One for All?

Measures of One’s Connection with Nature

Measuring people’s connection with nature poses some well-known methodological challenges that generally apply to opinion polls. Abstract psychological concepts, such as “inclusion of nature in one’s self,” are difficult to convey to laypersons and difficult for laypersons to grasp intellectually. Moreover, one’s connection with nature is probably even beyond a person’s own conscious awareness (Schultz et al., 2004). Technically speaking, it is rather contentious to assume that people can consciously reflect on their connection with nature and accurately assess its magnitude. The least we can say is, self-reflection on one’s connection is vulnerable to recollection and response biases – since the information is not readily available for retrieval (cf. Dillman, 2001). Furthermore, assessing one’s felt or imagined connection with nature also requires that different persons assign numbers in the same manner when they decide on the extent of their personal connection.

While these technical concerns apply to almost all currently available connection-with-nature measures, there is, to our knowledge, one instrument that avoids employing self-reflection: the so-called “implicit-association-with-nature” measure by Schultz et al. (2004). This measure utilizes differential response times to indirectly determine a person’s connection with nature. Thus, it does not require a person’s conscious reflection on his or her connection with nature. Unfortunately, implicit association with nature only showed moderate overlap with explicit connection measures (maximum r = .33) and virtually no link to nature conservation or ecological behavior (r = .06). In other words, implicit association with nature is probably not the anticipated panacea for the multiple methodological challenges of explicit measures (cf. Schultz & Tabanico, 2007). Due to its lack of convergence with the other, conventional connection-with-nature measures, its validity remains at best doubtful. Moreover, the conceptual shift into the attitude domain implied in the measurement procedure of the implicit-association-with-nature measure might have affected convergence negatively as well.

Connection with Nature as an Attitude

By employing the implicit association test with their measure, Schultz et al. (2004) conceptually converted connection with nature into some kind of attitude (cf. Wittenbrink & Schwarz, 2007). Traditionally, an attitude is an internal state that is linked with people’s evaluative responses toward a certain object (Eagly & Chaiken, 1993). Logically, the object of the connection-with-nature attitude is either nature or certain parts of nature. With such a conceptual conversion, connection with nature can theoretically be assessed as an attitude based on a person’s evaluative responses. In other words, we can anticipate that statements expressing one’s appreciation for nature or for certain aspects of nature will be suitable. Examples of evaluative statements that are already employed in existing connection-with-nature measures are: (a) “I have never seen a work of art that is as beautiful as a work of nature, like a sunset or a mountain range,” (b) “I really enjoy camping and hiking outdoors” (Clayton, 2003), and (c) “I recognize and appreciate the intelligence of other living organisms” (Mayer & Frantz, 2004).

Since attitude measures – based on evaluative responses – explicitly ask for the extent of a certain appreciation, they rely on more or less challenging self-reflections. Not surprisingly, such measures are also accused of being vulnerable to response biases (e.g., Wittenbrink & Schwarz, 2007). An alternative approach – apart from the implicit association test – for bypassing intellectually demanding self-reflection in the measurement of a person’s attitude is grounded in what Kaiser, Byrka, and Hartig (in press) call Campbell’s paradigm. Previously, this paradigm, which formally describes the axiomatic connection between an attitude and a corresponding class of behaviors, was successfully used for assessing environmental¹ and health attitudes (see Byrka, 2009; Kaiser, Oerke, & Bogner, 2007). Its core idea is universal and should thus, as Kaiser and his colleagues stress, be generally applicable to any kind of attitude.

When applied to connection with nature, Campbell’s paradigm requires the assumption that the extent of a person’s overall connection with nature can be indirectly derived from a systematic inspection of both (a) reports of past activities that imply a bonding with nature (e.g., “helping snails to cross the street”) and (b) responses to evaluative statements that reflect regard for nature, such as “carving a tree feels like cutting myself.” Accordingly, a person’s connection with nature is expected to be most obvious in the face of progressively demanding bonding activities and of increasingly challenging evaluative statements reflecting appreciation. Expectedly, the more tangible and mental obstacles a person overcomes in expressing esteem for nature or reporting bonding with nature, the more substantial this person’s connection is. Why would someone express a liking for gardening, get up early to watch sunrises, or admit to publicly mimicking a vulture’s way of walking if he or she did not feel connected with nature? Likewise, when the tiniest difficulty or resistance, such as anticipated public embarrassment, is enough to stop a person from expressing his or her respective feelings and thoughts, the connection is probably comparatively low. When a person, for instance, admits to mourning the loss of a pet, but simultaneously prefers indoor to outdoor sports and kicks disobedient dogs, he or she presumably is not deeply connected with nature. Conceptually, Campbell’s paradigm implies a formal link between a person’s extent of connection and the verbal and behavioral means to implement this connection and express appreciation of nature.

