! !
Grit and Mental Health: Is All Progress Beneficial?
Benjamin Holding
10407790Supervisors: Dr. Daiva Daukantaitė (Lund University, Clinical Psychology), Prof. Bertjan Doosje (University of Amsterdam, Social Psychology).
Thesis submitted in partial fulfillment (31ec) of a Master of Science (MSc) in Psychology (Res).
!
Abstract
Grit is a personality construct that measures the passion and perseverance to long-term goals (Duckworth, Peterson, Matthews & Kelly, 2007). It has an ability to predict increases in positive mental health and decreases in psychopathology. However, given that “not all progress is beneficial” (Sheldon & Kasser, 1998), it is possible that this relationship may change depending on the nature of the goal aspired towards. This study investigated the moderating effect of intrinsic (motivated by internal rewards) and extrinsic goals (motivated by external rewards) on the relationship between Grit and mental Well-Being (conceptualized by psychological well-being, satisfaction with life and positive affect frequency) and mental Ill-Being (anxiety, physical symptomology and negative affect frequency). The results found that an increase in likelihood of achieving intrinsic goals lead to the relationship between Grit and Well-Being decreasing. However, individuals showing a high level of intrinsic goal focus showed the greatest Well-Being overall, which changed even with equal levels of Grit. The results suggest that the relationship between Grit and mental health changes depending on the nature of the goals. The implications are discussed in terms of fulfillment of the basic psychological needs.
!
1. Introduction
1.1 Grit
Why do some people achieve more in life than others, even when they have equal ability? This has been a question of scientific interest for over a century, with many competing ideas. While researching eminent individuals of the time, Galton (1892) concluded that it is “ability combined with zeal and capacity for hard labour” (p. 33) that leads to
achievement. Modern intelligence research supports Galton’s emphasis on ability, showing that intelligence quotients (IQ) can predict future achievement in domains such as academic and career success (Kuncel, Hezlett, & Ones, 2001, 2004). However, this association is generally only found to be moderately correlated (Neisser et al., 1996).
Attempting to account for the gap between IQ and success is a huge research area in psychology, with multitudes of alternative traits and theories being suggested. Variations in personality, creativity, emotional intelligence, self-confidence and charisma have all been suggested to account for a person’s success in life (Duckworth, Peterson, Matthews, & Kelly, 2007). However, the predictive ability of these traits is often found to be small or moderate (Barchard, 2003; Baumeister, Campbell, Krueger & Vohs, 2003; Hirsh & Peterson, 2008).
Meanwhile, research continued to forget the main point that Galton made – that individuals required ‘zeal’. This idea of the importance of ‘hard labour’ has recently been reclaimed with a new single trait suggested to be essential to success no matter what the domain. Highly successful people have a shared quality – Grit.
Grit refers to a continuum of passion and perseverance to long-term goals (Duckworth et al., 2007). It is suggested that highly achieving and successful individuals show persistence over time rather than innate talent or ability. The Gritty individual sees
their aspirations.
Grit has the ability to closely predict high achievement, diligence and
perseverance through difficult external circumstances (Duckworth et al., 2007). This predictive ability is over and above related personality constructs such as
Conscientiousness, and does not include the focus on tidiness and orderliness (Duckworth et al., 2007; Reed, Pritschet, & Cutton, 2013). Grit is unique in that it
focuses on pursuing long-term goals, rather than constructs such as self-control, which is good at predicting short-term restraint but not the marathon like stamina that Gritty individuals possess (Duckworth et al., 2007). The gritty individual does not need positive feedback to keep persevering and will continue even if the process is unpleasant which makes it separate to constructs such as Need For Achievement (Duckworth, Kirby, Tsukayama, Berstein, & Ericsson, 2010).
Finally, Grit has been shown to be orthogonal or even slightly inversely
correlated with intelligence (Duckworth et al., 2007). Therefore, the predictive ability of Grit is not based on the predictive ability of intelligence and visa versa. Overall, Grit seems to be a novel construct that adds to the current level of understanding of individual differences in the motivation domain.
1.2 Mental Well-being
Increased levels of Grit suggest that an individual has a passion and commitment to a long term goal. This type of purpose serves to organize and plan one’s behaviour in life (McKnight and Kashdan, 2009). Indeed, Grittier individuals report that this consistency in behaviour forms a meaningful and self-defining directive (Hill, Burrow, & Bronk, 2014). According to Seligman (2002), becoming engaged and obtaining meaning in life is a pathway to happiness and mental health. This follows the humanistic Organismic Valuing Process theory (Rogers, 1951), suggesting that humans have an innate need for
personal growth and emphasizes the importance of becoming fulfilled in our lives and reaching ones full potential (also known as actualization; Rogers, 1961; Ryan & Deci, 2001; Sheldon, Arndt & Houser-Marko, 2003). Traditionally, this emphasis on personal growth was restricted to western ‘individualist’ cultures, however, with continued
globalization, this is increasingly found in non-western cultures as well (Murphy-Berman & Berman, 2003).
