Contribution
The first validation of the Loyola Generativity Scale (LGS) for Chilean teachers. It translates and adapts the scale, removes four items and proposes a three-factor structure in place of the two factors the authors attribute to the original. Its claim of interest to the record is that a “Generative Legacy” factor, motivation toward symbolic continuity, separates from “Generative Action”, concrete generative behaviour and commitment.
Key insight
Within a single generativity scale, wanting to leave something that lasts and doing generative things come apart as related but distinct factors (r = .448, p. 8). Generative Legacy “represented motivations related to symbolic permanence and societal contribution” (p. 8).
Assumptions
- The original LGS (p. 4): 20 items, answered on a 4-point scale from “never applies to me” to “very often applies to me”, covering concern for future generations, productivity, community involvement and perceived personal impact. The authors say it has two factors, Positive Generativity (17 positively worded items) and Generative Doubts (3 negatively worded), citing McAdams and de St. Aubin (1992, p. 1008). The record cannot check this against the 1992 paper.
- Ordinal items call for polychoric correlations and robust unweighted least squares rather than maximum likelihood (pp. 4–5).
- Sampling adequacy as a filter. Items with an MSA below .50 share too little variance and are removed (p. 5).
- Teaching is “inherently generative” (p. 2), the premise for choosing teachers.
Key results
- Sample (pp. 3, 5–6; Table 1). 430 elementary teachers with three or more years’ experience, in municipal and subsidized private schools in four regions (Metropolitan 39%, La Araucanía 26%, Maule 19%, Los Ríos 16%), from about 47,600 in those regions. Stratified by region and school type, then voluntary: a convenience sample. 76% women; 66% aged 30–49; 76% urban. Response rate unknown.
- Translation (p. 4). Forward translation, reviewed by three experts. No back-translation is mentioned.
- Split-sample check (p. 5; Table 2). The Solomon method split the sample into two halves of 215; ratio communality index S = 0.9789; KMO .8318 and .8143.
- Items (p. 5; Table 3). Item means ranged from 1.644 to 3.456.
- Four items were removed for sampling adequacy: GEN5, GEN9, GEN11 and GEN14. Their MSA values (.589 to .769) are above the stated .50 cut (p. 5); the lower bounds of their bootstrap intervals (.341 to .475) are the only ones below it, so that is presumably the rule applied, though the text does not say so.
- On the 16 remaining items: determinant 0.0019; Bartlett χ²(120) = 2645.4, p < .001; KMO .872 (bootstrap 95% CI .831–.905).
- Number of factors (pp. 6–8; Figure 1; Table 4).
- Eigenvalues 5.804, 1.674, 1.337, then 1.029. The scree plot’s elbow and the Hull method chose three.
- One factor: 36.3% of variance, χ²/df = 4.431, RMSEA = 0.089.
- Two factors: 46.7%, χ²/df = 3.137, RMSEA = 0.071.
- Three factors: 55.1%, χ²/df = 1.792, RMSEA = 0.043, CFI = 0.989, NNFI = 0.983, GFI = 0.991, AGFI = 0.985, RMSR = 0.041. Close-fit test p = 0.83 for RMSEA < .05.
- Four factors fit slightly better (61.5%, RMSEA = 0.032) but were rejected as a weak, poorly defined over-extraction.
- The three factors (p. 8; Table 5), standardized loadings.
- Generative Inefficacy: GEN2, GEN13, GEN15 (0.617–0.801). Reverse-coded items on low impact and lack of usefulness.
- Generative Action: GEN1, GEN6, GEN7, GEN12, GEN16, GEN17, GEN18, GEN19 (0.426–0.745). Creativity, productivity, teaching, community engagement.
- Generative Legacy: GEN8, GEN10, GEN20 (0.605–0.735).
- Correlations: Inefficacy with Action −.340, with Legacy −.295; Action with Legacy .448 (95% CI .370–.538).
- Reliability (p. 9; Table 6). ORION .795, .863 and .824; factor determinacy .892, .929 and .908.
- Not reported in the body. The item wordings, the factor means and the “consistently high levels of generativity across demographic groups” the abstract describes. These may be in the Supplementary Material.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | In Chilean teachers a 14-item LGS has three factors: inefficacy, action and legacy | moderate as a description of this sample; exploratory, single sample, no invariance test | pp. 6–9, Tables 4–5 |
| C2 | Generative Legacy, motivation toward symbolic continuity, is distinct from Generative Action | moderate: separate factors correlating .45; but each is three or eight items whose wording is not given | p. 8 |
| C3 | Generative Inefficacy is a new facet, reflecting the pressures of the Chilean school system | weak: it consists of the three negatively worded items the authors themselves list as the original “Generative Doubts”, so a wording factor is a rival explanation | pp. 8, 10 |
| C4 | The refined scale’s factor scores are reliable | moderate: ORION .80–.86, determinacy .89–.93 | Table 6 |
Concepts
- generativity: “the enduring concern for contributing to the well-being, development, and continuity of future generations” (p. 2).