Disposition to Connect with Nature

If a person is deeply connected with nature, we can expect this person to engage in all sorts of bonding activities and to appreciate nature in multiple ways. Such a compound of activities and of evaluations also involves people selecting various alternatives for realizing and for expressing their individual levels of connection with nature, leaving them with choices. Instead of gardening, people can talk to plants or deliberately watch clouds. They can prefer forest hikes to city strolls, or they can claim that watching animals is exciting. Any behavior that is performed, figuratively speaking, involves costs as it requires personal resources (e.g., time, effort, and courage): for instance, when one is about to publicly admit to some apparently bizarre and personally embarrassing statements. If one’s connection level can be implemented by a variety of different acts and be expressed by numerous verbal statements, we can anticipate a prudent selection of the behavioral means to its implementation and of the evaluative statements to its manifestation.

Naturally, people commonly prefer the more convenient, socially approved actions, over the more costly ones. Thus, we assume that people favor practicing less demanding actions over more challenging ones and prefer to make less controversial claims over ones that might raise an eyebrow. This becomes tangible in the probabilities of the occurrences of various behaviors and of different statements. This probability of a behavior or a statement is, in turn, anticipated to be a function of two components: (a) a person’s disposition to connect with nature and (b) the composite of all the figurative costs of a particular act or statement. Kaiser et al. (in press) propose the Rasch model for mathematically describing the aforementioned relationship. The classical Rasch model and, thus, the formal link between a person’s disposition to connect with nature and his or her probability of confirming a specific behavior or statement is depicted by the following formula (for more details, see e.g., Bond & Fox, 2007):

In this model, the natural logarithm of the ratio of the probability (p _ki ) of person k’s engagement or affirmation relative to the probability for nonengagement or denial (1 − p _ki ) of a specific activity or a specific statement i (implying a connection with nature) is given by the arithmetic difference between k’s disposition to connect with nature (θ _k ) and the composite of all the figurative costs involved in realizing a specific behavior or in acknowledging a certain statement i (δ _i ). Note that in this mathematical formalization, people are distinguishable with respect to the extent of their connection, and behaviors/statements are distinguishable by how demanding they are to realize or to admit.

Recalling previous behavior and expressing rather concrete personal preferences do not require any connection-with-nature-related self-reflection. In other words, behavioral self-reports and specific evaluative statements concerning nature are easy to answer as they primarily rest on recollection; as such they are probably not particularly vulnerable to response biases. Additionally, since we indirectly derive people’s dispositions to connect with nature from behavioral and evaluative statements, rather than from a direct exploration of the abstraction (e.g., one’s connection with nature), respondents usually fail to recognize the purpose of the measurement instrument. In summary, indirect measurement instruments based on Campbell’s paradigm are intellectually easy and not particularly vulnerable to forging.

Participants and Procedures

Participants were recruited by student mailing lists from the University of Zürich, Switzerland, mailing lists from sports clubs and music associations, and public forums on the Internet. In addition, links to our study were placed on the Internet pages of various Swiss newspapers. As incentives for completing the questionnaires, the inclusion in a lottery to win (a) an mp3 player (value: ~ US $80), (b) two vouchers for books (value: ~ US $20), and (c) three T-shirts (value: ~ US $15), and feedback on one’s personal results were offered.

Of the 2,935 persons who accessed the Internet questionnaire, 1,309 (response rate: 44.60%) completed it. Participants’ median age was 28 (M = 34.05, SD = 15.30; range: 18–80). The percentage of females was 45.16%. These figures imply that our sample cannot be regarded as representative. For the purposes of this research, however (i.e., scale development and the comparison of the strengths of relationships), it is sufficient that the participants reflect a wide range of diversity without indications of severe restrictions of range on the included variables.

