Introduction

Molecular Representations in Science Education

To think and communicate about scientific phenomena and their underlying structures and processes, a whole range of representations (e.g. diagrams, photographs, or symbols) are used, depending on the phenomenon and the purpose of a representation (Gilbert, 2005, 2008). At the molecular level, these representations are highly abstract and include various variants, such as very reduced forms of structural and molecular formulas or different symbols (Justi et al., 2009, Fig. 1). Although representations are extensively used in science classrooms, contemporary science education often struggles to support students in developing an elaborate understanding of molecular and structural representations (Taskin & Bernholt, 2014). As a result, many students do not see these representations as having any specific utility (Goodney, 2006), let alone as valuable tools to support problem-solving (Flener et al., 2010), knowledge acquisition (Mandl & Levin, 1989), or deeper insights into the relationships and processes underlying the phenomena discussed in science class (Rau, 2017).

Consequently, the ability to engage with molecular representations and to understand and evaluate them is crucial for the understanding of (molecular) scientific phenomena and therefore an important part of scientific literacy (Gilbert & Treagust, 2009). Moreover, learning from and with representations can enhance the understanding of science concepts (Tippett, 2016). Hence, several authors suggest that the understanding and evaluation of scientific representations should be fostered in students (Kozma & Russell, 1997, 2005; Nitz et al., 2014; Pande & Chandrasekharan, 2017). Besides the importance of the ability to develop, evaluate, and modify one’s own representations in order to engage in scientific reasoning (cf. modeling; Krell et al., 2017; Passmore et al., 2014), the more basic abilities to identify, understand, analyze, and interpret a given representation are also parts of students’ representational competence (e.g., Kozma & Russell, 2005; Nitz et al., 2014). For this, students need conceptual knowledge that is relevant to the respective representation and the symbols, signs, codes, etc. that constitute it (Anderson et al., 2013; Wu & Shah, 2004).

Difficulties of Students with Molecular Representations

Although complex representations of structures and processes at the molecular level are crucial for the understanding of structures, systems, and their relationships (Cooper et al., 2010), many previous studies have demonstrated significant challenges that students encounter with molecular representations, such as structural formulas (for an overview, see Taskin & Bernholt, 2014). These “short, reduced forms of precise specialized language” (Taasoobshirazi & Glynn, 2009, p. 347) require learners to know how and what information is encoded in these representations (Ainsworth, 2006). In the case of structural formulas, the format would consist of attributes such as letters, numbers, and lines and their arrangement. These attributes are primarily symbolic in nature, i.e. they are abbreviated characters with a specific meaning defined by convention (e.g., the letter ‘H’ for the element Hydrogen). The same applies to the names of many compounds (e.g. glucose; Fig. 1a), so that the “translation” of these names into an atomic composition has to be learned by hard, whereas so-called systematic names (e.g. carbon dioxide) carry this information with them. While students usually develop a good sense of this “primary notation”, a prototypical problem area for learners is that they do not have sufficient knowledge of “secondary notation” (Ainsworth, 2006, p. 186), i.e. the necessary reading and search strategies to extract information from a representation that is only implicitly conveyed (Taskin & Bernholt, 2014). Examples of this are the hiding of the chemical symbols of carbon or hydrogen atoms in the representation of hydrocarbon molecules (cf. Figure 1c vs. 1d). Accordingly, it is a necessary demand when learning to deal with complex molecular representations to be able to interpret, to construct, and to translate between corresponding representations, e.g. to relate compound names (Fig. 1a) to its chemical formula (Fig. 1b) to different representations about the structural constitution (Fig. 1c-e; Kozma & Russel, 2005). While these processes are partly based on understanding the chemical “sign language” and, thus, the meaning of lines and letters, successful use of these complex representations are also based on chemical knowledge as they require knowledge about the entities that these “signs” represent (e.g., taking the maximum number of bonds a specific chemical element can build into account) and the implicit chemical properties that can be derived from a specific representation (e.g., whether a specific molecule is likely to be soluble in water or not). This combination of conventions (e.g. for sequence: MgO instead of OMg) and content principles (e.g. valence, electronegativity, or conservation laws; Jacob, 2001; Canac & Kermen, 2016) comprises the chemical sign language, so that complex representations are a combination of symbols and models (Taber, 2009).