Having a purpose in life is one of the main aspects of Psychological Well-Being (PWB; Ryff & Singer, 2008). PWB is purported to be a general model of ‘positive’ mental health and positive functioning and shows a consistent ability to predict changes in mental health and illness (Jahoda, 1958; Keyes et al., 2012; Ryff, 1989, 2013). It is based on a core idea that health is more than just the mere absence of illness, and by promoting positive health one can decrease illness and boost longevity (Seligman, 2008). PWB is a single conceptualization of six core dimensions of positive mental health based on previous research and theory (Ryff, 2013). These dimensions are (1) self-acceptance, (2) positive relationships with others, (3) having control over our environment (mastery), (4) a sense of choice and volition in our behaviour (autonomy), (5) a sense of personal growth and (6) having a purpose in life (Ryff, 1989; Ryff & Singer, 2008).
If one considers Grit as a motivation similar to the growth motivation within the Organismic Valuing Process theory, it makes intuitive sense that a greater level of Grit would be associated with increased PWB. Especially, given that Gritty individuals tend to prefer to seek out and engage in goals considered to be meaningful (Von Culin,
Tsukayama & Duckworth, 2014). Evidence has also begun to support this theorization with Vainio & Daukantaite (2015) finding that increased Grit was related to greater PWB. This relationship was partially mediated by a sense of coherence (SOC), a psychological disposition relating to a sense that the world is comprehensible,
lives is an essential part of PWB, this mediation gives a partial explanation of the relationship between Grit and PWB (Pallant & Lae, 2002).
However, Seligman (2002) also suggests that happiness and mental health can be reached from another dimension – positive emotions and pleasure. This ‘hedonic’ approach to mental well-being, known as Subjective Well-Being (SWB), holds that an individual should attempt to maximize pleasure while minimizing pain and suffering (Diener, 1994). SWB operationalizes happiness by assessing the relative frequency of positive and negative emotions in combination with a measurement of life satisfaction. Unlike PWB there is no emphasis on purpose in life or in achieving anything specific. Instead individuals evaluate their lives as a whole in the moment without any specific goal criteria.
Despite a dearth of research, evidence points towards a relationship between Grit and positive emotions, with Gritty individuals showing greater levels of positive affect (Singh & Jha, 2008). Hill et al. (2014) point out three potential reasons to expect this relationship. Firstly, Gritty individuals are more likely to be less neurotic and more extroverted which are both associated with greater amounts of positive affect (see Duckworth and Quinn 2009; Steel, Schmidt & Shultz, 2008). Secondly, a base of positivity may help in individuals work towards long terms goals. This follows the theorisation that the purpose of positive affect is to help build physical, intellectual, and social skills (Fredrickson 2001). Thus, a positive platform may help support Gritty individuals when potential obstacles arise. Thirdly, it has been suggested that greater levels of SWB can drive changes in Grit which in turn enhances or maintain that well-being (Hill et al, 2014; Soto 2014; Specht, Egloff & Schmulke, 2013). It is possible therefore that by deepening or developing Grit, one may lead oneself towards greater satisfaction with life and well-being, which in turn acts to increase Gritty behaviour.
positive predictive relationship on overall mental wellbeing. However, Grit has also shown to be predictive of decreased levels of mental health problems. For example, Grit has been shown to negatively predict fear, sadness and anxiety (Sheridan, Boman, Mergler, Furlong, & Elmer, 2015; Singh & Jha, 2008) Grit has also shown to be associated with decreased depression and predictive of decreased suicide ideation in highly depressed individuals (Kleiman, Adams, Kashdan, & Riskind, 2013). This decrease may relate to the consistent future-based orientation that is a defining feature of Grit (Hirsch et al., 2006). Finally, Grit can also predict decreases in work related burnout (Salles, Cohen & Mueller, 2014).
While PWB and SWB started out as separate constructs, there has been an increasing move to attempt to combine them in order to gain a more complete understanding of positive mental health characteristics. Keyes (2005) suggests that mental health should be seen as consisting of two separate correlated unipolar
dimensions. One dimension consists of positive mental health constructs such as SWB and PWB. This acts to predict happiness, positive functioning and achievement of basic psychological needs. The other dimension consists of psychopathology and mental illness. This consists of states such as depression and anxiety. This dual-factor bifurcation of mental health allows greater understanding about the differences and antecedents between mental wellness and mental illness, which is being increasingly emphasised (Kelly, Hills, Huebner, & McQuillin, 2012; Seligman & Csikszentmihalyi, 2000).
