In this line of research, we view environmental attitude as a personal propensity to act in an environmentally protective way (Kaiser et al., 2010; Kaiser et al., 2007) or as people’s commitment to environmental protection. In this view, people’s attitudes represent a distinctly motivational concept (Allport, 1933). It is thereby unsurprising that there is a wealth of evidence, including behavioral observation, that environmental attitude is a significant predictor of manifest “behavior that harms the environment as little as possible” (Steg and Vlek, 2009, p. 309). For example, environmental attitude was found to be a significant predictor of sustainable travel behavior (e.g., Kaiser and Lange, 2021; Taube et al., 2018), energy consumption in a common’s dilemma (Kaiser and Byrka, 2015), green consumerism (Taube and Vetter, 2019), and vegetarian dish choices (Kaiser et al., 2020).

By contrast, attitude toward nature is viewed as a propensity that manifests in people’s appreciation for the natural environment (see Milfont and Duckitt, 2004). Accordingly, attitude toward nature can be derived from how people enjoy nature and use natural environments, such as visiting beautiful landscapes, engaging in outdoor sports, and other contemplative or restorative activities (e.g., Brügger et al., 2011; Tam, 2013). The particular attitude is more narrowly specified as an anthropocentric environmental attitude (and contrasted with an ecocentric attitude; see Thompson and Barton, 1994). More recently, the concept we refer to as attitude toward nature has been identified under various other names in the literature: for example, emotional affiliation with nature (Mayer and Frantz, 2004), personal connection to nature (Hinds and Sparks, 2008), relatedness with nature (Zelenski and Nisbet, 2014), environmental identity (Clayton, 2003), affinity toward nature (Kals et al., 1999), or people’s inclusion of nature in their self-concept (Schultz, 2002). Irrespective of the different names and conceptions, Brügger et al. (2011; see also Tam, 2013) corroborated a close convergence of measures of several of these concepts and their relevance for environmental protection. In this line of research, engagement in protecting the environment has typically been captured via self-reports (see, e.g., Clayton, 2003; Kals et al., 1999; Markowitz et al., 2012; Nisbet et al., 2008) but also via behavioral observations (see Whitburn et al., 2019) or observations of the consequences of behavior (i.e., energy consumption; see Frantz and Mayer, 2014).

In their research, Kaiser et al. (2013) were additionally able to distinguish attitude toward nature (i.e., the propensity to enjoy and use nature) from environmental attitude (i.e., the propensity to protect the environment). Simultaneously, they identified a strong positive relationship between the two attitudes of about r = 0.50 when corrected for measurement error attenuation (see also Kaiser et al., 2014). Such positive correlations have been reported reliably, although not always to this magnitude (see, e.g., Baldi et al., 2021; Davis et al., 2009; DeVille et al., 2021; Dutcher et al., 2007; Geng et al., 2015; Mackay and Schmitt, 2019).

Theoretically, it is typically expected that the personal benefits that come from using nature for restorative, recreational, spiritual, and other purposes create a dependency on natural environments to maintain the corresponding activities. Such a dependency expectedly renders people more concerned about and likely to put effort into environmental protection. In other words, the use and enjoyment of nature may, in fact, represent genuine reasons for people’s propensity to protect the environment. Such a view logically involves mediation: that is, the relevance of attitude toward nature for protecting the environment is transmitted via environmental attitude (Whitburn et al., 2020). Thus, strengthening people’s attitude toward nature by having them engage with and experience nature might present an avenue for additionally fostering their environmental attitude (e.g., Chawla, 1997; Dutcher et al., 2007; Ewert et al., 2005).

This line of primarily correlational research has yielded inconclusive results on the directionality of the relationship between these two attitudes and in a lack of evidence that substantiates a causal link between them (e.g., Kaiser et al., 2014; Whitburn et al., 2019). In addition, some studies have prompted skepticism about the mediating effects of environmental attitude, as attitude toward nature has been found to have a stronger effect on behavior than environmental attitude (see, e.g., Davis et al., 2009; Markowitz et al., 2012; Nisbet et al., 2008; Schultz, 2001: Study 3; Stets and Biga, 2003).