Fig. 1
figure 1

Different representations of Glucose (exact: α-glucopyranose) as (a) substance name, (b) chemical formula indicating the relative composition of a Glucose molecule, (c) planar valence structural formula with explicit indication of all carbon and hydrogen atoms, (d) skeletal formula with implicit representation of the carbon atoms and hydrogen atoms bonded to carbon, and (e) stereo valence skeletal formula with a planar ring structure and spatial arrangement of the substituents indicated by wedges or lines

Full size image

Establishing these connections between information that is represented differently is a particular challenge in learning, as developing an understanding of the individual pieces of information and their connection is not necessarily a linear process. Often, making connections is based on incomplete domain knowledge or results in inappropriate approaches based on intuition or inadequate conceptions (Ainsworth, 2006; Graulich, 2015; Rau, 2017). As a result, students often perceive molecular representations as an isomorphic, iconic depiction of the phenomenon at a specific point in time, and its use is merely a literal reading of a representation’s surface features without considering syntax and semantics (Cooper et al., 2010; Kozma & Russel, 2005).

Linking a specific representation to the relevant domain knowledge requires students to identify the relevant features of the representation to inform problem-solving or inference making. In addition to inadequate selection of information, the missing of two prerequisites for understanding representations in general — conceptual knowledge and knowledge about the mode and code of a representation (Anderson et al., 2013; Wu & Shah, 2004) — have been identified as major problem areas for students’ understanding of complex symbolic-textual representations (in this case, chemical formulas; Taskin & Bernholt, 2014).

The complexity and demands of these processing steps might induce a cognitive overload in students, especially when they have low prior knowledge or limited familiarity with a particular representation. In turn, higher levels of cognitive load may restrict their capacity to process the information effectively in working memory (Seufert & Brünken, 2006). As a consequence, students may feel insecure when dealing with complex representations and tend to avoid their use. For instance, Minkley et al. (2018) observed that students exhibit a strong preference for simple, symbolic representations over complex symbolic-textual representations. Moreover, the students demonstrated better test performance and experienced lower mental load when working with tests featuring simple symbolic representations, thus confirming the difficulties students face with symbolic–textual representations.

Overall, previous findings consistently demonstrate the difficulties students encounter when dealing with complex and abstract representations (Gilbert & Treagust, 2009; Rau, 2017; Taskin & Bernholt, 2014). Numerous studies have provided evidence of these difficulties in terms of student performance and test results. Also, the importance of supporting students in handling complex and abstract representations is widely acknowledged (Rau, 2017; Rodemer et al., 2021). However, less is known about the underlying cognitive and affective processes involved in dealing with complex representations.

Stress

In his well-established psychological model of stress development, Lazarus (1966) assumes that stress arises when “demands exceed the personal and social resources [of] the individual”. In this context, stress responses are considered as “psychological, physiological, and behavioral responses to an event perceived as relevant to one’s well-being with some potential for harm or loss and requiring adaptation” (Lazarus & Folkman, 1984; Salomon, 2013). Therefore, stress occurs when an individual perceives a relevant threat along with a lack of ability to cope with it. From a physiological perspective, stress is considered as an organism’s response to a stressor in order to restore homeostasis (de Kloet et al., 2005). As a result, stress triggers adaptive mechanisms to handle the stressor. The physiological responses to a stressor involve the activation of the hormonal (hypothalamic–pituitary– adrenal) and neuronal (sympathetic–adrenal–medullary) stress axes (McEwen, 1998), which can be measured by an increased cortisol concentration or changes in various heart rate variability parameters. For instance, when a person is confronted with a complex task where the demands exceed their resources, both heart rate and the frequency-domain heart rate variability measure LF/HF ratio (ratio between low (0.04–0.15 Hz) and high-frequency-bands (0.15–0.4 Hz) increase (e.g., Isowa et al., 2006; Malik et al., 1996) since these parameters represent sympathetic activity, which heightens under stress (Laborde et al., 2017). In contrast, the time-domain heart rate variability measure rMSSD (root mean square of successive differences between heartbeats) typically decreases under stress (e.g., Malik et al., 1996) as it represents parasympathetic activity that characterizes relaxation (Laborde et al., 2017).

