A Summary of Construct Validity Evidence for Two Measures of Character Strengths

Abstract

The VIA Inventory of Strengths has become the most widely used instrument in the world for measuring the construct referred to character strengths. However, several limitations were noted in its original development. In response, the VIA Assessment Suite for Adults was developed as a battery of instruments intended to address those gaps. The suite includes two inventories providing dimensional measures of the character strengths: the VIA Inventory of Strengths-Revised and the Global Assessment of Character Strengths. Short forms were also developed for each. So far, five reasonably sized samples of adults (total N = 7,924) have provided evidence for the empirical validity of some subset of these instruments, making them the most thoroughly vetted measures of character strengths available today. This article aggregates previously available and new findings on their construct validity. Evidence concerning substantive validity, structural validity, and external validity is reviewed, and in some cases aggregated across samples. The findings generally support the construct validity of the instruments evaluated according to all three standards, with exceptions noted. Recommendations are offered for their use in research and applied settings.

The VIA Classification of Character Strengths and Virtues (Peterson & Seligman, 2004) is a structural model that conceptualizes positive human traits in terms of 24 character strengths, as well as six superordinate variables developed conceptually that are referred to as the virtues (see Table 1). The model’s development was one of the seminal events in the founding of positive psychology. The success of the VIA model can be attributed to several factors beyond its primacy, though. First, its development involved input from more than 50 experts in positive human functioning as well as reviews of a variety of literatures, providing a broad foundation for the dimensions chosen for inclusion. This strategy was in contrast to earlier models of positive functioning that were generally the work of one person or at most several people. Second, the model introduced a concept referred to as signature strengths. These are defined as a subset of the strengths that the individual identifies as particularly central to his or her identity, the exercise of which is experienced as particularly energizing or fulfilling by the individual. The concept of signature strengths has played a central role in the application of the model to coaching, clinical, educational, and corporate settings (e.g., Linley et al., 2010; Merritt et al., 2019; Suldo et al., 2015; Uliaszek et al., 2022). Finally, these and other studies have connected character and signature strengths to flourishing and well-being in a variety of contexts (e.g., Blanchard et al., 2020).

Table 1. The VIA classification of character strengths and virtues.

Original virtue	Character strength
Wisdom	Creativity^a [originality, ingenuity]
& Knowledge	Curiosity^a [interest, novelty-seeking, openness to experience]
	Judgment & Open-Mindedness^a [critical thinking]
	Love of Learning^a
	Perspective [wisdom]
Courage	Bravery^a [valor]
	Perseverance^b [persistence, industriousness]
	Honesty^b [authenticity, integrity]
	Zest^a [vitality, enthusiasm, vigor, energy]
Humanity	Capacity to Love and Be Loved^c
	Kindness^c [generosity, nurturance, care, compassion, altruistic love, "niceness"]
	Social Intelligence [emotional intelligence, personal intelligence]
Justice	Teamwork^c [citizenship, social responsibility, loyalty]
	Fairness^c
	Leadership
Temperance	Forgiveness & Mercy^c
	Modesty & Humility
	Prudence^b
	Self-Regulation^b [self-control]
Transcendence	Appreciation of Beauty and Excellence [awe, wonder, elevation]
	Gratitude^c
	Hope [optimism, future-mindedness, future orientation]
	Humor [playfulness]
	Religiousness & Spirituality [faith, purpose]

Note. Adapted from Peterson and Seligman (2004, pp. 29–30). Copyright 2004 by the Values in Action Institute. Adapted with permission. Terms in brackets are variants of the character strength according to Peterson and Seligman (2004).

^a Unique marker of the evidence-based Inquisitiveness virtue (McGrath et al., 2018).

^b Unique marker of the evidence-based Self-Control virtue (McGrath et al., 2018).

^c Unique marker of the evidence-based Caring virtue (McGrath et al., 2018).

One of the most important contributors to the popularity of the model was the early introduction of inventories for measuring the 24 VIA character strengths. The first to be developed was the VIA Inventory of Strengths (VIA-IS; Peterson & Seligman, 2004), intended for use with adults 18 and over. This was followed shortly thereafter by the VIA-Youth (Park & Peterson, 2006), for youth ages 10–17. These inventories treat the 24 character strengths as personality traits (McGrath et al., 2020) capable of being measured as relatively enduring though potentially malleable attributes of the individual. The VIA-IS and VIA-Youth have also been used extensively to identify signature strengths in applied settings based on which scores are highest. The scale scores provide a basis for nomothetic evaluation of the results, while the signature strengths can be used as the basis for a more idiographic approach to interpretation. The VIA-IS and VIA-Youth have since been completed millions of times online and have been used in hundreds of studies.

In 2014, the VIA Institute on Character, the copyright holder for the two instruments, decided to fund revision of the VIA-IS. The reasons for this decision have been summarized elsewhere (McGrath, 2019), so the primary considerations will be listed here only in brief:

• All the items were positively keyed, which is considered undesirable by many testing professionals because of the potential for yea-saying (“all true”) or nay-saying (“all false”) response sets (McGrath et al., 2010).

• At 240 items (10 items per strength scale), the instrument was also too long for practical purposes in many contexts. For several years, a 120-item version of the VIA-IS was administered at the VIA Institute website while the VIA-IS was under revision. This version, called the VIA-120, consisted of the five items out of the original ten on each scale with the highest corrected item-total correlations in a sample of N = 458,854 who completed the VIA-IS online. However, users of the test were requesting even briefer versions.

• The original VIA-IS did not include measures of the six virtues. Some researchers have generated proxy virtue scores by aggregating scores across the scales comprising the virtues (e.g., Gustems-Carnicer & Calderón, 2016), but this strategy was never encouraged by the developers of the model and has never been empirically evaluated. Subsequent research on the latent structure of the 24 strengths has suggested a three-factor model that appears reliably across populations and measurement instruments (McGrath, 2015; McGrath et al., 2018), and that are consistent with the concept of higher-order virtues (McGrath, 2021). The primary markers for these latent variables are provided in Table 1. The popularity of the original virtue model and the empirical support for the revised model suggested the potential value of scales specifically targeting both virtue models.

• Problems were identified with the item content of several scales. These included questions asking about protected health information or other sensitive topics, and scales that included items reflecting markedly heterogeneous contents.

The VIA Assessment Suite for Adults

In response to these concerns, an extensive revision process was pursued that continued until 2019. This process ultimately resulted in the development of three measurement tools, collectively referred to as the VIA Assessment Suite for Adults (McGrath, 2019).

The premier instrument in this new set is the VIA-Inventory of Strengths-Revised (VIA-IS-R). Major differences between the VIA-IS and the VIA-IS-R include the following:

• The latter instrument consists of 192 items (eight items per strength), compared to 240 on the VIA-IS.