Schools serve a vital function in environmental education, as they educate and reach large numbers of students, particularly during formative periods of development (e.g., Otto et al., 2019). Schools provide opportunities for students to learn about the intricacies of environmental problems and to develop the competencies to combat these problems. Expectedly, people will consistently engage in environmental protection and, for example, avoid plastics only if they know about the detrimental consequences of plastic pollution and how to avoid plastics in their daily lives. Thus, it is crucial for students to acquire facts and develop competencies in order to learn how to protect the environment.

Previous research has demonstrated that environmental attitude’s supportive role in learning about environmental issues manifests in two distinct ways (Baierl et al., 2022a): (a) People’s environmental attitudes and the costs of a specific learning activity (e.g., reading a text with new information) jointly affect learning (e.g., the likelihood that a text will be read); and (b) environmental attitude additionally affects the rigor with which people learn. By doing so, environmental attitude affects the intensity with which students seize their learning opportunities, subsequently determining the amount of knowledge they gain and later retain. Thus, we propose that individuals with stronger environmental attitudes not only seize more learning opportunities but also engage more rigorously in learning (see Figure 1). Consequently, the varied levels of environmental attitude determine the diverse learning outcomes typically found in classrooms.

In line with Dutcher et al. (2007), we expect that developing an appreciation for nature and, consequently, a propensity to use and enjoy nature fuels the propensity to protect the environment (and not the reverse). In other words, environmental attitude mediates the effects of attitude toward nature on behavior and on the consequences of behavior (e.g., people’s CO₂ emissions or the amount of energy they consume). We argue that knowledge is the behavioral consequence that is easiest to measure. The more rigorously students seize learning opportunities, the more knowledge about environmental issues students gain and retain (see Figure 1).

By extending Baierl et al.’s (2022a) model on the mediating role of environmental attitude in the effect of attitude toward nature on learning about environmental issues (see Figure 1), we address an unanswered question about causality. The objective of this research is to make a modest contribution to addressing this question by testing whether environmental attitude mediates the effect of attitude toward nature on the consequences of learning: that is, the knowledge that people eventually acquire.

On the basis of the logic of a conventional mediation test (see Baron and Kenny, 1986; Kenny et al., 1998), we first explore whether those who are more committed to using nature—people with stronger attitudes toward nature—also know more about environmental issues (Path c in Figure 2). With a view to the mediating role of environmental attitude, we expect that attitude toward nature affects environmental attitude (Path a in Figure 2) and that environmental attitude in return affects knowledge about environmental issues (Path b in Figure 2). Once the two mediating Paths a and b are added to the analysis, the strength of the effect of attitude toward nature on knowledge about environmental issues becomes nonsignificant (or at least significantly reduced: Path c’ is substantially smaller than Path c in Figure 2).

As the prototypical attitude toward nature measure was initially developed for adults (see, e.g., Brügger et al., 2011; Kaiser et al., 2013), we additionally explored whether the indicators that have been used so far would also be suitable for adolescents and young adults—a target group that has been relatively neglected in the study of attitude toward nature and its measurement (e.g., Barrera-Hernández et al., 2020; Cheng and Monroe, 2012).

We used data from 1,486 students (M_age = 15.25, SD = 3.2) to address our research questions. Most were middle and high school students [n = 1,309; M_age = 14.44, SD = 1.82; 646 (49.4%) female and 592 (45.2%) male adolescents; 51 provided no information on gender], whereas 166 were university students [M_age = 22.29, SD = 3.23; 72 (43.4%) women and 89 (53.6%) men; five provided no information on gender]. For the total cohort, gender was almost evenly distributed, with 48.5% (n = 721) female and 46.2% (n = 686) male students.

The two attitude measures (i.e., environmental attitude and attitude toward nature) were administered to a total of 1,104 students [M_age = 14.43, SD = 2.19; 555 female (50.3%), 485 male (44.8%), and 42 who did not disclose a binary gender (3.8%)]. The remaining 382 students (M_age = 17.87, SD = 3.20; 43.5% female, 52.6% male, and 3.9% who did not disclose a binary gender) additionally participated in an educational intervention about the forest ecosystem and related sustainability issues. These students additionally received a test of environmental knowledge before and 6 weeks after the intervention to determine their gains in knowledge.

