BACKGROUND
Satisfaction with life (SWL) is defined as “a global assessment of a person’s quality of life according to his or her own chosen criteria” (Shin & Johnson, 1978, p. 478). To capture this evaluative judgment, Diener et al. (1985) developed the Satisfaction with Life Scale (SWLS), a five-item instrument assessing the cognitive dimension of subjective well-being. Since its development, the SWLS has become one of the most widely used measures in well-being research (e.g., Klonowska & Kowalczyk, 2024), with numerous cultural adaptations and psychometric evaluations across age groups, populations, and settings (Espejo et al., 2022; Pavot & Diener, 2009).
The original SWLS contains five items, each reflecting a different facet of global life evaluation. However, both theoretical reasoning and empirical analyses have shown that items 1 to 3 – “In most ways, my life is close to my ideal,” “The conditions of my life are excellent,” and “I am satisfied with my life” – consistently present the strongest factor loadings and item-total correlations (Diener et al., 1985; Pavot & Diener, 2009). In contrast, items 4 and 5 – “So far I have gotten the important things I want in life” and “If I could live my life over, I would change almost nothing” – are more reflective in nature and may be confounded by life stage or cultural expectations, potentially introducing measurement bias (Pavot & Diener, 2008; Vittersø et al., 2005). These psychometric issues are further compounded by evidence of scalar non-invariance in the five-item version across countries and demographic subgroups (Jang et al., 2017; Jovanović, 2016), which challenges the validity of cross-cultural comparisons – particularly in international public health surveillance.
This is especially important in the context of global health studies, where brief, psychometrically sound tools are needed to reduce cost and respondent burden without sacrificing measurement quality. As highlighted by Hays et al. (2017), removing even a single item from a large-scale health survey can save tens of thousands of dollars – with some estimates reaching up to $100,000 per item – making parsimony a critical concern. This consideration has also guided large initiatives such as PROMIS (Patient-Reported Outcomes Measurement Information System), developed by the National Institutes of Health to optimize population health measurement while minimizing instrument length and respondent fatigue (Cella et al., 2010; Hays et al., 2017).
To address these issues, Kjell and Diener (2021) proposed an abbreviated version of the SWLS consisting of the first three items. These were selected based on their psychometric superiority and conceptual alignment with the core construct of SWL. The authors demonstrated, using data from three independent samples (including Swedish and American adults, as well as a student subsample), that the SWLS-3 maintains excellent internal consistency, strong test–retest reliability, and good model fit in confirmatory factor analysis (CFA), along with measurement invariance across sex and time. The abbreviated version also retained construct validity, showing virtually identical correlations with indicators such as well-being, mental health, and social desirability, when compared to the original scale.
Further empirical support comes from Espejo et al. (2022), who validated the SWLS-3 in a Colombian adult sample, confirming its factorial structure, internal consistency, and invariance across gender within a culturally adapted version of the scale. Their results emphasize the scale’s suitability for use in public health and clinical assessment, particularly among individuals with limited formal education. Additionally, Neto and Neto (2024) tested the SWLS-3 in a combined Portuguese-speaking sample from Brazil and Portugal, confirming configural, metric, and scalar invariance in multigroup CFA models. The three-item version preserved the predictive associations of the original scale with key psychological constructs such as self-esteem, positive and negative affect, satisfaction with sex life, and romantic loneliness, demonstrating its applicability in multinational, cross-cultural research where balancing psychometric precision with practical constraints is critical.
Current study
Despite its growing international use, the psychometric properties of the SWLS-3 (Kjell & Diener, 2021) have not yet been evaluated in a Polish sample, leaving its applicability in this context uncertain. Therefore, we aimed to assess the structural validity, reliability, and measurement invariance across sex and between the Polish and original UK samples. In addition, we examined its convergent validity using constructs theoretically associated with SWL.
Given the scale’s brevity, classical CFA was not feasible, as three-item models are typically just-identified and lack sufficient degrees of freedom to assess model fit independently. To evaluate structural validity, we adopted a comparative model approach in which the three-item and five-item versions of the SWLS were each tested in combination with the WHO-5 Well-Being Index (Topp et al., 2015). This allowed us to assess the adequacy of the abbreviated form in a broader model that reflects the conceptual distinction and empirical association between SWL and psychological well-being. Although SWL constitutes the cognitive component of subjective well-being, and measures such as the WHO-5 primarily reflect affective and functional aspects, these components are theoretically related and typically show moderate to strong empirical associations (e.g., Skalski-Bednarz et al., 2024c). We hypothesized that (H1) the model including the SWLS-3 and WHO-5 would not demonstrate worse fit compared to the model including the full SWLS-5 and WHO-5. Additionally, to further examine the psychometric performance of individual items, we conducted an item response theory (IRT) analysis using graded response models (GRMs).