Regarding complex molecular representations, stress can arise due to the combination of difficulties (i.e., low resources) many students face with them (Taskin & Bernholt, 2014) coupled with their relevance for understanding molecular phenomena (Gilbert & Treagust, 2009). Consequently, students experience (relevant) demands that exceed their resources when they have to engage with complex molecular representations, which can lead to stress responses.

Concerning the influence that stress can have on cognitive functions, this can lead to further difficulties, as it follows an inverse U-function, where cognitive improvements occur only at moderate stress levels (Hanoch & Vitouch, 2004; Sapolsky, 2015; Yerkes & Dodson, 1908). Furthermore, the effect of stress on cognitive functions depends on the specific function in question: Stress results in a deterioration of memory retrieval (Roozendaal, 2002; Wingenfeld & Wolf, 2014) as well as in an improvement of memory consolidation (Diamond et al., 2007; Joëls et al., 2006; Roozendaal et al., 2009; Roozendaal & McGaugh, 2011). For example, higher cortisol concentrations are associated with poorer information processing as they hinder memory retrieval (Wolf, 2006), which can result in decreased performance. In this context, supporting students in processing molecular representations could have a balancing effect by helping to understand them and by mitigating stress reactions which, in turn, could prevent impairment of memory retrieval.

Self-Efficacy

Self-efficacy is a personal resource described as one’s belief in being able to cope with situational demands (Bandura, 1989) which describes the subjective certainty of a person to cope with a situation based on their own competencies. Bandura (1977) identifies four sources for the acquisition of self-efficacy, ranked by their potency: personal experiences, vicarious experiences, persuasion, and physiological reactions. In the context of this study, self-efficacy regarding complex molecular representations can be seen as the belief in one’s capabilities to understand and deal with molecular representation and can result e.g. from one’s own experiences with them and also from the perception of stress responses, when confronted with them.

Many studies suggest that high self-efficacy is associated with lower stress-related physiological responses (e.g., Minkley et al. 2014) and better performance in stress-related situations (for an overview, see Schönfeld et al., 2017; Talsma et al., 2018). Furthermore, there is clear evidence that high self-efficacy is positively related to challenge and better performance but negatively related to threat (e.g., Chemers et al., 2001; Putwain et al., 2014; Putwain & Symes, 2014). Therefore, self-efficacy should always be encouraged whenever possible.

Regarding complex molecular representations, which are challenging yet frequently used in science lessons, students repeatedly face demands that exceed their capabilities, which negatively affects their self-efficacy and may increase stress responses, further impacting their self-efficacy negatively. As a result, learning may become increasingly difficult, and students may avoid tasks involving complex molecular representations (Minkley et al., 2018). However, this negative correlation could be (partially) offset and interrupted through appropriate support formats. Thus, in the present study we investigate whether even short, very simple clues (which can be applied much more easily than clues that promote comprehensive and in-depth understanding of molecular representations) could lead to an improvement in self-efficacy, which in turn would cause the participants to perceive the tasks more as a challenge than as a threat and thus mitigate stress reactions.

Assistance in Task Processing

A comprehensive way to support students in understanding and working on tasks is scaffolding. Scaffolding is understood as didactic support that helps learners accomplish tasks they cannot do on their own (Wu & Pedersen, 2011). It assists learners in moving, step by step, from their current level of development to the level of development that is potentially possible for them, known as the zone of proximal development (Vygotsky, 1978). While knowledge and skills determine the actual development that someone can attain without assistance, potential development refers to the knowledge and skills that someone can attain with assistance. Thus, scaffolding is a process that reduces the probability of failure in a task a learner is attempting to accomplish (Maybin et al., 1992).

In the context of complex molecular representations, students need to identify relevant features (e.g., specific symbols or lines), link the selected information to the represented referent (e.g., specific chemical elements or bonds between atoms), and integrate this internal representation with their conceptual knowledge (e.g., deriving chemical properties of the displayed substance; Ainsworth, 2006). While experts can extract and process all this information fluently, students often struggle when dealing with complex molecular representations, but can be supported by making the necessary connections explicit (Rau, 2020). The use of specific highlighting techniques (Rodemer et al., 2021), prompts (Broman et al., 2018), or additional explanations (Caspari & Graulich, 2019) can guide the learner to focus on relevant parts of the representation or process the representation as intended.

