sa,1,m … sa,i,m … sa,n,m

1 sb,1,1 … sb,i,1 … sb,n,1

CO
N

TE
X

T 
se

gm
en

ts
 o

f c
la

ss
 

b

CO
N

TE
X

T 
se

gm
en

ts
 o

f c
la

ss
 

a
… … … … … …

CO
N

TE
X

T 
se

gm
en

ts
 o

f c
la

ss
 

b j sb,1,j … sb,i,j … sb,n,j

CO
N

TE
X

T 
se

gm
en

ts
 o

f c
la

ss
 

b
… … … … … …CO

N
TE

X
T 

se
gm

en
ts

 o
f c

la
ss

 
b

m sb,1,m … sb,i,m … sb,n,m

CO
N

TE
X

T 
se

gm
en

ts
 o

f c
la

ss
 

b



Kärner&et &al &
&

 
 
 

 | F L R !
!

26!

The restricted maximum likelihood method was chosen as an estimation method, as it is 
adequate for small data sets (Heck et al., 2010). The analysis was realised using the SPSS® Mixed 
ProcedureTM (SPSS, 2005), an adequate tool for assessing cross-classified multilevel longitudinal 
modelling (Heck et al., 2010). Predictor variables and covariates were standardised. The statistical 
analyses were performed using SPSS 23® (IBM®, Chicago, USA). 

4.3.3 Handling of missing values 

Note that missing values have been taken into account within the statistical analysis via 
multiple imputation (cf. Rubin, 1987; Raessler, Rubin & Schenker, 2008). Using the multiple 
imputation routine implemented in SPSS, five imputed data sets have been analysed where the 
corresponding estimates have been joined using the classical combining rules established by Rubin 
(1987) and Little and Rubin (2002). 

 

5.  Results 

5.1  Previous analysis 

Table 1 shows the Pearson product-moment correlations among the variables. The correlations 
show that students’ stress experience is significantly positive associated with age, AUCg, classroom 
demands at Level 2 (as single segments), and classroom demands at Level 3 (as class-related means). 
Further, SSE is significantly negative associated with academic self-concept and situational coping. 
SSE is not significantly associated with sex and LF/HF ratio, therefore these variables will not be 
considered as predictors in the multilevel analysis.  

 

Table 1 

Pearson product-moment correlations 

 

    (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  
(1) Sex                               
(2) Age –.16 ***                           
(3) A. self-concept –.35 *** –.20 ***                       
(4) AUCg –.07 ** –.12 *** –.07 **                   
(5) LF/HF ratio –.22 *** .24 *** .00   –.03                 
(6) Situational coping –.16 *** –.03   .47 *** –.01   .00             

(7) Class. demands 
(single segments) –.07 ** –.05 * –.13 *** .27 *** –.12 *** –.07 **     

  

(8) Class. demands 
(class mean) –.12 *** –.08 *** –.24 *** .50 *** –.07  –.12 *** .55 *** 

  

(9) Students’ stress 
experience .02   .19 *** –.35 *** .33 *** .02   –.41 *** .40 *** .31 *** 

Notes. *** p < .001, ** p < .01, * p < .05; Sex: 0 = male, 1 = female 

 



Kärner&et &al &
&

 
 
 

 | F L R !
!

27!

5.2  Multilevel analysis 

5.2.1 Null models 

We primarily examined the single contributions to the explained variance of the different 
levels of model #1. Therefore, we analysed the null model with SSE measures nested within students 
and within context segments (each at Level 2) and within classes (as Level 3 units) to check possible 
nesting-effects of the class affiliation. We found significant variances between persons and between 
context segments, but not between classes. Thus, the nesting-levels are the person-level and the 
context segment-level (cf. cross-classification) and not the class-level. But because of the significant 
class difference in general levels of classroom demands, we modelled the class-related means of 
classroom demands as a fixed effect at Level 3. The analysis of the null model with SSE measures 
nested within students and within context segments shows a significant between-students intercept 
variance of .412 (SE = .086, p < .001) and a significant between-context segment intercept variance of 
.152 (SE = .028, p < .001) (see Table 2). 

The null model of model #2 revealed a rho correlation of .721 (SE = .015, p < .001) (see Table 
2), that indicates a remarkable correlation between any two successive measurement points (cf. Heck 
et al., 2010). 

 5.2.2 Mixed-effects model (model #1) and fixed-effects model with AR1 (model #2) 

Table 3 provides the results of the analysis of both multilevel models. The main effect of the 
class-related means of classroom demands (Level 3) remains non-significant in both analysed models. 
The main effect of the academic self-concept on students’ stress experience remains negative and 
significant in model #2 but not in model #1. Baseline cortisol concentration has significant positive 
effects on SSE in both models. Limitations in the interpretability of the main effects will be mentioned 
below. Age as a control variable does affect SSE significantly positively in both models. The degree 
of classroom demands (Level 2) affects SSE positively and significantly in both models. At Level 1, 
situational coping has significantly negative effects on stress experience in both models. 

Beside the found main effects on students’ stress experience, we found significant interaction 
effects that can be interpreted as moderator effects (cf. Baron & Kenny, 1986; Hayes & Matthes, 
2009): academic self-concept and baseline cortisol concentration affect the relation between classroom 
demands and students’ stress experience. To illustrate the mentioned relations, the students were 
classified into three percentiles by their academic self-concept values as well as by their AUCg values. 
Concerning academic self-concept (selfc) Group aselfc has a self-concept group-mean value of M = 3.19 
(SD = .49), group bselfc has a mean value of M = 4.24 (SD = .18), and the self-concept mean value of 
group cselfc is 4.90 (SD = .29). The three groups differ significantly in their academic self-concept (p < 
.001, partial η2 = .798). Concerning AUCg Group aAUCg has a group-mean value of M = 2.93 (SD = 
1.55), group bAUCg has a mean value of M = 8.61 (SD = 2.32), and the AUCg mean value of group 
cAUCg is 19.97 (SD = 6.21). The three groups differ significantly in their areas under the curve values (p 
< .001, partial η2 = .767). Figure 3a illustrates the group-specific linear growth trajectories of the three 
self-concept groups, and Figure 3b illustrates the group-specific linear growth trajectories of the three 
AUCg groups: students who have low academic self-concept values show a stronger increase in stress 
experience with increasing classroom demands than students who have high self-concept values. On 
the other hand, students with high AUCg values show a stronger increase in stress experience with 
increasing classroom demands than students who have low AUCg values. 

To check the characteristics of main effects and interaction effects, we referred to the 
classification of interaction effects of Leigh and Kinnear (1980) (see also Bortz & Schuster, 2010). For 
both interaction effects (Figure 3a, Figure 3b) we found pure disordinal interactions (cf. crossed lines). 
This describes non-monotonic relations: the rank order of levels of one moderator factor (academic 
self-concept, AUCg) does not remain constant over all levels of the abscissa factor (classroom 



Kärner&et &al &
&

 
 
 

 | F L R !
!