To assess convergent validity, we included, apart from well-being, dispositional gratitude and awe, and three dimensions of dispositional forgiveness: toward oneself, toward others, and by God. These constructs have been consistently linked to SWL and represent emotionally grounded and meaning-oriented coping resources (Skalski-Bednarz et al., 2024a). Forgiveness is conceptualized within the stress-and-coping model of forgiveness (Worthington & Scherer, 2004) as an emotion-focused coping response to interpersonal transgressions or self-condemnation. It allows for the regulation of negative emotions and restoration of psychological balance. Specifically, forgiveness of others facilitates relational repair, self-forgiveness mitigates guilt and shame, and forgiveness by God may foster spiritual congruence and inner peace, particularly among believers (Skalski-Bednarz et al., 2024c). Gratitude and awe, in turn, are increasingly understood as forms of experiential secular spirituality, marked by embodied and emotionally charged responses to beauty, connection, or transcendence – distinct from institutional religiosity (Büssing, 2021). In the context of ongoing secularization in Europe, such experiential forms of spirituality may serve as alternative sources of existential meaning and psychological stability, often fueling greater SWL (Surzykiewicz et al., 2025). We therefore hypothesized (H2) that, for the purposes of assessing convergent validity, psychological well-being, gratitude and awe, and dispositional forgiveness (of others, self, and by God) would be positively associated with SWLS-3 scores.
PARTICIPANTS AND PROCEDURE
Participants
Participants were adults (18+), fluent in Polish, and provided informed consent. All identified as White Caucasian. A total of 506 individuals were recruited via snowball sampling from Warsaw university communities and broader networks. The participants were aged between 19 and 35 years (SD = 7.72); 58.1% identified as women, 39.7% as men, and 2.2% as other or undisclosed.
Procedure
Data were collected between March and May 2025 via an online survey administered on the Qualtrics platform. After providing informed consent, participants completed a survey comprising several validated measures. Satisfaction with life was assessed using the SWLS. In addition, the survey included a measure of psychological well-being, as well as indicators of experiential secular spirituality (gratitude/awe) and dispositional forgiveness. The average completion time was approximately 10 minutes. Participation was voluntary and anonymous. No missing data were observed, as all items required a response.
Measures
Satisfaction with life. SWL was assessed using the Satisfaction with Life Scale (SWLS) developed by Diener et al. (1985). The scale consists of five items assessing the cognitive evaluation of one’s life, as listed in the Introduction. Participants responded to each statement using a 7-point scale ranging from 1 (strongly disagree) to 7 (strongly agree). The study used the Polish version of the SWLS validated by Jankowski (2015). In addition to the full five-item scale, a shortened three-item version was also employed, consisting of the first three items of the original SWLS. Both the full (α = .80) and abbreviated versions (α = .78) showed acceptable internal consistency in the present sample.
Well-being. The World Health Organization Well-Being Index (WHO-5; Topp et al., 2015), in its Polish adaptation by Cichoń et al. (2020), was used to assess subjective well-being over the preceding two weeks. The scale comprises five items forming a single latent factor. A sample item is: “I have felt active and vigorous.” Responses were recorded on a 6-point scale ranging from 0 (at no time) to 5 (all the time). The internal consistency of the scale was high in the present sample (α = .86).
Gratitude and awe. Experiential secular spirituality was assessed using the Gratitude/Awe Questionnaire (GrAw-7) developed by Büssing et al. (2018), in the Polish translation by Konaszewski et al. (2022). The instrument focuses on feelings of gratitude and awe as they arise in everyday experiences and interactions with nature. Rather than capturing religious beliefs or practices, it emphasizes lived experiences such as being deeply moved, momentarily pausing daily activities, and responding with quiet attentiveness to meaningful moments. The GrAw-7 consists of seven items forming a single-factor structure. In the present study, the scale showed good internal consistency (α = .85). Responses were provided on a 4-point frequency scale ranging from 0 (never) to 3 (regularly). Sample items include: “I have a feeling of wondering awe” and “In certain places I become very quiet and devout.”