The 220-min educational intervention was designed as a student-centered, collaborative escape game in a typical classroom setting. With the goal of saving the forest, the students had to master the tasks embedded in nine workstations. The tasks were designed to help students learn facts about the forest ecosystem (e.g., flora and fauna biodiversity, succession of plant communities, and ecosystem services), sustainable behaviors to protect and conserve forests (e.g., recycling and considering eco-labels), and the environmental effects of land use and forest management strategies (e.g., deforestation and tree cultivation). For additional information regarding the intervention, see Fleissner-Martin and Bogner (2024).

The data were collected in schools and universities in Bavaria, Germany, in 2022 during regular lessons. The Bavarian Ministry of Education granted ethics approval for the survey. All principals of the participating schools were duly informed, and legal guardians provided written consent for all students below the age of 18 to take part in the study.

We measured environmental attitude, attitude toward nature, and environmental knowledge. Whereas the two attitude measures were administered to all participants (N = 1,486), the knowledge test was administered exclusively to participants in the educational intervention (n = 382). No items were excluded for any of the specific scale calibrations.

Environmental attitude reflects people’s propensity to protect the environment. The specific measure we employed is a variant, designed for adolescents (see Kaiser et al., 2007), of a well-established instrument for adults (see Kaiser and Wilson, 2004). This variant consists of 55 items: 15 opinions about environmental preservation, such as “We must set aside areas to protect endangered species,” and 40 self-reports of environmentally protective behaviors, such as “I collect and recycle used paper” (for more details, see Baierl et al., 2022a).

Participants answered the 55 polytomous items using either a 5-point frequency format ranging from 1 (never) to 5 (always) or a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). Of the 55 items, 22 were negatively formulated. All these negatively formulated items, which represent a lack of the propensity to protect the environment, were reverse-coded before we calibrated the Rasch scale.

The measurement instrument was calibrated using a dichotomous Rasch model (for more details, see Rasch, 1960; for recent accounts, see, e.g., Bond et al., 2021; Wilson, 2023). Therefore, the items had to be converted into a dichotomous format. Prior to calibration, it is recommended that the absolute number of response options be reduced to prevent unreliable measurement due to excessive measurement error (for details and supporting evidence, see Kaiser and Lange, 2021). This dichotomization is distinct from the untenable dichotomization of person scores after establishing an attitude scale as a continuous measure (DeCoster et al., 2009). For the dichotomization, the responses never, seldom, occasionally, strongly disagree, disagree, and not sure/neutral were coded as expressions of a lack of the propensity to protect the environment (0). The responses often, always, agree, and strongly agree were coded as expressions of the propensity to protect the environment (1).

Due to limitations in in-class time, approximately half of the sample (n = 718) were presented with only 25 of the 55 items that would have been available for measuring environmental attitude in principle. Therefore, 718 participants responded to fewer items by design. Such a reduction in items can negatively affect an instrument’s reliability by generally inflating the standard error of measurement. Nevertheless, the separation reliability indicated that the scale was effective in differentiating between students (rel = 0.73).

Similarly, the fit values reflecting the discrepancy between the Rasch model’s predicted and the observed responses were also encouraging. Specifically, mean square (MS_w) values weighted by the item variance fell between 0.84 and 1.21 (for more details, see, e.g., Wright and Masters, 1982). An MS_w value of 1.21 indicates 21% more response variation than the model predicted. It was reasonable to conclude that the Rasch-model-based expectations provided an adequate fit to the actual responses (for reference values and their justification, see Wright et al., 1994) and that the calibration of the scale was in line with previous such calibrations (e.g., Baierl et al., 2022a; Kaiser et al., 2007).

Environmental attitude measurements ranged from −4.07 to 4.40 logits (M = −0.50, SD = 0.91). Logits represent the natural logarithm of the ratio between environmentally protective and nonprotective answers such that a more positive logit value indicates a stronger attitude.