28!

demands); and the rank order of levels of the factor “classroom demands” does not remain constant 
over all levels of the factors “academic self-concept” and “AUCg”. Therefore, the main effects of 
academic self-concept, AUCg, and classroom demands cannot be sensibly interpreted. 

Table 3 also provides the variances of the random effects for model #1 and of the repeated 
measures effect for model #2. With regard to the random effects of model # 1 we found a significant 
random intercept between-students variance of .139 (compared to the null model’s random intercept 
variance between students of .412) and a significant random intercept variance between context 
segments of .045 (compared to the null model’s random intercept variance between context segments 
of .152). The variance components suggest that the SSE slopes vary randomly across context segments 
in relation to baseline cortisol concentration, while showing no significant variance in relation to 
academic self-concept. With regard to the repeated measures in model #2, Table 3 suggests a rho 
correlation of .565 (compared to the null model’s rho correlation of .721).  



Kärner&et &al &
&

 
 
 

 | F L R !
!

29!

Table 2 

Null models 

 

  Model #1 
Mixed-effects model   Model #2 

Autoregressive covariance 
        95% CI         95% CI 
Effect Estimate SE p LB UB   Estimate SE p LB UB 
FIXED EFFECTS                       
   Intercept .003 .100 .972 –.192 .199   –.023 .052 .657 –.124 .078 
RANDOM EFFECTS                       
   Residual variance .427 .015 < .001 .398 .456             
   Random intercept variance (Persons) .412 .086 < .001 .244 .580             
   Random intercept variance (Context segments) .152 .028 < .001 .096 .208             
REPEATED MEASURES                       
   AR1 diagonal             .983 .053 < .001 .879 1.087 
   AR1 ρ             .721 .015 < .001 .691 .751 

 

 

 



Kärner&et &al &
&

 
 
 

 | F L R !
!

30!

Table 3 
Mixed-effects model and fixed-effects model (AR1) 
   Model #1 

Mixed-effects model   Model #2 
Fixed-effects model (AR1) 

        95% CI         95% CI 
Effect Estimate SE p LB UB   Estimate SE p LB UB 
FIXED EFFECTS                       
   Intercept –.047 .059 .421 –.163 .068   –.039 .031 .209 –.100 .022 
LEVEL 3: CLASS                       
   Classroom demands (class mean) –.015 .071 .837 –.153 .124   .034 .039 .384 –.043 .111 
LEVEL 2: PERSON                       
   Age .228 .055 < .001 .120 .335   .216 .032 < .001 .153 .279 
   Academic self-concept –.108 .057 .060 –.220 .004   –.105 .034 .002 –.172 –.038 
   AUCg .238 .063 < .001 .115 .362   .247 .036 < .001 .177 .318 
LEVEL 2: CLASSROOM CONTEXT                       
   Classroom demands (single segments) .279 .034 < .001 .212 .346   .172 .022 < .001 .128 .216 
LEVEL 2: INTERACTIONS                       
   Academic self-concept × Classroom demands –.086 .018 < .001 –.121 –.051   –.060 .021 .004 –.101 –.019 
   AUCg × Classroom demands .127 .021 < .001 .087 .167   .110 .019 < .001 .072 .148 
LEVEL 1                       
   Situational coping –.283 .022 < .001 –.326 –.239   –.297 .024 < .001 –.344 –.250 
RANDOM EFFECTS                       
   Residual variance .370 .013 < .001 .344 .397             
   Random intercept variance (Persons) .139 .031 < .001 .078 .199             
   Random intercept variance (Context segments) .045 .010 < .001 .024 .065             
   Random slope variance (Academic self-concept) .001 .003 .675 .000 .006             
   Random slope variance (AUCg) .009 .004 .018 .002 .017             
REPEATED MEASURES                       
   AR1 diagonal             .562 .025 < .001 .513 .611 
   AR1 ρ             .565 .019 < .001 .527 .603 



Kärner&et &al &
&

 
 
 

 | F L R !
!

31!

 
 
Figure 3a. Interaction: Academic self-concept × Classroom demands 

 

 
 
Figure 3b. Interaction: AUCg × Classroom demands 

a)

aselfc: β = .44

bselfc: β = .41

cselfc: β = .27

Classroom demands (standardized)

St
ud

en
ts

’ s
tr

es
s 

ex
pe

ri
en

ce
 (s

ta
nd

ar
di

ze
d)

Academic self-concept
aselfc : M = 3.19 (± .49)
bselfc : M = 4.24 (± .18)
cselfc : M = 4.90 (± .29)

1 2–2 –1 0

1

2

–2

–1

0

cAUCg : β = .54

bAUCg : β = .21

aAUCg : β = .18

AUCg
aAUCg : M = 2.93 (± 1.55)
bAUCg : M = 8.61 (± 2.32)
cAUCg : M = 19.97 (± 6.21)

b)

Classroom demands (standardized)

St
ud

en
ts

’ s
tr

es
s 

ex
pe

ri
en

ce
 (s

ta
nd

ar
di

ze
d)

1 2–2 –1 0

1

2

–2

–1

0



Kärner&et &al &
&

 
 
 

 | F L R !
!

32!

5.2.3 Model comparisons 

To summarize the evidence referring to the different information and model selection criteria 
available for each imputed data set, we provide the median, minimum and maximum values for each 
information and model selection criterion. An inspection of the minimum and maximum values for the 
different criteria indicates that the autoregressive covariance structure is the overall preferred model 
specification (see Table 4 and Table 5). 

 

Table 4 

Model comparisons: Null models 

 

Information criteria 
Model #1 

Mixed-effects model 
  

Model #2 
Autoregressive covariance 

   Median Min. Max.    Median Min. Max. 
Restricted –2 Log Likelihood 4,372.6 4,333.7 4,394.1   4,248.1 4,227.5 4,268.4 
Akaike Information Criterion (AIC) 4,378.6 4,339.7 4,400.1   4,252.1 4,231.5 4,272.4 
Hurvich and Tsai's Criterion (IC) 4,378.6 4,339.7 4,400.1   4,252.1 4,231.5 4,272.4 
Bozdogan Criterion (CAIC) 4,398.4 4,359.6 4,420.0   4,265.3 4,244.8 4,285.6 
Schwarz Criterion (BIC) 4,395.4 4,356.6 4,417.0   4,263.3 4,242.8 4,283.6 

 

Table 5 

Model comparisons: Mixed-effects model and fixed-effects model (AR1) 

 

