Measuring Grit

Grit is a construct originating in positive psychology that can be defined as trait-level perseverance of effort and consistency of interest (Duckworth, Peterson, Matthews, & Kelly, 2007). It entails “working strenuously toward challenges, maintaining effort and interest over years despite failure, adversity, and plateaus in progress” (Duckworth et al., 2007, pp. 1087–1088). The construct was introduced by Duckworth et al. (2007) in light of a history of research investigating the relevant indicators for personal success. Based on biographical information, abilities such as perseverance, zeal, and integration toward goals seemed to be important for achievement in addition to talent or intelligence (Cox, 1926; Galton, 1892; Terman & Oden, 1947). Duckworth et al. (2007) integrated these findings by introducing a concept that taps these traits and narrows them down to two dimensions: perseverance of effort and consistency of interest. To achieve a certain level of mastery, deliberate practice and initial failures have to be tolerated and, while this process can take months or even years, the initial interest needs to be kept salient.

To measure grit, Duckworth et al. (2007) developed and validated a self-report questionnaire based on the construct definition of grit as described above. The scale was intended to be used for adolescents and adults who pursue goals in a variety of domains, thus as a domain-general measure (e.g., not just work or school; Duckworth et al., 2007). On these grounds, the authors selected and piloted items for their measure in the US, resulting in the 12-Item Grit Scale. They identified a two-factor structure for the scale. The model fit for a two-factor solution suggested room for improvement, but the scale showed high internal consistency and predictive validity for a variety of success outcomes. In a subsequent validation study, Duckworth and Quinn (2009) revisited the issue of model fit. The model fit improved substantially when two items were removed from each subscale, leaving eight items in total. The resulting 8-Item scale was named the Short Grit Scale (Grit-S). A second-order factor was included in the model, representing the superordinate latent construct grit. Even though they found differential associations with predicted outcomes for the two subscales, their pattern of findings supports the conceptualization of grit as a compound trait.

To our knowledge, only few studies have been published on grit in a European context so far (Akın & Arslan, 2014; Arslan, Akın, & Çitemel, 2013; Dumfart & Neubauer, 2016) and a validation study for measures of grit in German does not exist. For a fruitful and critical appraisal in the German-speaking scientific community, a validation of the German scale seems necessary.

This paper aims to provide evidence for the factorial structure as well as the validity of the German short form of the Grit Scale (Grit-S; Duckworth & Quinn, 2009). First, this will be done by replicating the factor structure, testing measurement invariance, and investigating the relationships with external criteria. Second, we will present findings on an issue that has been discussed in previous research (Duckworth & Quinn, 2009), but has not been attended to empirically yet: the domain specificity of grit. While grit has been shown to be relevant for success in several domains, it has not been investigated whether grit itself can be considered a domain-specific construct. Before we present our research in more detail, we will provide a short summary of the state of research on grit.

During the last decade a substantial body of research has found grit to be a relevant predictor of academic and vocational outcomes. Grit seems to explain additional variance in the investigated outcomes over and above intelligence, domain-specific skills, and other personality traits (Duckworth et al., 2007; Duckworth & Seligman, 2006; Duckworth, Tsukayama, & Geier, 2010; Zimmerman & Schunk, 2011). Further research by Duckworth and Quinn (2009) showed that grit was associated with fewer career changes and higher educational attainment. Results of a longitudinal study indicated that adolescents’ grit was positively associated with grade point average (GPA) and negatively associated with hours watching television. In addition, grit predicted retention rates for cadets at the United States Military Academy as well as final round attainment for Scripps National Spelling Bee competitors (Duckworth & Quinn, 2009). Moreover, Duckworth, Quinn, and Seligman (2009) found that grittier novice teachers had a more positive effect on the learning outcomes of their students than less gritty teachers. Finally, Robertson-Kraft and Duckworth (2014) showed that grittier novice teachers not only outperformed their colleagues with regard to student achievement but were also less likely to leave a teaching program prematurely. Altogether, grit seems to play a role in achievement and retention, both of which are highly relevant for success in life. Since the construct grit was introduced, the close relations to the personality trait conscientiousness and to the construct self-control have been addressed on a theoretical and empirical level. Grit correlates moderately with indicators of achievement and is highly correlated with conscientiousness and self-control (Credé, Tynan, & Harms, 2016; Duckworth & Gross, 2014). In the same vein, the meta-analysis by Credé et al. (2016) showed that the predictive power of grit is moderate and that other noncognitive skills even outperform grit in predicting achievements such as GPA (also see Dumfart & Neubauer, 2016).

Next to the question of validity in German-speaking countries, there has been some discussion about the domain specificity of grit. To date, however, this issue has only seldom been investigated. The Grit-S “attempts to assess behaviors that are reasonably stable across time and situation” implying that “the tendency to pursue long-term goals with passion and perseverance is relatively domain general” (Duckworth & Quinn, 2009, p. 173). It may be reasonably assumed but is yet to be shown that grit can vary between different interests or goals in life, for example, school and hobbies. Bandura (1994) highlighted the fact that personality traits and motivational constructs are often too general to predict specific behaviors sufficiently. In line with this proposition Wigfield (1997) stresses that the domain specificity of constructs marks an important research desideratum. While we are convinced that the validation of the scale is important, the domain-specific approach to grit marks an additional focal point of our study. Duckworth and Quinn (2009) proposed to test this assumption by asking respondents to answer items separately for particular contexts. We dealt with this desideratum by specifically asking for high school students’ grit in the context of school as well as their perseverance of effort for the subjects Mathematics and German in addition to domain-general grit. In other words we tested for domain-general grit as in previous studies and introduced a hierarchical approach, a domain-specific (school-specific) grit, and the even more specific perseverance of effort on the school-subject level. Due to the fact that a shift in interests within a given school subject is prevented by the need to follow the curriculum (i.e., topics change too frequently) we only assessed the subscale perseverance of effort for the respective school subjects. Accordingly, with this paper we want to give a first impetus on the possibility to assess school-specific grit and subject-specific perseverance of effort.

Present Research

Fleckenstein, Schmidt, and Möller (2014) translated the original 12-Item Grit Scale into German. The items were translated from English into German and then cross-checked through back translation. This was done by a native English speaker with years of experience working as a qualified editor in educational research. No significant deviations were found between the original and the final translated scale. The resulting scale was termed the BISS-12 Scale (Beharrlichkeit and Beständiges Interesse [perseverance and consistency of interest]; the German word Biss [English: bite] with a symbolic meaning close to grit). For this study we used the eight relevant items from the BISS-12 Scale in accordance with the Grit-S by Duckworth and Quinn (2009).

For this paper we first, wanted to investigate the factorial structure and measurement invariance of the German scale as well as the psychometric properties. We assumed a second-order factor structure with the superordinate latent construct grit. We used the data from Study 3 to cross-validate these findings. Our aim was to test the applicability of the German scale in a broader age range. We used samples of adults (Study 1 and Study 2) and a sample of younger high school students (Study 3). The original scale was successfully used in different kinds of samples (e.g., Adults: Von Culin et al., 2014; Adolescents: Robertson-Kraft & Duckworth, 2014; Children: Duckworth et al., 2009).

Second, we used external criteria to investigate the convergent and discriminant validity of the scale. As the theoretical proximity of grit and (1) self-control has been discussed thoroughly (Duckworth & Gross, 2014), we attempted to show the empirical distinctness of the two constructs in latent model tests. The two constructs share similar features and thus can be expected to correlate strongly. However, we still expected the two constructs to be empirically distinct: self-control refers to the inhibition of task-irrelevant impulses in a certain short-term situation, whereas grit means (1) staying interested in superordinate goals over long periods of time and (2) persisting to achieve those higher-order goals even when setbacks occur and alternatives to reach the superordinate goals have to be actively created (for an in-depth discussion of the distinction between grit and self-control, see Duckworth & Gross, 2014; Tangney, Baumeister, & Boone, 2004). Furthermore, we focused on the relationship of grit with other constructs such as (2) the Big Five that have been shown to correlate differentially with grit (Fleckenstein et al., 2014). As stated above, due to the close relatedness of grit and conscientiousness significant positive correlations were expected in contrast to the four other personality traits namely neuroticism, extraversion, openness to experience, and agreeableness. Next, since the belief in one’s own abilities seems to be an important prerequisite for achieving long-term goals we assumed (3) self-concept to be highly correlated with grit. Another indicator of convergent validity would be a positive relationship between grit and (4) GPA (Duckworth et al., 2009). In light of divergent validity we investigated the relation with school procrastination in Study 3. Students who procrastinate a lot tend to finish their work last minute and sometimes are not able to finish the tasks they aimed to do due to a lack of time (Patzelt & Opitz, 2014). Thus, we expected negative correlations between grit and (5) procrastination.

Finally we wanted to give a first impetus on the research desideratum in respect to the domain specificity of grit as discussed earlier. In this regard, we investigated school-specific grit and subject-specific perseverance of effort for the first time.

Study 1

The objective of Study 1 was to examine the factor structure of the BISS-8 Scale (see Table A1 in Appendix for the German version), in accordance with the corresponding findings for the Grit-S, using confirmatory factor analysis (CFA) and to test for measurement invariance for gender. Moreover, we investigated the relationship between grit and self-control in a latent model test.

Results

Psychometric Properties and Measurement Model

Means, standard deviations, corrected item difficulties, and item selectivities are provided in Table 1. The BISS-8 Scale showed good to excellent corrected item difficulties and item selectivities. The internal consistency was good: Cronbach’s α = .80.

Table 1 Means, standard deviations, item selectivities, and item difficulties for the BISS-8 Scale in Study 1, Study 2, and Study 3

		Study 1a (N = 525 university students)				Study 2a (N = 173 university students)				Study 3b (N = 271 high school students)
Item		M	SD	pcorr	rit	M	SD	pcorr	rit	M	SD	rit	pcorr
Notes. pcorr = Corrected item difficulties; rit = part-whole corrected item discriminations. *Item recoded; aResponses were made on a 5-point Likert scale ranging from (1) “not at all like me” to (5) “very much like me”. bMeans transformed from 4- to 5-point scale to facilitate comparison with means for the samples of Study 1 and Study 2. Items were presented in the German language.
Consistency of interest
1.*	I often set a goal but later choose to pursue a different one.	3.71	0.92	.58	.60	3.66	0.84	.56	.38	3.36	1.00	.41	.51
2.*	New ideas and projects sometimes distract me from previous ones.	3.23	1.05	.46	.54	3.08	0.88	.41	.40	3.13	1.04	.39	.46
3.*	I have been obsessed with a certain idea or project for a short time but later lost interest.	3.61	1.13	.57	.46	3.17	1.23	.46	.28	3.45	1.12	.38	.54
4.*	I have difficulty maintaining my focus on projects that take more than a few months to complete.	3.86	1.02	.63	.57	3.61	0.95	.55	.52	3.47	1.11	.42	.55
Perseverance of effort
5.	I finish whatever I begin.	4.00	0.81	.66	.53	4.03	0.76	.67	.49	3.61	1.02	.62	.58
6.	Setbacks don’t discourage me.	3.81	0.89	.61	.34	3.64	0.91	.56	.23	3.66	1.04	.31	.60
7.	I am a hard worker.	3.94	0.91	.65	.54	3.75	0.89	.59	.55	3.00	1.14	.54	.44
8.	I am diligent.	4.00	0.93	.67	.54	3.73	1.01	.59	.53	3.15	1.13	.55	.47

Table 1 Means, standard deviations, item selectivities, and item difficulties for the BISS-8 Scale in Study 1, Study 2, and Study 3

View as image HTML

The CFA for the two-dimensional higher-order model proposed initially by Duckworth and Quinn (2009) did not fit the data sufficiently: χ²(19) = 92.10, CFI = .92, TLI = .89, RMSEA = .09, SRMR = .08. Due to high modification indices, we allowed the correlation between item residuals of two items (Item 7 and Item 8). Because of the similarities of the two items, these modifications seemed to be theoretically sound. The modified model appeared to fit the data well: χ²(18) = 25.35, CFI = .99, TLI = .99, RMSEA = .03, SRMR = .03. All manifest variables and first-order factors loaded significantly (p < .001) and substantially (λ ≥ .45) on the first-order factors and on the second-order factor (see Figure 1).

Measurement Invariance

To test measurement invariance for gender, different levels were compared. Based on the guidelines for model comparisons presented earlier, the assumption of partial strict invariance was supported (see Table 2).

Table 2 Measurement invariance across gender in Study 1 and Study 3

Model		Parameters constrained	χ2	df	CFI	TLI	RMSEA	SRMR
Notes. FL = factor loadings; II = item intercepts; IRV = item residual variances. For model identification in models 1 and 2 (item intercepts freely estimated) latent means were fixed to zero. CFI = Comparative Fit Index. TLI = Tucker-Lewis index. RMSEA = root mean square error of approximation. SRMR = standardized root mean square residual.
Measurement invariance in Study 1
1	None (configural invariance)		44.311	36	.987	.980	.038	.040
2	First-order FL (metric invariance)		54.925	44	.983	.978	.039	.062
3	First- and second-order FL		54.973	45	.984	.981	.037	.062
4	First- and second-order FL, II		78.319	52	.959	.956	.056	.078
4a	First- and second-order FL, partial II		61.306	49	.981	.978	.039	.062
5	First-and second-order FL, partial II, IRV		79.206	57	.965	.966	.049	.134
5a	First- and second-order FL, partial II, partial IRV		63.549	53	.984	.983	.065	.065
Measurement invariance in Study 3
1	None (configural invariance)		45.577	36	.977	.965	.045	.050
2	First-order FL (metric invariance)		53.793	44	.977	.971	.041	.066
3	First- and second-order FL		53.912	45	.979	.974	.039	.067
4	First- and second-order FL, II		62.349	52	.976	.974	.039	.075
5	First- and second-order FL, II, IRV		69.327	60	.978	.979	.034	.089
5a	First- and second-order FL, II, partial IRV		67.052	59	.981	.982	.032	.76

Table 2 Measurement invariance across gender in Study 1 and Study 3

View as image HTML

External Criteria

We found strong correlations between grit and self-control, which are presented in Table 3. The correlations between self-control and the subscales consistency of interest (r = .61) and perseverance of effort (r = .65) did not differ substantially. To investigate whether the two scales still measure two distinct theoretical constructs, we compared a model with the items loading on one common factor with a nested model with the two theoretically postulated factors self-control and the second-order factor grit with its subdomains consistency of interest and perseverance of effort. The data showed a significantly better fit for the theoretically derived factors model, ∆χ² = 288; df = 5; p < .001.

Table 3 Internal consistencies of external criteria and manifest correlations with the BISS-8 Scale in Study 1, Study 2, and Study 3

	M	SD	α	r	p
Notes.aGrade point average was recoded for better readability (higher values mean better achievements). bScales of the BFI-K.
Study 1
Self-control	3.46	0.63	.87	.73	< .001
Study 2
Grade point averagea	2.20	0.49	–	.21	.005
Neuroticismb	2.90	0.72	.68	−.04	.591
Extraversionb	3.60	0.84	.83	.09	.262
Openness to experienceb	3.69	0.75	.74	−.11	.143
Conscientiousnessb	3.64	0.67	.74	.70	< .001
Agreeablenessb	3.21	0.77	.63	.06	.471
Generalized academic self-concept	3.85	0.79	.85	.16	.034
Generalized self-efficacy	3.59	0.43	.81	.31	< .001
Study 3
Year of schooling	8.15	1.03	–	−.08	.187
Grade point averagea	2.61	0.65	.83	.24	< .001
Procrastination	2.22	0.68	.90	−.47	< .001

Table 3 Internal consistencies of external criteria and manifest correlations with the BISS-8 Scale in Study 1, Study 2, and Study 3

View as image HTML

Study 2

The objective of Study 2 was to investigate the construct and criterion validity of the BISS-8 Scale. We expected it to be strongly related to the personality trait conscientiousness (Roberts, Lejuez, Krueger, Richards, & Hill, 2014). Moreover, grit was expected to be positively correlated with self-efficacy as well as self-concept, since the belief in one’s own abilities seems to be an important prerequisite for achieving long-term goals. Grittier students can also be expected to have achieved higher grades in school, which is why we asked them for their GPA.

Results

Table 1 shows the psychometric properties for the items of the BISS-8 Scale in our sample. The overall internal consistency was α = .72. The subscales were correlated at r = .45. Selectivity for one item was r_it < .30; however, removing this item did not lead to a significant increase in Cronbach’s α. The mean interitem correlation was r = .42.

The correlations of the BISS-8 Scale with the external validation criteria are shown in Table 3, a full correlation matrix is presented in Table 4. All in all, the directions of the relationships were according to expectations.

Table 4 Manifest correlations for the scales used in Study 2 and Study 3

Study 2	N	E	O	C	A	ASC	SE
Notes.aGrade point average was recoded for better readability (higher values mean better achievements). bScales of the BFI-K. *p < .05; **p < .01.
1. Grade point averagea	.16*	.04	.08	.19*	.02	.16*	.02
2. Neuroticismb (N)		−.13	.10	−.01	−.04	−.32**	−.43**
3. Extraversionb (E)			.21**	.12	.11	.14	.39
4. Openness to experienceb (O)				−.03	.13	.18*	.15*
5. Conscientiousnessb (C)					.00	.02	.19*
6. Agreeablenessb (A)						.01	.03
7. Generalized academic self-concept (ASC)							.37**
8. Generalized self-efficacy (SE)
Study 3	Grade point averagea			Procrastination
Year of schooling	.01			.28**
Grade point averagea				−.16**

Table 4 Manifest correlations for the scales used in Study 2 and Study 3

View as image HTML

The BISS-8 Scale strongly correlated with the Big Five subscale of conscientiousness. The other Big Five dimensions showed no statistically significant correlations with the BISS-8 Scale. In light of the ongoing debate on the predictive power of noncognitive traits mentioned earlier, we regressed GPA on grit. Grit emerged as significant predictor of GPA: β = −.22, p < .01, B = −.19, SE = .07, R² = .05. When adding conscientiousness to the regression neither grit nor conscientiousness were significant predictors of GPA: β_grit = −.13, p = .23, B_grit = −.11, SE_grit = .09; β_con = −.11, p = .32, B_con = −.08, SE_con = .08; R² = .05. As expected, there was a positive association between grit and general self-efficacy. The correlation with academic self-concept was lower but still significant.

Study 3

The objective of Study 3 was to replicate the findings regarding the CFA conducted in Study 1 in a younger sample and to get further indicators for the scale’s validity. Furthermore, we tested for the domain specificity of grit. In addition, we assessed subject-specific perseverance of effort. We expected school-specific grit to correlate higher with GPA than the domain-general measure of grit. Furthermore, subject-specific perseverance of effort is expected to be more closely related to the grades in the respective school subjects than domain-general and school-specific grit, respectively.

Moreover, grit was expected to correlate negatively with the students’ tendency to procrastinate, as gritty individuals should be inclined to finish their work on time and to not let themselves be distracted from the task at hand by more pleasurable activities.

General Discussion

The measurement of grit has been examined thoroughly in the US while it is still in very early stages of development in Europe. To our knowledge, validation studies for measures of grit in different languages and contexts in Europe do not exist. Thus, in our studies, we aimed to close this research gap by adapting and closely examining the German version of the most established grit measure: the Grit-S (Duckworth & Quinn, 2009). To take a step further, we adjusted the scale to give a first impetus on the domain specificity of grit.

The aim of the present research was to provide first indicators for the factorial structure, validity, and reliability of the BISS-8 Scale, a German adaptation of the Grit-S. Accordingly, we started off by performing a CFA for the BISS-8 Scale in Study 1. The two-dimensional higher-order measurement model yielded excellent fit indices for the BISS-8 Scale for both university and high school students. It was also shown to be invariant for gender. We were able to replicate these findings in Study 3. All in all, these are strong indicators for the structural validity of the BISS-8 Scale.

When investigating the convergent and discriminant validity of the BISS-8 Scale, we found significant correlations with a number of external criteria: positive correlations with GPA, conscientiousness, self-efficacy, and self-control as well as a negative correlation with procrastination supported the scale’s validity. The fact that, although strongly correlated, grit still seems to differ from the similar construct of self-control also empirically reinforces its relevance as a disjunct psychological construct.

Presenting findings on the domain specificity of grit for the very first time we were able to show that school-specific grit is more closely related to GPA than domain-general grit. Thereby, we were able to give a promising first impetus regarding the assumption of Duckworth and Quinn (2009) and the desideratum pointed out by Bandura (1994) and Wigfield (1997) in regard to the domain specificity of constructs as an important research desideratum. Additionally these findings can help to shed light on the lacking incremental validity of domain-general grit over and above conscientiousness – which we were able to corroborate with our findings from Study 2 – as mentioned by Ivcevic and Brackett (2014) and the meta-analysis by Credé et al. (2016). Both assumed grit to be a better predictor of more narrow goals. All in all, our results show that it is worthwhile to assess domain-specific grit in a school context. Further research concerning other domains could underpin our findings.