Testing and Extending Schwartz Refined Value Theory Using a Best–Worst Scaling Approach

The Schwartz Value Theory

Schwartz’s (1992) circular structure of values captures the conflicts and compatibilities among the motivations that values express. He divided the continuum into 10 broad, distinct values for scientific convenience (see Figure 1a). Adjacent values in the figure (e.g., conformity and tradition) express compatible motivations, such that a choice to pursue one value can simultaneously promote the attainment of the other. Values that are distant around the figure (e.g., security and stimulation) express conflicting motivations, such that choices to pursue one value block or interfere with attainment of the other.

OPEN IN VIEWER

Figure 1.

The original circular motivational continuum (a) of 10 basic human values and the 4 higher order values from Schwartz et al. (2012) and (b) for the refined values, four higher order values and underlying dynamic sources.

Evidence supporting the circular structure of values comes from hundreds of samples in over 75 countries using different measurement instruments (Schwartz, 2015). Support for the structure was primarily found by examining spatial relations among value items produced by nonmetric multidimensional scaling (MDS). As expected, items usually formed regions around a circle that represented each of the 10 basic values and located them in their theorized order, with minor deviations.

The 10 basic values in Figure 1a have been measured by at least six different instruments. These include the Schwartz Value Survey (Schwartz, 1992), the PVQ (Schwartz et al., 2001), the Picture-Based Value Survey for children (Döring, Blauensteiner, Aryus, Drögekamp, & Bilsky, 2010), the Schwartz Values Best–Worst Survey (Lee, Soutar, & Louviere, 2008), and two paired comparison methods (Bilsky et al., 2015; Oishi, Schimmack, Diener, & Suh, 1998). The bases for these instruments differ widely. Two measure values with rating scales (Schwartz Value Survey and PVQ) and three use forced choice methods, including q-sort (the Picture-Based Value Survey), BWS (Schwartz Values Best–Worst Survey), and pairwise comparisons.

Five of the six instruments ask respondents to consider how important each value is to them personally or to compare the personal importance of different values for themselves. The PVQ (Schwartz et al., 2001) asks respondents to consider how similar they are to brief descriptions of gender-matched individuals. Schwartz et al. (2001, p. 523) created the PVQ to be more concrete and less cognitively complex than previous instruments and to “differ substantially from the SVS in its format and judgment task to provide an independent test of the theory of value content and structure.” The main difference in this measure is that respondent’s values are inferred from their self-reported similarity to people described in each value portrait.

Refined Values Theory

The most recent extension to the basic theory of human values (Schwartz et al., 2012) sought to identify a more refined set of meaningful, conceptually distinctive values around the motivational continuum (see Figure 1b). To identify refinements, these researchers examined values with multifaceted definitions and looked for unidentified value constructs between adjacent basic values. Specifically, they suggested a division of self-direction into thought and action, power into dominance and resources, security into societal and personal, conformity into rules and interpersonal, benevolence into caring and dependability, and universalism into nature, concern, and tolerance. They also identified two new values—face and humility. Table 1 presents the conceptual definitions of the 19 refined values.

OPEN IN VIEWER

Table 1.

The 19 Values in the Refined Theory, Each Defined in Terms of Its Motivational Goal.

Value	Conceptual definition in terms of motivational goals
Self-direction–thought	Freedom to cultivate one’s own ideas and abilities
Self-direction–action	Freedom to determine one’s own actions
Stimulation	Excitement, novelty, and change
Hedonism	Pleasure and sensuous gratification
Achievement	Success according to social standards
Power–dominance	Power through exercising control over people
Power–resources	Power through control of material and social resources
Face	Maintaining one’s public image and avoiding humiliation
Security–personal	Safety in one’s immediate environment
Security–societal	Safety and stability in the wider society
Tradition	Maintaining and preserving cultural, family, or religious traditions
Conformity–rules	Compliance with rules, laws, and formal obligations
Conformity–interpersonal	Avoidance of upsetting or harming other people
Humility	Recognizing one’s insignificance in the larger scheme of things
Benevolence–dependability	Being a reliable and trustworthy member of the in-group
Benevolence–caring	Devotion to the welfare of in-group members
Universalism–concern	Commitment to equality, justice, and protection for all people
Universalism–nature	Preservation of the natural environment
Universalism–tolerance	Acceptance and understanding of those who are different from oneself

Note. Adapted from “Refining the theory of basic individual values,” by Schwartz et al. (2012, p. 669). Copyright 2012 by the American Psychological Association.