• The VIA-IS-R includes both positively and negatively keyed items on each scale.

• Items were selected based on a number of criteria. These included reliability statistics, results from item response theory analyses, ratings of item prototypicality for the scale construct (Broughton, 1984), and item readability statistics.

• Virtue scales were developed representing both the original six-virtue model and the evidence-based three-virtue model. Their addition raised the number of scales that can be scored from 24 to 33. The virtue scales can also be administered independently of the VIA-IS-R.

• Two short forms of the VIA-IS-R were developed simultaneously, both of which are 96 items long (four items per strength), which is consistent with the typical number of items on measures used in applied research (Flake et al., 2017). The VIA-IS-M (Mixed) version includes both positively and negatively keyed items like the VIA-IS-R. The VIA-IS-P (Positive) consists of items that are all positively keyed, reducing the cognitive load for the respondent. Virtue scales are also available for the short forms. The three versions as a group will be referred to here as the VIA-IS-R/M/P.

Items on the VIA-IS-R/M/P are completed on a 1 (Very Much Unlike Me) to 5 (Very Much Like Me) scale. Items are averaged so that all scale scores are on a 1–5 scale. Researchers and practitioners interested in collecting data using the VIA-IS-R/M/P are asked to request permission from the VIA Institute (https://www.viacharacter.org/www/Research/Conduct-a-Study). This requirement was implemented primarily to track the use of the instrument. No reasonable request is denied, and there is no cost for the inventory’s use.

In addition to the VIA-IS-R/M/P, the Assessment Suite includes a new inventory providing dimensional measures of the 24 strengths. The Global Assessment of Character Strengths (GACS-72; McGrath, 2019) is a 72-item questionnaire, three items per strength. Respondents first see a brief description of each of the 24 strengths. They then complete 24 items asking whether each strength is “essential” to them on a 1 (Very Strongly Disagree) to 7 (Very Strongly Agree) scale. These are followed by 24 items asking whether the strength is “natural and effortless” for them to use, and another 24 asking whether the strength is “energizing.” This wording comes from Peterson and Seligman’s (2004) discussion of how individuals they interviewed characterized their signature strengths. The 24 “natural and effortless” items can also be used as single-item indicators of each strength, bundled as the GACS-24. The two versions in combination can be referred to as the GACS-72/24.

The GACS-72 may be found in McGrath (2019) as well as on the VIA Institute website. It is freely available for researchers and practitioners to modify for their personal purposes and use without approval from the VIA Institute. The Institute will also provide data collection online for any of their measures in their original form at no charge.¹

The validation process

Flake et al. (2017) have suggested construct validation of scale scores consists of three elements. The first involves developing the instrument with issues of substantive validity in mind. The substantive validity of an instrument depends upon two factors. The first has to do with the degree to which the constructs to be measured are clearly defined. The second has to do with the degree to which items are chosen that effectively reflect those constructs. These are primarily conceptual issues.

The second and third elements of validation in this model are primarily empirical in focus. Structural validation has to do with the psychometric adequacy of instrument scores. Hussey and Hughes (2020) suggested four key indicators for structural validity: internal reliability, test-retest reliability, confirmatory factor structure, and measurement invariance. External validation is the term Flake et al. (2017) used to refer to the degree to which scale scores are related to other variables in ways that match expectations. They described several approaches to the evaluation of external validity, all of which represented aspects of convergent and discriminant validity.

Flake et al. (2017) and Hussey and Hughes (2020) raised concerns about the extent to which many studies rely on measures for which there is insufficient evidence of validity. In many cases an estimate of internal reliability is the only justification provided for using an instrument. Even this minimal evaluation is usually accomplished using coefficient alpha, a statistic that is considered questionable because of its unrealistic assumptions about the component items (e.g., McNeish, 2018; Sijtsma, 2009). These concerns are relevant to the current status of measurement in positive psychology research. Even articles introducing new measures of constructs in positive psychology often offer little more statistical support for their use than internal reliability metrics. For many measures, evidence of external validity is often missing all together.

Turning to the construct validity of the VIA Assessment Suite, several factors argue for the substantive validity of the new instruments. The original choice and definition of the 24 strengths was based on multiple literature reviews and substantial expert input. Extensive descriptions and nomological models are available for each of the 24 character strengths (Peterson & Seligman, 2004), the original six virtues (Dahlsgaard et al., 2005), and the three empirically derived virtues (McGrath et al., 2018). These sources were extensively used in the development of new items. Explicit definitions of the character strengths are even incorporated into the GACS-72/24. Item prototypicality ratings, item unidimensionality, and item-total correlations were all considered in the selection of items for the VIA-IS-R/M/P, which should contribute to their substantive validity as well.

That said, concerns can still be raised about the degree to which the items of the VIA-IS-R/P/M and GACS-72/24 provide adequate coverage for all the character strengths. Arbenz et al. (in press) asked psychologists to rate the breadth of the character strengths on a 1–9 scale. In every case the mean exceeded 5.0 (M = 6.13, range = [5.13, 7.09], raising questions about whether even eight-item scales would be sufficient for adequate coverage across all strengths. This concern reflects the practical obstacles to balancing competing ideals in psychological measurement. It would be desirable for scales to include items tapping all elements of a content domain. They should also be brief to avoid taxing respondents, particularly when the number of constructs being measured is large as in the case of the 24 character strengths. At the same time, it has been argued by some that the most important feature of multi-item scales is unidimensionality (e.g., Ng et al., 2017). What can be said about the substantive validity of the two instruments is that the items are likely to be representative of the underlying constructs but may not necessarily be comprehensive.

Moving on from substantive validity, data are now available from five samples of adults who completed the VIA-IS-R/M/P, four of which also took the GACS-72/24. These data can be used to evaluate the empirical validity of these instruments. The remainder of this manuscript will use results from these five samples to generate a comprehensive summary of the two instruments’ structural and external validity. The next section provides a description of the five currently available validation samples.

Description of the samples

The five samples collectively include 7,924 adults ages 18 and older. All data collection was approved by the Fairleigh Dickinson University Institutional Review Board or was judged exempt from review. Attention checks were implemented in all samples but the first.

The first sample will be called the VIA Derivation Sample (VD Sample; McGrath, 2019). Between October 2015 and March 2016, individuals who accessed the VIA Institute website and completed the English language version of the VIA-120 in return for personal feedback on their results were asked to volunteer to complete additional questionnaires for research purposes. This request generated a sample of 4,286 adults who responded to the entire set of materials. The sample was 77.7% female, with an average age of 45.6 years (SD = 13.1). They represented 83 different countries, though the majority were from the United States (50.9%), Australia (10.9%), Canada (7.4%), and the United Kingdom (6.0%). In addition to demographic variables, participants completed an additional 309 candidate items for the VIA-IS-R as well as the GACS-72, the SSS, and a 48-item questionnaire consisting of two behavioral acts representing each of the 24 strengths. This sample was used to select the items for the VIA-IS-R and its short forms.