Nevertheless, scaffolding by a tutor (e.g., Collins et al., 1989; Livengood et al., 2012) is a time- and resource-intensive process that may not always be feasible in classroom settings or for individual tasks. Even when written materials are used as scaffolds, the amount and complexity of conceptual information and processing demands should be carefully considered. In consequence, rather short and focused clues that help students “see something as requiring attention and decision making that they might otherwise overlook” (Reiser, 2004, p. 287) can balance the benefits and processing costs of providing additional support in dealing with complex tasks (Broman et al., 2018; Hermanns, 2020). Unlike simplifying task demands (e.g., using simplified representations or avoiding them altogether), the use of written clues empowers students to engage with complex tasks or those they find too difficult to complete. While the benefits of using written clues are well documented in supporting students’ problem-solving (Broman et al., 2018; Reiser, 2004), its effects on affective processes are less well researched.

Research Questions

In the present study, we aim to determine if and to what extent short text-based clues influence (1) performance/understanding, (2) self-efficacy, and (3) physiological and psychological stress responses when dealing with complex molecular representations.

The following research questions (RQ) were addressed in this study:

  1. RQ1:

    Do short text-based clues support students’ performance on tasks dealing with complex molecular representations?

  2. RQ2:

    Do short text-based clues support the self-efficacy related to the tasks they address?

  3. RQ3:

    Do text-based clues mitigate the stress response (psychologically and physiologically) during the processing of related tasks?

Methods

Design and Procedure

The data were collected during a one-day project on molecular biology at the teaching and learning laboratory of a university in Germany. The students participated in a one-day project focused on molecular biology techniques, which consisted of instructional and practical phases involving hands-on experiments, such as agarose gel preparation and loading. Upon arrival, the participants were welcomed and briefly introduced to the procedure of our experiment, which was conducted in accordance with the Declaration of Helsinki and approved by the local Ethics Commission. Pulse watches (V800, Polar) with corresponding chest belts (H7, Polar) were then distributed to the participants, and their correct use and adjustment were explained and approved.

The molecular biology project, which was not associated with our experiment, began thereafter. Data collection was carried out in two parts: the prior knowledge test, aimed at assessing prior knowledge regarding molecular representations, was administered in the morning, while the main test took place after the lunch break to separate it temporally from the prior knowledge test. Both tests occurred in a seminar room, where each participant was seated in front of a laptop separated from others by a screen. It was explained that the students were to answer a few questions to assess their prior knowledge. It was explicitly mentioned that this test was designed solely to gauge their existing knowledge to minimize stress levels during the prior knowledge test.

Before the prior knowledge test began, participants were asked to provide socio-demographic data (e.g., age, frequent medication, BMI), which were relevant for interpreting HRV data. Subsequently, students’ self-efficacy regarding molecular representations in the form of chemical formulas was assessed, and their subjective perception of stress was measured. Following that, the tasks for the prior knowledge test regarding molecular representations had to be completed. Finally, subjective stress perception was queried again, and the students left for their lunch break. At the beginning of the main test, students were once again asked about their subjective stress perception. This was followed by the main test comprising tasks with molecular representations. Students were randomly assigned to one of two groups, by being seated in front of laptops with different test versions installed (with and without clues). Both groups, ‘clues’ and ‘control’, (i.e., without clues), received identical tasks, with the only difference being that the ‘clues’ group was provided with clues for the respective task types. After the main test, students were again asked about their subjective stress perception and their self-efficacy regarding molecular representations. Additionally, participants in the ‘with clues’ group were asked to evaluate their perception of those clues. Finally, the pulse watches were stopped.

Participants

The participants (N = 136) were high school students from nine secondary schools in Germany, with a mean age of 17.7 years (SD = 1.6). All of them, or their parents respectively, gave written informed consent and were informed that participation in the study was voluntary and anonymous.