Attitude toward nature was measured with 21 self-reports of uses of nature for restorative and recreational experiences, such as “I take time to watch the clouds pass by,” and with 20 opinions expressing appreciation for nature, such as “I like the quiet of nature” (for the complete list of items, see Supplementary Table S4). Whereas evaluative statements are typical indicators of attitudes, they are often susceptible to some people’s socially desirable responses and to excessively high approval ratings (i.e., ceiling effects). By contrast, self-reported behaviors focus on past performance and are typically more difficult to endorse. People’s attitudes can also be inferred from records of past behavior because individuals often engage in behavior with intentions (to achieve specific goals). For example, a person might watch stars while going outside to smoke or, alternatively, to contemplate the night sky. By collecting behavioral self-reports that capture the extents to which people endorse particular goals (e.g., using and enjoying nature), it is thereby also possible to derive their corresponding attitudes.

Most of the items originated from Brügger et al. (2011) and, thus, from the adult version of the proposed measure. Four evaluative statements originated from Bogner and Wiseman (2006). Of the 41 items, 8 were adapted for broader use. For example, “As a child, I spent time in the woods” was reformulated as “As a child, I spent time outdoors.”

As with the environmental attitude items, participants answered the 20 polytomous items using either a 5-point frequency scale (i = 11) ranging from 1 (never) to 5 (always) or a 5-point Likert scale (j = 9) ranging from 1 (strongly disagree) to 5 (strongly agree). Each of the 21 dichotomous items was answered in a yes-no format. Participants could answer with not applicable if they could not provide an answer. Those answers were coded as missing values. Three items were negatively formulated and reverse-coded before the analysis. Once again, polytomous items were converted into a dichotomous format to prevent excessive measurement error. For the dichotomization, the responses never, seldom, occasionally, strongly disagree, disagree, and not sure/neutral were coded as expressions of a lack of the propensity to use and enjoy nature (0). The responses often, always, agree, and strongly agree were coded as expressions of the propensity to use and enjoy nature (1). Details about the calibration of the measurement instrument with the dichotomous Rasch model are presented in the Results section.

Environmental knowledge was measured with 25 knowledge items. These items were presented in a multiple-choice format with one correct answer and three distractors. The knowledge items covered facts about ecosystems and environmental issues, proper behavior that does not harm the ecosystem, and the ecological impact of specific technologies or landscape/forest management strategies. An example of facts is the following: “Abiotic factors are” (a) “water and light,” (b) “predators and competitors,” (c) “human beings and the industry,” and (d) “minerals and sustenance.” An example of behavior options is the following: “Identify the paper product that can be recycled by placing it in the paper trash.” The four options are: (a) “handkerchief,” (b) “egg crate,” (c) “public transport ticket,” and (d) “baking paper.” An example of an item that addresses ecological impacts is: “Identify the option that is not a consequence of deforestation.” The four options are: (a) “avalanche risk increases,” (b) “earthquake risk increases,” (c) “flooding increases,” and (d) “rockfall risk increases.” For more details about the knowledge items, see Fleissner-Martin et al. (2023).

We calibrated the environmental knowledge measure as a dichotomous Rasch scale. Environmental knowledge ranged from −2.35 and 3.61 logits (M = 0.53, SD = 0.89). This time, logits represented the natural logarithm of the ratio between correct and incorrect answers with more positive logit values indicating more knowledge. Additionally, the separation reliability indicated that the scale was able to differentiate students on the basis of their knowledge levels (rel = 0.71). Furthermore, mean square (MS_w) values weighted by the item variance between 0.85 and 1.25 confirmed that the scale was satisfactory and in alignment with previous impromptu measures of knowledge (see, e.g., Baierl et al., 2022b).

The 41 items were calibrated with the dichotomous Rasch model. The separation reliability indicated that the scale differentiated between people (rel = 0.87). Similarly acceptable were the fit values that reflected the discrepancy between the Rasch model’s predicted responses and the observed responses. Because our sample of 1,486 respondents was relatively large, we assessed item fit with Mean Square (MS_w) values weighted by the item variance. An MS_w value of 0.75 corresponds to 25% less than the expected amount of variation, and an MS_w value of 1.33 indicates 33% more variation in the data than what was predicted by the measurement model. MS_w values in this approximate range are considered a sensible threshold for instruments used in the scientific exploration of empirical relationships (see Wright et al., 1994). The item fit values for the current study ranged from 0.73 to 1.20 MS_w. For a complete list of items, their difficulties, and their MS_w values, see Supplementary Table S4.

Attitude toward nature values fell between −4.17 and 3.44 logits (M = −0.65, SD = 1.17), and items ranged from −1.69 to 2.27 logits. Logits represent the typical units of Rasch scales. Logits were used to measure both people’s attitudes and the item difficulties. Logits represent the natural logarithm of the ratio between affirmative and negative answers. More positive logit values correspond with the strength of people’s attitudes and the difficulty of an affirmative response to an item.

As measurements of people should be unbiased (i.e., fair; see, e.g., Fischer, 1987), instruments that are applied to make numerical comparisons between people should not unduly benefit members of certain groups. For example, bias occurs when men are more likely than women to provide affirmative responses to the items on a measurement instrument. To appreciate what is sometimes called specific objectivity in the proposed measurement instrument, we contrasted [with a test for differential item functioning (DIF)] the likelihoods of affirmative responses to the 41 items used in the measurement of attitude toward nature for the two gender groups (see Figure 3). The item difficulties were derived from calibrating two specific scales, one for female students (rel = 0.87, n = 721) and one for male students (rel = 0.85, n = 686). For male and female students, most of the item fit values fell within an acceptable range from 0.71 to 1.28 MS_w.

Of the 41 items, seven (17.1%) fell conspicuously outside the 95% confidence interval (the solid black lines) that we used to statistically factor uncertainty into the comparison of item difficulties (see Figure 3). Affirmative responses to items were either more difficult for females (above the confidence interval) or more difficult for males (below the confidence interval). Of the seven biased items, similar numbers benefitted female (i = 4) and male students (j = 3). In other words, biased items that fell above or below the confidence interval were relatively evenly arranged. Figure 3 illustrates this pattern and, thus, evidences a relatively gender-fair measurement of attitude toward nature. This conclusion was empirically backed by a Pearson correlation between the item difficulties for males and females of r = 0.94 (p < 0.001). Numerical details can be found in Supplementary Table S1.

To test our conclusion about gender fairness, we calibrated two different Attitude Toward Nature scales: one for female and one for male students. This time, the seven biased items were estimated separately for male and female students, whereas the difficulties of the other 34 items were fixed to equality for the two gender groups. Results showed no differences in the numerical depictions of people’s attitudes toward nature with Pearson correlations of r = 1.0 (n = 721) and r = 1.0 (n = 686) between the two measures: one ignoring and one considering DIF between male and female students.

In addition to gender, individuals in different phases of their adolescence might also benefit from differential difficulties in affirmative responses to items, for example, when individuals implement their personal propensities to use and enjoy nature restoratively or recreationally in different ways at different ages. Accordingly, we ran separate Rasch model tests for 11-13-year-olds (rel = 0.82, n = 349), 14-15-year-olds (rel = 0.86, n = 622), and 16-18-year-olds (rel = 0.88, n = 338). For 14-15-year-olds, the item fit values were in an acceptable range between 0.75 and 1.19 MS_w. By contrast, we found five items for 11-13-year-olds and two items for 16-18-year-olds that fell outside the acceptable range for fit (see Table 1).

When we contrasted 11-13-year-olds, 14-15-year-olds, and 16-18-year-olds in two DIF tests, we additionally found considerable DIF (see Supplementary Tables S2, S3). Of the 41 items, 25 were biased to the advantage of either 11-13-year-olds (i = 10) or 14-15-year-olds (j = 15). That is, 14-15-year-olds benefited slightly more from the selected items, as more items seemed better suited to reflect how 14-15-year-olds express their propensity to use and enjoy nature (for more details, see Supplementary Table S2). As age increases, verbally expressing appreciation for nature (but less so for actual outdoor activities) appears to be the preferred way to express attitudes toward nature.

Of the 41 items, 12 were biased to the advantage of 14-15-year-olds (i = 5) or 16-18-year-olds (j = 7). Again, the 12 biased items were relatively evenly arranged, with similar numbers benefiting 16-18-year-olds (i = 7) and 14-15-year-olds (j = 5: for more details, Supplementary Table S3). At older ages, contemplative outdoor activities appeared to have become the relatively preferred way to express one’s propensity to use and enjoy nature.

