A Five-Culture Validation of the Environmental Value-Bases Scale: A Measure of Instrumental, Intrinsic, and Relational Environmental Values

Abstract

Previous research identified three reasons for valuing nature (i.e., the basis for seeing nature as valuable and important): (1) valuing nature for what it gives to humans (instrumental), (2) valuing nature for its own sake (intrinsic), and (3) valuing nature because of the relationship between people and nature (relational). Of these, relational value-bases have been less studied, especially in non-Western cultures. Using a large sample (n = 2618), with participants from five distinct cultural regions (Japan, Taiwan, Italy, France, USA), the present research tests whether a three-factor framework of environmental value-bases generalizes to other cultures. Our findings demonstrate the configural and metric invariance of the recently validated Environmental Value-Bases Scale, indicating that the latent constructs generalize across sub-samples of the five regions and that the measure can be used to test associations between the value-bases and outcomes across cultures. However, we only found partial scalar invariance, suggesting (a) that caution is needed when comparing scale means between cultures and (b) that such tests are most appropriately performed using latent means. This research further contributes to the growing value-basis literature by comparing the latent means for each value-basis between and within each of the five regions and by demonstrating their associations with place attachment.

1. Introduction

Given the many environmental crises challenging the sustainability of the planet—including climate change, excess plastic waste, and biodiversity loss [1]—understanding predictors of pro-environmental outcomes is a global priority. One well-studied category of predictors of pro-environmental behavior is environmental values (see [2,3]). Values are overarching ideals or principles that transcend circumstance in guiding a variety of human behaviors and decision-making processes [2,4,5]. Environmental values, specifically, are the ideals or principles that an individual holds in reference to the health, status, and importance of the natural environment. Individuals endorsing pro-environmental values recognize the natural environment as important and valuable, and such endorsements predict pro-environmental outcomes (e.g., [6]).

However, there is nuance in the relationship between valuing the natural environment and pro-environmental outcomes [7]. Research has consistently shown that the reasons underlying people’s valuing of nature—dubbed environmental value-bases (VBs)—uniquely influence pro-environmental intentions and behavior (see [8]). Reviews of the literature and discourse surrounding the reasons people value nature have revealed at least three bases for valuing nature (see [9,10,11,12,13,14,15]): (1) valuing nature for its benefits to people based upon anthropocentric, means-to-an-end, or people-based reasons (i.e., an instrumental basis for valuing nature; Instrumental VBs); (2) valuing nature for its own sake based upon biocentric and ethical reasons or based upon its inherent value (i.e., an intrinsic basis for valuing nature; Intrinsic VBs); and (3) valuing nature as a participating member of the human–nature relationship or for cultural, collective, or community-oriented reasons (i.e., a relational basis for valuing nature; Relational VBs; see [8]).

Despite its long history (e.g., [5]), much of the quantitative research on VBs and many of the empirical measures that capture the different reasons for valuing the environment have neglected Relational VBs (see [8]). Even so, recent conceptual work has stressed the importance of relational reasons for valuing nature (see [16]), and global policy has begun to recognize the importance of the relationship between people and nature. For example, in 2020, the UN Secretary-General ended his speech on the state of the planet by stating, “Now is the time to transform humankind’s relationship with the natural world” [17], and land-mark policy documents, such as the Kunming–Montreal Global Biodiversity Framework [18] and the recent IPBES Transformative Change Assessment [19], have emphasized the role of people’s relationship with nature and place in addressing environmental crises (see [20] for a review of similar policy interest). Thus, there is growing recognition that relational elements of the human–nature relationship—including Relational VBs—are an important sustainability target [20]. There is also a clear need to include Relational VBs in quantitative investigations, since the way we relate to nature is central to why and how different cultures encourage sustainable practices [21]. For example, knowing which VBs different communities endorse is relevant to successfully matching policies to communities’ reasons for valuing nature, which may enhance the effectiveness of those policies in that the given community [22]. This matching inherently requires ways of documenting which VBs are most important in different contexts. Finally, beyond trends in the discourse and practical considerations for informing policy, empirical research using a recently validated measure of the three VBs has highlighted Relational VBs as an essential pro-environmental predictor [8]. Thus, additional investigations into the role of Relational VBs in predicting pro-environmental outcomes are well-warranted, especially in non-Western contexts where Relational VBs might be a more prevalent basis for valuing nature [13].

The primary purpose of this research is to facilitate such investigations by testing whether the three VBs manifest similarly in a globally diverse sample. Specifically, we aim to test whether the three-factor framework for understanding environmental VBs generalizes outside of the US context in which it was quantitatively validated. Accordingly, we test whether the latent constructs captured by the Environmental Value-Bases Scale are invariant across five cultural regions: two from Eastern cultures (Japan and Taiwan) and three from Western cutlures (Italy, France, and the USA). The cultural variability across these regions (e.g., differences in collectivism, which is found more frequently in Eastern than in Western countries [23,24]) allows us to robustly test whether the three-factor framework of environmental VBs generalizes under conditions of considerable cultural variability which may contribute to fundamental differences in the empirical structure of VBs. Thus, demonstrating the validity of this framework, which distinguishes between these three VBs, in regions with vastly different values fills two research gaps: (1) it documents, for the first time, the consistency of such distinctions across cultural regions and (2) advances our practical capacity to study these three VBs by allowing the common use of a single measure in different cultures, making research more comparable across contexts. With regard to the first gap, knowing whether environmental VBs can be treated as reflecting the same three underlying factors, even when using a diverse sample, is an important theoretical contribution that supports further investigating into the potential universality of such a three-factor framework. With regard to the second gap, we demonstrate how a common measure across different regions can be used to answer other research questions, including comparisons of the relative endorsement of the three VBs within and between different cultural regions. Additionally, building upon the considerable work demonstrating that people’s relationship with place [25] and planet [26] are important targets for sustainability [18,19,20], we also test whether and how environmental VBs predict place attachment.

1.1. Overview of Environmental Value-Bases (VBs)

In general, environmental VBs are the underlying reasons for individuals perceiving the environment as valuable, important, and ultimately worthy of protection (see [8]).

1.1.1. Instrumental Bases for Valuing Nature (Instrumental VBs)

One of the most commonly endorsed and frequently studied reasons for seeing nature as important is valuing nature because of its benefits to people [2,5,7,27,28]. This reason for valuing nature has gone by several terms, including anthropocentrism [29,30], enlightened self-interest [31], and the instrumental value of nature [9,11,12]. With their focus on how nature benefits people, Instrumental VBs are consistent with valuing nature for utilitarian reasons (as something useful) or as a means to an end, implying nature is an object that, therefore, can (a) be used by people and (b) can be replaced with other “means” so long as the same end is achieved (see [10,11,12]).

In the quantitative empirical literature, Instrumental VBs are largely consistent with constructs such as egoistic values [2] and egoistic concern for the environment [28], both of which involve valuing or caring about the environment because of self-centered concerns. Instrumental VBs are also captured by altruistic values [2] and altruistic concern for the environment [28] because both are rooted in valuing or caring about the environment because of people-centered concerns (see [8]).

1.1.2. Intrinsic Bases for Valuing Nature (Intrinsic VBs)

Another commonly endorsed and well-studied reason for seeing nature as important is valuing nature because of its inherent value [2,5,7,9,27,28]. This reason for valuing nature has most notably been discussed in terms of non-anthropocentrism (e.g., [8]), the inherent value of nature [12], and the rights of nature [32]. Intrinsic VBs are also consistent with valuing nature as an end in and of itself—implying that nature is irreplaceable [11]. This characterization often leads to Intrinsic VBs being portrayed as the foil to Instrumental VBs [11].

In the quantitative empirical literature, Intrinsic VBs are largely consistent with constructs such as biocentric values [2] and biospheric concern for the environment [28]. Both biocentric values and biospheric concern involve valuing or caring about the environment because of nature-focused concerns (i.e., for reasons that have little to do with people; see [8]).

1.1.3. Relational Bases for Valuing Nature (Relational VBs)

Finally, an increasingly recognized (but understudied) reason for seeing nature as important is valuing nature because of its role as a participating member of the community and the relationships that people and cultures form with it [9,13,14,33,34]. This reason for valuing nature is consistent with ideas such as the Land Ethic [31] or the Partnership Ethic [35]. Because Relational VBs focus on the relationship between people and nature, nature is irreplaceable from this perspective [12], similar to Intrinsic VBs; that is, if nature were to be replaced, the relationship would change, meaning the value would change (see [10,11]). However, the focus on the relationships between nature and people also means that Relational VBs are anthropocentric because they do have a focus on people [10]—at least through a purely philosophical lens (see [8] for empirical evidence that suggests Relational VBs are more associated with non-anthropocentrism than anthropocentrism). Thus, one could argue that Relational VBs are also similar, at least philosophically, to Instrumental VBs.

