Student study finds limited Grit-S invariance across South African language groups

A study of 959 first-year university students found that three South African translations of the short grit scale measured a broadly coherent psychological construct, but did not support direct comparisons of average grit levels across language groups. The researchers evaluated the English-Afrikaans, Nguni-Tsonga and Sotho-Makua-Venda versions of the Grit-S using confirmatory factor analysis and multigroup confirmatory factor analysis. Tests for configural, metric and scalar invariance assessed whether the translated forms could be compared in increasingly demanding ways across the three language groups.

In the pooled sample, the researchers’ predetermined two-factor model produced acceptable fit. Its comparative fit index was 0.97, its root mean square error of approximation was 0.047 and its standardized root mean square residual was 0.034. The model separated grit into passion and perseverance domains. However, one item—Question 2, “Setbacks don’t discourage me”—had an unfavorable loading of 0.15, indicating that it contributed weakly to its assigned factor in the tested model. The overall fit therefore coexisted with a poorly performing indicator, rather than establishing that every item operated as expected.

The multigroup results established configural and metric invariance but not scalar invariance. These findings support the conclusion that the three translations capture the same underlying construct and have comparable factor structures across the sampled language groups. The failure of the scalar-invariance model, reflected in a change in comparative fit index of −0.05, means that latent means across the groups cannot be treated as directly comparable. The study therefore supports a qualified form of measurement equivalence rather than full interchangeability, and the authors state that comparisons of grit levels should use latent-variable models that allow for partial invariance.

Reliability results added another limitation. Estimates from Cronbach’s alpha and McDonald’s omega indicated that the two domains were generally not reliable in the language-group samples, with the exception of the passion domain in the English-Afrikaans version, which had internal consistency above 0.70. Removing Question 2 improved reliability for the perseverance domain in the English-Afrikaans and Sotho-Makua-Venda translations, raising McDonald’s omega to 0.70–0.73. The authors consequently recommend revising or removing Question 2 in future iterations. Overall, the research identifies a workable but imperfect translation set: the factor structure and partial invariance support use of the Grit-S, while the non-invariant latent means and uneven reliability constrain comparisons among the South African language groups studied.

Subject of Research: Psychometric factor structure, measurement invariance and reliability of South African Grit-S translations among first-year university students

Article Title: South African translations of the Grit-S scale: factor structure, measurement invariance, and reliability across language groups

Article References: Young, K. (2026). South African translations of the Grit-S scale: factor structure, measurement invariance, and reliability across language groups. BMC Psychology. https://doi.org/10.1186/s40359-026-05833-3

Image Credits: AI Generated

DOI: 10.1186/s40359-026-05833-3

Keywords: Grit-S, psychological grit, South Africa, measurement invariance, confirmatory factor analysis, scale reliability, English-Afrikaans, Nguni-Tsonga, Sotho-Makua-Venda, university students, South, African