To test our conclusion regarding age fairness, we calibrated three different Attitude Toward Nature scales: one for 14-15-year-olds (rel = 0.82), one for 14-15-year-olds (rel = 0.86), and one for 16-18-year-olds (rel = 0.88). In these scale calibrations, the 12 items with the most extreme bias were estimated separately in the different age groups, whereas the difficulties of the remaining 29 items were fixed to equality for the three age groups. This time, item fit values across the three age groups were all in an acceptable range, with MS_w values between 0.72 and 1.21. Results showed marginal differences in the numerical depictions of students’ attitude toward nature with Pearson correlations of r = 1.0 (n = 349), r = 0.99 (n = 622), and r = 0.99 (n = 338) between the two measures: one ignoring and one considering DIF across the three age groups.

In our analysis of the 382 students who participated in the educational intervention, we followed the procedure for mediation tests proposed by Baron and Kenny (1986) (see Figure 2). Because of its superior technical features, we employed Preacher and Hayes’ bootstrapping method (see, e.g., MacKinnon et al., 2002; Preacher et al., 2007). Initially, we established that those with stronger attitudes toward nature also knew more about environmental issues (Path c in Figure 2) after the intervention (b = 0.23, 95% CI [0.13, 0.33], β = 0.23). Once we added the mediator (i.e., environmental attitude) by including the mediating Paths a (b = 0.42, 95% CI [0.34, 0.50], β = 0.46) and b (b = 0.41, 95% CI [0.29, 0.52], β = 0.37) in the analysis, the strength of the effect of attitude toward nature on knowledge about environmental issues (Path c’) was no longer significant (b = 0.06, 95% CI [−0.04, 0.18], β = 0.06: see Figure 4).

To gain a more profound understanding of the role of attitude toward nature in learning and the hypothesized relationships depicted in Figure 1, we initially verified that the intervention was beneficial for students. Apparent in the mean values, knowledge levels increased significantly from the assessment conducted before the intervention (M_pre = 0.51, SD = 0.77) to the assessment conducted 6 weeks after the intervention (M_retention = 0.76, SD = 1.03: p < 0.001; n = 382). In addition to students’ mean knowledge levels, the standard deviation of students’ knowledge also increased, thus indicating a disparity in learning. In other words, not all students benefited equally; some students learned disproportionally more than others. Indeed, students with increasing levels of attitude toward nature gradually gained more knowledge (β = 0.23): F_{(1, 381)} = 21.90, p < 0.001, R² = 0.05.

The relatively superior learning was also reflected by the correlations between students’ attitude toward nature and their preexisting (r = 0.19, p < 0.001, n = 382) and retained knowledge (r = 0.24, p < 0.001, n = 382), which—although not in an inference statistical sense—increased in strength in a descriptive sense after the intervention. Finally, we additionally corroborated that knowledge gains (i.e., the difference between retained and prior knowledge) were, in fact, controlled by students’ attitude toward nature (β = 0.10): F_{(1, 381)} = 3.92, p = 0.048, R² = 0.01.

Against the backdrop of the hypothesized mediating role of environmental attitude (see Figure 4), we revisited the attitude effects and replaced attitude toward nature with environmental attitude. At least on a descriptive level, we found that the measures of effect size (i.e., r and R²) became quantitatively larger than the equivalent effects with attitude toward nature. Evidentially, students with increasing levels of environmental attitude also gained linearly more knowledge (β = 0.41): F_{(1, 381)} = 76.31, p < 0.001, R² = 0.17. Whereas the connection between environmental attitude and knowledge remained relatively comparable for both retained knowledge (r = 0.39, p < 0.001, n = 382) and preexisting knowledge (r = 0.40, p < 0.001, n = 382), they almost doubled in magnitude. And once more, the relatively superior learning effect was again reflected in the knowledge gains (i.e., the difference between retained and prior knowledge) that were determined by students’ attitude, this time environmental attitude (β = 0.15): F_{(1, 381)} = 8.31, p = 0.004, R² = 0.02.