In the empirical literature, Relational VBs have only recently been quantified (see [8]). They have, however, been gaining increasing recognition in the field (e.g., “Nature as Culture” from the IPBES Nature Futures Framework [36], and other relational environmental value constructs; [37,38,39]). There have been some attempts to validate scales capturing such VBs. However, these existing attempts to measure Relational VBs have either (a) neglected to demonstrate predictive validity [12,40,41], (b) have not tried to measure all three VBs simultaneously [40], or (c) have had concerning flaws (e.g., treating clearly instrumental items—such as “Humans have the right to use nature any way we want”—as measuring Intrinsic VBs [41]). For these reasons, we believe it is fair to say that there had not been an empirically validated measure of the three VBs until recently.

Indeed, in the absence of validated scales, some researchers (e.g., [40]) have even taken an approach that implies that Relational VBs can be approximated with other measures (i.e., connection to nature [42]). Such measures fall short in three ways, however. First and foremost, the approximate measures were not designed to explicitly capture the relational basis for valuing nature (i.e., they were not intended to explain why people value nature). Second, they were not designed in concert with the other two VBs, thereby hindering comparisons among the three VBs (e.g., to determine which of the three VBs are most predictive of pro-environmental outcomes). Third, and perhaps most decisively, the current evidence suggests that the three most common measures of nature connectedness do not measure the same latent construct as Relational VBs (see [8]). Fortunately, these limitations were recently addressed by the measure that was designed to capture all three VBs and was subsequently validated in three US samples [8]. This scale was specifically designed to capture all three VBs so that they can be tested simultaneously to assess their distinct contributions to other outcomes.

1.2. Associations Between Environmental VBs and Pro-Environmental Outcomes

Empirical investigations into the relationship between environmental VBs and pro-environmental outcomes suggest that different reasons for valuing nature have unique effects on pro-environmental outcomes. When tested simultaneously with Instrumental VBs, Intrinsic VBs (e.g., biospheric concern, biocentric values) frequently have a significant, positive association with pro-environmental outcomes [43,44,45,46,47,48]. In contrast, when tested simultaneously with Intrinsic VBs, measures of Instrumental VBs focusing on the self (e.g., egoistic environmental concern, egoistic environmental values) frequently show a negative [43,47,48] or non-significant [44,45,46,48] association with pro-environmental outcomes. Moreover, measures of Instrumental VBs focusing on other people (e.g., altruistic environmental concern, altruistic environmental values) show inconsistent associations with the same outcomes. Specifically, in some studies, altruism-based measures of Instrumental VBs show non-significant associations [43,44,47,48], while others show negative [46,48] or even positive effects [43,45]. Thus, the literature generally indicates that (a) Intrinsic VBs are a strong, positive predictor of pro-environmental outcomes and (b) Instrumental VBs are, at best, an inconsistent predictor of pro-environmental outcomes.

Possibly because there had not been an empirically validated measure of Relational VBs until recently, only one paper has reported the association between Relational VBs and pro-environmental outcomes. In this study, Relational VBs demonstrated a consistent positive association with pro-environmental outcomes across all three studies—even when tested alongside Instrumental VBs and Intrinsic VBs [8]. In contrast, in the same models, Instrumental VBs showed a consistent negative association with pro-environmental outcomes, and Intrinsic VBs demonstrated an inconsistent association with pro-environmental outcomes [8]. Thus, these recent results not only reaffirm the differential effects of pro-environmental VBs (i.e., why we value nature matters), but also highlight that Relational VBs—which accounted for more than double the variance captured by Intrinsic VBs [8]—may be especially important pro-environmental VBs worthy of future study, including in other cultures. Accordingly, understanding whether the extant measures can be used to capture the same latent constructs across multiple cultural contexts is vital to supporting such research.

1.3. Primary Research Questions

While the Environmental Value-Bases Scale was developed in the US, the three VBs have been of interest across countries and cultures [9,13,14,33]. Yet, any time there is cross-cultural interest in a construct that has been validated in only one culture, there is the necessary question of whether that construct generalizes to other cultures or is merely idiosyncratic to the culture in which it was validated (see the early work on basic human values; [4,49]). This is no less true in this instance.

While some of the founding conceptual work proposing these three distinct VBs implies that they are globally applicable (see [9,14]), there are arguments that relational values, in particular, are inherently place-based and culturally bound—and therefore would not manifest similarly across cultures (see [50]). Furthermore, while distinguishable into relatively discrete conceptual categories, the three VBs exist on a continuum. Thus, in practice, their boundaries may not be as sharp as implied by the three-factor framework—even given the statistical evidence that they are [8]. This may be especially true across cultures with vastly different values. It therefore becomes necessary to confirm that these three factors remain distinguishable in diverse cultural contexts if one wishes to propose that these three ways of valuing nature apply globally. At present, however, no work has considered (a) whether the three-factor structure of environmental VBs implied by the conceptual literature (see [51]) exists in contexts other than the US, let alone (b) whether the same three-factor structure that applies to the US likewise applies to other vastly different cultural contexts (e.g., non-Western cultures).

Here, we address this gap and focus on whether the three-factor framework validated in the US extends to a more diverse sample of cultural regions: Japan, Taiwan, Italy, France, and the USA. Specifically, we investigate whether the three VBs can be captured effectively and with the same latent structure across different cultural regions by using muti-group Confirmatory Factor Analysis (CFA) to test the extent to which the latent constructs captured by Relational VBs, Intrinsic VBs, and Instrumental VBs remain invariant across the sampled cultural regions [52,53,54,55] (see [56]). We consider three types of measurement invariance: configural, metric, and scalar invariance (see [55]).

Configural invariance indicates that scale items form the same set of factors across different groups (i.e., all groups have the same number of factors comprised of the same sets of items). In our case, configural invariance would suggest that the items measure similar constructs across cultural regions [53,55] because, much like the evidence used to claim that Schwartz’s Basic Human Values are universal [4,49], all of the items tend to be situated in the same general structure across sub-samples. A failure to find configural invariance suggests that the Environmental Value-Bases Scale has a different factor structure in at least one cultural region and does not measure similar constructs [53]. For example, we would not likely find configural invariance if some regions reflected the expected three-factor structure while others reflected a one-factor structure where the three VBs are not distinguishable. Such a failure to find configural invariance would preclude further tests of invariance [55] and would also potentially suggest that the VBs employed outside the US are qualitatively different from the established three-factor framework. Conversely, evidence of a three-factor structure that is present across all regions would be largely consistent with the notion that these values are applicable to a wide range of contexts.

• Research Question 1: Does the Environmental Value-Bases Scale form the same three factors across cultural regions?

Metric invariance indicates that item loadings are similar enough across different groups that they can be treated as equivalent (i.e., constrained to be equal). Thus, not only would each factor have the same set of items across cultural regions (which would be established via the test of configural invariance), but metric invariance further implies that each factor is reflected by those items in an effectively identical manner because they have the same factor loadings in all groups [52,53]. This type of invariance is a prerequisite for comparing regression slopes [52,54] because it indicates that the latent construct is measured the same way in each group [53]. Put colloquially, whereas configural invariance points to the factor distinctions being shared across groups, metric invariance indicates that each of those factors has the same “shape” across groups. A failure to find metric invariance would suggest that the latent constructs assessed by the Environmental Value-Bases Scale are reflected differently in at least one cultural region. A failure at this step precludes any further tests of invariance [55]. Importantly, this test of metric invariance holds not only methodological and practical value but also represents an important theoretical question. That is, in conjunction with configural invariance, metric invariance represents a test of whether there is sufficient evidence to suggest that the same three latent constructs are captured similarly across groups and with the same structure. In other words, this analysis tests whether the latent constructs of Relational, Instrumental, and Intrinsic VBs generalize to contexts beyond the US.

• Research Question 2: Do the items forming the three factors have invariant loadings across cultural regions?

Finally, scalar invariance indicates that the scale item intercepts are similar enough across groups to be treated as equivalent—meaning that observed item scores are influenced by the same degree of response bias in each group (see [52]). This type of invariance is necessary for making mean comparisons between groups [54], especially when comparing observed composite means [52]. A failure at this step means that any detected group-mean differences across cultural regions might be due to artifacts such as cultural differences in systematic response biases rather than differences in the underlying latent construct [52]. In other words, a lack of scalar invariance would make it difficult to know whether any given mean differences among cultural regions are due to meaningful variations in the latent constructs or if they are, instead, due to some systematic bias (e.g., the way one culture uses scale points or methodological differences [52,54]). In the event of a failure to produce evidence of full scalar invariance, researchers can pursue partial scalar invariance by allowing some intercepts to be estimated freely [52,55]. However, it is important to note that, even with partial scalar invariance, comparing the observed mean (i.e., based on composites of observed variables) differences between groups is ill-advised, and more advanced comparisons using SEM—in order to compare differences using latent variables—are likely necessary if one wishes to interpret mean differences without concerns about the influence of systematic response biases (see [52]).

• Research Question 3: Do the items forming the three factors also have invariant intercepts across cultural regions?