Measures

The questionnaire consisted of seven instruments either originally developed in German or using translations of the original English instruments: Three explicit measures of connection with nature (i.e., Environmental Identity, Connectedness to Nature Scale, and inclusion of nature in one’s self), one indirect (i.e., Disposition to Connect with Nature), and one implicit (i.e., implicit association with nature) measure of connection, and two established scales for environmental concern and ecological behavior. For all items, “not applicable” was a response option when an answer could not be given. These answers were coded as missing values. Note that only data for those instruments for which participants had minimally answered 80% of the questions were retained.

Environmental identity involves 24 statements, such as “I think of myself as a part of nature, not separate from it” (Clayton, 2003). One’s personal point of view could be expressed with a 5-point Likert scale from 1 (strongly disagree) to 5 (strongly agree). Of all statements, 16.49% were found to be missing. The internal consistency of the identity measure was found to be α = .93 (N = 1,064). Person scores were conventionally calculated as mean values of the original 24 environmental-identity items.

Connectedness to nature (Mayer & Frantz, 2004) was assessed with 14 statements such as “I often feel a kinship with animals and plants.” Three of these statements were negatively formulated (expressing a lack of connectedness). For all questions, people could respond to a 5-point Likert scale from 1 (strongly disagree) to 5 (strongly agree). Across all statements, 11.42% were found to be missing. The internal consistency of the connectedness measure was found to be α = .80 (N = 1,121). Again, person scores were calculated as the arithmetic mean of the original 14 connectedness-to-nature items.

Inclusion of nature in one’s self is the third explicit connection-with-nature measure. This measure is based on only one item (see Schultz, 2001, 2002). By means of a series of seven differentially overlapping circles (labeled “self” and “nature”), participants could choose the one that best described how interconnected they felt with nature. This item was not answered by 10.74% of the participants. Since this measure is a single-item measure, its reliability could not be estimated from our data. Nevertheless, its 4-week test-retest reliability is reported to be r_tt = .84 (see Schultz et al., 2004).

Implicit association with nature was assessed with an Internet template developed by Schultz et al. (2004). It makes use of a specially designed version of the implicit association test (e.g., Greenwald, Nosek, & Banaji, 2003). The measure uses the response time difference that can be found when people link words from different categories. In Schultz and his colleagues’ version, people are compared with respect to how rapidly they associate either self-related words (i.e., one’s own first name) or other-related words (i.e., an unspecified first name) with nature-related words (such as “flower”) or with artifact-related words (such as “truck”; more details concerning the stimuli, the procedure, and the scoring algorithm are available upon request). The internal consistency of the implicit-association-with-nature measure turned out to be r = .67 (N = 734; details regarding this consistency estimate can also be found in Bruni & Schultz, 2010; Schultz & Tabanico, 2007).

A person’s disposition to connect with nature was assessed with 40 items (22 items originally developed by Beckers, 2005; see Table 1). Of these 40 items, 26 concerned behavioral self-reports presented with two different response formats: (a) for 17 behaviors, a 5-point frequency scale from 1 (never) to 5 (very often) was employed and (b) for nine behaviors, a dichotomous yes/no format was used. Based on Beckers’ (2005) research, we decided to recode the responses to the first 17 behaviors from a 5-point to a 3-point format by collapsing “seldom” and “occasionally,” as well as “often” and “very often.” “Never” was retained as “never.” The remaining 14 items contained evaluative statements presented with a yes/no format. Of the 40 items, three were negatively formulated, expressing a deficient disposition to connect with nature (see Table 1). Of all disposition items, 5.45% were found to be missing.