Of the participants, 57 were male, 77 were female, and 2 identified as diverse. Furthermore, 52 students indicated that they currently participate in chemistry classes at school. Some of the participants had to be excluded from further HRV data analysis: 28 due to erroneous measurements and an additional 15 due to a significantly elevated body mass index (BMI > 30) or frequent consumption of tobacco products. Thus, HRV data from 93 students were used for further analysis.

Tasks and Task Performance

In the prior knowledge test, which aims to identify students’ prior knowledge of molecular representations, seven tasks were designed to cover a broad spectrum of understanding and handling of complex molecular representations. Care was taken to design tasks that test pure subject knowledge as well as tasks that assess understanding of complex molecular representations. For example, students had to specify which chemical element is represented by a specific symbol or vice versa (e.g., N represents Nitrogen; subject knowledge). To test their understanding of molecular representations, the students were required to determine the number of hydrogen and carbon atoms in a given structural formula (such as 3-methyl-pent-2-ene) or identify which of four given chemical formulas corresponded to a specific structural formula (for an overview of the tasks see the electronic supplemental material).

For the main test, six tasks were designed to assess the understanding of complex molecular representations (for an overview of the tasks and clues see the electronic supplemental material). The difficulty of the tasks increases throughout the test, with the first task being the easiest and the last tasks being more difficult. In the first task, the students had to state how many hydrogen atoms are present in five compounds represented by chemical formulas. In the subsequent tasks, e.g. they had to indicate which structural formula corresponds to a given chemical formula, and vice versa. In the final tasks, the students had to decide which of the three shown structures can exist or determine the properties (nonpolar, basic, acidic) of three amino acids represented as structural formulas. Thus, in the prior knowledge test and also in the main test, on the one hand the pure knowledge in connection with molecular representations was assessed (e.g. tasks on the meaning of certain element symbols) and on the other hand their understanding (e.g. tasks on translating between specific representations of the same compound) in order to be able to eliminate prior knowledge differences between the two groups (clues vs. no clues), which would influence a possible effect of the clues.

As mentioned earlier, the students were randomly assigned to either the treatment group (clues) or the control group. Both groups received the same tasks, with the only difference being that the ‘clues’ group was provided with short written clues for the respective task types. In both the prior knowledge test and the main test, the molecular representations used can be classified as symbolic-mathematical with some verbal-textual elements (e.g., letters). Symbolic-mathematical representations “refer to schematized or iconic diagrams that apply signs, symbols, and mathematical conventions to depict entities, concepts, or processes” (Wu & Puntambekar, 2012, p. 756). They can be highly abstract, without any structural similarity to their target (e.g., mathematical formulas), or they can share structural features with the target, and also include some verbal-textual elements (e.g., structural formulas; Schnotz, 2001, Fig. 1c-f), as the ones we used in our test. These representations can be classified as more complex compared to purely symbolic-mathematical representations, which involve processing only abstract symbols without verbal-textual elements (Minkley et al., 2018).

For all tasks in both tests, performance expectations were prepared to assess performance. During the task development, they were discussed and revised by all three authors in several rounds. Points were awarded as follows: In the prior knowledge test, a total of 16 points could be achieved, with eight points assigned to the area of knowledge and eight points to the area of understanding of molecular representations. All correct answers were evaluated with one point each, and no partial points were awarded in the evaluation. The internal consistency of the scale was assessed using Cronbach’s Alpha. The result indicated acceptable to good internal consistency (α = 0.80). In the main test, one point was also awarded for each correct answer (without distribution of partial points), allowing a maximum of 12 points to be achieved. The internal consistency of this scale was acceptable (α = 0.76). After the points have been awarded according to this scheme, a task performance score was calculated by determining the percentage of achieved points relative to the maximum points. Therefore, the possible scores ranged from 0 to 100. For the total sample, the mean score in the main test was Mperformance = 56.53 (SD = 22.04). To assess the influence of the clues on performance independently of the sample, we then calculated the increase in performance by subtracting the percentage achieved in the prior knowledge test from the percentage achieved in the main test.

Clues

In designing the clues, we ensured that they assist in understanding complex molecular representations. Depending on the task, information was provided to enhance general understanding, highlight important aspects to consider, or guide a systematic approach. The clues were intended to provide students with coping strategies specific to complex molecular representations. Furthermore, we explicitly ensured that the clues are short and easy to understand, making them suitable for easy implementation in a classroom setting.