Together, these three research questions represent the first test of whether Relational, Instrumental, and Intrinsic VBs remain distinguishable across cultural contexts, and whether they manifest similarly between Eastern and Western cultures. Based on the initial validation of this scale (see [8]) and the state of extant research (see [33]), we expect the same three-factor structure to emerge across sub-samples despite their vast cultural differences. Thus, this research presents a critical theoretical step toward understanding the potential universality of environmental VBs (c.f., Schwartz’s investigations of the universality of human values; [4]). However, this research also fills a second, more practical gap in the literature. Specifically, a lack of common measurement is one of the factors identified as contributing to the lack of coherence that has been noted within the study of valuing nature (see [57] for a discussion of this gap). Thus, by potentially providing a measure of all three VBs that can be used across a wide range of cultural contexts, we fill this practical gap, thereby advancing the study of relational and pluralistic VBs toward a more coherent theoretical framework.

1.3.1. Contingent Research Questions

The above research questions test the theoretically and practically important question of whether the latent constructs reflecting Relational, Instrumental, and Intrinsic VBs are generalizable to contexts outside of the US. However, assuming we find evidence of measurement invariance, we have two additional areas of interest: (1) cross-cultural variation in endorsements of the VBs and (2) the relationship between the VBs and place attachment.

Cross-Cultural Variation

While quantitative cross-cultural comparisons in the endorsements of Relational, Intrinsic, and Instrumental VBs have not yet been explored, there is reason to expect the three VBs to vary both between and within cultures. First and foremost, just as there are cross-cultural variations in the importance of human values (e.g., [4,58]), there may be cross-cultural differences in the importance of environmental VBs. For example, focusing on Instrumental VBs and Intrinsic VBs (i.e., the instrumental–intrinsic dichotomy; [34]) to the exclusion of Relational VBs has been the prominent approach in Western cultures [8]. In contrast, the Nature Futures Framework [15,59]—which was derived from a global analysis of discourses surrounding valuing nature that deliberately incorporated the perspectives of indigenous communities—includes a relational perspective alongside the traditionally Western Instrumental and Intrinsic VB perspectives (see also [13,14,15,59]). The very fact that relational reasons for valuing nature have only emerged when considering a diverse representation of people across cultures and countries implies that there may be cultural variation in the relative importance attributed to the three VBs. That is, in the West, Instrumental and Intrinsic VBs seem to have been so important as to dominate the entirety of the Western discourse. In contrast, Relational VBs were important enough in other parts of the world that they were identified by a global review of the reasons that people value nature [9,13,14].

Second, in addition to conceptual reasons to expect variation in VB endorsements, empirical research provides both direct and indirect evidence to support cross-cultural variations. For example, a qualitative analysis of environmental values in Colombia indicated that relational values were far more prevalent amongst Colombian participants than either instrumental or intrinsic values, which were exceptionally low [34]. This provides direct evidence that relational values are more important in some cultures than others (and also that they may be differentially important within cultures, as noted below). Indirectly, research suggests that factors routinely differing between cultures, such as collectivism, are associated with endorsements of a broad range of values (e.g., [60]); if collectivism varies between cultures, then values likely will as well. Similarly, self-construals, which are likewise routinely recognized as culturally bound, are also significantly correlated with environmental values specifically (e.g., [61]). It therefore stands to reason that there may be cultural differences in the environmental VBs as well. Accordingly, assuming that we can establish invariance (i.e., partial or full scalar invariance), we planned to explore whether there were between-culture differences in the endorsements of Relational VBs, Instrumental VBs, and Intrinsic VBs.

Additionally, as implied above, the long-standing neglect of Relational VBs in Western science until the perspective of non-Western cultures was considered suggests that the relative importance of Relational VBs may be stronger in non-Western cultures (e.g., in our samples, Taiwan and Japan relative to the United States, France, and Italy). For example, collectivism, which tends to be higher in Eastern cultures [62], is more strongly associated with relational orientations than individualism [58], which tends to be higher in Western cultures [62]. Such a pattern, while indirect, suggests that Eastern cultures might more strongly endorse relational environmental VBs than other environmental VBs, whereas Western cultures might show a different within-culture pattern. Practically, if there are cultural differences in the endorsements of environmental VBs, policies—or more general persuasive appeals—emphasizing certain reasons for valuing nature might be accepted differently across cultures [22]. Therefore, understanding whether and how cultures differ in their endorsements of VBs (both at the between and within levels) is critical for informing how such values should be situated in national and international policy. Thus, assuming measurement invariance, we also plan to explore cross-cultural variations in the within-culture relative endorsements of the VBs.

• Research Question 4: Are there cross-cultural differences in endorsements of the VBs between (requires scalar invariance) and within cultures (requires metric invariance)?

While we anticipate the structure of the environmental VBs to hold across sub-samples, we expect mean differences across countries, with Relational VBs being higher in Eastern cultures (where collectivism is more prominent, as outlined above).

Place Attachment

Although Relational VBs are quantitatively underexplored, there are related constructs that have been examined more extensively. One such example is place attachment—the bond between people and the places with which they engage [63]—which is a popular topic of study within environmental social sciences and correlates robustly with pro-environmental behavior (see [25] for a meta-analysis evidencing the robust place attachment–pro-environmental behavior association). While there is no extant evidence linking place attachment to environmental VBs directly, given that place attachment is associated with a variety of pro-environmental constructs—including pro-environmental attitudes [64] and concern for the environment [65]—it is reasonable to expect measures of environmental VBs to predict place attachment.

Moreover, there is reason to expect Relational VBs to be more closely related to place attachment than Instrumental VBs and Intrinsic VBs. First, both Relational VBs and place attachment explicitly focus on the relationship between humans and the environment, whereas Instrumental VBs and Intrinsic VBs are more narrowly focused on either humans or the environment, respectively. Furthermore, based upon the association between VBs and constructs similar to place attachment (e.g., other place–self relationship constructs), there is also indirect evidence that Relational VBs and place attachment should be strongly associated. For example, past research has identified differential associations between the three VBs and nature connectedness—the closeness of one’s relationship with nature based upon the extent to which nature is included in one’s sense of self [20,66,67]. Specifically, while the original validation of the Environmental Value-Bases Scale showed that nature connectedness and Relational VBs are, indeed, distinct constructs, Relational VBs had a particularly strong association with nature connectedness—one that happened to be the strongest out of the three VBs [8]. Accordingly, to the extent that place attachment and nature connectedness represent similar environment–self constructs, we anticipate Relational VBs will be a stronger predictor of place attachment than Instrumental VBs and Intrinsic VBs.

Place attachment, however, is a multidimensional construct with two of its more commonly recognized dimensions being place identity—the affective attachment one has formed with a place—and place dependence—one’s sense that the place is able to support one’s achievement of goals and needs (see [68]). While we generally expect Relational VBs to be associated with both dimensions of place attachment, there might be differences across the dimensions. For example, the affective attachment emphasized in place identity is more similar to the notion of relationships with nature and might be more strongly associated with Relational VBs. In contrast, the emphasis on the achievement of goals and needs found in place dependence may reflect an underlying focus on how nature benefits oneself, which would be more consistent with Instrumental VBs.

Regardless of the ultimate nature of the associations, knowing whether and how place attachment and environmental VBs are related is relevant for both policy and theory. For one, in light of the growing recognition that the relationship between people and nature is a critical target for sustainability (see [20] for a review of policy documents with such emphasis; see also [17,18]), these analyses represent an important test of whether and how environmental VBs shape our relationship with place and the planet. Additionally, they will inevitably inform the discourse surrounding Relational VBs, given that a key conceptual assertion of the proponents of recognizing Relational VBs is that they are inherently place-based (e.g., [16]). To the extent that this assertion is true, one would expect particularly strong associations between Relational VBs and place attachment. Thus, assuming we can establish the necessary evidence of invariance, we explore whether there are unique associations between the three VBs and the two dimensions of place attachment.

• Research Question 5: How are the three VBs related to the two dimensions of place attachment? Are there particularly strong relationships with Relational VBs?

2. Materials and Methods

The data for this paper were collected as part of a larger, funded project focused on land-use change in five cultural regions (Japan, Taiwan, Italy, France, and the USA). Specifically, this broader project was a funded collaboration between social and natural scientists partially meant to understand how land-use decisions can affect the natural environment, as documented by natural scientists who have been tracking changes to the natural environment in the regions represented in the present study. Because of the natural science component, the research activities were tied to specific geographic sites suited for making geographic comparisons. This was a partial determinant of the regions selected. Thus, these specific cultural regions were chosen based on their fit with the larger project and not solely for the purposes of this paper. Still, they represent critical variations in geographic regions that differ in cultural beliefs and heritages represented across the world. Specifically, with two non-English-speaking Western cultures (Italy and France) and two non-English-speaking Eastern cultures (Taiwan and Japan), these cultural regions vary greatly in the degree of similarity to the US context where the scale was originally validated. Accordingly, their diversity allows us to test whether the three latent factors exist even under conditions of considerable cultural diversity.

2.1. Participants

A total of 2790 participants were recruited from the five cultural regions (see

Table 1. Exclusions by cultural region.

Cultural Region	Final N	Straight-Lined	Incomplete VBs
Taiwan	919	31	39
Japan	602	69	0
Italy	533	2	0
France	273	14	0
USA	291	2	15

for the breakdown of the final sample sizes and exclusions by cultural region). Researchers in each cultural region aimed to obtain a minimum sample size needed for CFAs within their region. General recommendations suggest that 200 to 300, or a 10:1 ratio of people per item, would be sufficient for our measure [69]. Variations in sample size per cultural region resulted from differences in resources and researchers using the broader survey for other purposes.