Table 1. Forty disposition-to-connect-with-nature items

		δ1	p1	δ2	p2
Note. δ represents the difficulty of an item expressed in logits; the more negative a logit value, the easier, and the more positive, the more difficult the particular item is. Logits stand for the natural logarithm of the engagement/nonengagement or endorsement/nonendorsement ratio, respectively, the natural log odds. p refers to the probability that the average person engages in the corresponding behavior or endorses a corresponding verbal statement. Some items have two difficulties and two probabilities: The first of the two stands for either “seldom” or “occasionally” engaging, and the second for either “often” or “very often” engaging. Items with only one difficulty and one probability were assessed using a yes/no format (representing engagement or endorsement, respectively). Items in italics were adopted from Beckers (2005). Shaded background represents a deficient connection with nature. Prior to the statistical analysis, the shaded items were reversed in their coding.
1.	I mimic animal behavior: for example, the way a vulture walks	1.26	.37	4.09	.03
2.	I get up early to watch the sunrise	−0.18	.71	3.35	.07
3.	I collect mushrooms or berries	0.34	.59	3.28	.07
4.	I mimic the sounds of animals	−0.23	.72	3.11	.08
5.	I cross meadows barefoot	−0.40	.75	3.08	.09
6.	I talk to plants	0.98	.43	2.80	.11
7.	I help snails cross the street	0.90	.45	2.46	.15
8.	I watch TV shows that have animals as the main characters	−1.24	.88	2.36	.16
9.	I have a CD or tape with recorded sounds of nature			2.15	.19
10.	I take time to watch the clouds pass by	−2.28	.95	2.06	.21
11.	I take time to consciously smell flowers	−1.60	.91	1.92	.23
12.	I consciously watch or listen to birds	−1.51	.90	1.88	.24
13.	I spend time in a park	−2.42	.96	1.79	.25
14.	I collect objects from nature such as stones, butterflies, or insects			1.69	.27
15.	I deliberately take time to watch stars at night	−2.57	.96	1.69	.27
16.	Indoor plants are part of the family			1.46	.32
17.	I would always prefer spending time with my friends to spending time alone in nature			1.42	.33
18.	Even when it is very cold or rainy I go out for a walk	−2.24	.95	1.35	.35
19.	I hike or run in nearby nature reserves or forests	−1.42	.89	1.08	.41
20.	I talk to animals	−1.40	.89	0.66	.51
21.	Carving a tree feels like cutting myself			0.47	.56
22.	If one of my plants dies, I reproach myself			0.43	.57
23.	If there is an insect, such as a fly, in my home, I try to catch and release it rather than kill it			0.41	.58
24.	The croaking of frogs is comforting			0.33	.60
25.	I enjoy gardening			0.32	.60
26.	I prefer living in a city			−0.04	.68
27.	I feel the need to be out in nature	−3.96	.99	−0.19	.71
28.	My favorite place is in nature			−0.28	.73
29.	Walking through a forest makes me forget about my daily worries			−0.93	.84
30.	As a child I spent time in the woods			−1.06	.86
31.	I personally take care of plants			−1.11	.86
32.	The noise of crickets gets on my nerves			−1.29	.88
33.	I prefer outdoor to indoor sports			−1.33	.89
34.	Listening to the sounds of nature makes me relax			−1.40	.89
35.	I prefer forest hikes to city strolls			−1.50	.90
36.	Pets are part of the family			−1.56	.91
37.	A cleared forest makes me miserable			−1.59	.91
38.	I mourn the loss of pets			−1.85	.93
39.	It makes me miserable to see a hedgehog that was hit by a car			−2.07	.94
40.	Watching animals is exciting			−2.09	.94

Table 1. Forty disposition-to-connect-with-nature items

View as image HTML

The New Ecological Paradigm (Dunlap et al., 2000) is a popular environmental-concern measure. The scale consists of 15 evaluative statements, such as “Plants and animals have as much right as humans to exist.” These statements were presented together with a 5-point Likert response format ranging from 1 (strongly disagree) to 5 (strongly agree). Seven items were negatively formulated (expressing an unconcerned attitude). Of all concern statements, 14.00% were missing. The internal consistency of the 15 concern items was α = .84 (N = 1,128). Once again, person scores were conventionally calculated as the arithmetic mean of the 15 original items.

For ecological behavior, we included 50 behavioral self-reports from Kaiser and Wilson (2004), such as “I buy meat and produce with eco-labels.” Of the 50 behaviors, 19 represented nonecological activities. Engagement in 18 behaviors was verified with a yes/no format and in 32 behaviors with a 5-point frequency scale ranging from 1 (never) to 5 (always). The responses to the latter set of behaviors were recoded into a dichotomous format by collapsing “never,” “seldom,” and “occasionally” into “unreliable ecological engagement.” “Often” and “always” were united into “reliable ecological engagement.” Of all possible behavior statements, 7.36% were found to be missing. The calibration of the behavior scale and the estimation of person scores – based on the classical Rasch model – were in line with previous calibrations of the same instrument: the General Ecological Behavior scale (see e.g., Kaiser & Wilson, 2004). All behavior items were found to acceptably fit the model, and the Rasch-model-based reliability-estimate of the measure turned out to be rel = .85 (N = 1,186).