The VIA Cross-Validation Sample (VC Sample; McGrath & Wallace, 2021) consisted of 631 U.S. residents who also completed the English version of the VIA-120 at the VIA Institute website, this time between August and October 2017, and subsequently agreed to participate in a study. Participants were 76.9% female, with a mean age of 41.9 years (SD = 13.1). They were administered the three VIA Assessment Suite instruments in their entirety (they were warned there would be some duplication of items with the VIA-120) as well as a 168-item questionnaire about the frequency of seven behavioral acts conceptually representing each of the 24 strengths. For the VC and subsequent sample, 17 attention items were added using the general format “select option x for this item.” Individuals who failed more than three attention items were excluded from the final analysis.

The Mechanical Turk Cross-Validation Sample (MC Sample; McGrath & Wallace, 2021) included 743 Amazon Mechanical Turk workers from the U.S. who completed the VIA Assessment Suite measures as well as the same behavioral acts measure completed by the VC Sample. Initial data collection occurred in June 2017. This sample was 49.7% female, with an average age of 34.4 (SD = 10.2). MC Sample members who completed all the measures and passed the attention check as described for the VC Sample were contacted three months later and invited to complete the VIA-IS-R again for purposes of evaluating test-retest reliability. The Time 2 sample included 474 workers. Bonferroni-corrected chi square tests revealed no significant differences between those who did and did not respond to the Time 2 invitation on any categorical demographic variable, with the same result for Bonferroni-corrected t tests of age and the 33 VIA-IS-R scores at Time 1. Members of the MC Sample received compensation for their participation even if they failed the attention check. The publication summarizing results for the VC and MC samples focused on the VIA-IS-R and its short forms, but statistics for the GACS in these samples will also be provided in the present article.

Fourth, between October and November 2019, a stratified sample of 1,765 U.S. residents was gathered in collaboration with Qualtrics, with the goal of matching U.S. adult Census statistics for gender, age, education, self-identified racial identity, and region of the country (McGrath et al., 2022). This Qualtrics Representative Sample (QR Sample) completed the VIA Assessment Suite instruments as well as several other questionnaires used in another study. The sample was 51.8% female, with an average age of 46.5 (SD = 17.0). Behavioral criteria specific to the character strengths were not included in this study. In consultation with Qualtrics, the number of attention items was reduced to five, with participants excluded if they failed more than two. Participants were compensated, with the amount varying depending on the difficulty associated with recruiting members of certain demographic subgroups.

Finally, a sample that will be called the German Cross-Validation Sample (GC Sample; Vylobkova et al., in press) consisted of 499 adults recruited via mailing lists and social media to complete the German-language versions of the VIA-IS-R and the original VIA-IS. The sample was 79% female, with an average age of 33.3 years (SD = 13.9). Where the previously listed studies involved completing the entire VIA-IS-R, the authors of this study used a planned missing data design. The protocol involved discrete administration of the VIA-IS and VIA-IS-R. However, full administration of both inventories was deemed too long (240 items for the VIA-IS and 192 for the VIA-IS-R). Instead, each participant was administered randomly chosen blocks of items from the two inventories in a sequence that ensured they were administered half the items from each of the 24 strength scales from each inventory. Unfortunately, the method of data collection does not allow for identifying which items were administered to which participants (V. Vylobkova, personal communication, June 7, 2022), so new short forms and virtue scales could not be computed, and additional item analyses could not be conducted.

The order of administration of the two inventories was also counterbalanced. Participants completed several other measures, one of which was used in this review. The Core Virtue Rating Form (Ruch et al., 2020) consists of six items, each of which represents one of the six virtues that were part of the original VIA Classification (see Table 1). The items begin with a short description of the virtue based on Peterson and Seligman (2004). The respondent then indicates how well the description fits their typical behaviors on a scale from 1 (not at all) to 9 (absolutely).

In return for participating, students received course credit. All participants received feedback on their results, and a small donation was made to a charitable organization for each completion of the study. Several variables (e.g., speed of completion, number of missing items) were used to exclude participants from the analyses.

Evidence for reliability

For each of the multi-item scales that comprise the VIA-IS-R/M/P and the GACS-72, two estimates of internal reliability were generated. Coefficient alpha was computed to provide a comparison with results for other instruments. Given the criticisms that have been leveled against that statistic, McDonald’s (1999) omega total was also computed.

Since the GACS-24 consists of single-item scales, an alternative strategy was needed to estimate reliability. Wanous and Hudy (2001) suggested such a method based on the correction for attenuation: $$r_{xx}=\frac{r_{xy}^2}{r_{yy}}.$$

In this formula, $r_{xx}$ is the reliability of a single-item measure, $r_{yy}$ is the reliability of a standard measure of the same latent construct, and $r_{xy}$ is the correlation between the two measures. In this case, $r_{yy}$ was estimated by computing coefficient alpha using the “essential” and “energizing” items for each strength. Alpha was used because omega could not be computed with only two items, and the concerns about alpha are reduced somewhat given the very high similarity in the two items used in the computation.

Internal reliability estimates for the first four samples were computed using the MBESS (Kelley, 2020) package in R (R Core Team, 2021), while reliability statistics from the article on the GC sample were computed with the psych package (Revelle, 2020). Table 2 provides mean reliability values across the samples weighted by sample size. Results were similar for the two reliability statistics. Specifically, 76% of mean reliability estimates for the VIA-IS-R were ≥ .80, while 100% were ≥ .70. This distribution is consistent with a large-scale review of reliability values across studies (Peterson et al., 2020). All reliability estimates for the GACS-72 also were .80 or higher.

Table 2. Mean internal reliability values across samples.