In Japan, Taiwan, France, and the US, participants were recruited through existing online panels where respondents complete various surveys in exchange for money or shopping reward points. Japanese respondents were recruited from Hokkaido, Miyagi, Tochigi, Niigata, and Ichikawa prefectures in Northern Japan. French respondents were recruited from the Gard and Hérault provinces in South France. Taiwanese respondents were recruited from Taipei and New Taipei City. US respondents were recruited from the Southeast of New Hampshire. In Italy, respondents were directly recruited from the Alpine region via social media, flyers posted in local businesses and clubs, and visits to these organizations in order to reach hard-to-reach populations in the Alpine region that are often excluded from research due to their location.

A total of 118 participants were excluded for straight-lining on all three subscales (i.e., providing identical non-midpoint responses on all items for a given subscale, including the reverse-coded item). Given the focus on invariance analyses in this paper, we also excluded 54 participants who had missing responses on any of the scale items. The final total sample comprised 2618 participants. The descriptive statistics by cultural region can be found in

Table 2. Demographics by cultural region.

Cultural Region	Proportion Men	Mean Age	(SD)	Median Education	(Range)
Taiwan	0.38	37.4	(12.2)	3	(1–5)
Japan	0.58	49.8	(12.0)	4	(1–5)
Italy	0.62	50.5	(14.2)	4	(1–5)
France	0.48	52.0	(16.5)	4	(1–5)
USA	0.55	58.2	(14.6)	4	(2–5)

Note. Education was coded using region-specific equivalents according to the following: 1—Less than High School; 2—High School Degree; 3—Some College; 4—College Degree; 5—Post-Graduate Degree.

.

2.2. Materials and Procedures

Surveys were approved by the necessary ethics board as required in each region and were administered online in all five cultural regions. The measures for the present study were among other measures included for the purposes of the larger project. Specifically, for the larger project, participants were directed to think about the area where they lived and, reflecting on that area, consider their perceptions of and preferences among land-use types and the state of the natural environment. In this survey, participants completed measures for the other research questions about land preference for different types of use (which are beyond the scope of the present study). After completing these measures, they completed the Environmental Value-Bases Scale [8] and the place attachment measure—the focal measures for this manuscript—and provided demographics.

2.2.1. Environmental Value-Bases Scale

The Environmental Value-Bases Scale contained 23 items plus 3 reverse-coded items not included in analyses but used to identify straight-line responders (see [8]). The scale had three subscales: Instrumental VBs (e.g., “Plants and animals are valuable because of the benefits they provide to people”), Intrinsic VBs (e.g., “Plants and animals are important because they are parts of a thriving ecosystem”), and Relational VBs (e.g., “Ecosystems are important because they are part of our community, like neighbors”). All items were individually standardized before being used in analyses. All three scales showed strong reliability (αs ≥ 0.95).

The scale was translated into each target language by members of the broader research team and back-translated to identify possible areas of mistranslation. Any such potential issues, however, were carefully discussed to preserve the original meaning of the items in the source language while ensuring that meaning was effectively conveyed in the target language, as is recommended practice [70]. The versions of the scale used in each cultural region can be found in Appendix A.

2.2.2. Place Attachment

To capture the two dimensions of place attachment, we used a six-item scale that has been validated across cultures [68], rated on a five-point scale from “Strongly Disagree” (−2) to “Strongly Agree” (2). Three items assessed place identity (e.g., “I identify strongly with the [region-specific focal area]”), and three items assessed place dependence (e.g., “I would not live anywhere else than in the [region-specific focal area]”). Both scales showed strong reliability (αs ≥ 0.86).

2.3. Analytic Approach

2.3.1. Tests of Invariance

Factor Structure

The originally validated scale had a three-factor structure, with the Intrinsic VB factor formed by two latent sub-factors (see [8]). However, given the complexity of the planned analyses in the present investigation, here, we tested a simplified three-factor structure applied across the five cultural regions (note that substantive conclusions regarding invariance tests were corroborated using the original factor structure). We compared this three-factor model to a one-factor model and three two-factor models that tested alternative pairings of subfactors: (1) Instrumental VBs and Intrinsic VBs forming one factor and Relational VBs forming a second, (2) Relational VBs and Intrinsic VBs forming one factor and Instrumental VBs forming a second, and (3) Relational VBs and Instrumental VBs forming one factor and Intrinsic VBs forming a second. These two-factor models were examined to ensure the complete set of possible combinations of factors was tested.

Multi-Group CFA

To test for the three types of invariance, we conducted a series of multi-group CFAs with varying constraints in the R package lavaan (version 0.6-18) [71] using robust estimation (“MLM” in lavaan). Because the chi-square test is known to be overly sensitive to non-meaningful differences, we instead used changes in fit indices to determine the meaningfulness of differences between invariance models, as is currently the suggested practice [54]. Sample weights were used in these analyses (see Section 2.3.3).

Fit Standards

Configural invariance is demonstrated by acceptable fit—CFI ≥ 0.90 [72,73], SRMR ≤ 0.08 [73], and RMSEA < 0.10 [72,74]—of the unconstrained model in a multi-group CFA [55]. Metric invariance is demonstrated by a non-meaningful change (∆CFI ≤ −0.010; ∆RMSEA ≤ 0.015; ∆SRMR ≤ 0.030) when comparing the unconstrained model to a model where item loadings are forced to be constant across groups (see [54,55]). Scalar invariance is demonstrated by a non-meaningful change (|∆CFI| ≤ 0.010; ∆RMSEA ≤ 0.015; ∆SRMR ≤ 0.010) when comparing the loading-constrained model to a model where both item loadings and intercepts are forced to be constant across groups (see [54]).

2.3.2. Testing the Three VBs Simultaneously

As noted in the original validation of this scale (see [8]), when testing associations between the VBs and other constructs, it is essential to pay attention to the unique associations between a given value-base and other constructs when controlling for the other VBs. This necessity is because scores on any measure of Instrumental VBs, for example, can be represented by the following symbolic equation:

$$\mathrm{I}\mathrm{n}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{u}\mathrm{m}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{a}\mathrm{l}\mathrm{V}\mathrm{B}={VN}_{gen}+{VN}_{\mathrm{I}\mathrm{n}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{u}\mathrm{m}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{a}\mathrm{l}}$$

where VN_gen is valuing nature, generally, and Instrumental VBs are valuing nature specifically for its benefits to people. Since both sources of variance are conflated in the aggregate measure, a significant raw correlation between Instrumental VBs and some outcome could be because the outcome correlates with the Instrumental VB’s variance, but it could also be entirely correlated with—and only with—the VN_gen variance. Thus, it is necessary to remove the effect of VN_gen in order to detect the unique effect of VN_Instrumental. This detection can be achieved when all three VBs are included in the same model, since the Type III sum of squares eliminates shared variance before assigning estimates to individual terms in the model (i.e., it statistically removes VN_gen, leaving us with only VN_Instrumental). For this reason, we test all three VBs simultaneously.

2.3.3. Weighting

Because the samples from each cultural region had different sizes, the data were weighted so that each region had an equal influence in the analyses. This weighting was performed by dividing the intended proportion for each sample (0.20) by the observed proportion in the sample for a given region. These values were supplied to both the lavaan, lmer, and ordinary regression analyses. The weights were also used in the calculation of all means and standard deviations where standardization was used.

3. Results

3.1. Configural Invariance (Research Question 1)

The configural model—which tested the unconstrained three-factor model—assessed whether the same factor structure was appropriate in all five cultural regions. The model fit the data well. The other models fit significantly worse compared to the three-factor configural model (ps < 0.001; see

Table 3. Comparison of models.

Model	df	χ2	CFI	SRMR	RMSEA	90% CI
Configural Three-Factor	1135	2583.76	0.924	0.049	0.071	[0.067, 0.074]
Metric Three-Factor	1215	2779.34	0.917	0.069	0.072	[0.068, 0.075]
Partial Scalar Three-Factor	1259	3064.16	0.907	0.071	0.074	[0.071, 0.078]
Scalar Three-Factor	1295	3565.47	0.887	0.076	0.081	[0.078, 0.084]
Relatl. vs. Instru./Intrin.	1145	3763.14	0.853	0.069	0.098	[0.095, 0.102]
Instru. vs. Relatl./Intrin.	1145	4223.98	0.839	0.071	0.103	[0.099, 0.106]
Intrin. vs. Relatl./Instru.	1145	4234.73	0.829	0.076	0.106	[0.103, 0.109]
One-Factor	1150	5254.42	0.766	0.086	0.124	[0.120, 0.127]

Note. As a reminder for the reader, Instrumental VBs were captured using items such as “Plants and animals are valuable because of the benefits they provide to people”; Intrinsic VBs were captured using items such as “Plants and animals are important because they are parts of a thriving ecosystem”; and Relational VBs were captured by items such as “Ecosystems are important because they are part of our community, like neighbors”. CFI = Confirmatory Fit Index; SRMR = Standardized Root Mean Residual; RMSEA = Root Mean Square Error of Approximation.

for all model fits), further indicating that the configuration (i.e., factor structure) was invariant across cultural regions. This comparison provides strong evidence of configural invariance, suggesting that the broad factor structure is consistent across the regions.