One particular task will be examined in more detail as an example: In this task, the students were required to determine the number of hydrogen atoms in five given chemical formulas (e.g., C2H5OH = 6 H atoms). The clues provided explanations for which symbol represents which chemical element (e.g., H = hydrogen) and how the indices following each element letter should be interpreted. Additionally, an example (H2SO4) was presented and used to illustrate the exact systematic procedure.

Measurements

Heart Rate (HR) and Heart Rate Variability (HRV)

The participants’ heart rate and heart rate variability were measured continuously during the prior knowledge test and main test using a chest belt with an ECG sensor (H7, Polar) and a storage device (V800 pulse watch, Polar) that wirelessly recorded the data. After the measurements, the data was transmitted to the R software to calculate several HRV parameters. These included the root mean square of successive differences (rMSSD), which is a common time domain measure of HRV that reflects parasympathetic activity (Hjortskov et al., 2004; Malik et al., 1996). Additionally, we calculated the LF/HF ratio as a frequency domain measure, which reflects the ratio between sympathetic activity (low frequency components; 0.04–0.15 Hz) and parasympathetic activity (high frequency components; 0.15–0.4 Hz) of the nervous system. The mean scores of HRV during the main test for the total sample were as follows: Mheart rate = 87.10 (SD = 12.46), MrMSSD = 56.24 (SD = 45.00), and MLF/HF ratio = 3.01 (SD = 1.66), respectively.

Perceived Stress

The subjective perception of stress by the participants was assessed using a visual analogue scale (Luria, 1975), adapted for stress measurement (Minkley et al., 2021). This scale consists of a 100 mm-long horizontal line labeled with ‘no stress’ on the left end and ‘maximum stress’ on the right end. Participants were asked to mark a point on the line that represents how stressed they felt at that moment. The level of subjective stress was determined by measuring the distance between the left end of the line and the participant’s marking, resulting in values ranging from 0 to 100 [mm].

Self-Efficacy

In order to measure self-efficacy regarding molecular representations, five items have been included in the questionnaire (e.g., I can solve challenging tasks involving chemical formula representations in class if I make an effort.). These items were derived from Althoff (2021), which, in turn, adapted them from instruments used to measure school-related self-efficacy (Jerusalem & Satow, 1999) and general self-efficacy (Jerusalem & Schwarzer, 1999). The students were asked to rate their agreement with the five statements on a four-point scale ranging from ‘strongly disagree’ to ‘strongly agree’.

Data Analysis

The software SPSS Statistics (Version 27.0, IBM) was utilized for data analysis. The data output of the Polar chest belt (in form of the sequence of heart beats) was matched individually to the two test periods (prior knowledge and main test) and processed separately by means of the R package RHRV (Rodriguez-Linares et al., 2011) and following the guidelines suggested by Martínez et al. (2017). First, the sequence of beats was used to generate the corresponding heart rate variability signal. In this process, adaptive thresholding is used to remove artifacts, i.e. data points that are not within acceptable physiological values (Vila et al., 1997). To deal with interruptions in the signal during data collection, an interpolated heart rate series is calculated based on linear interpolation with a sampling frequency of 4 Hz. The interpolated data is then analyzed to obtain individual rMSSD and LF/HF values for the two test periods. If there were too many measurement errors or interruptions, the data of that participant were excluded from further analysis.

Group differences regarding the performance, HRV, perceived stress, and self-efficacy data were analyzed using a repeated-measures ANOVA, with time as a repeated within-subjects factor (prior knowledge test vs. main test) and treatment (with clues vs. without clues) as the independent variable. In case of violated sphericity assumption a Greenhouse-Geisser correction was performed and adjusted p-values were reported. The effect size was measured using partial η2 and interpreted according to Cohen (1988): η2 ≥ 0.01 = small effect, η2 ≥ 0.06 = medium effect, η2 ≥ 0.14 = large effect.

