The Stanford Purpose Assessment (SPA) draws from Damon et al.’s (2003) conception of purpose, defined as “a stable and generalized intention to accomplish something that is at once meaningful to the self and consequential to the world beyond the self” (p. 121). This understanding led to the identification of three key dimensions of purpose (Damon & Colby, 2022; Malin, 2018): firstly, a long-term goal that holds personal significance and is future-oriented; secondly, a motivation that extends beyond oneself, rooted in a deep desire to make a positive impact on the world; and thirdly, a commitment to actively pursue the desired goal through actions that require time, energy, and personal resources.

Measures of purpose have evolved from this multidimensional concept. Initially, purpose research focused on qualitative studies (Damon, 2008; Damon & Malin, 2020) using the Stanford Youth Purpose Interview Protocol (Stanford Center on Adolescence, 2021; Malin et al., 2014). However, the past decade has seen efforts to develop and validate quantitative instruments despite challenges in capturing dimensions such as beyond-the-self motivation (Bronk, 2014).

The Claremont Purpose Scale (CPS) emerged from adapting three existing measures of goal-directedness, personal meaning, and a beyond-the-self orientation (Bronk et al., 2018). The Sense of Purpose Scale (SOPS) features three subscales based on literature review and existing scales: awareness of purpose, awakening to purpose, and altruistic purpose (Sharma et al., 2018). Both the CPS and SOPS focus not on the content of purpose but on an individual’s disposition or sense of purpose.

The Measure of Adolescent Purpose (MAP) also assesses a three-dimensional structure but aims to capture intention, engagement, and prosocial reasoning based on participants’ written goals (Summers & Falco, 2020). The Stanford Purpose Assessment (SPA) (Malin, 2022), developed from the aforementioned interview protocol and refined to its current version, includes space for participants to write an important goal, reasons for its importance, and actions being taken towards it, similarly to MAP. It then features nine closed-ended questions evaluating agreement with items across three factors: the significance of the stated goal, the beyond-the-self motivation, and the engagement in pursuing the goal through concrete actions. However, to date, the SPA has only been validated in one study with an American sample, highlighting the need for further research.

International studies indicate that the content and trajectories of purpose are shaped by sociodemographic factors, sociocultural context, and family values (Damon & Malin, 2020). Brazil, in particular, is characterized by acute socioeconomic inequalities that significantly affect people’s quality of life and ways of living. Additionally, its vast size results in notable cultural and identity differences across regions. The availability of validated purpose instruments like the SPA could facilitate national research and provide greater insight into life projects within the Brazilian context. Given that the SPA has only been administered to a single sample and that this is the first time a psychometric tool based on Damon’s theory has been applied in Brazil, we have chosen a more cautious approach in this study, opting for an exploratory factor analysis.

The translation and adaptation process of the original SPA (v1) into Brazilian Portuguese followed multiple stages, aligned with international guidelines (Borsa et al., 2012; International Test Commission, 2017). Initially, two independent translators rendered the instrument into Portuguese. Two of the authors independently reviewed the translated items, seeking the most appropriate linguistic and cultural adjustments. They then compared their suggested adjustments, discussing and resolving any discrepancies. Ultimately, they produced a unified version (v2). This version underwent evaluation by two experts, leading to minor revisions (v3). Once again, two of the authors reviewed the evaluations and compared the modifications. Subsequently, the target population (N=12) reviewed it, prompting slight modifications (v4). The adapted version was then back-translated into English by two new translators, different from those in the initial phase. The researchers synthesized this feedback (v5) and compared it with the original version (v1). They deemed this version grammatically and semantically equivalent to the original and received approval from the original instrument’s author. The final version is available in Appendix A.

The sample comprised 669 participants (52.9% female), aged 14 to 71 years (M = 30.2, SD = 14.5), from 22 Brazilian states. Data were collected from two independent studies. The first sample consisted of 144 teachers (75.6% female, M = 45.3, SD = 8.6) enrolled in a teacher training on purpose. The second sample, with 525 participants, included: 1) the general population (N = 157, 58% female, M = 38.7, SD = 16.7); 2) university students (N = 261, 35.2% female, M = 22.1, SD = 4.4); 3) public school students (N = 107, 57.9% female, M = 18.2, SD = 6.8). The general population was recruited through an online email and social media survey, while students were recruited at their educational institutions. The final sample excluded participants with attention test failures (N = 29), blank responses (N = 151), and missing SPA data (N = 11). Ethical standards were upheld throughout the study, with approval from the Human Research Ethics Committee of the University of São Paulo (CAAE 69798423.0.0000.5390). All participants provided informed consent before data collection.