3.2. Metric Invariance (Research Question 2)

The metric model—where loadings were constrained to be equal across cultural regions—tested whether the same latent constructs were captured in each cultural region. This model fit well (see Table 3). More importantly, it showed a non-meaningful change in fit (∆CFI = −0.007; ∆RMSEA = 0.001; ∆SRMR = 0.02) compared to the configural model. This comparison provides strong evidence of metric invariance, suggesting that the latent constructs are captured in the same way across the regions.

3.3. Scalar Invariance (Research Question 3)

The scalar model—where loadings and intercepts were constrained to be equal across cultural regions—tested whether the latent constructs were influenced by the same amount of systematic bias across cultural regions. This model fit acceptably (see Table 3). However, fit indices from the comparison to the metric model showed inconsistent evidence of scalar invariance (∆CFI = −0.030; ∆RMSEA = 0.009; ∆SRMR = 0.007). That is, two (RMSEA and SRMR) out of the three fit indices changed negligibly. One fit index (CFI), however, surpassed the threshold (<−0.010). Accordingly, we attempted to identify a model in which we could achieve partial invariance by releasing some intercepts in the model to be freely estimated for each cultural region rather than fixed to be the same.

To pursue partial invariance, we considered the items with the greatest variability (as determined by the interquartile range of each item intercept across groups) in the freely estimated intercepts from the metric model. In other words, we looked for items where the intercepts were notably different across cultural regions when allowed to vary freely (see

Table 4. Items with notably deviant intercepts based on scalar model.

Subscale	Item	IQR
Instrumental	Ecosystems are valuable because they contribute to modern medicine.	0.41
Relational	Nature is important because it strengthens our community.	0.35
Instrumental	Nature is important because it benefits people.	0.33
Intrinsic	Forests are important because they provide habitats for plants and animals.	0.32
Intrinsic	Plants and animals are important because they are parts of a thriving ecosystem.	0.28
Relational	Ecosystems are important because they are part of our community, like neighbors.	0.27
Intrinsic	Ecosystems are valuable because they support healthy lives for plants and animals.	0.27
Relational	Ecosystems are valuable because they are like family members.	0.25
Instrumental	Nature is valuable because it improves our quality of life.	0.24
Relational	Nature is important because of the relationships people form with it.	0.23
Intrinsic	Nature is valuable for its own sake.	0.21
Intrinsic	The environment has the right to flourish.	0.20
Instrumental	Plants and animals are valuable because of the benefits they provide to people.	0.20
Instrumental	Nature is important because it provides us with resources.	0.15
Intrinsic	Nature deserves to be healthy.	0.13
Relational	Plants and animals are important because they can teach people many lessons.	0.13
Relational	Nature is important because it is part of human spirituality.	0.10
Relational	Nature is important because it is inspirational.	0.08
Relational	Nature is important because it is embedded in people’s traditions.	0.06
Instrumental	The environment is important because it ensures human wellbeing.	0.06
Intrinsic	Ecosystems have the right to exist.	0.06
Instrumental	Biological processes in nature are valuable because they help clean water and air for people.	0.05
Instrumental	Nature is important because it provides us with food and water.	0.03

Note. For those interested in using the scale, the Appendix A contains full details for implementing them. Bolded items here were allowed to be free in the partial scalar model.

; inspection of the sub-samples with the most invariance for these items is reported in Supplemental Materials). From this information, we selected three items at a time (one from each factor) that had the highest variation in intercepts and iteratively released them either until the model reached the invariance threshold or until we had released more than half of the items for a given latent factor (i.e., when a majority of items were variant; see [52]). In the end, we released a total of nine items (noted in the Appendix A) and were able to identify a model that both (a) fit the data to standards (see Table 3) and (b) resulted in a negligible drop in fit (∆CFI > −0.001; ∆RMSEA = 0.003; ∆SRMR = 0.003). Thus, we found partial scalar invariance, suggesting that, while there may be some systematic bias captured within raw means, the latent variables are comparable using SEM.

3.4. Between- and Within-Culture Differences (Research Question 4)

Because we found evidence of partial scalar invariance, it was possible to test the mean differences across cultural regions. However, in cases of partial scalar invariance, it is only appropriate to compare latent means (rather than observed composites—i.e., averages of the raw observed variables; see [52]). Therefore, latent scores for each participant were exported from the fitted partial scalar model (using lavPredict() from lavaan) and subsequently used in a multi-level regression. The scores for each latent VB were standardized to allow for comparisons both between and within cultures. Thus, we conducted a multi-level analysis (with random intercepts clustered by participant) predicting standardized VB endorsements from cultural regions and VB types, as well as their interaction. Both categorical variables were dummy-coded.

The first-order model revealed an omnibus main effect of cultural region, F(4, 2497) = 19.24, p < 0.001. The nuanced comparisons with full statistics are reported in the Supplemental Materials. However, the simple pattern was that Taiwan, Italy, and the US tended to value nature (for any reason) the most, while France valued nature slightly less but much more than Japan, which valued nature the least. All three VBs were standardized individually, which gave them the same sample-level mean (i.e., zero), so the main effect of the VB type is uninformative and not reported here.

More informatively, the second-order model revealed a significant omnibus interaction between the value type and cultural region, F(8, 5272) = 59.82, p < 0.001 (see Figure 1). Table 4 displays the full collection of simple effects. In brief, for Relational VBs, Taiwan had the strongest endorsements, followed by Italy, the US, and France (which had similar endorsements), while Japan had dramatically weaker endorsements than all other cultural regions. For Instrumental VBs, the US had the strongest endorsements, followed by Italy, Taiwan, and France (which had similar endorsements), while Japan had dramatically weaker endorsements than all other cultural regions. For Intrinsic VBs, Italy had the strongest endorsements, followed by the US, France, and Taiwan (which had similar endorsements), while Japan had dramatically weaker endorsements than all other cultural regions. We explore the possible cultural explanations for these patterns in Section 4.

For the within-culture comparisons (see

Table 5. Between-culture differences in endorsements of the three VBs.

Comparison	β	df	t	SE	95% CI	p	Adj. p
Relational VBs
TWN–JPN	1.17	4520	23.64	0.05	[1.07, 1.26]	<0.001	<0.001
TWN–ITL	0.34	4338	6.78	0.05	[0.24, 0.44]	<0.001	<0.001
TWN–FRN	0.44	3430	7.30	0.06	[0.32, 0.55]	<0.001	<0.001
TWN–USA	0.42	3506	7.10	0.06	[0.30, 0.53]	<0.001	<0.001
JPN–ITL	−0.82	3951	−15.34	0.05	[−0.93, −0.71]	<0.001	<0.001
JPN–FRN	−0.73	3266	−11.70	0.06	[−0.85, −0.61]	<0.001	<0.001
JPN–USA	−0.75	3326	−12.21	0.06	[−0.87, −0.63]	<0.001	<0.001
ITL–FRN	0.09	3213	1.47	0.06	[−0.03, 0.22]	0.142	1.000
ITL–USA	0.07	3269	1.18	0.06	[−0.05, 0.20]	0.236	1.000
FRN–USA	−0.02	2928	−0.27	0.07	[−0.16, 0.12]	0.784	1.000
Instrumental VBs
TWN–JPN	1.13	4520	22.96	0.05	[1.04, 1.23]	<0.001	<0.001
TWN–ITL	−0.02	4338	−0.36	0.05	[−0.12, 0.08]	0.722	1.000
TWN–FRN	0.02	3430	0.34	0.06	[−0.10, 0.14]	0.731	1.000
TWN–USA	−0.13	3506	−2.24	0.06	[−0.25, −0.02]	0.025	0.253
JPN–ITL	−1.15	3951	−21.44	0.05	[−1.26, −1.05]	<0.001	<0.001
JPN–FRN	−1.11	3266	−17.82	0.06	[−1.24, −0.99]	<0.001	<0.001
JPN–USA	−1.26	3326	−2.60	0.06	[−1.39, −1.14]	<0.001	<0.001
ITL–FRN	0.04	3213	0.61	0.06	[−0.09, 0.16]	0.544	1.000
ITL–USA	−0.11	3269	−1.82	0.06	[−0.24, 0.01]	0.069	0.693
FRN–USA	−0.15	2928	−2.17	0.07	[−0.30, −0.02]	0.030	0.302
Intrinsic VBs
TWN–JPN	0.88	4520	17.79	0.05	[0.78, 0.98]	<0.001	<0.001
TWN–ITL	−0.36	4338	−7.11	0.05	[−0.46, −0.26]	<0.001	<0.001
TWN–FRN	−0.01	3430	−0.19	0.06	[−0.13, 0.11]	0.849	1.000
TWN–USA	−0.13	3506	−2.19	0.06	[−0.24, −0.01]	0.029	0.288
JPN–ITL	−1.24	3951	−23.05	0.05	[−1.34, −1.13]	<0.001	<0.001
JPN–FRN	−0.89	3266	−14.25	0.06	[−1.01, −0.77]	<0.001	<0.001
JPN–USA	−1.01	3326	−16.39	0.06	[−1.13, −0.89]	<0.001	<0.001
ITL–FRN	0.35	3213	5.49	0.06	[0.22, 0.47]	<0.001	<0.001
ITL–USA	0.23	3269	3.71	0.06	[0.11, 0.35]	<0.001	0.002
FRN–USA	−0.12	2928	−1.67	0.07	[−0.25, 0.02]	0.095	0.951

Note. Some βs are greater than 0 because the dummy-coded variables were not standardized; the slopes represent the literal standardized difference between two cultural regions (vs. a true standardized regression coefficient). Adjusted p-values are reported using a Bonferroni correction to account for the 10 dependent tests. Controlling for age and gender does not change substantive conclusions (based on effect size and adjusted p-values) regarding this table or the presence of the interaction.