Dispositional forgiveness. Dispositional forgiveness was measured with the Toussaint Forgiveness Scale (TFS; Toussaint et al., 2001) using the Polish adaptation by Charzyńska and Heszen (2013). The scale comprises nine items distributed across three dimensions: forgiveness of others (five items), self-forgiveness (two items), and perceived forgiveness by God (two items). In the present study, the internal consistency of the subscales ranged from questionable to high, with Cronbach’s α values of .74 for forgiveness of others, .65 for self-forgiveness, and .91 for forgiveness by God. The lower reliability of the self-forgiveness subscale may reflect its two-item length; for such scales, inter-item correlation is a more appropriate reliability index (Eisinga et al., 2013). In this study, r = .48 (p < .001), indicating acceptable internal consistency. A sample item is: “I have forgiven those who have hurt me.” Responses were provided on a 5-point scale from 1 (strongly disagree) to 5 (strongly agree).
Statistical analysis
All analyses were conducted in R (version 4.5.0). Preliminary analyses included descriptive statistics and bivariate correlations. Due to identification limitations of three-item models, the SWLS was tested within a broader framework including a related construct. Competing models were evaluated using standard fit indices: χ2, CFI, TLI, RMSEA, and SRMR, with thresholds of ≥ .90 (CFI/TLI) and ≤ .08 (RMSEA/SRMR) indicating acceptable fit (Byrne, 2016). Chi-square difference tests and parsimony were used to guide model comparisons. Measurement invariance across sex was tested via multigroup CFA (at the configural, metric, and scalar levels). Model differences were evaluated using ΔCFI (change in CFI) and ΔRMSEA, with values of ΔCFI ≤ .01 and ΔRMSEA ≤ .015 indicating invariance (Chen, 2007). Internal consistency was assessed via Cronbach’s α and McDonald’s ω. Convergent validity was tested using the average variance extracted (AVE; ≥ .50), and discriminant validity using the heterotrait–monotrait ratio of correlations (HTMT; < .85; Henseler et al., 2015). Additionally, item-level properties of the SWLS were examined using IRT; GRMs (Samejima, 1969) were estimated to evaluate item discrimination parameters (and threshold parameters, which indicate the level of the latent trait at which the probability of endorsing a given category or higher equals 50%). Correlations were used for associations between variables (with effect sizes interpreted according to Cohen, 1988). All tests were two-tailed with α = .05.
Given the relatively low complexity of the tested two-factor models and the large sample size (N = 506), the study was considered adequately powered for the planned SEM analyses, consistent with recommendations of at least 10-20 participants per estimated parameter (Kline, 2016).
RESULTS
At the preliminary stage, descriptive analyses and bivariate correlations were computed among all main variables included in Table 1. Inspection of skewness and kurtosis values indicated that the distributions of all variables were substantially normally distributed (i.e., all measures fall in the ±1 range; Kline, 2016). See Table 1 for details. Age was weakly but significantly negatively associated with both the SWLS-3, Pearson’s r(504) = −.11, p = .020, and the SWLS-5, r(504) = −.11, p = .026. Point-biserial correlations between sex and both versions of the SWLS were non-significant.
Table 1
Descriptive statistics and correlations among study variables
Confirmatory factor analysis
Two confirmatory factor models were estimated using robust maximum likelihood estimation to evaluate whether a theoretically derived three-item version of the SWLS adequately captures the core cognitive component of SWL. Both models included the WHO-5 as a concurrent latent construct to ensure model identification and to examine nomological relations. The first model specified SWL as a latent factor measured by the first three SWLS items (SWLS-3; see Figure S1 in Supplementary materials), whereas the second model specified SWL as a latent factor measured by all five SWLS items (SWLS-5; see Figure S2 in Supplementary materials). In both models, WHO-5 was specified as a latent factor measured by its five indicators, and the two latent constructs were allowed to correlate.
Model evaluation was based on global fit indices. As shown in Table 2, the model including the three-item version of the SWLS demonstrated better overall fit (CFI = .952, TLI = .930, RMSEA = .033, SRMR = .033) compared to the five-item version (CFI = .930, TLI = .907, RMSEA = .036, SRMR = .035). These results suggest that the abbreviated SWLS-3 provides a more parsimonious representation of the data without loss of model fit.
Table 2
Fit indices for the SWLS-5 and SWLS-3 measurement models
| Model | CFI | TLI | RMSEA | SRMR |
|---|---|---|---|---|
| SWLS-5 + WHO-5 | 0.930 | 0.907 | 0.036 | 0.035 |
| SWLS-3 + WHO-5 | 0.952 | 0.930 | 0.033 | 0.033 |
| WHO-5 | 0.990 | 0.981 | 0.054 | 0.021 |
Across both model specifications, SWL showed a small but statistically significant association with well-being of identical magnitude (r = .26, ps = .033-.034), indicating stable nomological relations irrespective of the number of SWLS items included. Furthermore, both models showed adequate factor loadings (i.e., λ > 0.30) comparable to each other.