Overall, we see that Grit shows a relationship with both approaches to mental health. Grit predicts increased is positive mental wellbeing and decreases in
psychopathology and mental health issues. However, while achieving one’s goals can be an important drive for some people, “not all progress is beneficial” (Sheldon & Kasser, 1998). It is possible that if individuals become passionate about the wrong goals, they may in fact cause themselves psychological harm (Sheldon, 2014).
1.2 Aspirations and Well-Being
Self-Determination Theory (SDT; Deci & Ryan, 1985) is a broad theory of human motivation. It attempts to explain tendencies to aspire towards certain goals and the effects this has on mental health. It focuses on understanding two competing sources of motivation and their respective roles in personality, cognition and behaviour. SDT posits that motivation can be split into two broad groups depending on the type of goal an individual aspires towards.
The first group is intrinsic goals, which includes such motivators as self-development, affiliation, and community contribution. These intrinsic goals reflect movement towards an individual’s inherent tendency for personal growth (Deci & Ryan, 2000). The second motivational group is those based on extrinsic goals, such as financial success, physical attractiveness, and social popularity (Kasser & Ryan, 1993, 1996). Extrinsic goals lead individuals outwards for rewards rather than the pursuit being satisfying in itself. Individuals pursuing extrinsic goals measure achievement upon the attainment of external signs of success, such as heightening one’s status in the eyes of others (Kasser & Ryan, 2001).
A growing body of evidence supports the theorization that these two aspirations categories are not equally beneficial for mental health. Kasser and Ryan (1993) showed that individuals who valued financial success over personal growth, relationships or community involvement reported lower PWB. Kasser and Ryan (1996) went on to examine more life goals, finding again that relative to extrinsic aspirations (e.g. wealth, fame and image), intrinsic aspirations (e.g. personal growth, community involvement and physical fitness) were positively associated with positive affect, vitality and
self-actualisation. A sub-theory of SDT was created based on these findings and PWB research, called the Basic Psychological Needs Theory (BPNT; Deci & Ryan, 2000),
suggesting that self-beneficial goals should provide the three essential needs of
autonomy, mastery, and positive relationships. Without such goals, there is a possibility that there will be distinct costs to mental health. It is concluded that only intrinsic goals provide these basic psychological needs (Deci & Ryan, 2000).
This could be taken as surprising, given that both goal orientations would give individuals a sense of purpose. However, there is an important difference in the routes they take to obtain this meaning. Extrinsic goals lead to self-worth via superiority (Kasser, 2002). Individuals attempt to achieve their aspirations by engaging in stressful interpersonal comparisons and then attempting to outperform others (Vansteenkiste, Simons, Lens, Sheldon & Deci, 2004). This particular process of obtaining self-worth creates a view of the world as being a competitive environment where in the struggle for recourses and power only the fittest survive. The contingent sense of self-esteem this requires has been suggested to be not only detrimental to mental health (Patrick, Neighbors, & Knee, 2004; Ryan & Deci, 2004) but also lead to damaging effects on individuals’ functional behaviour, with decrease cognitive ability, achievement and persistence (Vansteenkiste, Lens, & Deci, 2006).
This contrasts with individuals who have more intrinsic goal orientations. These individuals obtain self worth via genuine interests in positive social affiliations and helping others (Kasser, 2002). Intrinsically oriented individuals are concerned with their self-worth but also the welfare of others. As such, they are more likely to empathise and develop trusting and positive social relationships (Kasser, 2002). These individuals are also more likely to have a sense of control and autonomy (deCharms, 1968). This is linked to positive mental health outcomes and greater functional behaviours (Deci, Koestner, & Ryan, 1999).
The psychological differences between these aspiration motivations have been shown to have behavioural consequences. Kasser and Ryan (2001) found that
extrinsically orientated individuals have more conflicting, less happy and less trusting relationships with friends and partners. This fits with research showing that these individuals report putting less emphasis on affiliations or compassion (Kasser & Ryan, 1993; Schwartz, 1992), being less empathic (Sheldon & Kasser, 1995), and agreeing more often that they use their friends to gain advantages in life (Khanna & Kasser, 2004). Such individuals even explicitly report being more likely to objectify others and use them to obtain their own ambitions (Kasser, 2002). On other hand, intrinsic goals have shown to predict knowledge acquisition in students, feelings of competence and decreases in frustration (Deci & Ryan, 2000; Gottfield, 1985, 1990; Lepper et al.. 2005).