), given the degrees of freedom and the likelihood of obtaining significance even for trivial effects (|β| < 0.10), we based our interpretation on effect sizes rather than statistical significance. In Taiwan, Relational VBs were stronger than Instrumental VBs, which were stronger than Intrinsic VBs. In Japan, Relational VBs were stronger than Intrinsic VBs, which were stronger than Instrumental VBs. In Italy, Intrinsic VBs were stronger than Instrumental VBs and Relational VBs, the latter two of which were similar. In France, Instrumental VBs were slightly stronger than Relational VBs and Intrinsic VBs, which were both similar. Finally, in the US, Instrumental VBs were stronger than Intrinsic VBs, which were stronger than Relational VBs.

3.5. Place Attachment (Research Question 5)

We tested the associations among the three VBs and two dimensions of place attachment—place identity and place dependence. As shown in

Table 6. Within-culture differences in endorsements of the three VBs.

Comparison	β	df	t	SE	95% CI	p	Adj. p
TWN
Intrinsic–Instrumental	−0.13	5272	−4.23	0.03	[−0.18, −0.07]	<0.001	<0.001
Intrinsic–Relational	−0.40	5272	−13.42	0.03	[−0.46, −0.34]	<0.001	<0.001
Instrumental–Relational	−0.27	5272	−9.19	0.03	[−0.33, −0.21]	<0.001	<0.001
JPN
Intrinsic–Instrumental	0.13	5272	4.39	0.03	[0.07, 0.19]	<0.001	<0.001
Intrinsic–Relational	−0.11	5272	−3.66	0.03	[−0.17, −0.05]	<0.001	0.001
Instrumental–Relational	−0.24	5272	−8.05	0.03	[−0.30, −0.18]	<0.001	<0.001
ITL
Intrinsic–Instrumental	0.22	5272	7.33	0.03	[0.16, 0.28]	<0.001	<0.001
Intrinsic–Relational	0.31	5272	1.35	0.03	[0.25, 0.36]	<0.001	<0.001
Instrumental–Relational	0.09	5272	3.02	0.03	[0.03, 0.15]	0.003	0.008
FRN
Intrinsic–Instrumental	−0.09	5272	−3.15	0.03	[−0.15, −0.04]	0.002	0.005
Intrinsic–Relational	0.05	5272	1.71	0.03	[−0.01, 0.11]	0.087	0.260
Instrumental–Relational	0.14	5272	4.87	0.03	[ 0.09, 0.20]	<0.001	<0.001
USA
Intrinsic–Instrumental	−0.13	5272	−4.33	0.03	[−0.19, −0.07]	<0.001	<0.001
Intrinsic–Relational	0.15	5272	5.02	0.03	[0.09, 0.21]	<0.001	<0.001
Instrumental–Relational	0.28	5272	9.35	0.03	[0.22, 0.34]	<0.001	<0.001

Note. As with other tables, the dummy-coded variables comparing value type were not standardized. Thus, slopes represent the standardized difference between two VBs. Adjusted p-values are reported using a Bonferroni correction to account for the three dependent tests within each country.

, all measures were positively associated. All VIFs, however, were less than 5, suggesting that multicollinearity was not a concern.

Of interest were the unique associations between the VBs and each place attachment dimension (see

Table 7. Pseudo-correlations between values and place attachment when controlling for cultural region.

Variable	1.	2.	3.	4.	5.
Value-Bases
1. Relational VBs	—	0.72 **	0.71 **	0.12 **	0.21 **
2. Intrinsic VBs	0.69 **	—	0.80 **	0.04 *	0.14 **
3. Instrumental VBs	0.65 **	0.76 **	—	0.08 **	0.17 **
Place Attachment
4. Place Dependence	0.12 **	0.04 *	0.08 **	—	0.64 **
5. Place Identity	0.23 **	0.16 **	0.20 **	0.67 **	—

Note. * p < 0.05; ** p < 0.01. The cells of this table are standardized coefficients from a regression predicting each variable in the pair when controlling for cultural region. Therefore, the matrix is not perfectly symmetric. These pseudo-correlations were chosen rather than typical correlations expressly to highlight the comparison noted in Section 4.4.1. A similar correlation table is provided for each country in the Supplemental Materials.

for correlations). To test the unique associations, we conducted linear regressions predicting place identity and place dependence from the three VBs when controlling for cultural region (see

Table 8. The unique association between each value-basis and the two dimensions of place attachment.

Dimension	β	df	t	SE	p
Place Identity
Relational VBs	0.20	2594	6.96	0.03	<0.001
Instrumental VBs	0.10	2594	3.00	0.03	0.003
Intrinsic VBs	−0.06	2594	−1.81	0.03	0.070
Place Dependence
Relational VBs	0.17	2594	5.83	0.03	<0.001
Instrumental VBs	0.07	2594	2.13	0.03	0.033
Intrinsic VBs	−0.13	2594	−3.87	0.03	<0.001

). The regression predicting place identity from the three VBs when controlling for region revealed that Relational VBs were the strongest predictor of place identity. Specifically, place identity was positively predicted by Instrumental VBs and Relational VBs but not predicted by Intrinsic VBs. The effect of Relational VBs on place identity was roughly twice as strong as the effect of Instrumental VBs. The regression predicting place dependence also revealed that Relational VBs were the strongest predictor. Specifically, place dependence was positively predicted by Instrumental VBs and Relational VBs and negatively predicted by Intrinsic VBs. Like place identity, the effect of Relational VBs on place dependence was roughly twice as strong as the effect of Instrumental VBs.

4. Discussion

4.1. Invariance (Research Question 1–3)

The primary purpose of this research was to test whether a three-factor framework for understanding environmental VBs generalizes outside of the US context in which it was validated quantitatively (see

Table 9. Summary of research findings.

Research Question	Finding	Implication
Does the Environmental Value-Bases Scale form the same three factors across cultural regions?	Evidence of configural invariance implies that environmental VBs do form the same three factors across cultural regions.	The broad distinction between the three reasons for valuing nature seem to apply even across a great deal of cultural diversity.
Do the items forming the three factors have invariant loadings across cultural regions?	Evidence of metric invariance implies that the environmental VBs are reflected by the same underlying latent constructs across cultural regions.	The underlying (latent) construct reflecting each of the three VBs seems to take the same shape even within culturally diverse samples.
Do the items forming the three factors also have invariant intercepts across cultural regions?	Evidence of partial scalar invariance implies that intercepts are invariant enough for latent means to be modeled accurately by SEM analyses.	It is possible to compare latent means between countries, so long as one uses appropriate modeling techniques.
Are there cross-cultural differences in endorsements of the VBs between and within culture?	Yes, the cultural sub-samples vary in both their absolute endorsements of each VB (i.e., between cultures) and their endorsements of each VB relative to each other (i.e., within a culture).	While the three VBs may take the same shape and have the same structure across cultural contexts, their endorsements vary from one culture to the next—and within cultures.
How are the three VBs related to the two dimensions of place attachment? Are there particularly strong relationships with Relational VBs?	As expected, Relational VBs seem to be more related to place attachment than the other VBs.	Place attachment and Relational VBs are distinct but related constructs. Place attachment should not be used as a proxy for Relational VBs.

for a summary of findings). Overall, we found evidence that the Environmental Value-Bases Scale captures the same latent constructs in each cultural region, suggesting that treating Relational, Instrumental, and Intrinsic VBs as three separate constructs may indeed generalize outside of the US to non-English-speaking countries and non-Western countries. To our knowledge, this is the first quantitative evidence to suggest that the distinction between these VBs is shared across vastly different cultural regions. That is, while qualitative and conceptual work has asserted that these three VBs are shared across the globe (e.g., [9,13]), there have not yet been any investigations empirically documenting this assertion. Thus, until now, it was unclear whether the three VBs have a degree of shared structure across any cultures. We, however, demonstrate that the same set of items can be used to adequately capture the underlying latent constructs and, more importantly, that those latent constructs manifest in a common, comparable manner. The analyses here provide evidence suggestive of a shared structure across cultural regions and is a first step toward understanding the potential universality of these three VBs.

