Psychological grit has garnered much attention over the last 15 years and has widely been touted as the panacea to predicting student retention and success among traditional tertiary students (Duckworth et al., 2007; Wu et al., 2012; Cross, 2014; Salles et al., 2014; Reraki et al., 2015; Beyhan, 2016; Akos and Kretchmar, 2017; Broghammer, 2017; Lee and Sohn, 2017; Muenks et al., 2017; Pate et al., 2017; Salles et al., 2017; Mason, 2018; Rogalski, 2018; Terry and Peck, 2020; Sulla et al., 2022). In addition to this body of predictive validity, the Grit-S scale has demonstrated accepted construct validity and reliability levels among these student populations. Although noteworthy, the focus on traditional students precludes the generalizability of the psychometric properties of the Grit-S scale to other student populations (Xu et al., 2020). With the majority of the psychometric reports emanating from traditional, full-time students, or those enrolled in atypical educational settings (such as Ivy-League institutions or military academies), the importance of examining grit’s factorial stability in distance education (DE) cohorts cannot be overstated, especially considering the online impetus driven by COVID-19 (Karaman et al., 2019). Moreover, as Xu et al. (2020) rightly allude to, little is known about grit in DE and whether it can be reliably measured across culturally, ethnically and linguistically diverse student populations – characteristic of the South African DE student population. It is against this brief introduction that the current paper explores whether the Grit-S scale exhibits reasonable construct validity and reliability among postgraduate students in a diverse South African DE environment.

It is essential to first define grit in the context of this paper and then outline how the conceptualization of grit can be measured. The following section deals with both the definition and measurement of grit.

The two-factor structure of grit is commonly cited in educational research (Duckworth and Quinn, 2009; Arslan et al., 2013; Christensen and Knezek, 2014; Meriac et al., 2015; Reraki et al., 2015; Rojas, 2015; Stewart, 2015; Datu et al., 2016; Broghammer, 2017; Pate et al., 2017; Arco-tirado et al., 2018; Collaço, 2018; Rojas and Tyler, 2018). In the earliest and perhaps most seminal of these studies, Duckworth and Quinn (2009) showed that the two-factor structure of the Grit-S scale held an acceptable level of model fit among two cohorts of West Point cadets² (CFI ≥ 0.950, RMSEA < 0.07). More importantly, these results revealed that the two-factor model fit the data significantly better (when compared to a single-factor solution) as indicated by a significant chi-square difference, Δχ²(1) = 191.93, p < 0.001.

Within a HE setting, the two-factor solution exhibited good levels of fit among both traditional (CFI > 0.98, RMSEA < 0.05; Pate et al., 2017; Rojas and Tyler, 2018) and non-traditional students in the United States (CFI = 0.98, RMSEA = 0.06; Meriac et al., 2015), as well as among Filipino students enrolled in a private higher education institution (CFI = 0.97, TLI = 96, RMSEA = 0.05; Datu et al., 2016). Further studies conducted in various countries including Spain, Turkey, Italy, Poland, and the Netherlands have also shown that the two-factor model demonstrates acceptable levels of fit in translated versions of the Grit-S scale among tertiary students. These studies have reported CFI indices of at least 0.920, RMSEA of no more than 0.08, and SRMR indices of less than 0.08 (Arslan et al., 2013; Reraki et al., 2015; Saricam et al., 2016; Wyszyńska et al., 2017; Arco-tirado et al., 2018; Sulla et al., 2018; Xu et al., 2020).

Although limited, there is some support for the two-factor grit structure among DE students. Xu et al. (2020) compared one and two-factor solutions (for both the Grit-O and Grit-S scales) among DE students in the Netherlands using an item response theory (IRT) CFA (n = 2,027). The first model assumed uni-dimensionality, whereas the second tested the two-factor structure with passion and perseverance as separate dimensions. Concurrent with other findings in the field, the two-factor model of grit fit the data significantly better than a one-factor solution for both scales (Xu et al., 2020). Moreover, the two-factor Grit-S model fitted the data better than the earlier two-factor Grit-O model.

There are however more recent studies which have favoured a unidimensional structure of grit (Gonzalez et al., 2020; Postigo et al., 2021). These studies have reported model fit indices ranging between 0.89 and 0.98 (CFI) and RMSEA values between 0.12 and 0.05, respectively. Regardless of whether commonly-accepted thresholds have been met (CFI > 0.95 and RMSEA < 0. 06; Hu and Bentler, 1999), the findings by Gonzalez et al. (2020) emanate from samples outside of a HE setting (as do those by Postigo et al., 2021). Moreover, Postigo et al. (2021) developed a new measure of grit (i.e. the Oviedo Grit Scale / escala grit de Oviedo; EGO) and did not validate either of the grit scales by Duckworth et al. (2007) and Duckworth and Quinn (2009). As such, although these results highlight a lack of clarity which still exists on the dimensional aspect of the construct, their applicability in the current paper is limited.