While studies have generally focused on the relationship between aspiration type and positive mental health, Niemiec et al. (2008) found aspiration types showed a differential relationship with SWB as well as with what they described as mental ‘Ill- Being’ (a single construct consisting of the level of depression, anxiety, physical symptoms and negative emotions) at one year post test. Physical symptoms were included as mental illness shows a high level of association with physical health issues (Lamers, Westerhof, Bohlmeijer, & Keyes, 2013). A significant positive relationship was found between intrinsic aspirations and SWB, mediated by change in the satisfaction of the basic psychological needs. However, attainment of extrinsic aspirations was not associated with SWB. Instead, attainment of extrinsic aspirations was associated with an increase in Ill-Being. This suggests that aspiration type is not only associated with positive mental health, but also issues with mental and physical health problems as highlighted through the construct of ‘Ill-Being’.
Overall, the current state of research suggests that intrinsic motivation provides benefits to mental health, while extrinsic goals appear to increase psychopathology and potentially hinder positive mental health constructs such as PWB.
1.4 The Current Study
So far research suggests that Grit is positively associated with positive mental health, and negatively with mental illness. However, previous research has not taken account of relationship of differences in goals. Given the evidence showing that changes in types of goal aspirations can lead to different mental health outcomes, it is possible that the type of goal have an important impact on the relationship between Grit and mental health. More specifically, the association between Grit and positive mental health and mental health problems may be moderated by the type of goals orientation one has. Moderation is suggested over mediation, as it is not suggested that the predictive effect of Grit acts through aspiration type, but rather the specific goals can systematically change the direct relationship between Grit and mental health.
This study aims to investigate the relationship between Grit and the separate constructs of Well-Being (consisting of PWB, satisfaction with life and positive affect) and Ill-Being (consisting of anxiety, negative affect and physical symptomology). More specifically, it is aimed to investigate whether the ratio of intrinsic to extrinsic goals aspired to, and the likelihood of achieving these goals, moderates the relationship between Grit and Well-Being and Ill-Being.
1.5 Hypotheses
There are four main expected outcomes to this study:
Firstly, it is hypothesised that Grit will be positively predictive of Well-Being and negatively predictive of Ill-Being. Secondly, it is hypothesised that individuals who place more importance on intrinsic goals and report being more likely to achieve intrinsic goals will be predictive of greater Well-Being and decreased Ill-Being. Thirdly, the relationship between Grit and Well-Being will be moderated by the importance and likelihood
individuals place on achieving intrinsic goals. More specifically, the positive relationship between Grit and Well-Being will be stronger if individuals report placing more
importance on intrinsic goals and report themselves to be more likely to achieve intrinsic goals. Fourthly, the relationship between Grit and Ill-Being will be moderated by the importance and likelihood individuals place on achieving specific goals. Specifically, the negative relationship between Grit and Ill-Being will be stronger if individuals report placing more importance on intrinsic goals and report themselves to be more likely to achieve intrinsic goals.
2. Method
!
2.1 Design
The research used a cross-sectional correlational design. Grit was used to predict scores on the two factors of mental health: Well-Being (combination of positive affect, life satisfaction, and PWB) and Ill-Being (combination of anxiety, physical symptoms and negative emotions). The moderator variables were Intrinsic Importance Orientation and Intrinsic Likelihood Orientation as calculated from the Aspiration Index (Kasser & Ryan, 2001). These two variables represent a ratio of intrinsic to extrinsic aspirations, in terms of how important they see the goal and how likely they are to achieve it.
2.2 Participants
A sample of 384 participants was collected using an online survey. However, 24 participants were excluded from the study due to either incorrectly answering the
attention question (20 participants), or stating that English was not their first language (3 participants). After screening 361 participants remained (179 female, 181 male, 1 other). The participants ranged in age between 20-71, with a mean age of 35.82 (SD = 10.91). The majority of participants were from the United States of America (92.5% of total) or
India (5.8%). Participants were a mix of ethnicities with the most being white/Caucasian (77%), Asian (8.6%), African (7.8%) or Hispanic (4.2%).
Participants were recruited using Amazon.com’s Mechanical Turk (Amazon Inc., 2013). The Mechanical Turk (MTurk) website provides a platform where individuals are paid a small amount for completing a task. The MTurk is being increasingly used for behavioural research and has previously been used to study clinical and non-clinical psychological populations (Shapiro, Chandler & Mueller, 2013) and shows data quality that is both independent of compensation rates and more demographically diverse than standard internet and undergraduate samples (Buhrmester, Kwang, & Gosling, 2011). In the current study participants were paid $1 for successfully completing the study.
Participants were required to have an approval rate of 95% on the MTurk system, be native English speakers and be aged over 18. To ensure participants did not attempt to redo the study, the website was coded so that participants could only complete the questionnaire once.
