Reply to Guo et al. and Credé: Grit-S scale measures only perseverance, not passion, and its supposed subfactors are merely artifactors

We (1) propose that evidence linking grit and performance is mixed because the measure used to assess grit—the Short Grit (Grit-S) scale (2)—captures only perseverance, not passion, whereas the definition of grit encompasses both perseverance and passion (3). Our studies find that the combination of perseverance (measured through the whole Grit-S scale) and passion (measured through the passion attainment scale) predicted higher performance.

In their letters, Guo et al. (4) and Credé (5) suggest that the Grit-S scale should be treated as reflecting two factors. Credé (5) subsequently advocates for and Guo et al. (4) conducts separate analyses for each supposed subfacet of the Grit-S scale (called “perseverance of effort” and “consistency of interests”).

In this response, we provide additional evidence that the Grit-S scale is unidimensional. That is, there is no validity to the claims by Guo et al. (4) and Credé (5) that the Grit-S scale is composed of two subfactors. As a result, the analyses advocated by both letters are inappropriate. Instead, the Grit-S scale should be treated as unidimensional, which is how we conducted our original analyses (1).

A careful examination of the Grit-S scale items reveals two important points. First, closely reading the scale items highlights that the Grit-S scale captures perseverance alone and does not capture passion. Passion, defined as “a strong feeling toward a personally important value/preference that motivates intentions and behaviors to express that value/preference” (1), is not captured in any of the Grit-S scale items (Table 1, upper section). For example, Guo et al. (4) and Credé (5) claim that the items “I often set a goal but later choose to pursue a different one” and “New ideas and projects sometimes distract me from previous ones” capture passion. However, a careful reading of the items in the Grit-S scale reveals that there is not a single item in the Grit-S scale that reflects the theoretical concept of passion. Rather, each item in the Grit-S scale reflects perseverance. This is further supported by a recent study which finds no significant relationship between self-rated consistency of interests scores (i.e., the subfactor originally intended to measure passion) and other-rated passion (6).

Second, a careful examination of the Grit-S scale items also shows that all items for the perseverance of effort subfacet are positively worded, whereas all items for the consistency of interests subfacet are negatively worded (Table 1, upper section). It is well known that participants respond differently to positively and negatively worded items (7). For this reason, reverse-coded items are commonly evenly counterbalanced within rather than across factors. In the case of the Grit-S scale, this design flaw can lead to the spurious finding that the Grit-S scale loads on two factors. This does not, however, reflect a true underlying two-factor structure, but instead reflects a mere statistical artifact that produces a spurious multiple-factor structure [i.e., artifactors (8)].

Indeed, an independent research team recently published a paper which highlighted this flaw in the Grit-S scale construction, stating that “the two-factor structure of [the Grit-S scale] appears to be a method artifact, due to the use of both positively and negatively worded items” (9). Thus, it is most appropriate to collapse the whole Grit-S scale into a single measure of perseverance, which is how we employed this scale in our original analysis (1), and other authors have recently started doing as well (10).

To further demonstrate the Grit-S scale’s unidimensionality—that is, that it captures only perseverance, not passion—we wrote and tested two versions of the scale in which all items were positively worded (n = 958) or negatively worded (n = 781) (Table 1, lower sections; see ref. 11 for data and code). We then explored the factor structure of each measure by comparing the average variance extracted (AVE; i.e., the average variance that a latent factor explains in indicators as represented in the squared factor loadings) with the shared variance (SV) of the two latent factors [i.e., the squared latent correlation (12)]. A two-factor structure is present only if both AVEs are larger than the SV (13).

We first conducted a confirmatory factor analysis (CFA) on the responses to the all-positive version of the Grit-S scale (Grit-S-P) and modeled the two correlated subfacets [χ²(19) = 185.65, root mean square error of approximation (RMSEA) = 0.10, confirmatory fit index (CFI) = 0.96, standardized root mean square residual (SRMR) = 0.04]. Both subfacets correlated at r = 0.95 (P < 0.01). Thus, the SV was 0.90, while the AVEs for consistency of interests and persistence of effort were 0.55 and 0.47, respectively. Since the SV was greater than both AVEs, our results highlight that both factors do not have sufficient discriminant validity (12, 13) (Table 2, upper section).

We find similar results when conducting a CFA on the responses to the all-negative version of the Grit-S scale (Grit-S-N), modeling the two correlated subfacets [χ²(19) = 291.0, RMSEA = 0.14, CFI = 0.93, SRMR = 0.06]. This time, both factors correlated at r = 0.90 (P < 0.01), resulting in a SV of 0.81. The AVEs for consistency of interests and persistence of effort were 0.50 and 0.71, respectively. Once again, the SV was greater than both AVEs, highlighting that both factors do not have sufficient discriminant validity (12, 13) (Table 2, lower section).

These new results further demonstrate that both subfacets should not be treated as unique factors, but instead should be considered as overlapping facets loading on a single construct. These data make it clear that the Grit-S scale measures only perseverance, not passion. This empirical evidence, together with the theoretical reasons discussed above, highlight that one reason for the mixed findings in prior literature between the Grit-S scale and performance is that the Grit-S scale only captures perseverance, but does not capture passion (1).

In sum, the evidence presented here and in our original paper (1) reveals three points that render the analyses advocated by Guo et al. (4) and Credé (5) moot. First, the Grit-S scale measures perseverance only, and none of its items conceptually capture passion. Second, an apparent two-factor structure of the Grit-S scale is the result of a statistical artifact based on reverse-coded items; thus, it is most appropriate to collapse the whole Grit-S into a single measure of perseverance, which is how we employed the scale in our original analysis. Third, our original conclusion—that it is the combination of both perseverance and passion that predicts performance—remains valid (1).