Information criteria 
Model #1 

Mixed-effects model 
  

Model #2 
Autoregressive covariance 

   Median Min. Max.    Median Min. Max. 
Restricted –2 Log Likelihood 4,042.3 3,988.6 4,053.1   3,845.8 3,826.9 3,848.7 
Akaike Information Criterion (AIC) 4,052.3 3,998.6 4,063.1   3,849.8 3,830.9 3,852.7 
Hurvich and Tsai's Criterion (IC) 4,052.4 3,998.6 4,063.1   3,849.8 3,830.9 3,852.7 
Bozdogan Criterion (CAIC) 4,085.4 4,031.6 4,096.1   3,863.0 3,844.1 3,865.9 
Schwarz Criterion (BIC) 4,080.4 4,026.6 4,091.1   3,861.0 3,842.1 3,863.9 

 

6.  Summary of results and discussion 

In our study, we investigated 53 students at a German vocational training school over a course 
of 9 lessons, examining students’ stress response, considering psychological, physiological, and 
observational data. Via the continuous-state-sampling method we assessed almost 38 state 
measurements per person, adding up to a cumulative total of 2,014 measurements. To analyse effects 
on students’ stress experience, a cross-classified multilevel analysis was applied. 

We hypothesised that the baseline cortisol concentration should have been positively 
associated with students’ stress experience. The analysis shows a main effect of baseline cortisol 



Kärner&et &al &
&

 
 
 

 | F L R !
!

33!

concentration (operationalised by an individual’s area under the curve with respect to the ground) on 
SSE (sees Table 3). This main effect is consistent with existing research, because cortisol 
concentration during the day is an endocrinological indicator for chronic stress that may suggest a 
chronic elevation of HPA activation (Wüst et al., 2000; Takahashi et al., 2005). However, the found 
main effect cannot be reasonably interpreted due to the found pure disordinal interaction between 
AUCg and classroom demands (cf. Leigh & Kinnear, 1980).  

Contrary to our hypothesis, the LF/HF ratio as a time-varying somatic stress indicator is not 
significantly associated with SSE (see Table 1). This finding is not consistent with the existing 
literature that indicates that there should be a predominant status of the sympathetic nervous system 
that is activated under stress, accompanied by an underrepresented status of the parasympathetic 
nervous system that controls physiological resting functions (cf. Kemeny, 2003; Hjortskov et al., 
2004; Isowa et al., 2006). From a conceptual point of view, the non-significant correlation between 
SSE and LF/HF ratio could result from the general stress level in class. In reference to the rating scale 
from 0 (= “I fully disagree”) to 100 (= “I fully agree”) we found mean values lower than average for 
perceived time pressure (M = 17.44, SD = 25.01) and pressure to succeed (M = 20.62, SD = 25.43). 
However, existing studies show a synchrony of different stress systems only in highly stressful 
situations (Looser et al., 2010). From a methodological point of view, correlations between self-ratings 
and physiological measures are assumed to be quite lower than correlations between different self-
rating variables (cf. Campbell & Fiske, 1959).  

We hypothesised that academic self-concept should be negatively associated with students’ 
stress experience. The Pearson product-moment correlations show a significant and negative 
correlation (r = –.35, see Table 1) with the multilevel model #2 also revealing a significant negative 
effect. The main effect of academic self-concept on stress experience remains negative but non-
significant in model #1. However, the main effect of the self-concept cannot be sensibly interpreted 
due to the found pure disordinal interaction between self-concept and classroom demands. 

With regard to the situational coping that was operationalised by the items “I can cope with 
actual demands” and “I understand subject matter”, we found a significant and positive effect on stress 
experience conforming to our hypothesis. However, in that regard we have to consider potential 
effects of common method biases. The items related to both variables (students’ stress experience, 
situational coping) were measured by continuous-state-sampling method. With respect to Podsakoff et 
al. (2003) one has to consider possible method effects produced by item characteristics, as the way in 
which the items are presented to the students (scale format, scale anchors) was the same for the used 
state-items. Further, one has to take into account possible method effects produced by the 
measurement context. State-items of both variables have been measured concurrently with regard to 
time and location of measurement, using the same medium for measurement as well. Therefore, it 
seems to be worthwhile in future research to operationalise situational coping via observational 
methods instead of through self-rating items (cf. Roubinov, Hagan, & Luecken, 2012; Wuttke, 2000). 

As hypothesised, we found a main effect of classroom demands (Level 2) on students’ stress 
experience. However, the found main effect cannot be sensibly interpreted because of the found pure 
disordinal interactions (cf. Leigh & Kinnear, 1980). The main effect of classroom demands (Level 3) 
on SSE remains non-significant. 

Besides the previously discussed main effects on students’ stress experience, a significant 
person%situation interaction occurs between academic self-concept and the degree of classroom 
demands: students who have low academic self-concept values show a stronger increase in stress 
experience with increasing classroom demands than students who have high self-concept values (see 
Figure 3a). This finding is in line with our hypothesis. The “objective” degree of classroom demands 
is perceived as stressful in different ways depending on a subjective appraisal of potential coping 
resources. Following Lazarus and colleagues, a stressor is first interpreted with regard to its perceived 
relevance (indeed, it is a basic assumption of the current study that the students interpret the lessons as 
being relevant to them). If it is interpreted as relevant and threatening, the secondary appraisal entails 



Kärner&et &al &
&

 
 
 

 | F L R !
!

34!

an analysis of available resources and self-confidence in order to cope with the situation appropriately 
(cf. Lazarus, 1984; Lazarus & Folkman, 1984).  

In addition to the interaction between self-concept and classroom demands, the assumed 
person%situation interaction between baseline cortisol concentration and classroom demands also 
remains significant: students with high AUCg values show a stronger increase in stress experience 
with increasing classroom demands than students who have low AUCg values (see Figure 3b). With 
reference to the literature, high baseline cortisol values are associated with a chronic elevation in the 
level of HPA activation (cf. Wüst et al., 2000; Takahashi et al., 2005). In regard to the diathesis-stress 
model, physiological vulnerability—such as a chronic elevation of baseline cortisol concentration—in 
turn affects the relation between stressful context conditions and stress experience as a possible stress 
response. In other words: the higher the vulnerability, the greater the responsiveness to demanding 
conditions (cf. Belsky & Pluess, 2009).  

Our findings on person-situation interactions indicate that “objective” conditions in class may 
be experienced very differently depending on individual characteristics. Or with the words of Krohne 
(1990, p. 27): “[the] process of cognitively transforming objective characteristics into subjective 
representations is clearly codetermined by personality dispositions.” Furthermore, it seems to be 
important to take into greater consideration possible interaction effects between person- and situation-
related variables in learning research. Otherwise, there is the risk of overestimating main effects of 
person- and situation-related variables. 