The sample size was well over the recommended 138 participants shown through an a-priori statistical power analysis (Cohen, 1988; Soper, 2015). This was based on running a structural equation model (SEM) with 2 latent variables (Well-Being and Ill-Being) and 11 observed variables (Grit Scale, intrinsic importance and likelihood orientation based on the AI, positive/negative subscales of the PANAS, PSQ-15, Life satisfaction scale, PWB, SWB, anxiety and depression subscales of the HADS) with an expected effect size of r = .20, statistical power of .80 and an alpha level of .05. This sample size is also larger than often seen in published research (Kline, 2011), and well over the recommended 200 participants often suggested as a rule of thumb (Kenny, 2014). Therefore this sample size gives an SEM model enough power to detect significant effects that actually do exist.
2.3 Materials
The current study used eight questionnaire measurements:
1. The Short Grit Scale (Grit-S; Duckworth & Quinn, 2009) measures perseverance for long-term goals reflecting passionate interests or personally valued aims. This 8-item Likert scale questionnaire is rated from 1 (not at all like me) to 5 (very much like me). A sample item is “I finish whatever I begin”. The scale is scored by summing the 8-items, while taking account of certain reverse scored items. Higher scores represent greater levels of Grit. None of the items refer to any specific type of goal.
The Grit-S is an adaptation of the original 12-item Grit scale (Grit-O;
Duckworth, Peterson, Matthews & Kelly, 2007). The Grit-S has the same structure as the Grit-O with 2-factors – consistency of interest and perseverance of effort. The Grit-S has been shown to have stronger psychometric properties than the original but with increased brevity (Duckworth & Quinn, 2009). Grit-S shows adequate alpha (α) internal consistency (α = .73 to .83), one-year test–retest reliability (r = .68, p <.001; Duckworth and Quinn, 2009) and predicts achievements such as grade point averages, placing within competitions and completion of military training (Duckworth et al., 2007). The current study found good internal consistency (α = .87).
!
2. The Aspiration Index (AI; Kasser & Ryan, 1996) measures the importance and perceived likelihood of achieving different goals (e.g., financial success or social
affiliation). The 64-items questionnaire is rated using a 5-point Likert scale, ranging from 1 (not at all/very low) to 5 (very important/very high). An example item is “Your name will be known by many people”. Items can be summed to create 7 subscales: Self-Acceptance, Affiliation, Community Feeling, Physical Fitness, Financial Success, Attractive
based on these subscales – intrinsic (self-acceptance, affiliation, community feeling) and extrinsic aspirations (Physical Fitness, Financial Success, Attractive Appearance and Social Recognition) (Kasser & Ryan, 1996). This theorization has been supported by second-order factor analyses, with two factors explaining 77% of the total variance (Hammervold, 2014).
A number of ways to score this scale exist. This study created two values based on ratios between intrinsic and extrinsic goal aspirations. Firstly, the importance scores on subscales loading on the intrinsic and extrinsic factors separately were averaged for each participant. The average intrinsic importance score was then subtracted from the average extrinsic importance value to create an Intrinsic Value Orientation score (IVO; see Kasser and Ryan, 1996; Sheldon and McGregor, 2000). The same process was done for likelihood scores to give an Intrinsic Likelihood Orientation (ILO) score. A higher score indicates a focus towards intrinsic goals, while a lower score represents a focus towards extrinsic goals. It is important to create this type of corrected score, as the raw scores may reflect a general tendency for participants to view all goals and equally important, which is a tendency of individuals with greater positive mental health (Emmons, 1986).
The AI has demonstrated adequate internal consistency (α = .76 to .82; Kasser and Ryan, 1996) and predicts goal attainments at one-year post test (r = .51 to .67, p <.001; Niemiec et al., 2009). The current study found excellent internal consistency for items asking about importance of goals (α = .92) and the likelihood of achieving goals (α = .95).
3. The Satisfaction with Life Scale (SWLS; Diener, Emmons, Larsen, & Griffin, 1985) is a short 5-item scale designed to get an overall measurement of life satisfaction. A sample item is “The conditions of my life are excellent”. Each item is rated between 1 (Strongly
disagree) to 7 (Strongly agree), with higher scores representing greater satisfaction. Test-retest reliability with a one-month interval has been shown to be .84 (Diener et al. 1985) with good internal consistency of α= .87. Shevlin, Brunsden & Miles (1998) showed good construct validity with each item loading between .92 and .96 on a single factor. The current study found excellent internal consistency (α = .94).
4. The Psychological Well-Being Scale (PWB; Ryff & Keyes 1995) is an 18-item questionnaire measuring autonomy, environmental mastery, personal growth, self-acceptance, purpose in life and positive relationships. A sample item is “For me, life has been a continuous process of learning, changing and growth”. Items are scored on a 6-point Likert scale from 1 (strongly disagree) to 6 (strongly agree). The scale is scored by summing the 18-items, while taking account of certain reverse scored items. Higher scores represent greater levels of well-being. The Cronbach alpha reliability is α = .76 (Ryff & Keyes 1995) and shows good concurrent validity (Van Dierendonck, 2004). The current study found good internal consistency (α = .89).