Schwartz et al. (2012) located the 19 refined values within the existing motivational continuum and argued that these values meet the criteria he previously used to classify constructs as values (Schwartz, 1992). Specifically, he identified five characteristics of values on which earlier researchers had agreed: Values (a) are beliefs linked to emotions; (b) refer to desirable goals that motivate action; (c) transcend specific actions and situations; (d) serve as standards for evaluating actions, people, and events; and (e) form a relatively enduring hierarchical system ordered by importance (e.g., Rokeach, 1973; Vernon & Allport, 1931). Later, Schwartz (2006, 2015) added to these criteria, arguing that values (f) guide attitudes and actions based on trade-offs between the values that promote them and inhibit them; and (g) affect everyday decisions in a way that is rarely conscious.

Schwartz et al. (2012, 2017) tested the refined values theory in 10 countries with a new PVQ. Confirmatory factor analysis and distinct associations with attitudes, value-expressive behaviors, and sociodemographic variables supported the discrimination of 19 values. MDS supported the circular motivational structure of the 19 values. Unexpectedly, the theorized order of universalism and benevolence values reversed in each of 14 samples. The authors offered speculative explanations for this reversal and Schwartz and Butenko (2014) suggested that the findings might suggest a different theoretical order for the 19 values. One explanation pointed to the possible effect of measurement method, because such a reversal had appeared almost exclusively in studies of the original 10 values when using the PVQ. Thus, assessing the refined values theory with a different instrument might help clarify whether the reversal of universalism and benevolence values depends on method of measurement.

A New Measure of Refined Values: Best–Worst Scaling

Only one instrument is currently available to measure the 19 refined values: the Portrait Values Questionnaire–Refined (PVQ-R; Schwartz et al, 2012). This instrument asks respondents to rate how similar they are to gender-matched individuals who are described in terms of their important values. In total, there are 57 items (e.g., Hedonism item: “Enjoying life’s pleasures is important to him” ) for which respondents are asked “How much is this person like you?” on a 6-point rating scale (not like me at all, not like me, a little like me, moderately like me, like me, and very much like me). ¹ To provide an independent test of the theory of value content and structure, we developed a new measurement instrument that differed substantially from the PVQ-R.

We chose BWS to measure the refined values (Finn & Louviere, 1992; Louviere et al., 2015). BWS is an extension of the well-known “Method of Paired Comparisons” (David, 1988; Thurstone, 1927). Essentially, it is a method of multiple comparisons. The number of pairs one must compare for any set of J items is J(J−1)/2. This number increases geometrically with J, making the method much less attractive as J increases. BWS applies experimental design principles to form manageable numbers of comparison sets of fixed sizes (triples, quadruples, quintuples, etc.) rather than pairs.

Louviere et al. (2015) suggest that balanced incomplete block designs (BIBDs) can be used to form comparison sets of items. This approach has the advantages of creating sets of fixed sizes in which each item occurs and co-occurs with each other item equally often. The number of comparison sets increases approximately linearly in J. This makes BWS feasible for many research applications that use rating scales or similar methods. However, two limitations with using BIBDs to construct comparison sets can be problematic. First, BIBDs do not exist for all J. Second, some BIBDs (see Table S1[all supplemental files are available online at http://asm.sagepub.com/content/by/supplemental-data]), such as ones for J = 19 and J = 20 produce large numbers of comparison sets (e.g., 57 sets of 3 objects each) or large set sizes (e.g., 19 sets of 9 objects each).

BWS tasks ask participants to choose the two extreme items in each set (best/worst, most/least, etc.). Louviere et al. (2015) discuss the theory underlying BWS and the properties of the resulting scales (measures). Briefly, the theory is a multiple-choice extension of the theory underlying Paired Comparisons, known as Random Utility Theory (Thurstone, 1927). Each item has a latent (subjective) value that is not directly observed. With a sufficient number of BWS choices, it is possible to measure the mean subjective value of the item plus an associated random error. Thus, BWS yields additional discrete choice information from each comparison set.