One for All?

Based on the partial-credit Rasch model, we were able to successfully calibrate a Disposition to Connect with Nature scale. All 40 items fit the model prediction with reasonable mean square (MS) values between 0.86 and 1.16 (for reference values, see Wright & Linacre, 1994). Since our sample is relatively large, we relied on the MS-statistics in the assessment of item fit. MS-values reflect the relative discrepancy in variation between model prediction and observed data. They are not affected by sample size. For example, an MS-value of 1.20 indicates 20% more variation in the data than what was predicted by the Rasch model. Of the participants, only a tolerable number of 82 (6.26%) demonstrated poor fit with the model prediction (t > 1.96). And the Rasch-model-based reliability of our newly developed measure was accordingly also found to be good with rel = .89 (N = 1,309).

Convergent validity information was derived from the pattern of correlations between various connection-with-nature measures. In line with our hypothesis, the measurement-error-attenuation-corrected Pearson correlations of these connection-with-nature instruments revealed that four of the five measures substantially shared information with each other. This was indicated by consistently large effect sizes (.65 < r_corr < .79; see Table 2). By contrast, correlations with implicit association with nature were small (.10 < r_corr < .22).

Discriminant validity information becomes obvious in the correlations between the five connection-with-nature measures and environmental concern (measured with the New Ecological Paradigm). These effects were at most moderate in size for Disposition to Connect with Nature, Inclusion of Nature in Self, and implicit association with nature (.13 < r_corr < .39; see Table 2). However, Environmental Identity and the Connectedness to Nature Scale largely overlapped (r_corr = .57 and r_corr = .62, respectively) with environmental concern.

To further explore convergent and discriminant validity, we performed a maximum likelihood-based exploratory factor analysis with an oblique rotation (i.e., direct oblimin). We assumed the two factors (i.e., connection and concern) to be correlated, based on the corresponding and mounting empirical evidence (e.g., Mayer & Frantz, 2004). The use of maximum likelihood to extract factors was possible because multivariate normality can reasonably be assumed when using scale scores as the predominant type of data. A two-factor model fit the data reasonably well (N = 983): χ²(4) = 15.79, p = .003, compared to a one-factor model, χ²(9) = 115.50, p < .001: Δχ²(5) = 99.71, p < .001. Overall, 69.20% of the common variance of the six instruments was determined by two factors (see Table 3).

Table 3. Factor loadings and communalities (h²) of the five connection-with-nature measures and an environmental-concern scale

	Factors
	1	2	h2
Note. N = 983, bold figures indicate factor loadings > .60.
Disposition to Connect with Nature	.85	−.08	.66
Environmental Identity	.84	.09	.79
Connectedness to Nature Scale	.67	.17	.59
Inclusion of nature in self	.74	−.11	.48
Implicit association with nature	.11	.02	.02
New Ecological Paradigm	.04	.98	.99
Eigenvalues	3.16	0.99
Proportion of explained variance	52.72%	16.48%
Factor correlation		.46

Table 3. Factor loadings and communalities (h²) of the five connection-with-nature measures and an environmental-concern scale

View as image HTML

All communalities exceeded .40, except for the one for implicit association with nature (see Table 3). As can be seen in Table 3, and as theoretically anticipated, all connection-with-nature measures – except for the implicit-association-with-nature instrument – had nontrivial loadings of a > .60 on the same factor. All measures of people’s connection with nature (Disposition to Connect with Nature, Environmental Identity, Connectedness to Nature Scale, inclusion of nature in self, and implicit association with nature) loaded on the first factor with an eigenvalue of 3.16. Note that implicit association with nature loaded on this anticipated factor (a = .11) too, although not as clearly as the other connection-with-nature measures. The New Ecological Paradigm loaded on the second factor with an eigenvalue of 0.99. The factors were correlated with each other (r = .46).²