Participants in both studies completed the SPA, while those in the second study also responded to assessments concerning purpose, well-being, depression, anxiety, and stress levels.

Stanford Purpose Assessment (SPA). Developed by Heather Malin in 2022, the SPA is a mixed-method tool that includes three open-ended and nine closed-ended items. The open-ended items solicit narratives about a significant goal, why it is important, and steps being taken toward its achievement. The closed-ended items, rated on a 5-point Likert scale from 1 (completely disagree) to 5 (completely agree), evaluate three dimensions: meaning (e.g., “Doing this will give me a sense that my life has purpose”), beyond-the-self motivation (e.g., “My strongest motivation for doing this is a desire to do something good for the world”), and engagement (e.g., “In the past month, I have often taken action to accomplish this goal”). This investigation focuses exclusively on the closed-ended items.

Mental Health Continuum – Short Form (MHC-SF). The MHC-SF by Wagner de L. Machado and Denise R. Bandeira (2015) features 14 items that gauge well-being across three domains on a scale from 1 (never) to 6 (every day): emotional well-being (e.g., “Happy”), psychological well-being (e.g., “Likes own personality”), and social well-being (e.g., “Contribute to society”). As indicated in the original study conducted in Brazil, the score was calculated by summing the scores of all 14 items. In this study, the MHC-SF demonstrated excellent goodness-of-fit indices: CFI = .98; TLI = .97; RMSEA (90% CI) = .053 (.042 – .064); SRMR = .067. Cronbach’s alpha was .88.

Meaning in Life Questionnaire – Presence of Meaning (MLQ-P). Employing the MLQ-P’s 5-item subscale, participants rated their agreement from 1 (absolutely untrue) to 7 (absolutely true) on statements like “I understand my life’s meaning” (Damásio & Koller, 2015). The score was obtained by summing the values of the 5 items, one of which was reverse-scored. In this study, the MLQ demonstrated excellent goodness-of-fit indices: CFI = 1.00; TLI = 1.00; RMSEA (90% CI) = .000 (.000 – .041); SRMR = .016. Cronbach’s alpha was .89.

Depression Anxiety Stress Scale (DASS-21). The DASS-21, as adapted by Rose C. B. Vignola and Adriana M. Tucci (2014), uses a 4-point Likert scale to assess emotional distress across three 7-item subscales: depression (e.g., “I felt I wasn’t worth much as a person”), anxiety (e.g., “I felt I was close to panic”), and stress (e.g., “I found it difficult to relax”). The aggregate score for each subscale was calculated by summing the scores of the corresponding 7 items. In this study, the expected tri-factorial solution demonstrated excellent goodness-of-fit indices: CFI = 1.00; TLI = 1.00; RMSEA (90% CI) = .000 (.000 — .000); SRMR = .041. Cronbach’s alphas for depression, anxiety, and stress were .88, .84, and .87, respectively.

Bartlett’s sphericity test (2626.8, df = 36, p < 0.00001) and KMO measure (0.83) confirmed the interpretability of the item correlation matrix. Parallel analysis pointed to a single factor as the most representative of the data (refer to Table 1). Measures of Unidimensional Congruence (UniCo = 0.958 [95% CI = 0.839-0.983]) and Mean of Item Residual Absolute Loadings (MIREAL = 0.283 [95% CI = 0.242-0.320]) supported the data’s unidimensionality, despite the Explained Common Variance (ECV = 0.822 [95% CI = 0.798-0.853]) not doing so.