5. The Hospital Anxiety and Depression Scale (HADS; Zigmond & Snaith, 1983) is a 14-item scale used to assess anxiety and depression without referring to physical symptoms. A sample item from the anxiety subscale is “I feel tense or 'wound up”. Each item is scored from 0 (not at all) to 3 (most of the time), with some items reverse scored. The scale is sum-scored with higher scores represent greater levels of impairment. It is suggested that a score of above 10 represents probable clinical anxiety or depressive disturbance. The HADS shows good internal consistency (anxiety subscale α =.80, depression subscale α =.81) and adequate test-retest reliability over 6 weeks (anxiety and depression subscales r = 80; Herrmann, 1997) and found to have high concurrent validity (Bjelland, Dahl, Haug
& Neckelmann, 2002). The current study found good internal consistency for the anxiety subscale (α = .80) as well as for the depression subscale (α = .80).
6. The Public Health Questionnaire 15 (PHQ-15; Kroenke, Spitzer & Williams, 2002) measures somatic symptoms, asking whether over the past 4 weeks, individuals have suffered from problems such as “stomach pain”. The PHQ-15 has 15 items scored from 0 (not bothered at all) to 2 (bothered a lot). Higher scores represent a greater amount of somatic problems. The questionnaire has been validated in the general population (Kocalevent, Hinz & Brähler, 2013). The PHQ-15 shows adequate to good internal consistency (α = .78– .87; Zijlema et al., 2013), moderate test-retest reliability over 2 months (r =.54; Gulec, Gulec, Simsek, Turhan & Sunbul, 2012) and correlates well with other measures of somatic symptoms (Zijlema et al., 2013). The current study found good internal consistency (α = .85).
7. The Positive Affect Negative Affect Schedule (PANAS; Watson, 1988) measures positive and negative emotions. The PANAS contains two 10-item Likert scales for positive and negative affect. Participants are asked to rate the extent they have felt a certain emotion over the past week. Sample emotions include “interested” and
“irritable”. The scale is scored from 1 (very slightly or not at all) to 5 (extremely). Higher scores represent higher levels of positive or negative affect depending on the subscale. The scale shows good internal consistency (negative α = .85, positive α = .89) moderate test-retest reliability over 1 month (r = .57) and is well validated (Crawford & Henry, 2004). The current study found excellent internal consistency for the positive subscale (α = .93) and also for the negative subscale (α = .92). In the current study, individuals were asked to report how they felt in the past week.
2.4 Procedure
Participants were recruited through Amazon.com’s MTurk worker interface at
http://www.mturk.com. Once participants had found the study on the list of available tasks and decided to take part they were given a link to the study
(http://sgiz.mobi/s3/5c14a9b70a71). Participants were able to complete the study in their own home and were free to take up to one hour to finish it. The study initially provided a brief research synopsis and informed participants of their rights as per standard ethical protocol in psychological studies (see Appendix A). Informed consent was obtained by requiring participants to click a checkbox signifying that they
understood and consented to be part of the study. The questionnaire initially asked for demographic information before presenting the psychological measures. The order of the questionnaires was fixed (Grit-S, AI, SWLS, PWB, PHQ-15, PANAS, HADS). The questionnaire took approximately 20 minutes to complete, and participants were debriefed following the presentation of the final scale (see Appendix B).
2.5 Changes to Questionnaires
The questionnaires were adapted for use as an online survey. All questionnaires were put into a multiple-choice format. Within the HADS scale, an attentional check item was added to assess whether responses were being made with due attention, or being automated. The item “while watching TV, have you ever had a fatal heart attack?” was added after the 11th item of the HADS scale (adapted from Paolacci, Chandler &
Ipeirotis, 2010). Similar methods have been shown to increase statistical power and reliability (Oppenheimer, Meyvis & Davidenko, 2009).
2.6 Ethical Issues
Amsterdam) Psychology Ethics Committee (De Facultaire Commissie Ethiek). Participants were required to be over 18 years old to take part in the study.
2.7 Sensitivity Analysis and Data preparation
There was no missing data in the responses, and no evidence of unengaged responding. Q plots of normality showed little deviation from normality, apart from in the PHQ-15 which shows a left skew (skew = 1.51) and the negative emotion subscale of the PANAS which shows heavy tailing (skew = 1.87). The skew of these scales was lower than the cutoff of 3, which has been proposed to signify a major variation from the normal distribution (Kline, 2011). Therefore, untransformed variables were used for the analysis.