	VIA-IS-R		VIA-IS-M		VIA-IS-P		GACS-72
Strength/Virtue	α	ω	α	ω	α	ω	α	ω	GACS-24
Beauty	.85	.86	.76	.77	.82	.82	.88	.89	.79
Bravery	.81	.81	.73	.74	.72	.74	.82	.83	.71
Creativity	.88	.89	.80	.80	.81	.82	.86	.86	.73
Curiosity	.83	.83	.75	.76	.73	.74	.84	.84	.74
Fairness	.83	.83	.72	.72	.80	.80	.82	.82	.72
Forgiveness	.81	.82	.71	.72	.71	.72	.85	.85	.72
Gratitude	.82	.82	.70	.72	.80	.81	.86	.86	.73
Honesty	.78	.77	.75	.75	.68	.70	.81	.81	.75
Hope	.87	.87	.78	.78	.78	.78	.87	.87	.75
Humility	.78	.79	.61	.63	.63	.64	.83	.83	.71
Humor	.85	.84	.72	.70	.83	.84	.88	.88	.83
Judgment	.75	.75	.57	.59	.66	.67	.81	.81	.74
Kindness	.82	.82	.72	.72	.75	.75	.83	.84	.74
Leadership	.85	.86	.70	.70	.83	.84	.88	.89	.81
Learning	.85	.85	.81	.81	.77	.77	.85	.85	.73
Love	.91	.91	.84	.85	.80	.83	.86	.86	.75
Perseverance	.89	.89	.77	.78	.82	.84	.81	.82	.76
Perspective	.79	.79	.72	.73	.79	.80	.82	.82	.70
Prudence	.85	.85	.72	.72	.80	.80	.83	.84	.77
Self-Regulation	.86	.87	.76	.77	.78	.78	.80	.81	.73
Social Intelligence	.81	.81	.74	.74	.73	.74	.86	.87	.80
Spirituality	.87	.88	.84	.85	.82	.84	.92	.92	.82
Teamwork	.85	.86	.77	.79	.71	.72	.87	.87	.78
Zest	.88	.88	.83	.83	.84	.84	.88	.88	.82
Wisdom	.81	.81	.71	.71	.78	.78
Courage	.79	.79	.70	.71	.79	.78
Humanity	.82	.83	.62	.62	.78	.80
Justice	.78	.79	.68	.73	.72	.72
Temperance	.77	.76	.68	.68	.63	.66
Transcendence	.82	.82	.82	.82	.79	.80
Caring	.76	.74	.79	.82	.72	.71
Inquisitiveness	.83	.83	.76	.76	.79	.78
Self-Control	.87	.87	.75	.77	.76	.78

Note. Entries are the mean internal reliability value across five samples (for strength scales on the VIA-IS-R) or four samples (for all others) weighted by sample size. Unweighted means were also computed and never differed from the values presented by more than .05, and the mean difference was < .01. VIA-IS-R = VIA Inventory of Strengths-Revised; VIA-IS-M = VIA Inventory of Strengths-Mixed; VIA-IS-P = VIA Inventory of Strengths-Positive; GACS-72 = Global Assessment of Character Strengths-72; GACS-24 = Global Assessment of Character Strengths-24.

Unsurprisingly, results were weaker for the shorter measures. Only 21% of omega coefficients for the VIA-IS-M, 39% for the VIA-IS-P, and 21% for the GACS-24 were ≥ .80. These numbers increased to 85%, 88%, and 96%, respectively, when .70 was used as a minimum value, which Cortina (1993) suggested as the acceptable minimum for reliability. Finally, reliability estimates for the VIA-IS-M Judgment scale was slightly less than .60, which is the lowest recommended value for a reliability statistic according to other sources (e.g., Shrout, 1998).

The second pillar of structural validity (Hussey & Hughes, 2020) is test-retest reliability. As noted previously, members of the MC Sample provided three-month retest data for the VIA-IS-R. Test-retest reliability was estimated for each of the 33 scales of the VIA-IS-R/M/P inventories using the mixed model absolute agreement intraclass correlation for a single measure. McGrath and Wallace (2021, Table 8) provided all 99 estimates, so results will only be summarized here. Reliability was again strongest for the VIA-IS-R, with 76% at .80 or higher, and all were .70 or higher. In contrast, only 21% of VIA-IS-M correlations and 36% of VIA-IS-P correlations were ≥ .80. All equaled .70 or greater except in three cases: the VIA-IS-M Humor scale (.68), and the VIA-IS-P Humility (.65) and Judgment (.67) scales.

A third approach to exploring reliability in the context of the VIA-IS-R/M/P instruments is made possible by the 1–5 scaling of all the subtests resulting from the use of item averages for scores. McGrath et al. (2022) took advantage of this commonality to evaluate the alternate versions reliability of each of the VIA-IS-R/M/P scales across the three versions of the inventory in the QR Sample. For the present study, this analysis was extended to all four samples, and to the GACS-72/24 inventories as well. Mixed model absolute agreement intraclass correlations for a single measure for each scale may be found in Table 3. In this case, 91% of mean reliability estimates across the VIA-IS-R/M/P inventories were .80 or higher, and all were ≥ .70. All estimates across the GACS-72/24 scales exceeded .80. These findings suggest the three versions can be treated as essentially interchangeable instruments. This conclusion should be accompanied by the caution that the short forms were computed by extracting the short form items from the full item set, which would potentially exaggerate their convergence (see Smith et al., 2000).

Table 3. Summary of alternate versions reliability within samples.

Strength/Virtue	VIA-IS-R/M/P	GACS-72/24
Beauty	.90	.91
Bravery	.86	.84
Creativity	.90	.86
Curiosity	.84	.88
Fairness	.88	.87
Forgiveness	.79	.87
Gratitude	.87	.88
Honesty	.83	.86
Hope	.90	.88
Humility	.81	.87
Humor	.88	.91
Judgment	.79	.86
Kindness	.86	.87
Leadership	.90	.91
Learning	.79	.88
Love	.91	.87
Perseverance	.91	.85
Perspective	.87	.86
Prudence	.90	.88
Self-Regulation	.91	.85
Social Intelligence	.86	.90
Spirituality	.91	.93
Teamwork	.92	.90
Zest	.92	.90
Wisdom	.95
Courage	.81
Humanity	.90
Justice	.90
Temperance	.81
Transcendence	.85
Caring	.85
Inquisitiveness	.92
Self-Control	.93

Note. Entries are the mean alternative versions reliability across four samples (for all others) weighted by sample size. Unweighted means were also computed and never differed from the values presented by more than .01. VIA-IS-R/M/P = VIA-IS-Revised, -Mixed, and -Positive; GACS-72/24 = Global Assessment of Character Strengths-72 and -24; VD = VIA Derivation sample; VC = VIA Cross-Validation Sample; MC = Mechanical Turk Cross-Validation sample; QR = Qualtrics Representative sample.

Latent structural analyses

Tests of unidimensionality

In addition to internal reliability and test-retest reliability, Hussey and Hughes (2020) associated structural validity with a unidimensional factor structure and measurement invariance. Although some have argued that unidimensionality is an important if not essential feature for a well-functioning psychometric instrument (e.g., Feraco et al., 2022; Ng et al., 2017), others have argued this may be an unrealistic standard for useful measures of personality (Hopwood & Donnellan, 2010). This argument is particularly relevant to the VIA Classification constructs. Peterson and Seligman (2004) conceptualized some of the strengths as inherently multidimensional. As indicated by the text in brackets in Table 1, many of the strengths were given alternate labels to reflect this complexity. The virtue scales were also explicitly developed in a manner that included items from each of the strengths most closely related to the virtue, so unidimensionality was never intended as a consideration in their development (McGrath, 2019).