Even beyond the already acknowledged conceptual overlap of a connection with nature and environmental concern – recognizable in the correlation of the two factors: the Connectedness to Nature Scale and Environmental Identity – still revealed some substantial cross-loadings on the second factor: the concern factor (a = .17 and a = .09, respectively). Such an accentuated conceptual overlap, by contrast, did not appear with inclusion of nature in self and with the Disposition to Connect with Nature. Their cross-loadings happened to be negative (a = −.11 and a = −.08, respectively), which – in combination with the positive correlation of the two factors – numerically reduces the extent of the relationship with the second factor. No substantial cross-loadings were found with implicit association with nature. These results speak of the fact that the Connectedness to Nature Scale and the Environmental Identity measure also partly reflect people’s environmental concern and not exclusively connection with nature.

Predictive Validity

Predictive validity information for the newly developed Disposition to Connect with Nature scale was derived from a set of regression analyses anticipating ecological behavior. When Disposition to Connect with Nature was used as a single determinant, it explained 23.79% of the variance of General Ecological Behavior: β = .49, t(1,134) = 18.82, p < .001.

In a four-step hierarchical regression analysis, we first controlled for age and gender. Combined, they had a statistically significant effect, explaining 4.09% of the variance in ecological behavior. Both determinants turned out to be statistically significant (see Table 4). The New Ecological Paradigm, the environmental-concern measure, accounted for an additional 15.87% of the variance in ecological behavior when entered in Step 2. Entering Disposition to Connect with Nature in Step 3 increased the amount of explained variance by another 11.11%. Entering the remaining four connection-with-nature measures in Step 4 increased the explained variance by another 3.61%. However, only one of the newly entered variables (i.e., Environmental Identity) had a statistically significant influence on ecological behavior (see Table 4).³

In another model, we inverted Steps 3 and 4 to test the uniqueness of the disposition-to-connect-with-nature effect on ecological behavior. We found that our newly developed measure (even when entered in Step 4) could uniquely and significantly contribute 1.70% to explain ecological behavior: β = .21, t(963) = 5.02, p < .001. Overall, these findings speak to the supremacy of our newly developed, indirect disposition measure. It not only shares most of its variance with the other connection-with-nature measures, but also shares virtually no variance – apart from the empirically anticipated correlation on the conceptual level (e.g., Mayer & Frantz, 2004) – with the environmental-concern measure. It also accounts for extra variance in the prediction of ecological behavior.

Table 4. Hierarchical regression analysis predicting General Ecological Behavior

	Step 1			Step 2			Step 3			Step 4
	B	SE B	β	B	SE B	β	B	SE B	β	B	SE B	β
Note. N = 982.
*p < .001.
Age	0.01*	< 0.01	.15*	0.01*	< 0.01	.16*	< 0.01	< 0.01	.01	< 0.01	< 0.01	.04
Gender	−.30*	0.06	−.15*	−.13	0.06	−.07	0.02	0.06	.01	−.02	0.06	−.01
New Ecological Paradigm				0.71*	0.05	.41*	0.50*	0.05	.29*	0.38*	0.06	.22*
Disposition to Connect with Nature							0.33*	0.03	.39*	0.18*	0.04	.21*
Environmental Identity										0.42*	0.06	.30*
Connectedness to Nature Scale										−.11	0.07	−.07
Implicit association with nature										0.10	0.09	.03
Inclusion of nature in self										0.02	0.03	.02
∆R2	.04			.16			.11			.04
R2	.04			.20			.31			.35

Table 4. Hierarchical regression analysis predicting General Ecological Behavior

View as image HTML

Note that the results of the regression analyses concur with the findings of the factor analysis. Given that Disposition to Connect with Nature, Environmental Identity, inclusion of nature in self, and the Connectedness to Nature Scale share a substantial amount of variance (see Table 3), which can be collapsed into one single factor (representing connection with nature), we cannot expect all four individual indicators of people’s connection with nature to become significant in the prediction of a third variable. Necessarily, there is no variance left once the ecological-behavior variance is regressed on the common connection-with-nature factor (represented by the two most predictive of its indicators, Disposition to Connect with Nature and Environmental Identity).