Results

The short, text-based clues had a significant positive influence on performance. When prior knowledge was considered, there was a significant difference in the main test performance between both group. Students who received clues (n = 67) performed significantly better on the main test compared to the prior knowledge test, than those without them (n = 67; (F (1, 132) = 16.84, p < .001, ŋ2 = 0.113; Fig. 2). However, regardless of the group, the students performed significantly better in the main test than in the prior knowledge test (F (1, 132) = 117, p < .001, ŋ2 = 0.470). Furthermore, the majority of the students (71%) agreed or tended to agree that the clues were useful.

Fig. 2
figure 2

Change in performance from prior knowledge test to main test (∆ in percentage; n = 67; *** = p < .001)

Full size image

However, the short written clues had no effect on self-efficacy regarding dealing with complex molecular representations. In both groups, self-efficacy was in the medium range, around 2 (out of 4) but decreased significantly from around 2.5 before the two tests to around 1.9 afterwards (F (1, 130) = 97.19, p < .001, ŋ2 = 0.428; Fig. 3), with no interaction effect.

Fig. 3
figure 3

Change in self-efficacy from prior knowledge test to main test (Mean ± SE; nclues = 66, ncontrol = 66; *** = p < .001)

Full size image

Additionally, the different heart rate variability parameters (HR, LF/HF ratio, rMSSD) were similar between both groups during the test (Fig. 4).

Fig. 4
figure 4

Heart rate variability parameters during the main test (Mean ± SE; nclues = 58, ncontrol = 61)

Full size image

However, when the HRV data from the prior knowledge test were included, there was a significant interaction effect between the type of test (prior knowledge test vs. main test) and the treatment on the students’ rMSSD (control vs. clues; F (1; 117) = 6.297, p = .013, ŋ2 = 0.051; Fig. 5). Students who did not receive the clues experienced a significant decrease in their rMSSD from the prior knowledge test to the main test, indicating an increase in stress. In contrast, for those students who received the clues, the rMSSD level remained nearly the same during both tests, indicating a consistent stress level. However, there was no significant interaction effect for the other heart rate measurements (HR and LF/HF ratio).

Fig. 5
figure 5

Change in rMSSD from prior knowledge test to main test (Mean ± SE; nclues = 58, ncontrol = 61; * = p < .05)

Full size image

Subjective stress perception significantly decreased from the prior knowledge test to the main test in both groups (F (1, 132) = 20.207, p < .001, ŋ2 = 0.133; Fig. 6), with no significant interaction effect. However, there was a tendency for the students who received clues to perceive lower subjective stress during the main test compared to those who did not receive clues (F (1, 132) = 3.579, p = .061, ŋ2 = 0.026).

Fig. 6
figure 6

Change in subjective stress perception [mm] from prior knowledge test to main test (Mean ± SE; nclues = 66, ncontrol = 68; *** = p < .001)

Full size image

Discussion

The short written clues have a positive influence on test performance. They appear to assist students in correctly solving tasks related to complex molecular representations, regardless of their prior knowledge. This is remarkable in two aspects: First, the tasks used in this study cover a broad range of different affordances related to dealing with complex molecular representations. Students had to interpret, construct, and translate between different types of representations and to also link these representations to specific pieces of biological and chemical knowledge. Second, the clues implemented in the study were actually very short and concise and do not require any additional introduction but can simply be given with the tasks. This might be particularly relevant in biology classes, where the ability to read and understand complex molecular representations is assumed as a prerequisite, unlike in chemistry classes where it is explicitly introduced. However, further investigation is needed to determine the extent to which this effect persists across different tasks (e.g., focusing on specific problem types or specific requirements of dealing with complex representations) and across different types of clues (e.g., of different lengths or containing information at different levels of detail).

In contrast to our hypothesis, the clues have no influence on self-efficacy with respect to dealing with complex molecular representations. However, this is not entirely surprising, since self-efficacy is a relatively stable construct that cannot be easily changed in the short term (Bandura, 1997). It is, however, conceivable that recurring clues may indeed have an influence on self-efficacy, as students achieve better performance on tasks involving complex molecular representations through them, leading to a sense of accomplishment in dealing with such tasks, which in turn is an important source of self-efficacy (Bandura, 1977). As students did not receive immediate feedback on their performance on the tasks in the main test, they might have been unaware of their successful solving of tasks and, thus, this chain of effects may not have come into play. Unexpectedly, self-efficacy is lower in the main test, compared to the prior knowledge test in both groups. Since self-efficacy describes an individual’s subjective belief in their ability to handle a situation based on their own competencies, direct confrontation with that situation seems to influence the assessment. Thus, the decreased self-efficacy after the two tests may be attributed to the fact that our participants realized during the tests that they were less competent in dealing with molecular representations than they had previously assumed.