With these caveats in mind, one-factor confirmatory factor analyses were generated for the 24 strength scales in the four samples. These were conducted using the WLSMV estimator in the R lavaan package (Rosseel, 2012). Hu and Bentler (1999) evaluated a variety of rules for establishing good fit. Their results indicated using a squared root mean residual (SRMR) > .09 combined with a comparative fit index (CFI) < .95 as an indicator of poor fit. Out of 384 analyses (24 strengths from four inventories in four samples), 34 did not initially meet both these criteria, none of which involved the GACS-72. To address this issue, one error covariance was freed based on the largest modification index. This resolved the issue for 23 models, including all instances involving the VIA-IS-M and VIA-IS-P. The process was repeated, reducing the number of scales for which good fit was not achieved to two, the VIA-IS-R Bravery and Humility scales in the QR sample. In these cases, freeing a third error covariance resolved the issue. Overall, 47 out of 4,128 error covariances had to be freed to achieve good fit in all the models. Interestingly, 18 of the 39 freed covariances in the VC, MC, and RS samples involved consecutively administered items on a scale, suggesting item effects may have contributed to covariation not attributable to the shared factor even though items from each scale were administered at 24-item intervals to those samples. Table 4 provides the range of final estimates for the two statistics across the four samples.

Table 4. Tests of scale unidimensionality across samples.

	VIA-IS-R		VIA-IS-M		VIA-IS-P		GACS-72
Strength	SRMR	CFI	SRMR	CFI	SRMR	CFI	SRMR	CFI
Beauty	[.04, .06]	[.97, 1.00]	[.01, .07]	[.95, 1.00]	[.01, .02]	[1.00, 1.00]	[.00, .00]	[1.00, 1.00]
Bravery	[.06, .08]	[.95, .97]	[.03, .08]	[.95, .99]	[.00, .02]	[1.00, 1.00]	[.00, .00]	[1.00, 1.00]
Creativity	[.03, .06]	[.98, 1.00]	[.00, .05]	[.98, 1.00]	[.02, .04]	[.99, 1.00]	[.00, .00]	[1.00, 1.00]
Curiosity	[.04, .08]	[.95, .99]	[.01, .03]	[.99, 1.00]	[.00, .02]	[1.00, 1.00]	[.00, .00]	[1.00, 1.00]
Fairness	[.05, .07]	[.97, .98]	[.02, .03]	[.99, 1.00]	[.02, .05]	[.99, 1.00]	[.00, .00]	[1.00, 1.00]
Forgiveness	[.04, .06]	[.97, .99]	[.02, .04]	[.98, 1.00]	[.01, .03]	[.99, 1.00]	[.00, .00]	[1.00, 1.00]
Gratitude	[.06, .08]	[.95, .98]	[.01, .04]	[.99, 1.00]	[.00, .01]	[1.00, 1.00]	[.00, .00]	[1.00, 1.00]
Honesty	[.06, .08]	[.95, .96]	[.01, .05]	[.98, 1.00]	[.00, .04]	[.99, 1.00]	[.00, .00]	[1.00, 1.00]
Hope	[.06, .06]	[.97, .99]	[.01, .04]	[.99, 1.00]	[.00, .03]	[1.00, 1.00]	[.00, .00]	[1.00, 1.00]
Humility	[.06, .07]	[.95, .97]	[.01, .06]	[.96, 1.00]	[.01, .05]	[.95, 1.00]	[.00, .00]	[1.00, 1.00]
Humor	[.07, .09]	[.96, .96]	[.00, .01]	[1.00, 1.00]	[.01, .02]	[1.00, 1.00]	[.00, .00]	[1.00, 1.00]
Judgment	[.05, .06]	[.96, .98]	[.01, .04]	[.97, 1.00]	[.03, .04]	[.98, .99]	[.00, .00]	[1.00, 1.00]
Kindness	[.05, .06]	[.97, .99]	[.01, .04]	[.99, 1.00]	[.01, .04]	[.99, 1.00]	[.00, .00]	[1.00, 1.00]
Leadership	[.05, .08]	[.96, .99]	[.00, .05]	[.97, 1.00]	[.01, .03]	[1.00, 1.00]	[.00, .00]	[1.00, 1.00]
Learning	[.06, .07]	[.96, .99]	[.01, .07]	[.97, 1.00]	[.01, .03]	[.99, 1.00]	[.00, .00]	[1.00, 1.00]
Love	[.06, .09]	[.97, .99]	[.03, .07]	[.98, 1.00]	[.01, .03]	[1.00, 1.00]	[.00, .00]	[1.00, 1.00]
Perseverance	[.04, .08]	[.98, 1.00]	[.02, .07]	[.96, 1.00]	[.01, .01]	[1.00, 1.00]	[.00, .00]	[1.00, 1.00]
Perspective	[.05, .07]	[.95, .99]	[.01, .05]	[.97, 1.00]	[.01, .04]	[.99, 1.00]	[.00, .00]	[1.00, 1.00]
Prudence	[.05, .07]	[.97, .99]	[.02, .05]	[.97, 1.00]	[.03, .05]	[.99, 1.00]	[.00, .00]	[1.00, 1.00]
Self-Regulation	[.06, .09]	[.95, .98]	[.02, .03]	[.99, 1.00]	[.02, .05]	[.98, 1.00]	[.00, .00]	[1.00, 1.00]
Social Intell	[.06, .08]	[.96, .96]	[.01, .03]	[.99, 1.00]	[.00, .02]	[1.00, 1.00]	[.00, .00]	[1.00, 1.00]
Spirituality	[.05, .08]	[.97, .99]	[.00, .02]	[1.00, 1.00]	[.02, .04]	[.99, 1.00]	[.00, .00]	[1.00, 1.00]
Teamwork	[.08, .09]	[.96, .97]	[.01, .02]	[1.00, 1.00]	[.05, .07]	[.95, .97]	[.00, .00]	[1.00, 1.00]
Zest	[.07, .08]	[.97, .98]	[.03, .05]	[.98, 1.00]	[.04, .09]	[.96, .99]	[.00, .00]	[1.00, 1.00]

Note. VIA-IS-R = VIA Inventory of Strengths-Revised; VIA-IS-M = VIA Inventory of Strengths-Mixed; VIA-IS-P = VIA Inventory of Strengths-Positive; GACS-72 = Global Assessment of Character Strengths-72; CFI = comparative fit index; SRMR = squared root mean residual.

Measurement invariance

The evaluation of measurement invariance across subgroups is treated as a sequential process. The first step examines the configural invariance of a factor structure across subgroups. Evidence of invariance suggests the model is equally applicable across the populations represented by the groups. Step 2 tests metric invariance by fixing corresponding loadings to be equal across groups. In this case, invariance provides a basis for allowing comparisons of covariation in the subpopulations. Finally, scalar invariance is evaluated by fixing intercepts for corresponding variables to be equal across groups. Scalar invariance provides evidence that mean differences in the two subpopulations represent differences in the underlying factor(s) rather than resulting from extraneous causes.