Reliability
The SWLS-5 demonstrated good reliability (Cronbach’s α = .80; McDonald’s ω = .81), while the abbreviated SWLS-3 showed a comparable level of internal consistency despite its brevity (α = .78; ω = .79). For completeness, reliability estimates were also computed for the WHO-5, which exhibited high internal consistency (α = .86; ω = .86).
Item response theory analysis
To further examine the item-level properties of the scale, GRMs were estimated. The results indicated that the retained items (Items 1-3) exhibited relatively high discrimination parameters (a = 2.49-3.61), with Item 3 showing particularly strong discrimination. In contrast, the removed items (Items 4 and 5) demonstrated lower discrimination (a = 1.90-1.95). Overall, this pattern of item parameters provides additional support for the psychometric rationale of the abbreviated SWLS-3, particularly in terms of item discrimination, suggesting that the retained items provide more precise measurement of the underlying construct.
Measurement invariance
Measurement invariance across sex was tested for a two-factor model including the WHO-5 and either the three- or five-item version of the SWLS. As shown in Table 3, the SWLS-3 model demonstrated full configural and metric invariance, with minimal changes in fit indices. At the scalar level, fit did not deteriorate (RMSEA decreased from .072 to .055; ΔCFI = .002). Although the absolute change in RMSEA (.017) slightly exceeded the conventional cutoff, it reflected an improvement rather than a decline in fit. Nevertheless, as a sensitivity check, a partial scalar model freeing the intercepts of the SWLS-3 items was additionally examined. This modification did not lead to a meaningful improvement in model fit (Δχ2(6) = 7.53, p = .270; ΔRMSEA = .004; ΔCFI = .001). Therefore, the fully constrained scalar model was retained. The SWLS-5 model showed somewhat weaker baseline fit but satisfied all invariance criteria. Overall, both versions of the SWLS demonstrated scalar invariance across sex.
Table 3
Measurement invariance across sex and samples for SWLS-3 and SWLS-5
[i] Note. ΔCFI and ΔRMSEA represent absolute differences between nested models. The configural model for SWLS-3 between original and current samples was just-identified (df = 0) and therefore not informative; thus, it is not reported. SWLS-3, SWLS-5 – threeandfive-item Satisfaction with Life Scale; WHO-5 – World Health Organization Well-Being Index.
Additionally, measurement invariance across samples (original UK sample; Kjell, 2020 vs. current PL sample) was examined for the SWLS-3. The configural model was just-identified and therefore did not provide an informative test of model fit. Metric invariance was supported; however, full scalar invariance was not achieved, as the change in RMSEA was large (ΔRMSEA = .075). Inspection of modification indices revealed only negligible values (MI ranged from 0.005 to 0.020), indicating no substantial localized sources of misfit; therefore, partial scalar invariance was not pursued.
Convergent and discriminant validity
Bivariate associations were examined using Pearson’s r. As expected, SWLS-3 showed a strong positive correlation with the full version of SWLS-5. Furthermore, both SWLS-3 and SWLS-5 showed positive associations with well-being (WHO-5), indicating moderate relationships. In addition, both versions of the scale were positively related to gratitude and awe (GrAw-7), as well as to self-forgiveness, forgiveness of others, and divine forgiveness (TFS), with effect sizes ranging from small to moderate. Detailed correlation coefficients are presented in Table 1. Differences between dependent correlations were tested using Steiger’s Z, indicating no significant differences (zs ranging from –1.17 to –0.27, ps ranging from .241 to .787).
Convergent validity was further supported by AVE values exceeding recommended thresholds for all constructs (> 0.50). Discriminant validity was examined using the HTMT, which indicated adequate discriminant validity between SWL and well-being (< 0.85). A summary of AVE and HTMT values for both models is presented in Table S1 in Supplementary materials.
Discussion
The findings of the present study confirm that the Polish version of the SWLS-3 is a psychometrically sound instrument for assessing global SWL. As in previous validations (Espejo et al., 2022; Kjell & Diener, 2021; Neto & Neto, 2024), the three-item scale demonstrated good internal consistency, strong item-total correlations, and favorable model fit indices, while offering a more parsimonious representation with comparable structural clarity to the original SWLS-5.