So far, our findings draw quite a negative picture of class-related demands. However, one has 
to consider that challenges can also encourage an individual’s development. For example, 
experiencing mastery over a demanding situation powerfully impacts an individual’s self-efficacy 
beliefs (Bandura, 1977). Further, the concept of the “zone of proximal development” suggests the 
importance of overcoming barriers for an individual’s competency development within a social 
context (Vygotskii, 1978). Nonetheless, it remains a pivotal endeavour for teachers and learning 
researchers to design and implement teaching and learning arrangements where learners can gain such 
experiences of mastery within a protected and supportive environment (cf. Sembill, 1999). 

 

7.  Limitations of the study 

The design of our study has advantages as well as limitations. The naturalistic setting and the 
high frequency of data sampling offer a high degree of ecological validity (cf. Sembill, Seifried, & 
Dreyer, 2008) further ensured by the situational specificity of data sampling (Hormuth, 1986; Wild, 
2001). On the other hand, the naturalistic setting bears the risk of confounding variables that may bias 
the results. Therefore, for further research a combination of naturalistic settings and laboratory 
experimental conditions seems to be worthwhile. For example, specific class-related stressors could be 
identified in naturalistic learning contexts, to be validated within later laboratory experiments. 

Another potential limitation of our study appears in regard to the collection of saliva samples 
as the basis for measurement of cortisol concentrations. Indeed, all participants were instructed to 
adhere strictly to the prescribed sampling times, but sampling times could not be strictly monitored. 
Further research would necessitate employing methods of electronic medication event monitoring (e.g. 
Medication Event Monitoring Systems [MEMS®], Aardex Ltd., Schweiz; see also Rösler et al., 2010; 
Kudielka, Borderick, & Kirschbaum, 2003). 

Furthermore, our analysis only focuses on a relatively narrow scope within the discussion of 
stress among students. We considered only students’ academic self-concept; the baseline cortisol 
concentration; the LF/HF ratio, as an indicator for heart rate variability; and as indicators for 
classroom demands, the amount of student-centred learning and the levels of cognitive challenge 



Kärner&et &al &
&

 
 
 

 | F L R !
!

35!

during education. Taking account for the wide range of other stress- and coping-related variables, 
further research should also focus on other relevant traits (e.g. emotion-focused and problem-focused 
strategies, behavioural inhibition and behavioural activation), on additional situational characteristics 
(such as variations in domain-specific learning contents or didactic methods), and on other dependent 
variables in terms of psychological, physiological, and behavioural states. 

 

8.  Conclusions and implications for future research 

In conclusion, a deeper knowledge about the complex interrelations between traits, states, and 
continuous changing context conditions seems to be essential for a more holistic understanding of 
learning at school and for the identification of crucial aspects for an evidence-based design and 
implementation of teaching and learning arrangements that are able to meet individual’s requirements 
in an optimal way. Thus, further learning process research should focus more intensively on causal 
relationships and regulatory feedback mechanisms between (1) psychological and physiological 
characteristics; (2) situational context conditions during education; (3) psychological, physiological, 
and behavioural states; and (4) learning outcomes such as domain-specific competencies. 

From a methodological point of view, there are some implications for further research on such 
complex interrelations. Our analysis only took account for unidirectional relations (personal and 
situational variables → states). However, in line with the transactional model of stress and coping 
(Lazarus, 1984; Lazarus & Folkman, 1984) one must assume reciprocal relations: expressed emotions 
and behaviours, in turn constituting new and modified situations and context conditions. Furthermore, 
we only took linear relationships into account. However, particularly in stress research, empirical 
evidence shows non-linear relations between demanding conditions and neurophysiological processes 
that are described by inverted U-shaped curves (e.g. Diamond et al., 1992; Hebb, 1955; Yerkes & 
Dodson, 1908). Another aspect worth considering in future research is that possible interaction effects 
between personal characteristics and situational conditions during class themselves may be affected by 
variables—such as school climate or aspects of class composition—that are located on higher levels 
and that account for variations in the nature of interaction effects on lower levels. But in that case, a 
thorough examination of possible context effects on the basis of aggregated context data seems 
worthwhile. However, the consideration of characteristics of “micro-segments” of the classroom 
context that are constituted by process-related characteristics might be more precise than an exclusive 
consideration of possible effects of aggregated data (e.g. class-related means for classroom demands).  

With regard to high frequency assessments of classroom conditions, future research should 
focus on a “multidimensional mapping” of context conditions and on an analysis of corresponding 
effects on students’ learning, experience, and achievement. In the current analysis, we only considered 
classroom demands as one possible dimension of the context. But one can imagine considering 
multiple dimensions in the same way, for instance looking at learning objectives, learning contents, 
didactic methods, and media as core dimensions of the Berlin model of didactics (summarizing Klafki, 
1995 and Zierer & Seel, 2012) or considering classroom organisation, student orientation, and 
cognitive activation as dimensions of instructional quality (Wagner et al., 2013). Such research could 
finally result in empirically validated “multidimensional cubes” that characterise the antecedents and 
effects of successful classroom learning. 

From a conceptual point of view, it will be a challenge for future applied learning process 
research to translate its empirical and theoretical findings into appropriate recommendations for 
classroom practitioners leading to an evidence-based design and implementation of teaching and 
learning arrangements. 



Kärner&et &al &
&

 
 
 

 | F L R !
!

36!

Keypoints  

 Academic self-concept and baseline cortisol concentration affect the relationship between 
classroom demands and students’ stress experience. 

 Cross-classified multilevel longitudinal modelling is a capable method for investigating 
associations between traits, states, and “objective” situational context conditions in 
naturalistic learning processes. 

 It seems to be important to take into greater consideration possible interaction effects 
between person- and situation-related variables in learning research. Otherwise, there is a 
risk of overestimating corresponding main effects of person- and situation-related 
variables. 

 A deeper knowledge about the complex interrelations between traits, states, and continuous 
changing context conditions seems to be essential for a more holistic understanding of 
learning at school and for the identification of crucial aspects for an evidence-based design 
and implementation of teaching and learning arrangements. 

Abbreviations 

ATI: aptitude treatment interaction; AUCg: area under the curve with respect to the ground; 
HF: high frequency (domain); HPA axis: hypothalamic-pituitary-adrenal axis; Hz: Hertz; LF: low 
frequency (domain); SC: situational coping; SSE: students’ stress experience 

Acknowledgments 

The authors would like to thank the anonymous reviewers for their critical, but very helpful 
and constructive comments and suggestions, especially with regard to the statistical analysis and to the 
line of argument. Further, the authors would like to thank Douglas Boatwright (Berlin, Germany) for 
the professional proof-reading of our manuscript. 

References 

Achtenhagen, F. (1978). Beanspruchung von Schülern. Methodisch-didaktische Aspekte [Students’ 
stress. Methodical and didactical aspects]. BMBW Werkstattberichte, Bundesminister für 
Bildung und Wissenschaft. Regensburg: Aumüller Druck KG. 