In the present study, invariance was evaluated applying the unidimensional models evaluated in the previous section across genders and age groups. Cases with missing data for the demographic variable were omitted, as were respondents who identified as transgender from the gender analyses. The age analyses were based on a median split of participants within each sample into younger and older subgroups. For purposes of consistency with the previous analyses, error covariances needed to achieve good fit for the unidimensional model were allowed to vary. In addition, models were evaluated for good fit using the same criteria used for unidimensionality (SRMR ≤ .09, CFI ≥ 0.95).

Table 5 summarizes the results for both gender and age. For each strength scale, four values are provided. The first is the number of samples out of four in which the test of configural invariance failed, that is, basic factor structure seemed to vary across the two groups. The second number is the number of samples in which only configural invariance was demonstrated, that is, the test of configural invariance was passed but not the test of metric invariance. The third number is instances where metric invariance was demonstrated based on equivalent factor loadings across groups. The final number indicates how frequently scaler invariance was achieved. Again, scalar and metric invariance required good fit in prior models as well.

Table 5. Tests of measurement invariance across samples.

	Gender				Age
Strength	VIA-IS-R	VIA-IS-M	VIA-IS-P	GACS-72	VIA-IS-R	VIA-IS-M	VIA-IS-P	GACS-72
Beauty	0/0/0/4	0/1/0/3	0/0/0/4	0/0/0/4	0/0/1/3	0/0/2/2	0/0/0/4	0/0/0/4
Bravery	0/0/2/2	0/0/1/3	0/0/0/4	0/0/0/4	0/0/2/2	0/0/2/2	0/0/0/4	0/0/0/4
Creativity	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/1/3	0/0/0/4	0/0/0/4
Curiosity	0/1/1/2	0/0/0/4	0/0/0/4	0/0/0/4	1/0/1/2	0/0/1/3	0/0/0/4	0/0/0/4
Fairness	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4
Forgiveness	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4
Gratitude	0/1/0/3	0/0/0/4	0/0/0/4	0/0/0/4	0/1/0/3	0/0/0/4	0/0/0/4	0/0/0/4
Honesty	0/1/0/3	0/0/0/4	0/0/0/4	0/0/0/4	0/0/2/2	0/0/0/4	0/0/0/4	0/0/0/4
Hope	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4
Humility	0/2/0/2	0/1/0/3	0/0/1/3	0/0/0/4	0/2/0/2	0/0/0/4	0/0/1/3	0/0/0/4
Humor	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4
Judgment	0/0/2/2	0/0/2/2	0/0/1/3	0/0/0/4	0/0/1/3	0/0/0/4	0/0/0/4	0/0/0/4
Kindness	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4
Leadership	0/0/1/3	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4
Learning	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4
Love	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4
Perseverance	0/0/0/4	0/0/1/3	0/0/0/4	0/0/0/4	0/0/0/4	0/0/1/3	0/0/0/4	0/0/0/4
Perspective	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/1/3	0/0/0/4	0/0/0/4	0/0/0/4
Prudence	0/0/0/4	0/0/1/3	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4
Self-Regulation	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/1/3	0/0/0/4	0/0/0/4	0/0/0/4
Social Intelligence	0/0/1/3	0/0/0/4	0/0/0/4	0/0/0/4	0/0/1/3	0/0/0/4	0/0/0/4	0/0/0/4
Spirituality	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4
Teamwork	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/1/0/3	0/0/0/4
Zest	0/0/0/4	0/0/0/4	0/0/0/4	0/0/0/4	0/0/1/3	0/0/1/3	0/0/0/4	0/0/0/4
# < 3	4	1	0	0	4	2	0	0

Note. Entries are the number of samples demonstrating non-invariance/configural/metric/scalar invariance The bottom row is the number of strengths in that column where scalar invariance was demonstrated in less than three of the four samples. VIA-IS-R = VIA Inventory of Strengths-Revised; VIA-IS-M = VIA Inventory of Strengths-Mixed; VIA-IS-P = VIA Inventory of Strengths-Positive; GACS-72 = Global Assessment of Character Strengths-72.

Out of 768 tests of configural invariance, only one demonstrated poor fit (the VIA-IS-R Curiosity scale for age in the MC Sample). Because scalar invariance is required for comparisons of means across populations to be clearly interpretable, it is considered particularly important. For both gender and age, four VIA-IS-R strengths failed to achieve scalar invariance in at least three samples: Bravery, Curiosity, Humility, and Judgment for gender; and Bravery, Curiosity, Honesty, and Humility for age. None of the scales of the VIA-IS-P or GACS-72 failed to meet this standard. In fact, scalar invariance was demonstrated for every GACS-72 scale in every sample, and in all but four VIA-IS-P analyses. The pattern demonstrates a connection between number of items per scale, reverse keying, and invariance outcomes. Even so, the results were strongly supportive of the interpretability of mean differences on all but a few scales. In addition, at least metric invariance was demonstrated in the majority of samples for every scale, suggesting relationships with other variables can be interpreted across ages and genders.

Evidence for external validity

As noted previously, three of the samples (the VD, VC, and MC Samples) also completed a behavioral measure of the character strengths and virtues. These criteria were developed using the act frequency approach to measuring personality variables (Buss & Craik, 1983) by consensus among the researchers involved in each study. The GC Sample also completed an item for each of the six virtues in the original VIA Classification. These measures can be used to gauge the convergent and discriminant validity of the Assessment Suite instruments, but certain caveats should be raised about doing so. First, the term “convergent validity” is being used broadly here, with no implication that the criteria used in the earlier studies converge with the strength scales on a common latent variable. The virtue scales used in the GC Sample are at best seen as potential correlates of the strength scales. The behavioral scales used in the other samples similarly should not be assumed to be convergent measures in the absence of any evidence for their validity. This point raises a second concern, which is that none of these “criterion” measures have themselves been externally validated. Third, both the strength scales and the act scales tend to demonstrate substantial intercorrelations, limiting the potential for achieving discriminant validity. The virtue measure in the GC Sample is also problematic as a criterion in that the original association between strength and virtue was developed conceptually by Peterson and Seligman (2004). Several studies have now raised questions about some of these assignments (see Ruch et al., 2020, for a review). In particular, Humor seems to have more to do with the virtue of Humanity, Hope may fit better as an exemplar of Courage, and Leadership and Honesty each seems interstitial between virtues, as will be discussed below.