Following the methodology of Niemiec et al. (2008), a Principle Components Analysis (PCA) with Oblimin rotation was initially conducted to assess whether the data supported an analytic strategy of bifurcating mental health into separate Well-Being and Ill-Being constructs. It also allows for the mental health data to be reduced to two variables. Well-Being was expected to consist of PWB, SWLS and PANAS-positive subscale. Ill-Being was expected to consist of PHQ-15, HADS, depression and anxiety, and PANAS-negative subscale. However, the HADS depression subscale showed high cross loading to both factors and was removed from the PCA model in order to create simple structure.
After removing depression, the PCA revealed two distinct factors. Factor 1 was defined by high loadings of three variables: (1) PANAS positive subscale (.90), (2) Satisfaction with Life (.74), and (3) Psychological Well-Being (.77). Factor 1 accounted for 53.31% of the total variance with an eigenvalue of 3.20. No other variable loaded onto this factor at a level greater than .3. This factor was interpreted as the ‘Well-Being’ construct.
Factor 2 was defined by high loadings of (1) PANAS negative subscale (.83), (2) HADS anxiety subscale (.76) and (3) the PHQ-15 physical symptoms scale (.88). Factor 2 accounted for an additional 19.03% of the total variance with an eigenvalue of 1.14. Therefore, this factor was interpreted as a general construct of ‘Ill-Being’.
Overall, the PCA was interpreted as supporting the creation of two mental health constructs, based on scree-plots (Cattell, 1966), the eigenvalue-greater-than-one rule of thumb (Kaiser, 1960), and a Monte Carlo Parallel Analysis simulation (Horn, 1965). Standardized scores of Well-Being and Ill-Being resulting from the PCA analysis for were used for correlations, Analysis of Variance and comparison of means.
2.8 SEM Preparation
The moderation analysis was conducted using Structural Equation Modeling (SEM) in AMOS 21 (Arbuckle, 2012). SEM is a multivariate analytical technique used to investigate proposed structural relationships. This method uses elements from factor analysis and multiple linear regression to analyze potential causal dependencies between measured variables and hypothesized latent constructs (Byrne, 2010).
Within the SEM there are two types of model. The first is the measurement model, which represents how the indicator variables come together to represent the latent variables. The second is the structural model, which represents how the theorized latent variables are related to other constructs (Byrne, 2010). The structural model then tests the proposed relationships simultaneously, which allows measurement of the effects of each variable while controlling for the effects of others.
A number of assumptions are required when using SEM. The variables must be normally distributed, show linear relationships, have no missing data, relationships should be non-spurious and data used for variables must be interval data (Byrne, 2010).
While moderation can be investigated with standard regression analyses, SEM provides a method to calculate predictive abilities with reduced bias. Regression analyses do not estimate reciprocal effects or account for measurement error, while SEM controls for measurement error within the variables, allows for the measurement of reciprocal effects and assesses the overall fit of the data to the specified model (Peyrot, 1996). The ability to estimate model fit is beneficial as it provides information about the suitability of the specified model and support for model based theories (Byrne, 2010).
It has been previously suggested that a pure moderator should not be associated with either the predictor or the DV (Baron & Kenny, 1986) with such relationships reported to lead to decreases in statistical power (Alasuutari, Bickman & Brannen, 2008). However, others report that this multicolliniarity problem is unfounded and that small levels of intercorrelations do not adversely affect power (Aguinis, 2004). Nevertheless, it is suggested that centering variables when moderating should be used to decrease any potential multicolliniarity problems (Alasuutari et al., 2008).
In the current study, the theorised SEM model contains three sets of pathways that feed into the latent dependent variables (DVs) of Well-Being and Ill-Being. The first pathway set is the relationship between Grit and the DVs, the second is the relationship between the aspiration moderator (IVO and ILO separately) and the DVs, and the last is the interaction of these two (calculated as Predictor X Moderator as mean centered standardised variables) and the DV. A moderation hypothesis is supported if an interaction path is significant (see Figure 1).
! Figure 1. The proposed structural moderation model.
!
To test the measurement model a Confirmatory Factor Analysis (CFA) was used to assess whether the relationship of the observed indicator variables to Well-Being and Ill-Being loaded sufficiently on their respective factors and showed acceptable model fit. The data were used to investigate the stability of the two factor structures and provide theoretical support for their use (Cole and Maxwell, 2003).