CFA revealed that the SWLS-3 combined with the WHO-5 demonstrated no deterioration in model fit, with marginally better fit indices compared to the full SWLS-5 model, thereby supporting our first hypothesis. This is consistent with findings from Kjell and Diener (2021), who reported that the three-item version provided comparable or slightly more favorable fit across multiple samples. Among the removed items, item 5 (“If I could live my life over...”) has been particularly noted for its susceptibility to retrospective bias and culturally contingent interpretation (Pavot & Diener, 2008), more so than item 4, which retains some reference to current life conditions.
Consistent with prior work (Espejo et al., 2022; Kjell & Diener, 2021; Neto & Neto, 2024), our results also support scalar measurement invariance across sex, indicating that men and women appear to interpret and respond to the SWLS-3 items in a comparable manner. In line with this, no significant associations were found between sex and SWLS-3 scores in the present sample. This level of invariance supports the use of SWLS-3 in comparisons of mean scores across sex, an essential feature for epidemiological and health survey research (Jovanović, 2016). Interestingly, unlike the SWLS-5, which in some cross-cultural studies has failed to reach scalar invariance, the abbreviated version has been suggested to reduce differential item functioning (Kjell & Diener, 2021). However, in the present study, we observed scalar invariance for the SWLS-5, suggesting that both versions are comparable across sex within the Polish sample. In contrast, although metric invariance was supported across the original (UK; Kjell, 2020) and current (PL) samples, the lack of scalar invariance indicates that while the underlying construct is represented similarly across countries, differences in item thresholds may limit direct comparisons of mean scores and should be interpreted with caution. However, metric invariance supports cross-country comparisons of associations, as commonly pursued in cross-cultural research (e.g., Cieciuch et al., 2018).
Convergent validity was also supported, as the SWLS-3 showed significant positive associations with psychological well-being, gratitude and awe, and the three dimensions of dispositional forgiveness. These findings align with the conceptual framework that views gratitude and awe as core components of experiential secular spirituality, which is positively associated with SWL (Büssing, 2021; Konaszewski et al., 2022). Forgiveness, particularly of self and others, similarly has been conceptualized as a coping resource associated with emotional balance and greater SWL (Worthington & Scherer, 2004). The somewhat weaker association with divine forgiveness may reflect the influence of unmeasured religious variables, as this resource tends to be more salient among those who believe in a personal, intervening God (Fincham, 2022). Prior research has shown that its associations vary by culture and denomination – for example, predicting unique variance in depressive symptoms among Georgian but not Polish adults (Skalski-Bednarz et al., 2025). Moreover, divine forgiveness may interact with self-forgiveness, potentially strengthening their association with psychological outcomes under specific conditions (Skalski-Bednarz, 2024).
The SWLS-3 also met recommended thresholds for AVE and demonstrated discriminant validity with well-being (WHO-5), echoing prior results that the three-item version correlates strongly with the full version while retaining conceptual distinctiveness (Kjell & Diener, 2021; Neto & Neto, 2024). The very high intercorrelation between SWLS-3 and SWLS-5 supports the use of the abbreviated form as a practical alternative, although the substantial item overlap between the two versions should be considered when interpreting this association.
It is also important to consider the practical advantages of the SWLS-3. As Hays et al. (2017) noted, shortening scales in large surveys can substantially reduce cost and burden—both financial and cognitive. Given that the additional psychometric benefits of the longer version appear to be limited (Kjell & Diener, 2021), as further supported by our IRT findings indicating higher item discrimination for the retained items, the use of SWLS-3 in public health research and cross-national monitoring may be a practical alternative, particularly when survey burden is a concern.
Nonetheless, some limitations remain. The cross-sectional nature of this study precludes the evaluation of test–retest reliability and longitudinal invariance, both of which were strongly supported in Kjell and Diener’s (2021) multi-wave data. Future research should also explore invariance across age and educational strata, as performed by Neto and Neto (2024), who found full invariance across age groups. Although age was negatively correlated with SWLS-3 scores in our sample, this effect was weak and may reflect sampling bias toward younger participants. Additionally, the use of snowball sampling limits the representativeness of the sample and restricts the generalizability of the findings. Prior findings suggest that SWL remains relatively stable across adulthood, with only negligible-to-small age-related differences in observed scores (Swami et al., 2025).
To conclude, the SWLS-3 appears to be a reliable, valid, and efficient instrument. Consistent with previous research, it demonstrates satisfactory psychometric properties while reducing respondent burden. Future cross-cultural and longitudinal validations may provide further evidence for its use as a concise measure of SWL.
Supplementary materials are available on the journal’s website.