With these caveats in mind, Table 6 summarizes the results for the external validity analyses. For each instrument, the convergent validity column provides the mean correlation weighted by sample size between the strength scale and the corresponding criterion across the four samples. To evaluate the discriminant validity of the strength scales, the number of other criterion scales that correlated more highly with the strength scale than the expected criterion was summed across the samples. For example, the entry of 6 for the VIA-IS-R Gratitude scale means that across the four samples there were six criterion measures that correlated more highly with the VIA-IS-R scale than the expected criterion. A value of 0 therefore means no unexpected strength-criterion correlation was larger than the expected one in any sample. At the bottom of the columns are means as well as the number of strength scales where an unexpected correlation was larger than the expected correlation in at least one sample. The columns for the VIA-IS-R are based on all four samples. All other columns exclude the GC Sample.

Table 6. External validity analyses across samples.

	VIA-IS-R		VIA-IS-M		VIA-IS-P		GACS-72		GACS-24
	Conv Val	Disk Val	Conv Val	Disk Val	Conv Val	Disk Val	Conv Val	Disk Val	Conv Val	Disk Val
Beauty	.67	0	.66	0	.67	0	.64	0	.60	0
Bravery	.51	0	.47	0	.49	0	.35	1	.33	1
Creativity	.66	1	.67	0	.66	0	.56	0	.55	0
Curiosity	.48	10	.52	5	.47	4	.44	4	.41	4
Fairness	.41	3	.39	4	.40	4	.34	1	.31	0
Forgiveness	.49	1	.47	0	.52	0	.48	0	.43	0
Gratitude	.67	6	.66	8	.66	15	.63	1	.57	1
Honesty	.45	2	.36	3	.42	0	.43	0	.39	0
Hope	.73	3	.75	0	.71	0	.61	0	.59	0
Humility	.36	0	.39	0	.35	0	.30	11	.29	3
Humor	.61	5	.61	0	.66	0	.65	0	.63	0
Judgment	.46	2	.42	3	.44	3	.35	1	.32	3
Kindness	.51	0	.48	0	.50	0	.45	1	.43	0
Leadership	.63	2	.62	0	.66	0	.59	0	.56	0
Learning	.58	0	.57	1	.59	0	.55	0	.50	0
Love	.60	0	.59	0	.62	0	.61	0	.56	0
Perseverance	.54	0	.52	0	.49	0	.49	0	.45	0
Perspective	.64	0	.62	0	.59	0	.51	0	.48	0
Prudence	.46	0	.44	0	.45	0	.33	0	.30	0
Self-Regulation	.51	0	.50	0	.46	2	.37	0	.35	0
Social Intelligence	.63	0	.64	0	.60	0	.50	0	.49	0
Spirituality	.64	0	.54	0	.62	0	.53	0	.52	0
Teamwork	.40	10	.43	9	.42	8	.43	10	.39	10
Zest	.76	0	.78	0	.76	0	.58	0	.57	0
# >0		11		7		6		8		6
M	.56	1.88	.55	1.38	.55	1.50	.49	1.25	.46	.92

Note. VIA-IS-R = VIA Inventory of Strengths-Revised; VIA-IS-M = VIA Inventory of Strengths-Mixed; VIA-IS-P = VIA Inventory of Strengths-Positive; GACS-72 = Global Assessment of Character Strengths-72; GACS-24 = Global Assessment of Character Strengths-24; Conv Val = convergent validity; Disk Val = discriminant validity. Unweighted mean correlations were slightly larger on average, with a mean difference of .04.

Convergent validity results were quite consistent across the VIA-IS-R/M/P inventories, with mean correlations around .55 for all three versions and a minimum mean value of .35. Prior research suggests correlations of .30 or higher represent relatively large correlations in practice (e.g., Brydges, 2019; Hemphill, 2003), while McGrath et al. (2020) used several lines of evidence to justify the conclusion that correlations ≥ .60 provide evidence of “likely redundancy” (p. 130). Mean correlations for the GACS-72 and -24 were somewhat smaller. Discriminant validity results were generally good, with some problems noted for Curiosity, Gratitude, and Teamwork. Interestingly, for these two strengths every problem with discriminant validity except six (five for Gratitude and one for Teamwork) involved the MC Sample; the reason for this finding is unknown.

There were several interesting results for the GC Sample given concerns that have been raised about the virtue placement for some of the strengths in the original model. If Hope was considered a Courage strength, as previous research has suggested, the discriminant validity value for the Hope scale would drop to 0. In contrast, the Humor scale unexpectedly correlated best with the Wisdom item on the virtues scale. Research mentioned earlier (Ruch et al., 2020) suggested Leadership is interstitial across the virtues Wisdom and Knowledge, Courage, and its hypothesized placement as a Judgment variable. In the GC sample, the correlations of Leadership with Wisdom and Knowledge and Courage both exceeded the expected correlation between Leadership and Justice. Similarly, Honesty correlated more with self-perceived Justice than its expected correlate Courage, as previous research has suggested. It also correlated more with Temperance, which was a new finding.

Mean convergent validity correlations were lower for the two versions of the GACS. Comparing internal reliability and external validity results for the GACS-72/24 suggests a classic example of the attenuation paradox (Loevinger, 1954; Smith et al., 2000), in which the use of more similar items (often in the process of developing a shorter scale) can optimize reliability and unidimensionality while sacrificing content coverage, thereby reducing validity coefficients. However, all means were still .29 or greater, with the Humility and Teamwork scales demonstrating the poorest convergent validity. Consistent with findings for the VIA-IS-R/M/P scales, for these two strengths 33 of 34 discriminant validity failures involved the MC Sample.

Discussion

Given the number of analyses, a summary of the findings is worthwhile:

1. The VIA-IS-R and GACS-72 scales were the most reliable across the different inventories, with mean values approaching or surpassing .80. This is not surprising given the former has the most items per scale and the latter has the most consistent items in content. The values for the shorter instruments were weaker. Most still exceeded .70 on average, though users may consider the values for Humility and Judgment on the VIA-IS-M and VIA-IS-P less than desirable or even unacceptable. Test-retest reliability results were similar.

2. Alternate versions reliability across the VIA-IS-R/M/P inventories was quite good. The results in Table 3 suggest that, for example, a study comparing pretest to ‌posttest data on strengths and virtues might use different short forms for the two administrations as a means of reducing practice effects. However, this finding can be faulted in that it was based on generating short version scores from administration of the VIA-IS-R. A better strategy would involve administering the three as discrete instruments, though doing so would probably prove burdensome to the participants. A future study might employ an administration strategy involving planned missing data used as in the case of the GC Sample to make such an analysis more practical.

3. It was possible to achieve unidimensionality after freeing a relatively small set of error covariances. None were needed for the GACS-72 scales, and no more than one for any of the VIA-IS-M and -P scale. These findings are particularly surprising given that unidimensionality was one of only several factors considered in item selection (McGrath, 2019).