Louviere et al. (2015) discuss ways to represent and analyze BWS choices. A simple method is to use the observed choice counts to estimate the subjective values. This is consistent with using the marginal totals from contingency tables to represent statistical effects, where one side of the table is the observed choices and the other side is the J items of research interest (Louviere & Woodworth, 1983). A number of other parametric and nonparametric methods can be used to estimate the subjective values of a set of J items from BWS choices. All of these methods provide approximately the same information about the items, but some are more efficient statistically.

The more consistent participants are in choosing particular items as best or worst, the larger the differences in the resulting measures of the subjective values of the items. These measures can be treated as ipsatized differences (i.e., Best minus Worst choice counts), odds ratios (Best counts divided by Worst counts) or other measures such as square root ratios (i.e., the square root of Best counts divided by Worst counts). Perfectly consistent choosers will produce counts that are exactly the inverse of one another. For simplicity and ease of understanding, we use BWS scores based on the simple calculation of counts, subtracting the number of times each value item is chosen least important from the number of times it is chosen most important. Such BWS scores typically correlate highly with corresponding conditional logit choice model estimates (Louviere et al., 2015; McFadden, 1974) for aggregates of individuals.

As mentioned, BWS relies on well-tested and long-standing Random Utility Theory and provides a type of choice-based measurement that has several advantages over rating scale tasks:

We followed the BWS approach to develop a new Best–Worst Values instrument, the Best–Worst Refined Values scale (BWVr), that differs markedly from the existing PVQ-R instrument by (a) asking directly about self-value importance rather than similarity to another and by (b) using choice frequencies to indicate item importance rather than averages of rating scales. The development of a new instrument to measure the refined values also provided an opportunity to examine whether a proposed new value fits within the circular structure.

Animal Welfare as a Personal Value

Sneddon, Lee, Ballantyne, and Packer (2016) argued that the growing public concern for the treatment of animals should be considered a personal value rather than a more situation-specific attitude. The literature cited below suggests that there is a distinct animal welfare value, defined as empathic concern for the welfare of all animals. We argue that an animal welfare value fulfils the seven features of values previously described. Specifically, an animal welfare value:

We propose that the animal welfare value is a facet of the broad universalism value. Schwartz (1992, p. 12) defined the goal of universalism values as the “understanding, appreciation, tolerance, and protection for the welfare of all people and for nature.” The refined value theory distinguishes three facets of universalism Schwartz et al. (2012, p. 669):

One might argue that animal welfare values are part of the universalism–nature value that emphasizes concern for the environment and unity with nature. However, the welfare of individual animals is sometimes at odds with concern for the natural environment. Protecting the natural environment reflects a holistic perspective of a world constructed of biological collectives, systems, species, and communities (Callicott, 1980; Malamud, 2011; Sagoff, 1984; Sarkar, 2005). A holistic view of the environment is, by definition, largely indifferent to the suffering of individual animals. Conserving populations, communities, and biodiversity is the objective, but individual animals are expendable (Callicott, 1980; Sarkar, 2005; Soulé, 1985). For instance, people who place a high value on animal welfare would oppose culling animals (native or introduced) to help preserve ecological integrity, because doing so would cause suffering to animals (Callicott, 1980; Sarkar, 2005; Soulé, 1985; Varner, 1998). The distinctly different views about hunting as a means of population control illustrate the opposition between animal welfare and conservation organizations. Animal welfare groups, such as the Humane Society of the United States ² and PETA, ³ actively protest against hunting and promote humane approaches to population control (e.g., fertility control and relocating animals). In contrast, when asked to state their position, conservation organizations, such as the Sierra Club, Wildlife Fund, and the Wilderness Society, either support or refuse to condemn hunting (Sagoff, 1984).

In order for the distinction between universalism–nature and universalism–animals to be worthwhile, they must relate differently to a set of attitudes and behaviors. The finer distinction should yield stronger associations. We expected universalism–animals to relate more strongly than universalism–nature to positive attitudes toward animals, pet ownership, and donations to animal welfare organizations. We expected universalism–nature to relate more strongly than universalism–animals to proenvironmental attitudes and donation to environmental organizations. Sagiv, Sverdlik, and Schwarz (2011) reported positive associations between universalism values and donation behavior, but we found no studies that distinguished among the types of organizations to which people donate.