- Generative Legacy: the authors’ third factor, motivation toward “symbolic permanence and societal contribution”.
- Generative Action: concrete prosocial behaviour and commitment.
- Generative Inefficacy: doubts about one’s generative impact, measured by reverse-coded items.
- MSA, Hull method, ORION, FDI: item sampling adequacy; a fit-against-parsimony rule for the number of factors; reliability of factor scores; correlation of factor scores with the latent factor.
Connections
- McAdams and de St. Aubin 1992 (LIT-1777), the work this note stands in for. The paper gives the scale’s format and cites the 1992 abstract: the LGS “showed strong positive associations with reports of actual generative acts” (p. 4, quoting p. 1003). It does not report the 1992 samples or statistics.
- Kruse and Schmitt 2012, read alongside, restates the model the scale was built from and uses the scale in an intervention.
- Kotre (1984), Outliving the Self, is in the references, the source of the phrase “outliving the self” behind generativity’s symbolic-immortality motive. Not held by the record.
Bearing on the record
- The legacy question (QUESTION-301). Its first answer
criterion is a test that separates the readings: measure felt
continuity with one’s legacy, and ask whether concern follows that
measure or caring alone. As inference:
- A candidate measure exists. A three-item Legacy factor, about symbolic continuity, separates from an eight-item Action factor in this sample. If its items are about outliving oneself and Action’s about doing for others, the two could be pitted against each other: does generative action track the legacy motive?
- The paper does not run that test. It reports only the correlation, .45. Action and Legacy are both self-reports, so the correlation cannot say which drives which.
- The items are not printed, so the record cannot tell whether Legacy measures a self extended in time (“my work will live on”) or a concern for the future that is not about oneself. That difference is the one between the thesis’s reading and Scheffler’s.
- The triviality objection (CLAIM-492). A legacy factor separable from action is a criterion of the temporal self’s extent that might come apart from caring. It is a candidate only: one sample, exploratory, items unseen.
- The owner’s thesis (CLAIM-491). Nothing changes yet. The thesis would predict that generative action rises with the legacy motive; the data are compatible with that and with its reverse.
- A dated note on QUESTION-301, jointly with Kruse and Schmitt: the proposed text is in the Kruse draft. No note is needed on CLAIM-491.
- No THEORY filed.
- No ML instruction.
Limitations
- Called confirmatory, done exploratory. The number of factors was chosen from scree plot, eigenvalues and the Hull method, solutions of one to four factors were compared, and “rotated factor variances” are reported (pp. 6–9). That is exploratory factor analysis, run in FACTOR. The split halves were used only to check equivalence, not to confirm a structure found in one half on the other.
- Two items disappear. Sixteen items survive the MSA screen and enter the matrix tests (Bartlett’s df of 120 fits 16 items). The final solution has 14 (Table 5; “the 14 retained items”, p. 9). GEN3 and GEN4 are not mentioned again, and no reason is given.
- A wording factor is a rival reading (an inference from Table 3). The six items with the lowest means are GEN2, GEN5, GEN9, GEN13, GEN14 and GEN15 (1.64 to 2.21), the pattern of negatively worded items. MSA removed three of them; the other three are the whole of Generative Inefficacy. That also sits oddly with the authors’ account of three negatively worded items in the original.
- The MSA rule is misstated. The text says items with MSA below .50 were removed, but the four removed items have MSA between .589 and .769; only their bootstrap lower bounds fall below .50 (Table 3).
- Sample: convenience, teachers only, cross-sectional, self-report; no measurement invariance test; no criterion measure of generative behaviour.
- Unrelated literature in the introduction, such as generativity in machine learning and a study of autism (Hocaoglu and Güvendir, cited for generativity across groups).
- Typos. “Relaibility” (Table 6 title); “No binary” and “Ungraduated” (Table 1); “47.600” for 47,600 (p. 3). The education rows of Table 1 sum to 404 of 430 (94%), so a category is missing.
- Generative AI (Microsoft Copilot) was used in writing, as the authors declare (p. 12).
Open questions
- What do GEN8, GEN10 and GEN20 say? If they are about being remembered or outliving oneself, the Legacy factor is a measure of the extended temporal self; if about the future’s good, it is not.
- Does Legacy predict generative behaviour, measured by McAdams and de St. Aubin’s acts checklist, over and above Action?