Ultimately, in demonstrating the cross-cultural generalizability of the latent constructs captured by this scale, this research paves the way for rigorous investigations into both the magnitude of environmental VBs across the globe—similar to recent investigations looking at which cultures are most disconnected to nature and which are most connected [75]—and the ways in which environmental VBs shape pro-environmental outcomes—similar to the original validation of this scale. Given Relational VBs’ clear relevance to sustainability [8,9,13,76], this is a valuable step forward, one which corroborates the primary assertion of the major qualitative reviews of the environmental values discourse [9,14]: there are at least three distinguishable reasons that people value nature across the globe. Moreover, given the clear variation in VB endorsements within each cultural region, this research provides an important tool to measure and assess the three VBs (as discussed in Section 4.4).

4.2. Differences in VBs Between and Within Cultures (Research Question 4)

We tested whether there were differences in the endorsements of the three VBs both between and within the cultural regions included in our sample. Most notably, we found that Relational VBs were endorsed most strongly in the Taiwanese sample (one of the two Eastern cultures) and least strongly in the Japanese sample (the other Eastern culture). At first, this may seem surprising, as one might have expected Japanese participants and Taiwanese participants to have similar (and high) endorsements of Relational VBs given that they both come from Eastern regions. However, a recent analysis of global levels of nature connectedness has shown that Taiwan has comparatively high levels of nature connectedness, whereas Japan was toward the bottom of the sample, which included more than 60 countries [75]. In light of this finding, the pattern observed here is less surprising.

Moreover, it is worth considering the low Relational VBs in the Japanese sample in light of the other VBs, which were likewise low. That is, it was not just that the Japanese sample scored low on only Relational VBs, but rather on all VBs. It is, therefore, potentially useful to consider distinctions in other cultural dimensions beyond individualism/collectivism that may explain the low VB endorsements overall. For example, some research suggests that, compared to France [23,77] or the USA [23], Japan tends to more strongly endorse masculine cultural values. Such masculine cultural values are potentially inconsistent with endorsing environmental values of any kind because of the stereotype of environmentalism as a primarily feminine concern (see [78]). Thus, this dimensions of Japanese culture (rather than collectivism), may explain the overall low environmental VBs in the Japanese sample. Accordingly, given the within-culture pattern observed in the Japanese sample, it seems that the low Relational VBs are potentially a byproduct of other cultural processes driving down overall VB endorsements, making the within-culture comparisons more informative.

Indeed, upon closer inspection, the within-culture pattern is more consistent with our initial expectations, with both the Taiwanese and Japanese samples being the only two cultural regions where Relational VBs were rated the highest relative to the other VBs. Thus, even though the Japanese sample’s absolute level of endorsement for Relational VBs was the lowest across cultural regions, it still showed a stronger relative orientation toward Relational VBs in comparison to the other two VBs, more in line with what one might have expected of participants from an Eastern culture. Even so, the findings from the Japanese sample do highlight an important point: researchers may want to carefully consider the validity of prevailing assumptions regarding cross-cultural environmental values—and, more importantly, be mindful of relying on stereotypes to make predictions about other cultures.

It is also informative to consider the other two VBs as well. Instrumental VBs—which are a notoriously problematic (i.e., often negative) predictor of pro-environmental outcomes (see [8] for a brief review)—were highest in the US sample. Incidentally, they were also the strongest VBs endorsed amongst US participants. Thus, not only did the US sample seem to endorse this problematic VB the most, but it also seemed to dominate the VB hierarchy in this cultural context. It is also interesting that Intrinsic VBs were endorsed only slightly less than Instrumental VBs in both the US sample and the French sample (which also endorsed Instrumental VBs the most). Both patterns are consistent with the noted presence of an instrumental–intrinsic dichotomy that has dominated Western environmental psychology and academic discourse [8,34]. Both patterns are also consistent with the strong emphasis on individualism within both cultures [23,24,77] which is a strong driver for utilitarian-consistent orientations toward power and achievement values [60]. Such values (i.e., power and achievement) may very well be important predictors of seeing nature as valuable mostly for its instrumental benefits.

Finally, it is interesting to note that, of all the cultural regions, Intrinsic VBs were highest in the Italian sample and were also these participants’ strongest VB. However, given that Italy used a different sampling strategy relative to the other cultural regions, it is possible that this effect is due to a self-selection bias (whereby more environmentally inclined individuals—who ostensibly show a greater appreciation for nature’s intrinsic values—were more likely to participate in the survey). For this reason, we avoid over-interpreting this finding and recommend further investigation with a non-convenience sample.

4.3. The Association Between VBs and Place Attachment (Research Question 5)

We tested whether the three VBs were differentially associated with two dimensions of place attachment: place identity and place dependence. We found that all three VBs were uniquely associated with both dimensions of place attachment. However—as one might expect given the shared emphasis on human–environment relationships found in Relational VBs and place attachment—Relational VBs showed the strongest positive association with place identity and place dependence. Yet the strength of the Relational VB–place attachment association (which was small by typical standards) indicates that the two constructs are distinct but related. This is an important theoretical observation—as past research has attempted to use place attachment as a proxy for Relational VBs (see [16])—and is largely consistent with the evidence in the original validation of the scale showing that nature connectedness and Relational VBs are related but distinct constructs [8]. Thus, while both Relational VBs and place attachment reflect one’s relationship with their surroundings, they are different constructs and should not be used interchangeably.

Like Relational VBs, Instrumental VBs were also positively associated with place attachment, yet this relationship was quite weak. It makes sense that people’s relationships with place would be influenced by instrumental considerations. For one, as noted in the introduction, place dependence emphasizes the benefits a place has for oneself [68]. Thus, the association between Instrumental VBs and place dependence is not surprising. However, it was not just place dependence that was associated with Instrumental VBs; Instrumental VBs were also associated with place identity, which is less obviously aligned with self-focused reasons for attachments to place [68]. Yet this still makes sense given (a) the literature on relationships between people and nature (see [20,76]) and (b) that place identity focuses on the affective attachment one has to a place [68]. For example, drawing on the parallel between human–nature relationships and interpersonal relationships [20,76], research shows that people often form relationships with those they feel will help them improve themselves, and people often come to identify with those with whom they form relationships (see [79] for discussions of both). Thus, recognizing the instrumental value of nature might contribute to our attachment to both place dependence and place identity as two dimensions of our relationship with place.

While the association between Instrumental VBs and place attachment can be partly explained, the unexpected association between Intrinsic VBs and place attachment is more puzzling. While Intrinsic VBs were positively associated with both dimensions of place attachment on their own (i.e., only controlling for cultural region), they showed no association with place identity and a negative association with place dependence after controlling for Instrumental VBs and Intrinsic VBs. One possible explanation for this unexpected association, however, may be that Intrinsic VBs are non-anthropocentric to such a degree as to become anti-anthropocentric. That is, because we ultimately found that higher endorsements of values that fail to acknowledge people (i.e., Intrinsic VBs) were associated with feeling less dependent on place, these findings are largely in line with suggestions that Intrinsic VB-based and rights-based arguments about nature might reify a separation between people and nature [8,80,81]. To rely on a human–nature relationships-as-relationships analogy [20,76], solely valuing the benefits that one’s relationship partner receives without acknowledging the benefits that one receives oneself or that the relationship receives itself is not likely to promote an attachment to the relationship. Likewise, focusing on the benefits that nature receives from being healthy without acknowledging one’s own benefits from nature or the benefits to one’s relationship with nature is not likely to promote attachment to the planet or place.

Even so, it should still be noted that the pattern we have observed is still largely inconsistent with the literature, which has shown strong associations between nature connectedness and Intrinsic VBs [8,27,28]. Yet the present effect was demonstrated in a large multi-cultural sample, suggesting that it might generalize to other cultural contexts, although this finding should be confirmed in future research. Thus, if nothing else, the present findings suggest that, despite their conceptual similarities, nature connectedness and place attachment are dissimilar in how they relate to Intrinsic VBs.

4.4. Recommended Usage of the Environmental Value-Bases Scale

The present research provides four new versions of the Environmental Value-Bases Scale in Traditional Chinese, Japanese, French, and Italian. This, hopefully, will make it possible for more researchers to make use of the scale and will contribute to rectifying the lack of coherence within the literature (see [57]). Therefore, we provide a handful of recommendations below considering the present findings.

4.4.1. Testing All Three VBs Simultaneously

The analyses with place attachment presented in Section 3.5 also serve to highlight the usefulness of a scale specifically designed to capture the three environmental VBs simultaneously. Specifically, as noted in the data analysis section, this is because one should always control for the three VBs to ensure that any uncovered associations are unique to the value-basis of interest and not just a general valuing of the environment. The distinction of specific VBs is important both when VBs are the outcome and when they are predictors (see [8]). This necessity is illustrated well by the comparison between the associations reported in Table 6 (which reports the effects without controlling for the other two VBs) and Table 7 (which reports the effects when controlling for the three VBs simultaneously): had we not deliberately tested the unique associations, we would have erroneously concluded that Intrinsic VBs were positively associated with place attachment because of their significant bivariate association with place attachment. Accordingly, the fact that this scale not only adequately captures Relational VBs across five separate cultural regions but also captures Instrumental and Intrinsic VBs as well makes it a valuable tool for studying these phenomena globally. Given the above considerations, even if only one VB is of interest, future researchers using the scale should be sure to include all three VB subscales and control for them in any analyses to properly isolate the construct of interest.