This indication of unidimensionality led to the following fit indices for the model: χ² = 311.855, df = 27; p < 0.05; RMSEA = 0.126 (95% CI = 0.010-0.141); CFI = 0.936 (95% CI = 0.915-0.961); TLI = 0.914 (95% CI = 0.886-0.948); BIC = 428.959 (95% CI = 324.667-506.907). The significant chi-square and the RMSEA’s indication of a poor model prompted us to evaluate a three-dimensional model, which presented acceptable fit indices: χ² = 18.800, df = 12; p > 0.05; RMSEA = 0.029 (95% CI = 0.000-0.037); CFI = 0.998 (95% CI = 0.997-1.000); TLI = 0.995 (95% CI = 0.992-1.001); BIC = 253.008 (95% CI = 244.293-257.097). With all three-dimensional model indices being acceptable and the BIC lower than the unidimensional model, we adopted the multidimensional model in line with Malin (2022).

Table 2 details item factor loadings, internal reliability indices for each factor (composite reliability, Cronbach’s alpha, and McDonald’s omega), and factor score estimates for the representativeness of the latent trait and effectiveness of factor estimation (FDI), accuracy (Orion marginal reliability), and replicability (H-index). Items demonstrated suitable factor loadings, with values exceeding 0.3 on theoretically anticipated factors.

While “meaning” and “beyond-the-self motivation” factors showed acceptable internal consistency indices, the “engagement” factor did not. The factor structure’s replicability (H-index; Ferrando & Lorenzo-Seva, 2018) indicated the potential non-replicability of the ‘engagement’ factor in future studies (H < 0.80). Furthermore, the FDI and Orion marginal reliability suggested that scores from this factor might lack reliability, possibly impacting its association with external measures. In contrast, the “meaning” and “beyond-the-self motivation” factors displayed satisfactory H-index, FDI, and Orion marginal reliability. After excluding item 6 from the instrument, recalculating indices improved all indices except composite reliability, yet not to acceptable levels.

Our evaluation of convergent validity involved 424 participants who filled out external measures (refer to Table 3). In line with expectations, the three SPA factors correlated positively with presence of meaning (MLQ-P) and well-being (MHC-SF) and negatively with depression (DASS-21). However, while negative correlations of SPA factors with anxiety and stress were anticipated, only the “engagement” factor showed significant negative correlations. Notably, the significant negative correlation between meaning and anxiety was almost negligible.

Considering the magnitude of these correlations is crucial (Cohen, 1977). The observed correlations ranged from weak (|0.10 – 0.30|) to moderate (|0.30 – 0.50|). Additionally, the three SPA factors exhibited positive intercorrelations.

Since the engagement factor failed to achieve expected reliability and the factor correlations did not match anticipated magnitudes, we applied the Rasch model for polytomous data from Item Response Theory to assess item alignment. Table 4 details general reliability indicators for both persons and items for each factor, including item overall difficulty, separation index, reliability, and fit measures (infit and outfit). Reliability indices for the three factors were satisfactory for items (reliability > 0.81; separation index > 1.53) but not for persons, suggesting that the factor estimates might not replicate this sample’s latent trait level with a comparable instrument. The performance deviations (infit and outfit) indicated that the factors’ performance conformed to expected cutoffs.

However, when analyzing performance deviations for individual items, three items fell outside the anticipated cutoffs, as shown in Table 5, which also presents each item’s difficulty indicators. Item spa5 (BTS motivation) exhibited an overfit to the Rasch model, indicating a response pattern better than expected (Infit MNSQ = 0.86, ZSTD = −2.24; Outfit MNSQ = 0.86; ZSTD = −2.23). In contrast, items spa6 and spa2 from the engagement factor showed a misfit, indicating a more significant response variability than expected. This misfit was only evident in the infit for item spa2 (Infit MNSQ = 0.85, ZSTD = 2.68; Outfit MNSQ = 0.91; ZSTD = −1.52), while for item spa6, it appeared in both infit and outfit (Infit MNSQ = 1.15, ZSTD = 2.74; Outfit MNSQ = 1.11; ZSTD = 1.90).

Additionally, the items’ difficulty exceeded the latent trait level, as the mean theta for all three factors was higher than the mean item difficulty (item location). This discrepancy is more apparent in Figure 1(a-c), which displays the item-person map for each factor. Generally, item difficulty did not exceed 0.5 logits, whereas most participants’ latent traits ranged from 0 to +4 logits. The meaning factor, depicted in Figure 1(a), showed the most significant discrepancy between theta and delta and, accordingly, the lowest person reliability index, as indicated in Table 4.