The Present Research

We conducted a series of three studies to validate Schwartz et al.’s (2012) refined values with a new instrument and to situate the animal welfare value within universalism. Study 1 assesses the theoretical structure of the refined values with data from the new instrument and tests the distinctiveness of the universalism–animals value in an adult sample in two countries (Australia and the United States). Study 2 assesses the structure with data in a diverse sample of American immigrants (United States). Study 3 examines the test–retest reliability of the new instrument and the convergent and discriminant validity of this measure with the PVQ-R instrument and value-expressive behavior.

Study 1

Instruments

As noted earlier, we developed a new instrument (the BWVr) to measure the Schwartz et al. (2012) refined values plus the proposed new animal welfare value (universalism–animals) as an alternative to the PVQ-R. We followed the BWS approach outlined in the BWS section. First, we designed value items (e.g., Hedonism item: “Taking advantage of every opportunity to enjoy life’s pleasures”) appropriate for assessing personal importance rather than comparing other with self. In comparison, the refined values PVQ-R items describe a person in terms of his or her values (e.g., Hedonism item: “Enjoying life’s pleasures is important to him”).

Next, we sought an experimental design that produced a manageable number of subsets for 20 values (19 refined values plus a new universalism–animals value). A desirable experimental design was found for 21 objects or, in this case, items (see Table S2 in supplemental materials). ⁴ This design produced 21 sets with five value items in each set. Across all 21 sets, each item appears five times and each pair of items appears together once. Figure 2 illustrates one set of five items, including the items for self-direction–action, achievement, security–personal, conformity–interpersonal, and universalism–tolerance. Each respondent received 21 randomly ordered sets. The full BWVr instrument is available in the supplemental materials.

OPEN IN VIEWER

Figure 2.

BWVr question example Set 1 of 21.

Participants were asked to choose both the most and the least important value in each of 21 sets of five values. We calculated respondents’ scores for each value by subtracting the number of times they chose a value as least important from the number of times they chose the same value as most important. We divided these scores by five, the number of times each value appeared across all of the BWS sets, so resulting scores ranged from −1 to +1 (see Louviere, Lings et al., 2013). Zero is the midpoint of the latent scale, with scores above zero indicating greater importance and scores below zero indicating less importance.

To capture past donation behaviors, we used the Bekkers and Wiepking (2006) method. Respondents reported whether they had donated in each of 12 ways in the past year (e.g., door-to-door solicitation at home, collection in a place of worship). They then reported whether they had donated in the past year to each of 14 charitable or nonprofit sectors (e.g., animal welfare, environmental, religious, and spiritual organizations) and if they were current pet owners. All responses were on a yes or no scale.

To measure environmental attitudes, we used Dunlap, Van Liere, Mertig, and Jones’s (2000) 15-item New Ecological Paradigm (NEP) scale. To measure attitudes to animals, we used Herzog, Betchart, and Pittman’s (1991) 20-item scale. Responses were obtained on 5-point scales (1 = strongly disagree and 5 = strongly agree). Cronbach’s alpha was .77 for the NEP and .87 for the animal welfare scale.

Results

Columns 1 to 4 of Table 2 show the means and standard deviations of the value scores obtained in Study 1. In both the Australian and U.S. samples, benevolence–dependability and benevolence–caring were the most important values and power–dominance was the least important value. All 20 value scores showed reasonable variation, as indicated by their standard deviations.

OPEN IN VIEWER

Table 2.

Means and Standard Deviations of the Refined Values: Study 1 (Australia and United States), Study 2 (U.S. immigrants) and Study 3 (United States).

Refined values	Study 1 (N = 784)						Study 3 (N = 277)
	Australia (n = 385)		United States (n = 399)		Study 2 (N = 439)		8-Week test–retest
	M	SD	M	SD	M	SD	r
Self-direction–thought	.06	.36	.25	.38	.18	.39	.58
Self-direction–action	.21	.34	.36	.36	.32	.39	.62
Stimulation–novelty	−.03	.35	.07	.38	.07	.43	.73
Stimulation–excitement	−.08	.38	−.05	.42
Hedonism	.08	.37	.16	.43	.18	.45	.66
Achievement	−.28	.44	−.01	.42	.07	.50	.73
Power–dominance	−.55	.35	−.54	.36	−.48	.36	.63
Power–resources	−.44	.44	−.39	.46	−.33	.48	.66
Face	−.34	.36	−.37	.34	−.29	.37	.47
Security–personal	.28	.31	.23	.33	.24	.34	.59
Security–societal	.39	.32	.21	.32	.23	.33	.59
Tradition	−.25	.44	−.32	.46	−.30	.47	.73
Conformity–rules	.00	.38	−.23	.36	−.25	.39	.68
Conformity–interpersonal	−.20	.38	−.41	.38	−.36	.42	.53
Humility	−.14	.34	−.17	.33	−.15	.35	.60
Benevolence–dependability	.50	.36	.42	.39	.42	.37	.58
Benevolence–caring	.47	.34	.44	.35	.41	.35	.61
Universalism–concern	.16	.38	.20	.40	.18	.40	.68
Universalism–nature	.01	.37	−.02	.39	−.04	.43	.80
Universalism–animals	−.01	.41	.00	.40	−.09	.46	.61
Universalism–tolerance	.16	.38	.16	.42	.14	.40	.84

(1) Of these, which are the most and least important to you, as guiding principles in your life?

Figure 3a displays the two-dimensional space, partitioned into the 20 refined values. The space represents the associations among the items well (Kruskal’s Stress 1 measure was .12 and the dispersion accounted for was .99). The location of the refined values around the circle in Figure 3a corresponds to the theoretical order (see Figure 1b) with one small deviation, benevolence–caring and benevolence–dependability are reversed. Moreover, the universalism and benevolence values conform to the originally theorized order rather than reversing as they did in Schwartz et al. (2012). The observed order permits combining adjacent values to reconstitute the 10 basic values of the original theory. This adds support to the assumption that the refined values form the same motivational circle as the original 10 values and makes it possible to compute scores for the 10 basic values.

OPEN IN VIEWER

Figure 3.

Multidimensional scaling analysis of the refined values based on the (a) pooled within-sample covariance matrix for Study 1 (N = 784). Stress 1 = .12, dispersion accounted for = .99, Tucker’s coefficient of congruence = .99 and (b) sample correlation matrix for Study 2 (N = 439). Stress 1 = .13, dispersion accounted for = .98, Tucker’s coefficient of congruence = .99.

Study 2

The aim of this study was to replicate the findings of Study 1 with participants from a different population (a diverse sample of American immigrants). Again, we tested the theoretical structure of the refined values and the distinctiveness of the new universalism–animals value. We revised the instrument to include only one stimulation item. To maintain the 21 items needed for the balanced experimental design, we included an unrelated item. This enabled us to remove one item in subsequent analyses to reduce the ipsatized nature of the scale.

Study 3

This study assessed the test–retest reliability of the values measured with the BWVr over 8 weeks and compared the measurement of values using the BWS instrument with the measurement of refined values by the PVQ-RR and value-expressive behaviors. The PVQ-RR is the currently recommended revised instrument for operationalizing the refined values theory (Schwartz, 2017). We assessed the convergent and discriminant validity of the values through a multitrait–multimethod approach (Campbell & Fiske, 1959), as well as the correlated traits-correlated methods (CTCM) confirmatory factor analysis approach (Jöreskog, 1974; Widaman, 1985).

Convergent and Discriminant Validity

Table S3 (in supplemental materials) presents the multitrait–multimethod matrix of the correlations among the 20 values measured by the BWVr and the PVQ-RR. For this analysis, we partialled each individual’s responses to the PVQ-RR items on his or her mean rating of the 60 items to control for individual differences in scale use, as is usual practice (see Schwartz, 1992, 2006). Using Campbell and Fiske’s (1959) approach, we found support for the convergent validity of the 20 values: The 20 single trait–multimethod correlations ranged from .38 to .77 (see Table S4), and all differed significantly from zero (p < .001). The following evidence also supported discriminant validity of the values:

Each of the single trait–multimethod correlations on the diagonal was higher than the correlations of that value with other values, regardless of method.

Each value measured with one method correlated more highly with the same value measured with the other method than with the other values measured with its own method.

The pattern of interrelationships among values was similar for each method.

We also applied the CTCM confirmatory factor analysis approach (Jöreskog, 1974; Widaman, 1985) to examine this issue further. In this analysis, we used the data on everyday behaviors (see Table 3), which enabled a comparison across three approaches. For confirmatory factor analyses, it is appropriate to use the raw PVQ-RR scores in order to avoid problems with ipsatized scores (see Fischer, 2004; Schwartz, 2016).

OPEN IN VIEWER

Table 3.

Confirmatory Factor Analysis Results and Scale Characteristics for Behavioral Scales—Study 3.

Value-expressive behavior	Scale M	Scale SD	χ2	# Items	Significance
Self-direction–thought	3.32	0.73	3.06	5	.69
Self-direction–action	3.05	0.60	3.03	5	.69
Stimulation	2.47	0.73	2.58	3	.11
Hedonism	2.58	0.64	0.05	3	.82
Achievement	3.28	0.75	0.20	3	.66
Power–dominance	2.01	0.73	0.21	3	.66
Power–resources	1.76	0.73	1.52	3	.22
Face	2.50	0.80	2.34	4	.31
Security–-personal	3.49	0.78	5.72	5	.34
Security–societal	1.97	0.86	0.01	3	.95
Tradition	2.15	1.05	0.56	3	.46
Conformity–rules	4.05	0.77	1.56	3	.21
Conformity–interpersonal	3.23	0.71	2.31	4	.32
Humility	3.28	0.75	2.01	4	.37
Benevolence–caring	3.74	0.74	5.25	4	.07
Benevolence–dependability	4.03	0.61	4.22	4	.12
Universalism–concern	2.68	0.90	2.24	4	.33
Universalism–nature	2.54	0.94	0.14	3	.71
Universalism–animals	2.27	1.02	0.49	4	.78
Universalism–tolerance	3.23	0.84	0.69	4	.71

Note. N = 383. When there were only three items, two of the error variances were set to be equal to obtain the degrees of freedom needed to examine the significance of the confirmatory factor analysis.

It is problematic to examine the CTCM model across the 20 values simultaneously. We therefore estimated four models, using subsets of the values around the circular motivational continuum shown in Figure 1b. The subsets were (a) benevolence and universalism values; (b) self-direction, stimulation, and hedonism values; (c) achievement and power values; and (d) face, security, tradition, conformity, and humility values. Table 4 presents the results. Although the chi-square statistics were significant, the other fit indexes suggested a reasonable fit in each case. ⁶ Thus, the CTCM model provides a good basis for examining the three methods. The trait variance accounted for 60% of the variability in the data. This indicates that the traits were somewhat more important than the methods used, but the methods did have an impact. Of particular interest are the correlations between the methods (i.e., the PVQ-RR, the BWVr, and the behavior scales), which varied from .33 to .77. This suggests that the BWS approach of the BWVr taps the same underlying dimensions as the other two methods.

OPEN IN VIEWER

Table 4.

Correlated Traits-Correlated Methods Analysis Results for Study 3.

	χ2	Significance	GFI	CFI	RMSEA	BW-PVQ correlation	BW-behaviors correlation	PVQ-behaviors correlation
Benevolence to universalism	342.93	<.01	0.91	0.94	0.08	0.43	0.35	0.49
Self-direction to hedonism	116.22	<.01	0.95	0.95	0.08	0.33	0.44	0.59
Achievement to power	23.88	.03	0.99	0.99	0.05	0.77	0.51	0.46
Face to humility	494.25	<.01	0.88	0.86	0.08	0.69	0.71	0.59

Note. N = 383. CFI = comparative fit index; GFI = goodness-of-fit index; RMSEA = root mean square error of approximation; BW = best–worst; PVQ = Portrait Values Questionnaire.

General Discussion

The current research provides the first support for the 19 values in the refined values theory (Schwartz et al., 2012, Schwartz, 2017) using a measurement approach other than the PVQ-R or PVQ-RR. The BWVr uses choice frequency rather than averages of rating scales, and direct self-description rather than judgment of similarity to self. Results of the multimethod–multitrait analysis, based on the BWVr and the PVQ-RR confirmed the convergent and discriminant validity of each of the refined values based on the BWVr.

The studies also supported the distinctiveness of a new universalism–animals value whose motivational basis differed from the other universalism facets and from all the other 16 values. Within the broad universalism value, two subsets of facets differed most, the two human focused values (universalism–concern and universalism–tolerance) versus the two nonhuman focused values (universalism–animals and universalism–nature). However, there was clear evidence that the motivations underlying the universalism–nature and universalism–animals values differed. Although these two values share a common broad concern, they can conflict when protecting animals is at odds with protecting nature (e.g., culling animals to protect the natural environment in national parks). The significantly different associations of these two values with relevant attitudes and behaviors in Studies 1 and 2 attest to their distinctiveness.