Spanning over a decade, grit and its dimensions have shown to be reliable across a range of student populations. In the earliest of these studies, Duckworth and Quinn (2009) reported an overall reliability estimate of 0.83 for the Grit-S scale among Ivy League undergraduates in the United States (n = 139). Further analyses revealed reliability coefficients of 0.79 and 0.78 for the passion and perseverance domains, respectively. Additional research emanating from contact-based institutions in the United States has attested to the reliability of the Grit-S scale among undergraduates, with Cronbach’s alpha coefficients ranging from 0.73 to 0.84 (Akos and Kretchmar, 2017; Rojas and Tyler, 2018; Luthans et al., 2019).

Reraki et al. (2015) conducted a study with undergraduate students in Turkey and found similar results, reporting an overall Cronbach’s alpha of 0.88 for the Grit-S scale. More recently, Arco-tirado et al. (2018) and Sulla et al. (2018) conducted studies with undergraduate students in Spain and Italy, respectively, and found that the Grit-S scale also produced reliable results, with Cronbach’s alpha coefficients of 0.75 and 0.76. In addition to these European statistics, research emanating from Korea has attested to the reliability of the Grit-O scale among a sample of undergraduates (α = 0.75; Hwang et al., 2017).

From a global south perspective, the Grit-S and Grit-O scales have also shown to produce reliable results among undergraduate students in contact-based universities in South Africa (Urban and Pendame, 2015; Mason, 2018). In the most recent of these South African studies, Mason (2018) revealed that the Grit-O scale, along with its subscales of passion and perseverance, exhibited good reliability among first-year students in undergraduate programmes at a university of technology (α = 0.81, α = 0.80, and α = 0.78, respectively). The Grit-S scale, by similar token, demonstrated acceptable levels of reliability among final-year undergraduates enrolled in a traditional university in South Africa, producing a reliability coefficient of 0.71 (Urban and Pendame, 2015). Clearly, there is substantial evidence to support the reliability of the grit scales among various undergraduate cohorts (see Table 1 for a chronological record of the reliability coefficients derived from previous research in the higher education sector).

Given the paucity highlighted above, the need to investigate the psychometric rigour among distance learning students is evident. The current paper aims to psychometrically evaluate the Grit-S scale among a sample of postgraduate distance learning students in South Africa by examining the inherent factor structure and the derived reliability coefficients. The following question prompted the current analysis;

An English survey consisting of two sections (i.e. the Grit-S scale and a demographic section) was administered online during the 2017 academic year. Of the 8,689 potential participants in the target population (i.e. postgraduate students who enrolled for their qualification for the first time in 2017), 837 students submitted complete responses. A final response rate of 9.63% was calculated.

The majority of the research participants were female (n = 587; 70.1%) and between the ages of 25–29 years (n = 203; 24.3%). This was followed by 18.6% of the sample who indicated that they were between the ages of 30–34 years (n = 156). Regarding the ethnicities of the research sample, over half was African (n = 495; 59.1%), while over a quarter was white (n = 228; 27.2%). A fair majority of the research participants were first-language English-speakers (n = 237; 28.3%).³ Afrikaans and isiZulu were also prominent among the sample, comprising 14.9% (n = 125) and 14.5% (n = 121), respectively. These were followed by Setswana- (n = 62) and Northern Sotho-speaking participants (n = 59). A detailed description of the sample can be found in Young (2020).

Subsequent to analysing the demographic attributes, the Grit-S scale’s factorability was assessed (using the KMO statistic and Bartlett’s test of sphericity) followed by an exploration of the underlying dimensions in the scale (using exploratory factor analysis; EFA). This underlying structure was then supplemented using a partial confirmatory factor analysis (PCFA; Gignac, 2009) and the factors tested for internal consistency using Cronbach’s alpha coefficients. All analyses were performed in SPSS (version 28) and are presented below.

The findings of the EFA showed that the same two factors (i.e. passion and perseverance), previously identified by Duckworth and colleagues (Duckworth et al., 2007; Duckworth and Quinn 2009), were evident in the Grit-S scale when applied to a sample of postgraduate, DE students in South Africa. These two factors, consistency of interest and perseverance of effort, collectively explained 57.2% of the variance. Although typical of social science research where extracted factors usually explain between 50 and 60% (Peterson, 2000), these results are slightly higher than those reported in the current discourse. Alhadabi et al. (2019) indicated that 47.7 and 51.0% of the total variance in the data could be explained by the same two-factor structure among Omani and American students, respectively. Collaço (2018) results, on the other hand, indicated that the two-factor solution explained only 42.84% of the total variance among a sample of college students, almost 15% less than that reported in the current paper. Overall, the higher percentage of total variance explained in the current paper may suggest that the two factors are more meaningful and useful in explaining the relationships among the variables among a sample of postgraduate, DE students in South Africa.