Adam, E. K. (2006). Transactions among adolescent trait and state emotion and diurnal and 
momentary cortisol activity in naturalistic settings. Psychoneuroendocrinology, 31 (5), 664–679. 
doi: 10.1016/j.psyneuen.2006.01.010 

Agelink, M. W., Malessa, R., Baumann, B., Majewski, T., Akila, F., Zeit, T., & Ziegler D. (2001). 
Standardized tests of heart rate variability: normal ranges obtained from 309 healthy humans, and 
effects of age, gender, and heart rate. Clinical Autonomic Research, 11 (2), 99–108. doi: 
10.1007/BF02322053 

Baert, S., Casier, A., & De Raedt, R. (2012). The effects of attentional training on physiological stress 
recovery after induced social threat. Anxiety, Stress, & Coping: An International Journal, 25 (4), 
365–379. doi: 10.1080/10615806.2011.605122 

Bandura, A. (1977). Self-efficacy: Toward a Unifying Theory of Behavioral Change. Psychological 
Review, 84 (2), 191–215. doi: 10.1037/0033-295X.84.2.191 



Kärner&et &al &
&

 
 
 

 | F L R !
!

37!

Baron, R. M. & Kenny, D. A. (1986). The Moderator-Mediator Variable Distinction in Social 
Psychological Research: Conceptual, Strategic, and Statistical Considerations. Journal of 
Personality and Social Psychology, 51 (6), 1173–1182. doi: 10.1037/0022-3514.51.6.1173 

Beck, K. & Sczesny, C. (1993). Zur Psychophysiologie des Lernens in der Computerinteraktion. 
Grundprobleme, Forschungsresultate, Untersuchungsperspektiven. Bericht zum 
Forschungsprojekt „Prozeßanalyse computergestützen Lernens“ [To the psychophysiology of 
learning within human-computer interaction. Basic problems, research findings, perspectives of 
investigation. Report to the research project “Process-analysis of computer-assisted learning”]. 
Nürnberg: University Erlangen-Nürnberg. 

Belsky, J. & Pluess, M. (2009). Beyond diathesis stress: Differential susceptibility to environmental 
influences. Psychological Bulletin, 135 (6), 885–908. doi: 10.1037/a0017376 

Bloom, B. S., Engelhart, M. D., Furst, E. J., Hill, W. H., & Krathwohl, D. R. (1956). Taxonomy of 
educational objectives: The classification of educational goals. Handbook I: Cognitive domain. 
New York: David McKay Company. 

Bong, M. & Skaalvik, E. M. (2003). Academic Self-Concept and Self-Efficacy: How Different Are 
They Really? Educational Psychology Review, 15 (1), 1–40. doi: 10.1023/A:1021302408382 

Bortz, J. & Schuster, C. (2010). Statistik für Human- und Sozialwissenschaftler [Statistics for the 
humanities and social sciences]. Springer: Berlin. 

Bryk, A. S. & Raudenbush, S. W. (1992). Hierarchical Linear Models: Applications and Data 
Analysis Methods. London: Sage Publications. 

Buchanan, T. W. & Lovallo W. R. (2001). Enhanced memory for emotional material following stress-
level cortisol treatment in humans. Psychoneuroendocrinology, 26 (3), 307–317. doi: 
10.1016/S0306-4530(00)00058-5 

Campbell, D. T. & Fiske, D. W. (1959). Convergent and discriminant validation by the multitrait-
multimethod matrix. Psychological Bulletin, 56 (2), 81–105, doi: 10.1037/h0046016 

Caplan, R. D., Cobb, S., & French, J. R. P. Jr. (1979). White collar work load and cortisol: Disruption 
of a circadian rhythm by job stress? Journal of Psychosomatic Research, 23 (3), 181–192. doi: 
10.1016/0022-3999(79)90003-5 

Carver, Ch. S. & Scheier, M. F. (2008). Perspectives on Personality. Boston: Pearson.  
Cobas (2010). Leistungsdaten von Elecsys® und cobas e® Geräten hinsichtlich der Bestimmung des 

Cortisolwertes [Performance data for Elecsys® and cobas e® equipment concerning the 
identification of salvia cortisol concentration]. (W. Mayer, Immumed GmbH, Karlstrasse 46, 
80333 Munich, Germany). Technical manuscript. 

Cronbach, L. J. & Snow, R. E. (1977). Aptitudes and Instructional Methods. A Handbook for Research 
on Interactions. New York: Irvington Publishers. 

Csíkszentmihályi, M. & Larson, R. (1987). Validity and Reliability of the Experience-Sampling 
Method. The Journal of Nervous and Mental Diseases, 175 (9), 526–536. doi: 
10.1097/00005053-198709000-00004 

Danev, S. (2010). QHRV-Lite System Guide & User’s Manual. Medeia Ltd. 
Deci, E. L. & Ryan, R. M. (1985). Intrinsic Motivation and Self-Determination in Human Behavior. 

New York: Plenum Press. 
Diamond, D. M., Bennett, M. C., Fleshner, M., & Rose, G. M. (1992). Inverted-U Relationship 

between the Level of Peripheral Corticosterone and the Magnitude of Hippocampal Primed Burst 
Potention. Hippocampus, 2 (4), 421–430. doi: 10.1002/hipo.450020409 

Dickerson, S. S. & Kemeny, M. E. (2004). Acute Stressors and Cortisol Responses: A Theoretical 
Integration and Synthesis of Laboratory Research. Psychological Bulletin, 130 (3), 355–391. doi: 
10.1037/0033-2909.130.3.355 

Ellenbogen, M. A. (2012). Introduction to the special section on biopsychosocial moderators of the 
stress response. Anxiety, Stress, & Coping: An International Journal, 25 (4), 359–364. doi: 
10.1080/10615806.2012.691621 

Fekedulegn, D. B., Andrew, M. E., Burchfiel, C. M., Violanti, J. M., Hartley, T. A., Charles, L. E., & 
Miller, D. B. (2007). Area under the curve and other summary indicators of repeated waking 



Kärner&et &al &
&

 
 
 

 | F L R !
!

38!

cortisol measurements. Psychosomatic Medicine, 69 (7), 651–659. doi: 
10.1097/PSY.0b013e31814c405c 

Goetz, T., Preckel, F., Zeidner, M., & Schleyer, E. (2008). Big fish in big ponds: A multilevel analysis 
of test anxiety and achievement in special gifted classes. Anxiety, Stress, & Coping: An 
International Journal, 21 (2), 185–198. doi: 10.1080/10615800701628827 

Goldstein, H. (1994). Multilevel Cross-Classified Models. Sociological Methods Research, 22 (3), 
364–375. doi: 10.1177/0049124194022003005 

Goldstein, H., Healy, M. J., & Rasbash, J. (1994). Multilevel time series models with applications to 
repeated measures data. Statistics in Medicine, 13 (16), 1643–1655. doi: 
10.1002/sim.4780131605 

Hall, J. E. (2011). Guyton and Hall Textbook of Medical Physiology. Philadelphia: Saunders Elisever. 
Hayes, A. F. & Matthes, J. (2009). Computational procedures for probing interactions in OLS and 

logistic regression: SPSS and SAS implementations. Behavior Research Methods, 41 (3), 924–
936. doi: 10.3758/BRM.41.3.924 

Hebb, D. O. (1955). Drives and the C.N.S (Conceptual Nervous System). The Psychological Review, 
62 (4), 243–254. doi: 10.1037/h0041823 

Heck, R. H. & Thomas, S. L. (2009). An Introduction to Multilevel Modeling Techniques. New York: 
Routledge. 

Heck, R. H., Thomas, S. L., & Tabata, L. N. (2010). Multilevel and Longitudinal Modeling with IBM 
SPSS. New York: Routledge. 

Hill, P. W. & Goldstein, H. (1998). Multilevel Modeling of Educational Data with Cross-
Classification and Missing Identification for Units. Journal of Educational and Behavioral 
Statistics, 23 (2), 117–128. doi: 10.3102/10769986023002117 

Hjortskov, N., Rissén, D., Blangsted, A. K., Fallentin, N., Lundberg, U., & Søgaard, K. (2004). The 
effect of mental stress on heart rate variability and blood pressure during computer work. 
European Journal of Applied Physiology, 92 (1–2), 84–89. doi: 10.1007/s00421-004-1055-z 

Hormuth, S. E. (1986). The sampling of experience in situ. Journal of Personality, 54 (1), 262–293. 
doi: 10.1111/j.1467-6494.1986.tb00395.x 

Hox, J. (2002). Multilevel Analysis. Techniques and Applications. London: Lawrence Erlbaum 
Associates. 

Isowa, T., Ohira, H., & Murashima, S. (2006). Immune, endocrine and cardiovascular responses to 
controllable and uncontrollable acute stress. Biological Psychology, 71 (2), 202–213. doi: 
10.1016/j.biopsycho.2005.04.002 

Kärner, T. (2015). Erwartungswidrige Minderleistung und Belastung im kaufmännischen Unterricht – 
Analyse pädagogischer, psychologischer und physiologischer Aspekte [Underachievement and 
stress in vocational business education – analysis of pedagogical, psychological, and 
physiological aspects]. Frankfurt a. M.: Lang. 

Kemeny, M. E. (2003). The Psychobiology of Stress. Current Directions in Psychological Science, 12 
(4), 124–129. doi: 10.1111/1467-8721.01246 

Kendall, P. C. (1981). Anxiety: States, Traits—Situations? In A. Furnham & M. Argyle (eds.), The 
Psychology of Social Situations: Selected Readings, (pp. 208–220). Oxford: Pergamon Press. 

Klafki, W. (1995). Didactic analysis as the core of preparation of instruction. Journal of Curriculum 
Studies, 27 (1), 13–30. doi: 10.1080/002202795027010 

Kouzma, N. M. & Kennedy, G. A. (2004). Self-Reported Sources of Stress in Senior High School 
Students. Psychological Reports, 94 (1), 314–316. doi: 10.2466/pr0.94.1.314-316 

Kristof-Brown, A. L., Zimmerman, R. D., & Johnson, E. C. (2005). Consequences of individuals’ fit 
at work: A meta-analysis of person-job, person-organization, person-group, and person-
supervisor fit. Personnel Psychology, 58 (2), 281–342. doi: 10.1111/j.1744-6570.2005.00672.x 

Krohne, H. W. (1990). Personality as a Mediator Between Objective Events and Their Subjective 
Representation. Psychological Inquiry, 1 (1), 26–29. doi: 10.1207/s15327965pli0101_7 



Kärner&et &al &
&

 
 
 

 | F L R !
!

39!

Kudielka, B., Borderick, J. E., & Kirschbaum, C. (2003). Compliance with saliva sampling protocols: 
Electronic monitoring reveals invalid cortisol daytime profiles in noncompliant subjects. 
Psychosomatic Medicine, 65 (2), 313–319. doi: 10.1097/01.PSY.0000058374.50240.BF 

Lazarus, R. S. (1966). Psychological Stress and Coping Process. New York: McGraw-Hill Book 
Company.  

Lazarus, R. S. (1984). On the Primacy of Cognition. American Psychologist, 39 (2), 124–129. doi: 
10.1037/0003-066X.39.2.124 

Lazarus, R. S. (1999). Stress and Emotion. A New Synthesis. New York: Springer. 
Lazarus, R. S. & Folkman, S. (1984). Stress, Appraisal, and Coping. New York: Springer. 
Leech, N. L., Barrett, K. C., & Morgan, G. A. (2008). SPSS for Intermediate Statistics. Use and 

Interpretation. New York: Taylor & Francis. 
Leigh, J. H. & Kinnear, T. C. (1980). On interaction classification. Educational and Psychological 

Measurement, 40 (4), 841–843. doi: 10.1177/001316448004000405 
Littell, R. C., Pendergast, J., & Natarajan, R. (2000). Tutorial in biostatistics: Modelling covariance 

structure in the analysis of repeated measures data. Statistics in Medicine, 19 (13), 1793–1819. 
doi: 10.1002/1097-0258(20000715)19:13<1793::AID-SIM482>3.0.CO;2-Q 

Little, R. J. A. & Rubin, D. B. (2002). Statistical Analysis with Missing Data. New Jersey: Wiley. 
Looser, R. R., Metzenthin, P., Helfricht, S., Kudielka, B. M., Loerbroks, A., Thayer, J. F., & Fischer, 

J. E. (2010). Cortisol Is Significantly Correlated With Cardiovascular Responses During High 
Levels of Stress in Critical Care Personnel. Psychosomatic Medicine, 72 (3), 281–289. doi: 
10.1097/PSY.0b013e3181d35065 

Luria, R. E. (1975). The validity and reliability of the visual analogue mood scale. Journal of 
Psychiatric Research, 12 (1), 51–57.  doi: 10.1016/0022-3956(75)90020-5 

Malik, M., Bigger, J. T., Camm, A. J., Kleiger, R. E., Malliani, A., Moss, A. J., & Schwartz, P. J. 
[Task Force of The European Society of Cardiology and The North American Society of Pacing 
and Electrophysiology] (1996). Heart rate variability: Standards of measurement, physiological 
interpretation, and clinical use. European Heart Journal, 17 (3), 354–381. 

Michaud, K., Matheson, K., Kelly, O., & Anisman, H. (2008). Impact of Stressors in a natural context 
on release of cortisol in healthy adult humans: A meta-analysis. Stress, 11 (3), 177–197. doi: 
10.1080/10253890701727874 

Nezlek, J. B. (2007). A Multilevel Framework for Understanding Relationships Among Traits, States, 
Situations and Behaviours. European Journal of Personality, 21 (6), 789–810. doi: 
10.1002/per.640 

Peugh, J. L. & Enders, C. K. (2005). Using the SPSS Mixed Procedure to Fit Cross-Sectional and 
Longitudinal Multilevel Models. Educational and Psychological Measurement, 65 (5), 717–741. 
doi: 10.1177/0013164405278558 

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), 879–890. doi: 10.1037/0021-9010.88.5.879 

Pruessner, J. C., Kirschbaum, C., Meinlschmid, G., & Hellhammer, D. H. (2003). Two formulas for 
computation of the area under the curve represent measures of total hormone concentration 
versus time-dependent change. Psychoneuroendocrinology, 28 (7), 916–931. doi: 
10.1016/S0306-4530(02)00108-7 

Raessler, S., Rubin, D. B., & Schenker, N. (2008). Incomplete Data: Diagnosis, Imputation, and 
Estimation. In E. D. de Leeuw, J. J. Hox, & D. A. Dillman (eds.),  International Handbook of 
Survey Methodology, (pp. 370–386). New York: Lawrence Erlbaum. 

Raudenbush, S. W. & Bryk, A. S. (2002). Hierarchical Linear Models. Applications and Data 
Analysis Methods, Second Edition. London: Sage. 

Rieser, S., Fauth, B., Decristan, J., Klieme, E., & Büttner, G. (2013). The connection between primary 
school students’ self-regulation in learning and perceived teaching quality. Journal of Cognitive 
Education and Psychology, 12 (2), 138–156. doi:10.1891/1945-8959.12.2.138 



Kärner&et &al &
&

 
 
 

 | F L R !
!

40!

Rösler, U., Gebele, N., Hoffmann, K., Morling, K., Müller, A., Rau, R., & Stephan, U. (2010). 
Cortisol – ein geeigneter physiologischer Indikator für Belastungen am Arbeitsplatz? [Cortisol – 
a useful physiological parameter for work-related stress?] Zeitschrift für Arbeits- und 
Organisationspsychologie, 54 (2), 68–82. doi: 10.1026/0932-4089/a000011 

Rost, D. H., Sparfeldt, J. R., & Schilling, S. R. (2007). DISK-GITTER mit SKSLF-8. Differentielles 
Schulisches Selbstkonzept-Gitter mit Skala zur Erfassung des Selbstkonzepts schulischer 
Leistungen und Fähigkeiten [Scale for assessing Self-concept of School related Achievement and 
Skills]. Göttingen: Hogrefe. 

Roubinov, D., Hagan M. J., & Luecken L. J. (2012). If at first you don't succeed: the neuroendocrine 
impact of using a range of strategies during social conflict. Anxiety, Stress, & Coping: An 
International Journal, 25 (4), 397–410. doi: 10.1080/10615806.2011.613459 

Rubin, D. B. (1987). Multiple Imputation for Nonresponse in Surveys. Hoboken, NJ: Wiley-IEEE. 
Saxbe, D. E. (2008). A field (researcher’s) guide to cortisol: tracking HPA axis functioning in 

everyday life. Health Psychology Review, 2 (2), 163–190. doi: 10.1080/17437190802530812 
Schunk, D. H. & Zimmerman, B. J. (eds.) (1994). Self-Regulation of Learning and Performance. 

Issues and Edicational Applications. Hillsdale: Lawrence Erlbaum. 
Scollon, Ch. N., Kim-Prieto, Ch., & Diener, E. (2003). Experience Sampling: Promises and Pitfalls, 

Strengths and Weaknesses. Journal of Happiness Studies, 4 (1), 5–34. doi: 
10.1023/A:1023605205115 

Seidel, T., Dalehefte, I. M., & Meyer, L. (2001). Videoanalysen – Beobachtungsschemata zur 
Erfassung von „Sicht-Strukturen“ im Physikunterricht [Video-based analysis – observation 
patterns for assessing conditions in classroom in physics education]. In M. Prenzel, R. Duit, M. 
Euler, M. Lehrke, & T. Seidel (eds.), Erhebungs- und Auswertungsverfahren des DFG-Projekts 
„Lehr-Lern-Prozesse im Physikunterricht – eine Videostudie“, (pp. 41–58). Kiel: IPN-
Materialien. 

Seifried, J. (2004). Fachdidaktische Variationen in einer selbstorganisationsoffenen Lernumgebung. 
Eine empirische Untersuchung im Rechnungswesenunterricht [Subject-didactical variations in a 
self-organized learning environment. An empirical study in accounting education]. Wiesbaden: 
Deutscher Universitäts-Verlag. 

Sembill, D. (1984). Modellgeleitete Interaktionsanalysen im Rahmen einer forschungsorientierten 
Lehrerausbildung – am Beispiel von Untersuchungen zum „Kaufvertrag“ [Modell-driven 
interaction analysis within the scope of research oriented teacher education by the example of 
the topic “purchase agreement”]. Dissertation, Georg-August-Universität Göttingen, Germany. 

Sembill, D. (1999). Selbstorganisation als Modellierungs-, Gestaltungs- und Erforschungsidee 
beruflichen Lernens [Self-organisation as idea for modelling, design, and research of vocational 
learning]. In T. Tramm, D. Sembill, F. Klauser, &  E. G. John (eds.), Professionalisierung 
kaufmännischer Berufsbildung. Festschrift zum 60. Geburtstag von Frank Achtenhagen 
[Professionalisation of commercial vocational education. Festschrift to the 60th birthday of 
Frank Achtenhagen], (pp. 146–174). Frankfurt a. M.: Peter Lang. 

Sembill, D. (2004). Abschlussbericht an die Deutsche Forschungsgemeinschaft im Rahmen des 
Schwerpunktprogramms „Lehr-Lern-Prozesse in der kaufmännischen Erstausbildung“ [Final 
report to the German Research Foundation within the scope of the priority programme 
“Teaching and Learning Processes in Business Education”]. http://www.uni-bamberg.de/file-
admin/uni/fakultaeten/sowi_lehrstuehle/wirtschaftspaedagogik/Dateien/Forschung/Forschungs-
projekte/Prozessanalysen/DFG-Abschlussbericht_sole.pdf. Accessed 10 July 2013.  

Sembill, D., Rausch, A., & Kögler, K. (2013). Non-cognitive facets of competence – Theoretical 
foundations and implications for measurement. In K. Beck & O. Zlatkin-Troitschanskaia (eds.), 
From Diagnostics to Learning Success: Proceedings in Vocational Education and Training, (pp. 
199–212). Rotterdam: Sense. 

Sembill, D., Seifried, J., & Dreyer, K. (2008). PDAs als Erhebungsinstrument in der beruflichen Lehr-
Lern-Forschung – Ein neues Wundermittel oder bewährter Standard? Eine Replik auf Henning 



Kärner&et &al &
&

 
 
 

 | F L R !
!

41!

Pätzold [PDAs as a tool for data collection in adult education research – New panacea or 
approved standard? A reply to Henning Pätzold]. Empirische Pädagogik, 22 (1), 64–77.  

Sembill, D., Wolf, K. D., Wuttke, E., & Schumacher, L. (2002). Self-Organized Learning in 
Vocational Education – Foundation, Implementation, and Evaluation. In K. Beck (ed.), Teaching-
Learning Processes in Vocational Education. Foundations of Modern Training Programms, (pp. 
267–295). Frankfurt a.M.: Peter Lang. 

Shavelson, R. J., Hubner, J. J., & Stanton, G. C. (1976). Self-Concept: Validation of Construct 
Interpretations. Review of Educational Research, 46 (3), 407–441. doi: 10.1037/0022-
3514.45.1.173 

Shedletsky, R. & Endler, N. S. (1974). Anxiety: The state-trait model and the interaction model. 
Journal of Personality, 42 (4), 511–527. doi: 10.1111/j.1467-6494.1974.tb00690.x 

Sleight, P. & Bernardi, L. (1998). Sympathovagal Balance. Circulation, 98 (23), 2640. doi: 
10.1161/01.CIR.98.23.2640 

Snow, R. (1989). Aptitude-Treatment Interaction as a framework for research on individual 
differences in learning. In P. Ackerman, R. J. Sternberg, & R. Glaser (eds.), Learning and 
Individual Differences, (pp. 13–59). New York: Freeman. 

Spangler G., Pekrun, R., Kramer, K., & Hofmann, H. (2002). Students’ Emotions, Physiological 
Reactions, and Coping in Academic Exams. Anxiety, Stress, & Coping: An International Journal, 
15 (4), 413–432. doi: 10.1080/1061580021000056555 

Spielberger, C. D. (1989). State-Trait Anxiety Inventory: Bibliography (2nd ed.). Palo Alto, CA: 
Consulting Psychologists Press. 

SPSS (2005). Linear Mixed-Effects Modeling in SPSS: An Introduction to the MIXED Procedure. 
Armonk, New York: IBM. Retrieved from http://www.spss.ch/upload/1126184451_Linear-
%20Mixed%20Effects-%20Modeling%20in%20SPSS.pdf.  

Stadler, T., Evans, Ph., Hucklebridge, F., & Clow, A. (2011). Associations between the cortisol 
awaking response and heart rate variability. Psychoneuroendocrinology, 36 (4), 454–462. doi: 
10.1016/j.psyneuen.2010.07.020 

Sweller, J. (1988). Cognitive Load During Problem Solving: Effects on Learning. Cognitive Science, 
12 (2), 257–285. doi: 10.1207/s15516709cog1202_4 

Sweller, J. (1994). Cognitive Load Theory, Learning Difficulty, and Instructional Design. Learning 
and Instruction, 4 (4), 295–312. doi: 10.1016/0959-4752(94)90003-5 

Sweller, J., Ayres, P., & Kalyuga, S. (2011). Cognitive Load Theory. New York: Springer. 
Takahashi, T., Ikeda, K., Ishikawa, M., Kitamura, N., Tsukasaki, T., Nakama, D., & Kameda, T. 

(2005). Anxiety, reactivity, and social stress-induced cortisol elevation in humans. 
Neuroendocrinology Letters, 26 (4), 351–354. 

Twisk, J. W. R. (2006). Applied Multilevel Analysis. A Practical Guide. Cambridge: Cambridge 
University Press. 

Vygotskii, L. S. (1978). Mind in Society: The Development of Higher Psychological Processes. 
Cambridge: Harvard University Press. 

Wagner, W., Göllner, R., Helmke, A., Trautwein, U., & Lüdtke, O. (2013). Construct validity of 
student perceptions of instructional quality is high, but not perfect: Dimensionality and 
generalizability of domain-independent assessments. Learning and Instruction, 28, 1–11. doi: 
10.1016/j.learninstruc.2013.03.003 

Wild, K. P. (2001). Die Optimierung von Videoanalysen durch zeitsynchrone Befragungsdaten aus 
dem Experience Sampling. In S. v. Aufschnaiter & M. Wenzel (eds.), Nutzung von Videodaten 
zur Untersuchung von Lehr-Lern-Prozessen. Aktuelle Methoden empirischer pädagogischer 
Forschung [Using video data for investigating teaching and learning processes. Current methods 
of empirical pedagogical research], (pp. 61–74). Münster: Waxmann. 

Wild, K. P. & Krapp, A. (1996). Die Qualität subjektiven Erlebens in schulischen und betrieblichen 
Lernumwelten: Untersuchungen mit der Erlebens-Stichproben-Methode [The quality of 
emotional experiences in vocational and company learning environments: Studies with the 
experience-sampling-method]. Unterrichtswissenschaft, 24 (3), 195–216. 



Kärner&et &al &
&

 
 
 

 | F L R !
!

42!

Wüst, S., Federenko, I., Hellhammer, D. H., & Kirschbaum, C. (2000). Genetic factors, perceived 
chronic stress, and the free cortisol response to awakening. Psychoneuroendocrinology, 25 (7), 
707–720. doi: 10.1016/S0306-4530(00)00021-4 

Wuttke, E. (2000). Lernstrategien im Lernprozeß. Analysemethode, Strategieeinsatz und 
Auswirkungen auf den Lernerfolg [Learning strategies in learning processes: Method of analysis, 
use of strategies, and effects on learning]. Zeitschrift für Erziehungswissenschaft, 3 (1), 97–110. 
doi: 10.1007/s11618-000-0007-6. 

Yeh, Yu-chu (2012). Aptitude-Treatment Interaction. In N. M. Seel (ed.), Encyclopedia of the 
Sciences of Learning, (pp. 295–298). New York: Springer. 

Yerkes, R. M. & Dodson, J. D. (1908). The Relation of Strength of Stimulus to Rapidity of Habit-
Formation. Journal of Comparative Neurology and Psychology, 18 (5), 459–482. doi: 
10.1002/cne.920180503. 

Zierer, K. & Seel, N. M. (2012). General Didactics and Instructional Design: eyes like twins. A 
transatlantic dialogue about similarities and differences, about the past and the future of two 
sciences of learning and teaching. Springerplus, 17, 1–15. doi: 10.1186/2193-1801-1-155.