The findings are compatible with the idea that many people perceive nonhuman animals as sentient beings with humanlike feelings and cognitions. The universalism–animals value expresses a motivation to enhance or protect the welfare of nonhuman animals with which people empathize. In contrast, universalism–nature expresses a motivation to enhance and protect the natural environment for the welfare of present and future generations. Although universalism–animals refers to “all animals,” the empathic concern that uniquely underlies this value may be greater for animals that are more like us (e.g., mammals) and less for other species, especially species considered dangerous to humans (e.g., venomous snakes and sharks). The animal species that elicit empathy from people and to which this value applies may also differ across cultures. These are topics for future studies. Further research should also examine whether people who prioritize universalism–nature have more concern for animals at the species level as an integral part of an ecosystem than those who prioritize universalism–animals.

The universalism–animals value promises greater precision and predictive power for understanding animal welfare attitudes and behavior, a growing societal concern (Eurobarometer, 2007; Herzog & Golden, 2009; Phillips, 2008; Regan, 1987; Rollin, 2010; Singer, 1990; Varner, 1998; Verbeke, 2009). The importance of issues related to animal welfare in society and the improvement in predictive ability from the inclusion of a universalism–animals value should offset the respondent costs of adding this item.

In the MDS analyses in all three studies, there were a few deviations from the theorized order of values around the circle. In each case, the reversal of order occurred between values that are adjacent in the circle or at most separated by one other value according to the theory. Moreover, in no case did a value from one higher order value emerge in the spatial region of a different higher order value. Schwartz and Sagiv (1995) found a similar degree of variation when they ran MDS analyses on subsamples obtained by splitting larger samples. When we ran MDS analyses on randomly chosen subsamples of the data in the three studies, only one deviation replicated across all subsamples, the reversal of benevolence–dependability and benevolence–caring which led to the location of benevolence-caring adjacent to universalism–concern. This may reflect the motivational affinity between these two values. Both emphasize helping others in need, whether close others (benevolence) or those outside one’s in-group (universalism).

The order of the universalism and benevolence values in all the samples studied here corresponded to the theorized order. This contrasts with the reversal in Schwartz et al.’s (2012) findings from the PVQ-R for 10 countries. This sheds doubt on Schwartz et al.’s (2012) explanations for the reversal of these values they observed. They suggested that the most likely cause was changes in the meaning of the values due to recent historical developments or to splitting the two values into five narrower values. However, they also considered the possibility of a methodological artifact. In the current study, we did not find the reversal with the PVQ-RR data in Study 3. We therefore also examined newly available MDS analyses of unpublished PVQ-RR data from 42 samples in 21 countries. ⁷ Benevolence and universalism exhibited the same theorized order found in our studies in 25% of these samples but reversed in 38% of the samples. In another 39% of samples, one or both of the benevolence value facets were located toward the center of the value space. Apparently, there is no stable ordering of these two values. This leaves the explanation of the order of benevolence and universalism in the refined value theory as a challenge for future investigation.

Our results add to growing evidence that one can legitimately partition the circular motivational continuum into broader or narrower sets (Schwartz et al., 2012). The MDS results permitted partitioning into the 20 refined values, the 10 original values, and the 4 original higher order values. The optimal partitioning depends on one’s research objectives.

These three studies used data from online, panel surveys in the United States and Australia. Although panels can provide quality data, it would be valuable to measure the refined values with the BWVr in face-to-face surveys. It is also important to test this instrument in samples from different countries and in different languages.

Together, these studies provide strong support for Schwartz et al.’s (2012) refinement of the theory of basic human values. Applying the newly developed best–worst scale, a substantially different method of measurement than previously used to assess the theory, we replicated its theorized circular motivational structure of values. The studies also validated the new BWVr as a useful way to measure the 19 values. In addition, the studies suggest that it is possible to distinguish a 20th value (universalism–animals) that has low to moderate correlations with the other universalism values. Its relationships with attitudes and behavior that are relevant to animal welfare and to the environment distinguish it from universalism–nature.