4. Using the final unidimensional models, most tests of measurement invariance indicated scalar invariance was achieved in a majority of the samples. There were three scales for which this standard was not met in the case of both gender and age on the VIA-IS-R: Bravery, Curiosity, and Humility. Judgment also demonstrated problems with gender on the VIA-IS-R and VIA-IS-M, the two instruments with reverse-keyed items. The majority of samples did achieve metric invariance in all cases except the Humility scale, suggesting covariation comparisons between demographic subgroups can still be meaningful. The VIA-IS-P and GACS-72 achieved scalar invariance in all cases.

5. Evidence for convergent and discriminant validity was generally good, though problems with the latter were noted for the Curiosity, Gratitude, and Teamwork scales. Convergent validity for the GACS-24 was on average only slightly lower than that for the GACS-72, and discriminant validity results were slightly better. These findings have to be interpreted with particular caution given that none of the criterion measures used for these analyses has been validated in their own right. Unfortunately, most studies that involve measures of the 24 strengths focus exclusively on well-being and flourishing as an outcome (e.g., Blanchard et al., 2020; Proyer et al., 2011), which offers no basis for judging discriminant validity at all.

Of course, this study has its limitations. Some of the statistics gathered for this article were reviewed in the VD sample in the process of selecting the items for the VIA-IS-R/M/P. However, item selection was based on a balancing of considerations, including some such as item response theory results that played no role in the present study. The extent to which the resulting scales meet all standards for adequate validity is therefore worth evaluating. These statistics also played no role in the development of the GACS-72/24 items.

Second, two of the samples were recruited through the VIA website. These are individuals who demonstrate a preexisting interest in the VIA Classification. Their decision to complete a lengthy questionnaire in return for personal feedback, as well as their willingness to respond to an extensive battery of additional items for research purposes, carves them out as specialized samples. These concerns provided the rationale for paying two samples of adults to participate in the research, and in one case to attempt a representative sampling of population demographic categories (McGrath et al., 2022; McGrath & Wallace, 2021). Of course, the types of individuals accessible through online panels are hardly representative of the population in general. That said, the present samples may well be no poorer than many of the adult samples used in personality research.

Given the variety of instruments discussed here, it is worth concluding with a few recommendations for their use. If the primary goal is to maximize both reliability and external validity (a common preference in practice), the VIA-IS-R is probably the best choice, though research with a broader array of outcome measures would be helpful for verifying this recommendation. It offers the greatest breadth of item contents and demonstrates consistently desirable psychometric qualities.

If length is the paramount consideration, the GACS-72, or even the GACS-24, provides a credible choice. If brevity is a key consideration while still wanting to achieve better content coverage, the two short forms should be considered. The VIA-IS-M in general performed more poorly than the other two versions of that inventory. However, as noted earlier, some test users object to unidirectional item keying because of concerns about yea-saying and nay-saying response sets. The better option for these users is the VIA-IS-R, but when that instrument is impractical because of its length the VIA-IS-M offers a reasonably acceptable alternative.

An implicit theme of this article has been that no instrument is or can be perfect. There will always be tensions between the varying considerations for evaluating psychological measures: maximizing unidimensionality, variation in item difficulties, content coverage, convergent validity, and discriminant validity among others. number of standards. That said, the VIA-IS-P may well be the best single choice for balancing issues of structural validity, external validity, content coverage, cognitive load, and length. The English language version of the instrument currently administered by default on the VIA Institute website is the VIA-IS-P. As translations of this instrument are completed (Bates & McGrath, n.d.), they will be substituted for translations of the VIA-120.

It should be noted that other multi-item measures of the VIA strengths are available for adults. Ng and associates developed both an alternate measure of the 24 strengths called the Comprehensive Inventory of Virtuous Instantiations of Character, or CIVIC (Ng et al., 2018), as well as a short form of the original VIA-IS (Ng et al., 2017). In both cases, they emphasized scale unidimensionality as the primary criterion for item selection. They have also introduced a forced choice version of the CIVIC in an attempt to reduce socially desirable responding about character strengths (Ng et al., 2021). Several inventories measuring the 24 character strengths have been developed using items from the International Personality Item Pool (Goldberg et al., 2006), a freely available database of over 3,000 personality items. One version was developed with the intention of optimizing statistics from item response theory analyses (du Plessis & de Bruin, 2015), while another focused primarily on unidimensionality and invariance across English and German versions (Partsch et al., 2022). At least five other measures using a single item for each character strength have been developed besides the GACS-24 (Cosentino & Solano, 2012; Furnham & Lester, 2012; Ruch et al., 2014; Vanhove et al., 2016; Vie et al., 2016), two of which were developed for use in military settings. Ruch et al. (2020) also described several measures of the six original virtues specifically and virtuousness in general.

It is also worth noting that several measures have been developed for purposes other than measuring the trait level of the strengths. The Overuse, Underuse, Optimal-Use of Character Strengths (Freidlin et al., 2017) builds on the Aristotelian concept of the golden mean for virtues by asking the respondent to indicate what percent of the time they use each strength in a manner that suggests insufficient use, excessive use of the strength, or a desirable level of use. The Partners Strengths Questionnaire (Kashdan et al., 2018) allows measurement of appreciation for the strengths of one’s partner. The Strengths Use Scale (Govindji & Linley, 2007) was developed for purposes of coaching to gauge the extent to which the respondent uses each of the strengths in daily life. Lamson and McGrath (2021) have even developed a text analysis tool for the 24 strengths.

All of these instruments are free to use. As noted earlier, the VIA Institute requires permission to use the VIA-IS-R/M/P, but this is primarily to track research, and no reasonable request is rejected. There is clearly no shortage of alternatives to the VIA Assessment Suite tools, and some of these instruments serve specific purposes not addressed by the measures in the suite. The VIA-IS-R/M/P and GACS-72/24 are distinct from these others in the existence of five samples explicitly collected for purposes of validation, including latent structural analyses and scale-specific correlates used for external validation. These features make them the most thoroughly vetted measures of the character strengths available. The Institute is also actively involved in translation of these measures into other languages, with 13 completed to date.

Author’s note

Robert McGrath is a Senior Scientist for the VIA Institute on Character, the copyright holder for the instruments examined in this manuscript.

Acknowledgments

I am grateful to the many coauthors who helped make this work possible. The data and syntax used in analyses described in this article are available upon request to the author.

Additional information

Funding

The research summarized in this article was funded in part by the VIA Institute on Character.

Notes

1 The Assessment Suite includes a third measure called the Signature Strengths Survey (SSS; McGrath, 2019), which uses a checklist that allows respondents to self-identify signature strengths. This instrument is unusual in that it generates a binary outcome for each strength, it is infrequently used, and has been less extensively vetted than the other two. For these reasons, it was omitted from this review of the instruments that use more typical item formats. The statements made in this paragraph about the GACS-72/24 apply to the SSS as well.