4.4.2. Implications of Configural Invariance

The test of configural invariance indicates that future researchers can reasonably rely on the three-factor framework tested here when using any one of the five scale versions included in the present paper. While this simplified factor structure represents a minor deviation from the original validation of the scale, it will hopefully make the scale easier for researchers to implement.

4.4.3. Implications of Metric Invariance

Given that the test of metric invariance supported the metric model, the scale can be used as one would any other scale (i.e., by creating composites through simply averaging raw item scores for a given subscale) so long as the researcher is not using the scale to make between-country comparisons. In other words, for simple cases of using regressions and similar analyses to (a) predict outcomes from the VBs or (b) predict one of the VBs from continuous variables (or non-country categorical variables such as gender), there is no need for latent variable modeling. That is, the evidence presented here suggests that the scale can be used to compare associations between the VBs and other constructs across cultural regions [52,53,54]. Thus, researchers can test for (in)consistency in associations across cultural regions and be confident that the differences in associations are meaningful and not due to different latent constructs being captured in each cultural region.

4.4.4. Implications of Partial Scalar Invariance

Given the partial scalar invariance, however, one cannot currently use simple composites to compare differences between cultures or countries [52]. This is because the partial scalar invariance indicates that scores on those composites partly reflect meaningful differences in the latent construct and partly reflect some degree of systematic response bias not attributable to the latent construct. Therefore, using such composites would run the risk of two groups with identical latent scores appearing significantly different simply because they have different response biases. That is, any significant differences in observed means could easily be due to systematic artifacts (e.g., some cultures tending to avoid answering with extreme values; [82,83,84]) rather than differences in the latent construct. In light of this, it is instead necessary to use latent variable modeling, which has been shown to be able to still estimate the latent construct largely accurately even under conditions of partial invariance (see [52]). Future research, however, could investigate whether changes to the scale items could achieve full scalar invariance while maintaining metric invariance—although, with the many latent variable modeling tools presently available, this is not especially necessary.

4.5. Broader Implications

The evidence presented here suggests, first and foremost, that there do seem to be three distinguishable reasons people tend to value nature and which share enough similarity across cultures that they were supported even when using a culturally diverse sample. Until now, such evidence had not been documented. Given that past research has shown how policy success and acceptability are critically influenced by their match with individuals’ reasons for valuing and caring about nature [22], this research highlights that future policy and practice will benefit from considering the level of VB endorsements of stakeholders (e.g., constituents). Policies and practices centered around Relational VBs will likely be more successful in places where they serve as the primary basis for valuing nature. Conversely, policies and practices centered around Instrumental VBs will likely be poorly received in places where nature is valued on a primarily relational basis. This, of course, requires policy-makers (and those informing policy-makers) to have a clear understanding of which VBs communities hold. Fortunately, the measure presented here seems to hold the potential to serve that very purpose given its utility despite this wide range of cultural contexts.

Additionally, policies—and global entities such as the UN [17]—have recently begun emphasizing improving the human–nature relationship as a concrete policy target (see [20] for a review of this trend; [1,18,19,85,86]). Given that Relational VBs are an inherent part of the human–nature relationship [76], it is likely that policies and grass-roots campaigns aimed at promoting such VBs will likely begin to emerge (see [87] for an example of a similar campaign). It will, therefore, become necessary to evaluate the impact of such policies and campaigns. This can only be performed with valid, generalizable quantitative scales. Fortunately, again, the Environmental Value-Bases Scale represents a useful tool for such evaluation efforts.

4.6. Limitations and Future Directions

4.6.1. Pro-Environmental Behavior

Although we would have liked to reassess the relationship of our Environmental Value-Bases Scale with a measure of pro-environmental behavior, there was no such measure in the present study. The current analyses were performed on data collected as part of a larger project investigating land-use change as part of a funded grant, thereby limiting the number of questions that could be placed in the survey. However, in the original validation of this measure, Relational VBs consistently predicted pro-environmental behavior in three US samples, indicating that Relational VBs had the strongest positive effect on pro-environmental behavior, followed by Instrumental VBs, whereas Instrumental VBs had a consistent negative effect [8]. This previously demonstrated predictive validity, alongside the present research showing that environmental VB subscales capture the same latent constructs across our five samples, suggests that similar associations with behaviors should, in theory, be found across the cultural regions—and, if they are not found, it would simply mean that the association between Relational VBs and pro-environmental behavior depends on the cultural context (since we found strong evidence of metric invariance here). Put simply, a scale that measures the same latent constructs in multiple cultural regions is not invalidated by potential cross-cultural differences in the relationship between that latent construct and the given outcomes, so long as there is at least one cultural context where a meaningful relationship exists (which has already been demonstrated). Given this fact, the absence of a measure of pro-environmental behavior in this dataset does not limit the value of the present findings or the utility of the scale.

4.6.2. Methodological Considerations

Item Presentation

There are a few methodological limitations to consider. First, considering the length of the overall survey, we chose not to randomize all the items presented as part of the Environmental Value-Bases Scale. Instead, unlike the original validation of the scale, we presented the items as blocks to limit participant fatigue. Therefore, it must be acknowledged that this arrangement of items may have created artificially weaker correlations between items from different scales. The correlations between factors, however, were like those found in the original validation, suggesting that presenting the items in blocks did not contribute to a noticeable difference in the degree of separation (or overlap) between factors (see also the correlation tables in the Supplemental Materials).

Sampling Strategy

Second, because this data was collected as part of a larger funded project bringing together natural and social scientists (which created strict geographical requirements), the samples collected were not selected to represent the larger culture regions in which they were located. Therefore, the samples may not generalize to their broader cultures. Even so, these analyses still hold cross-cultural value. Additionally, with these geographical requirements, it was also necessary for some sub-samples (i.e., Italy) to be recruited using more active means in order to recruit hard-to-reach samples. This means that there were differences in recruitment strategies which may have resulted in artificial differences between Italy and the other countries (as noted above). Thus, this could be one explanation for the notable difference in the relative pattern of Intrinsic VBs for the Italian sample. All other mean comparisons, however, are not attributable to recruitment differences, as the remaining sub-samples all used comparable recruitment methods. Still, the samples we used do require generalizing from subregions to the entire culture. Thus, it would be valuable for future research to test for variations within these cultural regions.

Despite these two methodological limitations, this research still holds great value. While there is considerable interest in research on values across academic fields, the quantitative research on relational environmental VBs is in its infancy. The present findings, therefore, provide a useful first step toward investigating these phenomena cross-culturally. Specifically, while the Japanese sub-sample might not be perfectly representative of all of Japan as a culture, it is both (a) more representative of Japanese culture than a sub-sample from outside of Japan and (b) less representative of the Western cultures frequently found in quantitative research. And yet, even when including this sub-sample, we found metric invariance. In other words, the three-factor structure of environmental VBs in four non-US and non-English speaking samples is consistent with the notion that such values are potentially globally generalizable (granted this evidence is not in any way meant to be framed as conclusive). That is, if the true reflection of reality had been that the three distinct latent constructs could not be generalized, we would not have found metric invariance. Thus, the evidence of metric invariance is a promising—even if only preliminary—indication that these VBs may generalize across diverse cultural regions. Therefore, while this research is limited to the specific regions in which these sub-samples were collected, the present findings suggest that further cross-cultural investigations of the three VBs are well-warranted and, in fact, needed. The test of metric invariance indicates, as noted throughout this paper, that the latent constructs reflecting the three VBs are applicable across a culturally diverse set of sub-samples, suggesting that the three-factor framework is, indeed, useful outside of Western, English-speaking countries such as the US. That said, there are numerous other cultures and cultural dimensions worthy of study in future research seeking to confirm the globality of these three VBs. Furthermore, as seen here, there is a clear variability in the endorsements of the VBs both across and within cultural regions, suggesting that such heterogeneity is likely in other contexts, including possible within-cultural differences (e.g., differences in these values between rural and urban populations, which were not considered in the present research).

5. Conclusions

With the rapid increase in publications on relational values [9,13,14,33] and with global policies increasingly referencing a desire to improve the human–nature relationship (e.g., [1,18,85]; see [20]), it is necessary to understand whether such Relational VBs are culturally bound—and therefore idiosyncratic to each culture—or, instead, if they have a degree of common structure across cultures. Supporting the latter, the present research offers preliminary evidence to suggest that Relational, Instrumental, and Intrinsic VBs are each formed by a latent construct that is shared even across five culturally distinct sub-samples. This not only provides a valuable theoretical insight—that these environmental VBs might generalize cross-culturally—but also paves the way for future cross-cultural research investigating the nature of environmental VBs. For example, here we show that relative endorsements of the three VBs vary across countries and have distinct associations with place attachment. Thus, the multi-group validation of the Environmental Value-Bases Scale provides a useful tool that allows for cross-cultural investigations of the relationship between relational values for pro-environmental outcomes and for gauging the effectiveness of global efforts to improve the human–nature relationship.