Results further revealed that the two-factor solution proposed by Duckworth et al. (2007) and Duckworth and Quinn (2009) demonstrated good levels of fit among the postgraduate DE sample. Index calculations revealed an NFI of 0.98, a TLI of 0.98, a CFI of 0.99, and an RMSEA of 0.04. Based on the corresponding residual correlation matrix, SRMR was calculated at 0.02. Support for this two-factor solution among tertiary students is not novel, with concurrent results emanating from around the globe. Most recently, Schmidt et al. (2021) reported adequate fit of the two-factor solution⁵ among university students in the Czech Republic; CFI = 0.94, TLI = 0.91, RMSEA = 0.08, SRMR = 0.05. Broghammer (2017); Karaman et al. (2019); and Meriac et al. (2015) also reported satisfactory indices for the two-factor solution among undergraduate students in the United States: GFI ≥ 0.96; CFI ≥ 0.96; TLI ≥ 0.94; RMSEA = 0.06; SRMR = 0.04. Interestingly, Fosnacht et al. (2018) only achieved adequate fit for the two-factor model when Item 2 (Setbacks do not discourage me; CFI = 0.97, TLI = 0.95, RMSEA = 0.06) was removed, somewhat consonant with that discussed next.

The reliability analysis indicated that the two dimensions as well as the Grit-S scale in its entirety demonstrated acceptable levels of reliability among the sample (α > 0.62). However, as seen in Table 5 above, the analysis suggested the removal of Question 2 (Setbacks do not discourage me), resulting in a Cronbach’s Alpha coefficient of 0.79 for the perseverance domain and an overall Cronbach’s Alpha of 0.77. Interestingly, this finding is not isolated; a number of studies have noted concerns with this item among different student populations (Tyumeneva et al., 2014; Bowman et al., 2015; Li, 2015; Broghammer, 2017; Wyszyńska et al., 2017; Arco-tirado et al., 2018; Fosnacht et al., 2018). In the earliest of these studies, Bowman et al. (2015) revealed that ‘Setbacks do not discourage me’ reduces the internal consistency of the perseverance domain among a population of undergraduates in the United States (n = 417). According to these authors (preliminary) analyses showed that the removal of this item does not affect subsequent results. Broghammer (2017) also reported a reduction in reliability with the inclusion of this item among first-time entering students in the United States (n = 544).

Although it is unclear whether the lower reliability coefficient is an artefact of the negatively-phrased item, the substantial gain in reliability cannot be ignored. This may be due, in part, to the cognitively laborious nature of these types of items which require additional mental resources to process, often cause confusion, and do not perform as intended (Chyung et al., 2018). There also remains the possibility that this item is irrelevant to the construct being measured (i.e. perseverance of effort) among the South African sample of postgraduate DE students, despite its face validity. Resultantly, regardless of the suggestions by Furr and Bacharach (2014) to retain items that are intraneous to the core construct (so as to avoid construct under-representation), its exclusion could be warranted in this setting and others which explore grit among postgraduate students studying at a distance in South Africa. However, this is stated tentatively and requires further exploration.

A number of implications, limitations and future directions are inherent in the findings above. Firstly, the results imply that grit (and its two inherent dimensions) can be measured accurately within DE settings, across a diverse population of postgraduate students in South Africa (i.e. in terms of ethnicities, home languages and cultures). However, the measure has yet to be validated among an undergraduate DE population in South Africa. Considering the sheer size of the undergraduate population at UNISA which accounts for approximately 30% of the entire student body in the public HE system (i.e. 316,000 students; Department of Higher Education and Training, 2023), results from such a study may yield fruitful insights into the scale (and the construct) from a national perspective. Secondly, the survey was administered and validated in English, undoubtedly adding to the plethora of psychometric evidence on the English Grit-S scale. However, considering that English is only the sixth most common home language spoken inside South African households, future studies should explore the psychometric rigour of translated versions of the Grit-S scale - paying particular attention to the official languages of the country. Moreover, the measurement invariance across these languages should be explored. With so few studies reporting the measurement invariance of grit across different language and cultural groups (Li et al., 2023), future endeavours in this regard may suggest (or refute the idea) that the theoretical understanding of grit is limited to a single culture. Thirdly, having partially established the construct validity of the scale among postgraduate DE students in the country, future research endeavours should seek to confirm this two-factor solution using a confirmatory approach (i.e. using CFA), as suggested by Gignac (2009). And lastly, educational practitioners should explore grit’s predictive validity in the same domain. Initiatives in this area may yield interesting findings into grit’s role in determining retention and success among DE students in the country.

Although there is much debate regarding the factor structure of grit – a critique recognised by the developers themselves (Duckworth et al., 2021) – the current paper provided novel insight into how the Grit-S scale and its intraneous dimensions perform among a sample of postgraduate distance education students in South Africa. Following the suggested criteria (Hu and Bentler, 1999) and acceptable thresholds (Field, 2009), it seems reasonable to conclude that the Grit-S scale is a valid tool and can reliably measure passion and perseverance among postgraduate students at a South African DE institution.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

This study, involving humans, was approved by the University of South Africa’s (UNISA) Research Permission Sub-committee (RPSC) and the Department of Psychology’s Ethics Review Committee. The study was conducted in accordance with the local legislation and institutional requirements. The participants provided their informed consent to participate in this study.

All authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication.