Examining Mindset and Grit in Concurrent and Future Reading Comprehension: A Twin Study

Abstract

Non-cognitive factors have gained attention in recent years as potential intervention targets for academic achievement improvement in students. Two notable facets, intelligence mindset and grit, have been of particular interest. Both have been shown to consistently improve educational outcomes, although little work has focused on reading ability. As such, we examined the relation between both grit and mindset on current, future, and change in reading comprehension ability in a twin sample. We used data from 422 twin pairs (171 monozygotic pairs, 251 dizygotic pairs) drawn from the Florida Twin Project on Reading, Behavior and Environment (Taylor et al., 2019). The racial composition of the sample included 1.00% American Indian or Alaska Native, 2.25% Asian, 13.25% Black or African American, 22.63% Hispanic, 1.00% Native Hawaiian or Other Pacific Islander, 56.13% White, and 3.75% more than one race. The household income of the sample at time 1 was 16.15% below $25,000, 18.06% $25,000–49,999, 36.34% $50,000–99,999, and 29.45% $100,000 or more and closely align with the overall composition reported for the state of Florida (United States Census Bureau, 2021). Twins were on average 13 years old when the questionnaire and first reading ability measure were collected, and on average 15 years old when the second reading ability measure was collected. Weak and moderate positive correlations were found between both mindset and grit and with each reading ability score and neither were significantly related to change in reading ability. Twin modeling suggested little to no common genetic or environmental influences between mindset and grit to reading ability. In total, our results do not lend support to the notion of mindset or grit being a mechanism of change for reading ability.

Academic achievement, particularly reading ability, has been a topic of great interest over the past few decades—largely attributable to the lasting negative effects that early reading difficulties present (Francis et al., 1996; Scarborough, 2001; Shaywitz et al., 1999; Torgesen & Burgess, 1998). Early reading difficulties have been linked to lower rates of high school completion and higher rates of conduct issues at age sixteen (Fergusson & Lynskey, 1997). In addition, early reading difficulties predict reading ability in both fourth and eleventh grade, such that students who display reading difficulties in first grade tend to remain poor readers as they age through school (Juel, 1988; Scarborough, 2001). Previous research has also demonstrated the importance of early reading ability on later reading ability and academic achievement, including standardized test performance (Daneman & Carpenter, 1980). Taken together, this body of research emphasizes the importance of early reading mastery for later academic achievement, as well as the potential consequences, including lower high school graduation rates, behavior problems, socioeconomic difficulty, and adverse health outcomes (DeWalt et al., 2004; Fergusson & Lynskey, 1997; Hernandez, 2011; Juel, 1988; Miles & Stipek, 2006). As such, reading is an important domain to focus on for improving academic achievement and in turn, other important life outcomes as well.

Mindset and Grit: Important Non-Cognitive Factors

Attempts to understand why some students read better than others have led researchers to focus on a wide range of factors, including non-cognitive factors. Non-cognitive factors are those defined as “patterns of thought, feelings, and behaviors” (Borghans et al., 2008) that continue to develop throughout one’s lifetime. Unlike traditional cognitive abilities (e.g., verbal fluency, executive function), non-cognitive factors are comprised of a mixture of attitudinal, personality, and motivational factors. Non-cognitive factors, such as motivation, self-efficacy, and certain personality traits have been linked to educational success (Blackwell et al., 2007; Laidra et al., 2007; Chamorro-Premuzic & Furnham, 2003; Duckworth et al., 2007; Ivcevic & Brackett, 2014; Noftle & Robins, 2007; Poropat, 2009). For example, personality traits such as conscientiousness have been linked to academic success (Noftle & Robins, 2007), predicting higher grades in both high school (Bratko et al., 2006; Ivcevic & Brackett, 2014; Preckel et al., 2006) and university samples (Chamorro-Premuzic & Furnham, 2003; Noftle & Robins, 2007; Poropat, 2009). Motivation has also been linked to increased reading ability (Lee & Johnson-Reid, 2016; Wigfield & Guthrie, 1997). A systematic review of the association between academic self-efficacy and broad academic performance also demonstrates a moderate positive relationship between the two (Honicke & Broadbend, 2016). Indeed, there are a wide array of non-cognitive factors discussed across the literature, but here we will focus on two of the most prominent non-cognitive factors in the educational literature: intelligence mindset and grit.

Intelligence mindset, or simply “mindset”, has received a considerable amount of attention in the literature as a non-cognitive factor associated with academic achievement. Carol Dweck (1999, 2007) identified two distinct ways that students tend to view their own intelligence, in what she called a “fixed mindset” and a “growth mindset.” Individuals who hold a fixed mindset tend to believe that their intelligence is an inborn trait, and there is not much that can change a person’s inherited intelligence level. On the contrary, individuals who hold a growth mindset hold the view that their intelligence is malleable and can be improved upon with effort and time. A growth mindset has been shown to predict better academic outcomes, including higher reading scores, compared to a fixed mindset (Andersen & Nielsen, 2016; Blackwell et al., 2007; McCutchen et al., 2016; Yeager et al., 2019), although some studies have failed to find any such relationship (Aditomo, 2015; Wilson, 2016; Foliano et al., 2019). Despite inconsistencies, there is evidence that a growth mindset may serve as a buffer for minority students and help combat the typically-negative effects that low socioeconomic status (SES) has on students (Claro et al., 2016; Sisk et al., 2018). If a child believes that their hard work and effort will amount to positive changes and results, then they may be willing to put forth more effort in improving their performance in school, including their reading ability. Additionally, students with growth mindsets may view challenges as opportunities to grow and learn, while students with fixed mindsets may view challenges as threats to their perceived intelligence (Dweck, 2010). In other words, students may feel more inclined to work through challenges if they believe that their hard work will pay off. Conversely, students may put forth less effort if they feel constrained by their current level of intelligence.

The non-cognitive factor of grit has also received considerable attention by educational researchers and schools, and has been linked to both academic and personal success (Duckworth et al., 2007). Duckworth and colleagues (2007) define grit as trait-level perseverance and passion for long-term goals. Grit has two main components: consistency of interest and perseverance of effort (Duckworth & Quinn, 2009). The former refers to keeping a constant goal, or keeping “eyes on the prize,” while the latter refers to consistently working towards that goal, despite any challenges along the way. Grit has been linked to success in an intense military training course, a national spelling bee, occupational retention, and even marital longevity (Eskreis-Winkler et al., 2014). Similarly, Strayhorn (2014) found that for a group of Black males attending a predominantly White college, individuals who scored higher in grit earned higher grades, even after controlling for year in school, degree aspirations, and prior achievement. Although there seems to be a link between “grittier” individuals and success, it is unclear whether grit displays those same predictive qualities in the context of reading ability. Although studies on grit and reading ability are lacking, one study found a positive relationship between teacher’s grit scores and student’s standardized reading examination scores (McCain, 2017). Several other studies have failed to find a relationship between grit and broader academic achievement (MacCann & Roberts, 2010; Ivcevic & Brackett, 2014; Browne, 2017; McCain, 2017; Muenks et al., 2017), including one intervention involving teachers (Wilson, 2016). Despite mixed findings, grit is generally thought to be related to goal completion, particularly when those goals require sustained stamina and effort, and may help predict which students will persevere in school despite challenges and setbacks.

An important similarity between mindset and grit is the focus on the student’s perspective of their own competence. Mindset relates to how people perceive their own intelligence, while grit relates to how people view their ability to work through challenges. In support of this similarity, mindset and grit have been shown to be moderately correlated with one another (Kench et al., 2016; West et al., 2016). Due to the purported malleable quality of such characteristics (e.g. a student’s beliefs about themselves), the question arises as to whether changing a student’s mindset or grit may result in educational improvements, specifically reading ability.

Interventions with Mindset and Grit

Interventions have been implemented in schools that focus on growth mindsets, although most have measured either broad academic achievement (GPA) or have incorporated reading ability into a broader measure (Dweck, 2010; Perkins-Gough, 2013; Paunesku et al., 2015; Rattan et al., 2015; Rimfeld et al., 2016). For example, in a study of seventh-grade students, standardized test scores in math and reading were significantly higher for individuals who were assigned to a mentor who conveyed a growth mindset along with the ability to overcome challenges with effort and persistence (Good et al., 2003). Another intervention by Paunesku et al. (2015) highlighted the effectiveness of a 45-minute growth mindset intervention for improving course completion – including an English course - among a large sample of high school students. Another study by Andersen and Nielsen (2016) demonstrated the effectiveness of a parental mindset intervention for improving reading ability in second-grade students. Interestingly, a recent randomized-controlled trial by Foliano and colleagues (2019) found no significant effects of a growth mindset intervention on literacy scores in a large sample of year 6 students. However, another recent randomized control trial found positive effects on GPA resulting from a short mindset intervention (Yeager et al., 2019).

Less research has been published on grit-focused interventions (Credé et al., 2017; Muenks et al., 2017), although some report that grit interventions are in the process of being developed (Perkins-Gough, 2013; Bashant, 2014). One of the few known interventions was carried out by Alan, Boneva, and Ertac (2016) in a sample of 1700 children in Turkish elementary schools. Their randomized intervention, aimed at enhancing grit in students, resulted in improved performance on both math and Turkish standardized exams following the intervention. Although it is theorized that interventions designed to foster grit may be of particular interest to schools, there are not any published grit-focused interventions that we are aware of that specifically evaluate the effects of improving grit on reading achievement scores. Nevertheless, grit remains a construct of interest regarding successful educational and vocational outcomes (Duckworth et al., 2007).

Although there is conflicting, and lacking, efficacy evidence, using non-cognitive factors like mindset and grit in interventions is thought to improve cognitive skills like reading ability through helping students cope with and work through academic challenges. Although the current study is not an intervention study, we were interested in using a correlational approach to contribute to the literature. We reasoned that if mindset and grit are to be thought of as potential intervention targets to increase reading ability, then one might expect that at the very least both should be correlated with concurrent reading ability (e.g., as a first step in understanding causality; Gardner, 2000). Although certainly an important first step, mindset and grit are often considered to be malleable factors (see Duckworth et al., 2007; Dweck, 2008, 2010), or factors which can be intervened on to result in changes on a different factor, including a variety of academic outcomes (e.g., Alan et al., 2016; Good et al., 2003; Paunesku et al., 2016), and reading ability (Andersen & Nielsen, 2016). If this was the case, then mindset and grit should be associated with not only concurrent student reading ability, but also future reading ability and changes in reading ability across time. As such, the major aim of this study was to examine the relation of mindset and grit with concurrent and future reading ability, as well as change in reading ability over time.

Mindset, Grit, and Twin Models

In support of this major aim, twin modeling allows for an expansion of our understanding of the relation of mindset and grit with reading ability. Twin modeling allows us to quantify the role of genetic, shared environmental (i.e., non-genetic aspects shared by siblings, such as the home and school environments) and nonshared environmental (i.e., non-genetic aspects not shared by siblings, such as friends) influences underlying individual differences on a characteristic, and among characteristics. If mindset and grit are to be thought of as malleable factors, it is often presumed that these characteristics should therefore be environmentally influenced (e.g., Blackwell et al., 2007), and incorrect assumed that they cannot be genetically influenced. However, the results of twin modeling only describe “what is” in a given normally varying sample, not “what could be” due to intervention. Therefore, no matter the findings from twin modeling, no conclusions can be drawn about the malleability of a given trait.

What can be known from twin modeling is the underlying nature of how traits are correlated with each other, which can give some clues as to how an intervention might be effective. If the correlation between mindset and grit to reading ability is characterized by shared genetic influences, it would point to the possibility of a transactional model of gene-environment correlation (Tucker-Drob & Harden, 2017). The transactional model predicts the opposite of a common, but incorrect, intuition that genetic influences mean immutability. Instead, the transactional model predicts that individuals differentially select and evoke learning experiences due to their genetically-influenced characteristics related to learning. These learning experiences then shape their academic achievement, such as reading ability, as well as reinforce the characteristics that led them towards selecting and evoking those very learning experiences. Given this, if mindset and grit are genetically driven characteristics, which are also shown to share genetic influences with reading ability, they should then have the greatest potential to effect reading ability through systematic and sustained environmental exposure (Tucker-Drob & Harden, 2017). In other words, the transactional model would predict that if mindset and grit shared genetic influences with reading ability, then we should be considering long-term and integrated mindset and grit intervention strategies.

On the other hand, if mindset and grit are shown to have common shared, or possible nonshared, environmental influences with reading ability, this would support the common theories in educational research for the role of non-cognitive aspects such as mindset and grit, in academic achievement (e.g., Blackwell et al., 2007; Wigfield & Eccles, 2000). That is, environmental experiences, typically experienced in the home or in school (i.e., the shared environment; see Rimfield et al., 2016) but also could be through individual experiences (i.e., the nonshared environment), are the driving mechanism for the potential effects of mindset and grit on reading ability. Finding common environmental influences between mindset and grit with reading ability would support the idea that shorter-term intervention strategies, deployed in homes or the school if through the shared environment, or deployed individually, if through the nonshared environment, might work.

To date, there have only been two studies examining the role of genetic and environmental influences on mindset and grit (Rimfeld et al., 2016; Tucker-Drob et al., 2016). These studies investigated the extent to which genetic and environmental influences on mindset and grit are shared with concurrent academic achievement, using a combined measure of reading and math achievement scores (and science, for Rimfeld et al., 2016). Overall, these studies have pointed to a strong genetic component underpinning both mindset and grit, with the nonshared environment playing a supportive role (Rimfeld et al., 2016; Tucker-Drob et al., 2016). Both studies noted the lack of shared environmental influence on both grit and mindset, which is contradictory to previous educational research focusing on factors shared by twins (e.g., school, family) influencing non-cognitive factors (Tucker-Drob et al., 2016). Nonshared environmental influences may include aspects such as motivation (Tucker-Drob et al., 2016), which helps to explain its role in both mindset and grit.

The work done to date using behavioral genetic approaches to examine the relation of mindset and grit to achievement has only looked at academic achievement at one timepoint, and none have specifically focused on reading ability. As such, in this study we take our correlational findings a step further by decomposing the association of mindset and grit with concurrent, future, and change in reading ability, into genetic and environmental influences. Elucidating the nature of such relations will help us to understand the constructs of mindset and grit more thoroughly.

Purposes of the Present Study

For the present study, we explored five research questions. First, what is the relation between mindset and grit? We predicted mindset and grit would be positively correlated. Second and third, what is the relation between mindset, grit and student reading ability—both at an initial time point (i.e., concurrent reading ability), and two years later (i.e., future reading ability)? We expected mindset and grit would be positively correlated with concurrent and future reading ability. Fourth, do either mindset or grit predict change in student reading ability over time? We expected that mindset and grit would predict variance in future reading ability above and beyond concurrent reading ability. And finally, how do genetic and environmental influences contribute to and explain these factors and their relations? We expected to see genetic and nonshared environmental influences, but no shared environmental influences, underlying the relation between mindset, grit, and reading ability.

Methods

Participants

Participants were drawn from the Florida Twin Project on Reading, Behavior, and Environment (Taylor et al., 2019). The project is now exempt but was formerly maintained under IRB approval at Florida State University. The Florida Twin Project on Reading, Behavior, and Environment examined achievement, home environment, and behavioral factors on a large twin sample at 3 time points. Data for the current study are from the second and third time points only, which we will refer to as “time 1” and “time 2” for ease of understanding. For the current study, data were available from 171 monozygotic (MZ; 53.80% female pairs), and 251 dizygotic (DZ; 39.84% same-sex female pairs, 28.69% same-sex male pairs, 31.47% opposite-sex pairs) twin pairs. The twins were approximately 13 years old (M = 13.25yrs, SD = 2.45 yrs, range = 7.43–18.13yrs) at time 1, and 15 years old at time 2 (M = 15.22yrs, SD = 2.51yrs, range = 9.64–20.40yrs). The racial composition of the sample included 1.00% American Indian or Alaska Native, 2.25% Asian, 13.25% Black or African American, 22.63% Hispanic, 1.00% Native Hawaiian or Other Pacific Islander, 56.13% White, and 3.75% more than one race. The household income of the sample at time 1 was 16.15% below $25,000, 18.06% $25,000–49,999, 36.34% $50,000–99,999, and 29.45% $100,000 or more. The demographic and household income characteristics of the study sample closely align with the overall composition reported for the state of Florida (United States Census Bureau, 2021). All available data were used, but complete data were not available for the entire sample due to the voluntary nature of the questionnaires and the method of data collection employed (see Table 1 for the sample size available for each measure)¹.

Table 1.

Descriptive Statistics of mindset, grit, and both time points of reading comprehension

Measure	n	Mean	SD	Min	Max	Skew	Kurtosis
Time 1 Mindset	768	4.29	1.02	1.00	6.00	−0.24	−0.40
Time 1 Grit	768	3.48	0.64	1.50	5.00	−0.07	−0.35
Time 1 Reading Comprehension	847	534.05	50.04	316	660	−0.55	0.75
Time 2 Reading Comprehension	597	557.74	43.74	365	694	−0.16	1.23

Note. The sample size reflects individuals.

Procedure and Measures

Families in the Florida Twin Project on Reading, Behavior, and Environment project were mailed questionnaire packets to complete at home 3 times between 2012 and 2017. Data for the current study came from only the last two waves of data collection. Data on mindset and grit were collected only during the second wave (called “time 1” in the current study), and reading ability data were from that same point and the last data collection point (called “time 2” in this study). Twin zygosity was determined when families first enrolled in the project through a five-item questionnaire assessing physical similarities between the twins (Lykken, Bouchard, McGue & Tellegen, 1990). Both mindset and grit were measured through self-report questionnaires completed by the twins individually, and reading ability was assessed in the home by a parent administering a standardized protocol.

Mindset.

Twins were asked to rate the extent to which they either agreed or disagreed with 8 items from Dweck’s (1999) Theory of Intelligence scale. This scale contains 4 questions pertaining to an entity belief (“fixed mindset”), and 4 questions pertaining to an incremental belief (“growth mindset”). Incremental belief questions have been reversed score, and a mean score was calculated based on all 8 items, with a higher score indicative of a “growth mindset”. Reliability in this sample was adequate, Cronbach’s alpha = .87.

Grit.

Grit, defined as trait-level passion and perseverance for long-term goals (Duckworth et al., 2007), was assessed using the 12-item Grit Scale, in which subjects self-reported the extent to which they either agreed or disagreed with statements related to the two construct dimensions: Consistency of Interests (e.g. “I become interested in new pursuits every few months) and Perseverance of Effort (e.g. “I finish whatever I begin”). In line with more recent literature, the mean of only 8 items were used in the current analysis, comprising the Short Grit Scale (Grit-S) (Duckworth et al., 2009) with higher scores indicating more grit. The Grit-S scale has consistently shown superior reliability and predictive validity as compared to the original Grit Scale. Reliability in this sample was adequate, Cronbach’s alpha = .72.

Reading Comprehension.

The Gates-MacGinite Reading Test-Fourth Edition (GMRT; MacGinitie & MacGinitie, 2006) Reading Comprehension subtest was used to measure reading ability. This subtest is grade level specific, which each level consisting of passages that must be read and the comprehension questions answered. This section was administered to each subject by an adult caregiver, and each subject was allotted 35 minutes to complete this task. Each student’s raw score was totaled, with higher scores indicating more correct answers. This total was then converted to the extended scale score using the publisher conversion tables. The extended scale score allows for the comparison of score across different levels of the test. The published reliability is high for this test, with KuderRichardson Formula 20 reliability coefficients between .90-.96.

Data Analysis Plan

First, we conducted descriptive statistics on raw data. Subsequently, for the first research question we conducted Pearson correlations on data that had been residualized for wave specific age, wave specific age-squared, and sex, reflecting the data used in the subsequent steps. Then, we conducted three hierarchical multiple regression analyses, controlling for family-level nesting, all of which included time 1 age, time 1 age-squared and sex as demographic control variables. The first hierarchical multiple regression included mindset and grit as predictors of time 1 reading comprehension, to determine the amount of variance each construct contributed to concurrent reading ability, related to the second research question. The second hierarchical multiple regression included mindset and grit as predictors of time 2 reading comprehension, to determine the amount of variance of each construct contributed to reading ability two years later, related to the third research question. The third hierarchical multiple regression included mindset, grit and time 1 reading comprehension as predictors of time 2 reading comprehension, related to the fourth research question. This allowed us to account for reading comprehension at baseline while determining whether mindset or grit predicted variance in reading comprehension at time 2 over and above reading comprehension at time 1, in other words, change in reading ability.

As a preparatory step for the twin modeling related to the fifth research question, all data were then residualized for wave specific age, wave specific age-squared, and sex (McGue & Bouchard, 1984), and then z-scored so that all variables could be compared on the same scale. We then calculated twin intra-class correlations (ICCs) and cross-twin cross-trait correlations (CTCTs), by zygosity, for each measure. By comparing the magnitudes of the ICCs between monozygotic (MZ) twin pairs, and dizygotic (DZ) twin pairs, we get a preliminary indication of the genetic, shared and/or nonshared environmental influences on each measure. Genetic influences represent the additive genetic influences inherited from parents (also called “heritability”), shared environmental influences represent any environmental influences that make twins more similar (e.g., home income level; attending the same school) and nonshared environmental influences represent any environmental influences that make twins less similar (e.g., nonshared friends and teachers) as well as measurement error. Additive genetic influences are indicated on a trait when MZ twin ICCs are higher than those of the DZ twins. On the other hand, the extent to which MZ twin intra-class correlations are less than twice the magnitude of DZ twin intra-class correlations indicates shared environmental influences on that trait. Lastly, nonshared environmental influences are implicated when MZ twin pairs are not perfectly correlated with one another. The CTCTs can be interpreted in a similar fashion. All analyses to this step were conducted in SAS 9.4.

Structural equation modeling was conducted to assess the univariate genetic and environmental influences on all variables, as well as to examine the genetic and environmental influences on the covariation among mindset, grit, and the reading comprehension measures. Specifically, four trivariate and two quadvariate Cholesky decomposition models were run to partition the covariation among the variables into a series of biometric latent factors representing additive genetic (A), shared environmental (C), and nonshared environmental (E) influences among the measures. This study was not preregistered, however, data and study materials are available, and inquiries should be directed to Dr. Sara Hart. More details on the trivariate and quadvariate Cholesky decomposition models can be found in the supplemental materials.

Results

Descriptive Statistics

Table 1 displays the descriptive statistics, which indicated that the participants rated their average mindset and grit slightly higher than the midpoint of the scale. Correlations among mindset, grit, and reading ability scores are displayed in Table 2. The two waves of reading ability were moderately correlated with each other (r = .57), as can be expected, but surprisingly, mindset and grit were weakly correlated with each other (r = .19). This would suggest that these two non-cognitive factors, although thought to be similar constructs, are not strongly related in this sample.

Table 2.

Pearson Correlations between mindset, grit, and both waves of reading comprehension

	Time 1 Mindset	Time 1 Grit	Time 1 Reading Comprehension	Time 2 Reading Comprehension	Time 1 Age	Gender
Time 1 Mindset	1.00 n = 768	0.18 p = .000 n = 757	0.20 p = .000 n = 751	0.20 p = .000 n = 494	−0.02 p = .613 n = 768	0.02 p = .636 n = 768
Time 1 Grit	0.19 p < .0001 n = 757	1.00 n = 768	0.02 p = .595 n = 751	0.036 p = .419 n = 494	−0.08 p = .019 n = 768	−0.08 p = .021 n = 768
Time 1 Reading Comprehension	0.24 p < .0001 n = 751	0.08 p = .036 n = 751	1.00 n = 847	0.72 p = .000 n = 559	0.62 p = .000 n = 847	−0.10 p = .004 n = 847
Time 2 Reading Comprehension	0.21 p < .0001 n = 494	0.11 p = .049 n = 494	0.60 p < .0001 n = 559	1.00 n = 565	0.47 p = .000 n = 565	−0.061 p = .148 n = 565

Note. Correlations above the diagonal are on raw data. Correlations below the diagonal calculated using data regressed on age, age-squared, and gender. The sample size reflects individuals.

Mapping onto the correlation results, the hierarchical multiple regression analyses indicated that mindset was a statistically significant predictor of both time 1 and time 2 reading comprehension, whereas grit was not statistically significantly related to reading in these models (Table 3)². However, when time 1 reading comprehension was added into the model as the autoregressor (i.e., controlling for initial reading comprehension performance), mindset was no longer a statistically significant predictor of (change in) reading comprehension^3,4.

Table 3.

Hierarchical multiple regressions.

	Model 1: Time 1 Reading Comprehension				Model 2: Time 2 Reading Comprehension				Model 3: Time 2 Reading Comprehensiona
	b	se b	t-value	p	b	se b	t-value	p	b	se b	t-value	p
Intercept	85.72	66.88	1.28	.201	128.61	85.86	1.50	.135	81.02	65.93	1.23	.220
Demographics
Time 1 Age	54.47	10.02	5.43	.000	56.70	13.49	4.20	.000	25.21	11.82	2.13	.014
Time 1 Age-squared	−1.65	.37	−4.43	.000	−1.86	.51	−3.63	.000	−.91	.44	−2.09	.007
Sex	−9.07	2.87	−3.17	.001	−7.42	3.89	−1.91	.057	−2.07	3.53	−.59	.585
Non-cognitive factors
Time 1 Mindset	7.55	1.27	5.96	.000	5.06	1.66	3.05	.002	1.84	1.48	1.25	.301
Time 1 Grit	1.75	2.30	.76	.445	1.28	2.12	.60	.548	1.56	1.75	.89	.247
Time 1 Reading Comprehension									.56	.07	8.15	.000

Note.

^*p < .05

^**p < .01. Sex coded as 1 = female and 2 = male. Model 1 n = 734. Model 2 n =479. Model 3 n = 473.

^aThis model represents the change in reading comprehension performance.

Turning to the twin modeling results, the intraclass correlations (Table 4) indicated that genetic, shared environmental, and nonshared environmental influences should be seen on, and among, mindset, grit, and reading comprehension. An exception is the CTCT relation between grit and both waves of reading comprehension. For both, the MZ twin CTCT was lower than the DZ twin correlation, suggesting that there were no genetic influences in common between grit and reading comprehension. The univariate results from structural equation modeling support this, with some differences between the non-cognitive factors and reading comprehension (Table 5). All measures had statistically significant genetic influences. The shared environmental influences were statistically significant for both waves of reading comprehension; however, these were statistically non-significant in both mindset and grit. Further, mindset and grit had high and statistically significant nonshared environmental influences (which includes measurement error).

Table 4.

Intraclass and Cross-Twin Cross-Trait Correlations

Measure	Zygosity	Time 1 Mindset Twin 2	Time 1 Grit Twin 2	Time 1 Reading Comprehension Twin 2	Time 2 Reading Comprehension Twin 2
Time 1 Mindset Twin 1	MZ	0.49 p < .0001 n = 304	-	-	-
	DZ	0.32 p < .0001 n = 448	-	-	-
Time 1 Grit Twin 1	MZ	0.15 p = .009 n = 302	0.40 p < .0001 n = 300	-	-
	DZ	0.13 p = .006 n = 451	0.25 p < .0001 n = 454	-	-
Time 1 Reading Comprehension Twin 1	MZ	0.19 p = .0008 n = 306	−0.02 p = .798 n = 303	0.67 p < .0001 n = 342	-
	DZ	0.09 p = .047 n = 443	0.10 p = .041 n = 446	0.51 p < .0001 n = 502	-
Time 2 Reading Comprehension Twin 1	MZ	0.23 p = .0010 n = 200	0.003 p = .965 n = 198	0.57 p < .0001 n = 223	0.80 p < .0001 n = 224
	DZ	0.14 p = .019 n = 294	0.12 p = .037 n = 296	0.40 p < .0001 n = 336	0.57 p < .0001 n = 338

Note. MZ = monozygotic twin pairs, DZ = dizygotic twin pairs. Correlations were calculated using data regressed on age, age-squared, and gender. Sample size reflects individuals.

Table 5.

Univariate Genetic (A), Shared Environmental (C) and Nonshared Environmental Estimates (E), with 95% Confidence Interval (in parentheses)

Measure	A	C	E
Time 1 Mindset	.35 (.12; .57)	.15 (−.04; .32)	.50 (.44; .59)
Time 1 Grit	.42 (.17; .67)	.03 (−.16; .21)	.55 (.48; .66)
Time 1 Reading Comprehension	.29 (.13; .44)	.38 (.23; .51)	.34 (.29; .39)
Time 2 Reading Comprehension	.57 (.43; .72)	.27 (.12; .42)	.17 (.14; .21)

Note. Statistical significance is denoted when the 95% confidence interval does not bound zero.

Although not elaborated upon in this section, results from the multivariate twin models are included in the supplemental materials.

Discussion

The aim of this current study was to determine if either mindset or grit was related to concurrent, future, and change in reading ability scores over time. We also set out to determine whether these associations could be explained by shared genes and/or environmental factors. Overall, we found minimal evidence that mindset was associated to concurrent and future reading ability, although what small relationship did exist owed to nonshared environmental factors. While prior research had shown grit to relate to academic achievement (e.g., Duckworth & Quinn, 2009), we found no evidence that grit was associated with reading ability.

The first research question was related to the association between mindset and grit. As both mindset and grit are non-cognitive factors that focus on a student’s perspective of their own competence, we hypothesized they would be positively correlated. This hypothesis was supported, although the overall size of the correlation was small. Previous work has shown that there is a moderate correlation between these two constructs, ranging from .18-.41 (Duckworth & Eskreis-Winkler, 2013; Kench et al., 2016; West et al., 2016). The strength of the correlation that we found maps onto those found in work with students of a similar age (see Kench et al., 2016).

The second and third research questions addressed whether mindset and grit would be positively associated with concurrent and future reading ability. We predicted that they would be, in spite of the somewhat mixed findings with their relation to reading (Wilson, 2016; McCutchen et al., 2016; Andersen & Nielsen, 2016; McCain, 2017) but the generally consistent findings that both are associated with general academic achievement (Good et al., 2003; Blackwell et al., 2007; Aronson et al., 2002; Richardson et al., 2012; Paunesku et al., 2015). Both hypotheses were partially supported. Mindset was moderately correlated with both waves of reading ability (although slightly higher between the concurrent measures, as would be expected), and these correlation coefficients were similar in range to that found in other work (Andersen & Nielsen, 2016; Good et al., 2003). On the other hand, grit was only weakly correlated with both waves of reading ability, although both correlation coefficients were statistically significant. It is important to note that much of the previous work linking grit to positive academic outcomes has used cross-sectional data with general educational outcomes, such as educational attainment in adults, university grade-point averages, standardized exam scores, and self-reported course grades (Credé et al., 2017; Duckworth et al., 2007; Strayhorn, 2014; Eskreis-Winkler et al., 2014), whereas our study was comprised of longitudinal, reading-specific data. Although grit has been linked to broad measures of academic achievement, it may not be particularly useful in predicting specific constructs like reading ability. Reading comprehension is a complex skill that involves the orchestrated functioning of many different cognitive constructs (e.g., decoding, linguistic comprehension), to which grit may be unrelated. In other words, the factors that foster grit may be separate from the more specific cognitive factors that underlie reading comprehension.

We note that our lack of support for grit, in particular, could also be due to the age of our sample when grit was measured. Duckworth’s (2007) linking of higher grit with college GPA may not translate in the same way to middle school-aged students focusing on reading, specifically. Grit, defined by Duckworth as passion and perseverance for long-term goals, may simply not be developed until one matures. Post-secondary education often involves long-term goal setting and itself can be viewed as a step towards longer-term goals. As such, one might expect that students may continue to develop grit as they age through school. Indeed, it has been noted by Duckworth and Eskreis-Winkler (2013) that grit increases with age, becoming more refined as individuals understand their lifelong goals. Younger students may also not define long-term goals in the same way as one might later in high school or college. For example, because of the wording on the measure (“I have overcome setbacks to conquer an important challenge”) it is possible that the students in our sample simply did not view reading as an important challenge worthy of “conquering.” In turn, they may not have viewed reading itself as a “goal.” It may be beneficial for future research to incorporate questions regarding each child’s goals in order to understand the types of activities and skills that students are motivated to put effort towards. It also may be helpful to understand how important students feel reading is to their academic performance. If reading is not thought of as a “goal” – something worthy of eliciting passion and perseverance towards - grit may not be an appropriate measure for school interventions related to improving reading ability. Likewise, if students hold fixed views of their intelligence, they may also not view reading as a skill worthy of persevering through, since they may feel constrained by their current level of intelligence. Reading is a skill that is intertwined in many aspects of everyday life, and the methods used to teach reading skills differ as students age through school. As school level increases, reading is less often explicitly taught; rather, it is integrated into other school lessons and tasks. Students may therefore not view reading as a separate entity from other aspects of their education, making it a difficult skill to improve with measures such as mindset or grit.

For the fourth research question, we hypothesized that mindset and grit would predict variance in future reading ability above and beyond concurrent reading ability (i.e., change in reading ability). We did not find support for this hypothesis: the only predictor of future reading ability was the student’s previous reading ability.

Both mindset and grit have gained considerable attention as possibly malleable constructs (Duckworth et al., 2007; Dweck, 2008, 2010). Accordingly, both constructs have been picked up by the educational community as “easy” to change characteristics with the potential for downstream educational achievement impacts, despite having mixed empirical support as intervention targets for educationally relevant outcomes (Paunesku et al., 2015; Foliano et al., 2019; Yeager et al., 2019; Credé et al., 2017; Sisk et al., 2018; also see Moreau et al., 2018). For both factors, we noted that less work has been done specifically on interventions leading to impacts on reading ability. As mentioned earlier, although we did not conduct an intervention, we reasoned that if mindset and grit were to be potential malleable factors to change reading ability, then at the very least they should be correlated with reading ability. We went one step further and reasoned that mindset and grit should also be correlated with change in reading ability. Although not conclusive, if non-cognitive factors are a mechanism of change for reading ability, we would have expected to see mindset or grit significantly predicting change in reading ability. Therefore, our results do not lend support to the notion of mindset or grit being a mechanism of change for reading ability.

We note that especially for grit, conflicting definitions and interpretations of how grit is characterized brings up questions about how much we would expect grit to be changeable, and in turn, how much it might be useful for intervention purposes. Grit was initially characterized as a personality trait (Duckworth et al 2007) which would make the extent of its malleability inherently unclear. However, grit has since be described as important for intervention purposes (see Perkins-Gough, 2013), suggesting it as something malleable that should be encouraged for students to develop in order to better contend with the various academic challenges faced in their academic trajectory (see Yeager & Walton 2011). In fact, there are recent reports of grit interventions being developed for students for these purposes (Perkins-Gough, 2013). At the moment it remains unclear to us how much we should expect grit to be changeable and in turn how much we should expect it to facilitate more positive performance trajectories for students. To our knowledge, this is also unclear in the literature.

Although grit may be linked to successful, long-term educational and vocational outcomes, its placement in the broader context of non-cognitive factors impacting elementary-aged children’s’ academic achievement remains inconclusive. As noted in a review of non-cognitive skills (Tucker-Drob & Harden, 2017), skills such as academic interest and self-perceived ability yield the strongest link to academic achievement in younger children. Additionally, conscientiousness is a personality trait that not only predicts student academic success (Rimfeld et al., 2016), but also is positively related to grit (Credé et al., 2017; Duckworth & Gross, 2014; Rimfeld et al., 2016). Our finding failed to link grit with improvement in reading comprehension ability, but this may further indicate the differences between conscientiousness and grit as constructs. Grit still may be an important factor in determining which children will persevere through challenges over the long-term, but may be less pertinent to short-term, reading-focused interventions for elementary school children.

For our final research question, we proposed that if we found a correlation between mindset and grit with reading ability, then understanding the underlying genetic and environmental association would help us better understand the nature of that correlation, giving us a better idea of how mindset and grit could serve as agents of change in reading ability. Although our findings would not have told us what would happen due to an intervention, finding that the relation between mindset and grit with reading ability was due to shared environmental influences would give some evidence that a mindset or grit intervention might be successful on reading ability. Based on previous research findings (Tucker-Drob et al., 2016; Rimfeld et al., 2016), we expected to see common genetic and nonshared environmental influences underlying the correlation between mindset and grit with reading ability. We minimally supported our hypothesis in that we found that mindset and reading ability shared common nonshared environmental influences. It was reasonable to expect to see nonshared environmental influences at play, as mindset and grit are considered malleable factors that are both attitudinal and contextual, thus they can also be thought of as individual-specific factors. Therefore, it may be the case that mindset and grit interventions should be focused on individual contexts, as is done in the social psychological intervention literature (Yeager & Walton, 2011; Cohen et al., 2017) and, as such, we would expect mindset and grit to operate through the nonshared environment. Exactly what these unique, or nonshared environmental contexts actually are in the behavioral genetics literature, have been difficult to find (Neiderhiser et al., 2007).

In addition, we acknowledge that there are a variety of noncognitive and personality factors that are of importance and could be explored in the context of student achievement. For example, Richardson et al. 2012 highlight several non-intellective correlates of student grade point average and among the strongest were self-efficacy, conscientiousness, need for cognition, effort regulation, and strategic approaches to learning. Focusing on noncognitive factors and behavioral genetics approaches, the relative importance of mindset and grit, and the magnitude of their importance in supporting general achievement compared to other non-cognitive factors (Tucker Drob et al. 2016) and reading ability specifically remains weak (Malanchini et al. 2019) or inconclusive. For these reasons additional research is needed employing behavioral genetics approaches in order to better understand noncognitive factors and their role in predicting reading ability. It is important to note that we focused on mindset and grit because of their popularity and emphasis in recent literature and applications for intervention purposes.

Overall, the results lend very minimal support for mindset, and no support for grit, as potential targets for direct interventions to improve reading ability. It appeared that, despite the student’s level of grit or their mindset orientation, the best predictor of their future reading scores were their initial reading scores. This provides additional support that intervening directly on reading ability, although certainly more difficult, will benefit students more than the quick mindset or grit interventions. With that being said, there still may be clinical importance to even small effects – especially for students who fall just below thresholds or otherwise have specific characteristics.

In terms of practical utility of mindset and grit for predicting reading achievement, the effects we observed are small and, on the surface, do not lend much confidence in their ability to be applied for future forms of intervention. We will discuss a few findings from previous literature in order to clarify how our results fall in line with previous research and to expand on what we believe the next steps should be in future research on these topics. First, Sisk et al. 2018 found small but significant effects of mindset for subgroups for students of lower SES and at-risk backgrounds. They argue that small effects of mindset bring up the point as to whether it is useful to implement policy based on mindset. We believe this generates two questions: what size of effect is useful enough and for whom? And is it worth it to implement policies with mindset or grit? It could be worthwhile but at this point more research is needed to uncover for whom mindset and grit intervention can be beneficial and the ways we can benefit students by further tailoring intervention to their needs.

When thinking about what effect size is meaningful, Yeager et al. 2019 mention additional work that provides evidence (Dynarski et al, 2017; Slavin, 1986) that meaningful effect sizes for field work should be considered differently than for studies conducted in the lab. They argue that an effect size of d = .20 should be viewed as a large since this is a realistic educational intervention effect size to show in a year for students’ standardized test performance. Although we only measured mindset and grit and observed the change in their relations with reading over time, based on the size of our phenotypic correlations we are in line with Yeager et al.’s suggested effect size. We believe future research should be more considerate about what qualifies as meaningful effect sizes for lab studies versus field studies and the conventions might need to be reevaluated.

Additionally, Yeager et al. (2019) show that when examining the overall mindset intervention effect on student grades it was not significant in a national sample of about 12,000 students. However, they show for low achieving students there were significant effects of mindset intervention on overall grades and a variety of domain specific grades. Moreover, these effects for low achieving students were strongest when the context of the school and peer norms aligned with the messaging students received in the mindset intervention. Likewise, Foliano et al. (2019) show no overall effect of mindset intervention on reading and math grades in a sample of students in the UK. However, they only examined the effect overall and it was unclear if subgroup analysis was available for lower SES and at-risk students. Taken together, we argue that these findings emphasize the importance of contextual factors around the individual and that they should be further considered when we examine the efficacy of educational interventions. As such, in the case of Foliano et al. (2019) there is likely to be a contextual and conceptual difference when thinking about what constitutes lower SES and at-risk individuals in the UK compared to those in the US.

We see the utility of implementing a more contextual approach in Norris et al. (2020) as they argue that meaningful response to treatment effects for reading ability can depend on child individual and contextual factors. They describe this as the aptitude by treatment interaction or the child by instruction interaction (Cole & Dale, 1986; Kanfer & Ackerman, 1989; Speece, 1990; Connor et al., 2011b). Further, Yeager et al. (2019) demonstrates that differences in students’ experiences and needs means that different forms of intervention might be more or less effective. Collectively, we believe this suggests that relying on just overall effect heterogeneity can be misleading. We propose that future work should focus on further tailoring intervention to be in line with more precise forms of education and intervention to meet different students’ needs (Cohen et al., 2017; Hart, 2016). In addition to considering more targeted and precise forms of education intervention, future research should incorporate experimental and behavioral genetic approaches (see Burt et al., 2018) to better understand the intervention effects of mindset and grit on student achievement from a developmental perspective and through the lens of genetic and environmental influences. In general, we conclude that future research would benefit from examining mindset and grit interventions more thoroughly in the context of reading ability, although we suggest that these happen in tandem with a reading intervention (see recent work by Wanzek et al., 2021), and should not be short-term.

Limitations

This study is not without limitations. First, because our phenotypic results show weak correlations between mindset and grit and reading ability the lack of strong evidence for genetic and environmental influences is not surprising. With weak phenotypic correlations it makes it difficult to better capture genetic and environmental influences. Although we would not expect to show very large phenotypic correlations with reading ability based on previous research it is also important to note that our final sample might have been a bit underpowered compared to other twin studies investigating noncognitive factors and student achievement measures (i.e., Rimfeld et al. 2016, Tucker-Drob et al. 2016, Malanchini et al., 2019). Future behavioral genetic approaches should also examine noncognitive factors and student achievement in larger samples.

Second, conscientiousness and grit have been found to be two highly related constructs (Credé et al., 2017; Duckworth & Gross, 2014; Rimfeld et al., 2016), with some studies suggesting that grit may actually be a component of conscientiousness (Rimfeld et al., 2016; Schmidt et al., 2018). Conscientiousness has also been linked to higher overall achievement (Chamorro-Premuzic & Furnham, 2003; Ivcevic & Brackett, 2014; Noftle & Roins, 2007). Indeed, Rimfeld and colleagues (2016) conducted a large twin study examining the contributions of both grit and other personality traits to a broad measure of academic achievement and found that grit added little to that prediction after controlling for other personality traits, such as conscientiousness. Since our study did not contain a measure of conscientiousness, we could not replicate such findings. We caution researchers that although grit seems like a logical target for school interventions, the underlying mechanism may be more directly rooted in conscientiousness. Some have gone as far as to say that grit suffers from the jingle-jangle fallacy, in that grit is not different than conscientiousness other than by name (Crede, 2018, also see Crede, Tynan, & Harms, 2017; Duckworth & Quinn, 2021).

Moreover, previous work (Malanchini et al. 2019) suggests that the perseverance type noncognitive factors (e.g., conscientiousness, perseverance, mindset, grit) are less predictive of reading achievement than the openness and curiosity type non-cognitive factors (e.g., need for cognition, intellectual self-concept and identification). Future work might also consider further exploring the importance of a variety of these noncognitive factors in the context of reading ability.

In addition, it came to our attention post-hoc that there is conflicting evidence in the literature on whether the grit construct makes sense as the combination of passion and perseverance. Some argue that the combination is more predictive of performance (Jachimowicz et al., 2018), others finding that the perseverance component is more predictive (Muenks et al., 2017; Morrell et al., 2021), others arguing that the combination of subcomponents does not make sense because it is problematic measurement-wise (Crede, 2018; Crede, 2019), and further, that the subcomponents are more predictive of performance compared to their combination (Crede et al., 2017). Rimfeld et al. used the subscale items from the short-grit scale and show that the perseverance of effort subscale tended to be a better than the consistency of effort subscale, however their phenotypic correlations were also small. Based on this conflicting literature, we acknowledge that our combination of the subcomponents of grit as a potential limitation. We aimed to address this by running our same planned analyses with the subscale measures of grit and include those full results in supplemental information. We feel this neither further clarified the issue nor provided stronger evidence for grit as a predictor of reading ability. If anything, the twin modeling approach suggests that the reliability of the subscales are not adequate as the Cronbach’s alpha would suggest. To help in further clarifying this issue we encourage future work to continue to investigate grit and its predictive power based on the combined subcomponents compared to when they are separate.

It is also important to note that we only included measures of mindset and grit at time 1 and not at time 2, thus we have no way of knowing how much students’ attitudes on these measures changed naturally over the course of the two years between waves. For this reason, we are unable to determine if students increased, decreased, or had no change in their mindset or grit scores on their own (without intervention) over the two waves. This would be critical for uncovering whether any potential changes in mindset and grit from time 1 to 2 were related to potential change in reading over that period.

Finally, the self-report nature of the study could have also led to errors in reporting. The univariate twin models indicated high nonshared environmental influences for mindset and grit, which is partly indicative of measurement error. Future studies may consider methods for reducing measurement error, including using multiple informants or multiple waves of measurement, to allow for latent factors to be created.

Conclusions

This study aimed to examine the relationship between mindset and grit with both concurrent and future reading ability in a twin sample. Overall, we found very little relation between either mindset or grit with concurrent reading ability, or change in reading ability over time. In addition, the majority of the genetic and environmental influences were independent to each factor—indicating there was not much overlap in these influences among these measures. We also found that the nonshared environmental influences were high for mindset and grit, possibly indicative of measurement error. However, because these are believed to be malleable factors, the high nonshared environmental influences could also indicate meaningful individual differences. Overall, we found that the best predictor of future reading ability was the child’s previous reading score, indicating that if mindset and grit are important factors for improving achievement, they may not be as important or effective for reading ability interventions. These findings highlight the importance of conducting future research on both grit and intelligence mindset within varying educational domains (e.g. reading, math) to determine if either of those constructs may serve as useful targets for interventions. We also suggest that future work should aim to better understand how to separate variance of individual differences from plain measurement error where twin studies are involved—because highly contextual and malleable factors like mindset and grit are expected to be high in terms of environmental influences. Future work should employ twin studies in order to further clarify the influence of systematic individual differences from error variance.