The perception of stress during the processing of tasks in the main test was lower compared to that during the prior knowledge test tasks in both groups. However, it tended to be lower in the students who received the clues compared to the students in the control group without clues. The overall decrease in stress perception between the prior knowledge test and the main test could be attributed to a habituation effect: In the first (pre-) test, the students had already become familiar with the type of tasks involved, which might reduce the anticipatory stress typically experienced before tests in the main test.

Regarding the various heart rate parameters, the results are divergent: The values during the main test are comparable in both groups. However, when examining the change from the prior knowledge test to the main test, the rMSSD in the treatment group remains relatively stable, while it significantly decreases in the control group, indicating an increase in stress. Consequently, students who do not receive any help experience higher stress levels in the main test compared to the prior knowledge test, while those who receive the clues maintain a similar stress level.

This difference between psychological and physiological stress responses can be interpreted in several ways. It is conceivable that a habituation effect may have played a role. Since our study design involved comparable prior knowledge test and main test tasks, students may have become accustomed to the main test, resulting in a reduction of perceived stress (specifically, the stress factor ‘uncertainty about the threat and one’s resources’ decreased from the prior knowledge test to the main test). However, the physiological results indicate a different trend. Among students who did not receive the clues, the rMSSD (a measure of relaxation) decreased, causing increased stress levels from the prior knowledge test to the main test. It appears that the prior knowledge test (which was announced as such) was not taken as seriously as the main test (the stress factor ‘relevance of a threat’ increased from the prior knowledge test to the main test). These divergent results highlight the inconsistency between the assessment of psychological stress response (subjective perception of stress) and physiological stress responses (e.g. heart rate variability). Previous studies have already indicated that physiological responses to stress, such as decreased heart rate variability and increased cortisol release (Rensing et al., 2006), are often not systematically related to an individual’s perception of stress (for an overview, see Campbell & Ehlert, 2012; Minkley et al., 2021). Campbell and Ehlert (2012) discuss various factors, including different measurement protocols, mediating factors, and interindividual differences in the degree of correspondence between physiological and psychological responses as possible reasons for this inconsistency. In addition to these more formal reasons, they also suggest the concept that physiological and psychological stress responses may represent different aspects of the overall stress response.

In summary, our study demonstrates that providing short written clues improves test performance but has no immediate effect on self-efficacy — at least within our experimental setting. Furthermore, the presence of clues significantly mitigated the increase in stress levels from the prior knowledge test to the main test, although there were no significant differences in physiological stress responses during the main test, between the students who received the clues and those who did not. These findings highlight the benefits that short written clues can have as a supportive tool in educational contexts.

Limitations and Implications

There are some limitations to consider. Firstly, we only administered a relatively small number of tasks and clues, and therefore, we did not systematically test the effects of different types of clues for different types of tasks. Since the compilation of the tests was intended in such a way that the selected tasks address several of the problem areas of students frequently mentioned in the literature, the analysis had to be limited to the test level. If the test were to focus on specific problems or problem types, a more differentiated analysis of the interplay between self-efficacy expectations and stress experience would be possible and would thus open up the possibility of a situational perspective. Therefore, it would be valuable in future studies to examine the effect using different types of tasks and various forms of clues.

Secondly, due to the requirements of the ethical commission, we had to clearly inform the participants that the data would be collected anonymously. This likely reduces the experience of stress and physiological stress responses, as the stress factor ‘relevance of the task’ is automatically diminished when it is clear that the results will have no consequences and cannot be attributed to the individual. As a consequence, the findings reported here might not be easily transferable to real classroom and/or high-stakes test situations, e.g., exams.

Nevertheless, our results demonstrate the advantages that short written clues can have. They can be easily integrated into teaching or different tasks, which is particularly relevant in biology classes where there is little to no time allocated for systematically introducing the understanding of complex molecular representations.