To investigate whether the CFA shows acceptable fit, one should use a
combination of standardized root mean square residual (SRMR), comparative fit index (CFI) and root mean square error of approximation (RMSEA; Hu & Bentler, 1999). SRMR is a representation of the average value across all standardized residual variances arising from fitting the variance-covariance matrix of the model to the data (Byrne, 2010). CFI represents the level of covariation within the data. RMSEA attempts to estimate how well the data would fit with the general population, rather than the measured sample. In a well-fitting model, SRMR should be close to or below .08, CFI should be close to or above .95, and RMSEA should be close to or below .06 (Hu and
Bentler, 1999). However, it has been suggested that SRMR and RMSEA below .10 can be used to indicate adequate fit (Browne & Cudeck, 1993).
Following confirmation of the two-factor structure of the DVs, the model fit of the structural path model can be investigated. Assessment of model fit in structural models is also performed with the CFI, SRMR and the RMSEA with similar recommendations for values.
When investigating group moderators, the measurement model can additionally be tested for configural and metric model invariance (Millsap & Meredith, 2007). Invariance requires that the indicator variables load with equal strength to their
respective latent factor despite changes in IV or moderating variables. To assess whether configural invariance exists, the model fit indices are compared between groups. This is done to assess whether the same model configuration holds for both groups. To assess metric model invariance, a chi-squared difference test between each goodness-of-fit χ2 (chi-square) test statistic obtained during configural model testing can be computed. This is done to assess that factor loadings, intercepts and residual variance between groups is invariant.
3. Results
3.1 Demographic analysis
Preliminary analyses were performed to investigate whether demographic variables (Gender and Nationality) were related to Grit, PWB, SWB, satisfaction with life, anxiety, positive and negative emotions, physical symptoms, IVO or ILO.
A series of Benjamini-Hochberg False Discovery Rate (Benjamini & Hochberg, 1995) corrected t-tests found four significant gender differences. Differences were found on Intrinsic Value Orientation, indicating that women reported significantly higher scores on IVO (M = 9.43, SD = 2.94) than men did (M = 8.40, SD = 2.91, t(358) =
-3.33, p = .001, d = 0.35). Suggesting women place more importance on intrinsic aspirations. ILO also showed a significant difference with women reporting higher likelihood of achieving intrinsic aspirations (M = 7.56, SD = 2.38) than men (M = 6.83, SD = 2.25, t(358) = -2.96, p = .003, d = 0.32). Similarly, women reported significantly higher scores on the PHQ-15 (women: M = 5.40, SD = 5.00; men: M = 4.31, SD = 3.81, t(358) = -2.33, p = .02, d = 0.25), suggesting that women report more physical ailments than men. Finally, Women reported greater levels of anxiety (M = 7.84, SD = 3.89) than men (M = 6.66, SD = 3.88, t(358) = -3.33, p = .001, d = 0.30).
Secondly, a further set of False Discovery Rate corrected t-tests were used to compare the results of the majority of participants from the United States of America with the results of those who held other nationalities. Two differences were found within participant aspiration orientations. Differences were found on IVO, indicating that the American Group reported a significantly higher scores (M = 9.10, SD = 2.87) than other nationalities did (M = 6.48, SD = 3.10, t(359) = -4.55, p < .001, d = 0.88). This suggests that the American group reported a greater focus on intrinsic goals, than those from other countries.
Secondly, significant differences were found on Intrinsic Likelihood Orientation between groups. The American group reported higher ILO scores (M =7.31, SD = 2.33) than other nationalities (M = 5.78, SD = 2.00, t(359) = -3.31, p = .001, d = 0.71). This suggests again that the American group reported a greater likelihood to achieve intrinsic goals over extrinsic goals.
Given the differences between nationality groups, the SEM analysis was also run on a purely USA sample. The pattern of the results followed the complete sample, except where explicitly stated.
! !
3.2 Descriptive Statistics
Table 1 presents means, standard deviations, and Pearson’s correlations between the observed variables used as indicators of the Well-Being and Ill-Being latent constructs examined in the SEM analysis: Standardized scores of Well-Being and Ill-Being (created through PCA analysis), Grit, and the IVO and ILO aspiration type values.
Table 1.
The Mean Scores, Standard Deviations and Correlations of All Variables
M (SD) 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 1. Age 35.82 (10.91) .20** .27** .23** .06 -.10 -.07 .12* .06 -.12* -.15** -.01 2. Grit 3.46 (0.75) - .11* .25** .47** -.36** .35** .50** .34** -.36** -.40** -.20** 3. IVO 0.69 (2.96) - .76** .30** -.19** .13* .48** .16** -.19** -.21** -.07 4. ILO -0.92 (2.34) - .46** -.32** .31** .57** .28** -.28** -.37** -.18** 5. Well-Being 0.00 (1.00) - -.37** .79** .86** .83** -.37** -.53** -.20** 6. Ill-Being 0.00 (1.00) - -.42** -.52** -.16** .85** .85** .83** 7. SWLS 22.85 (8.13) - .64** .42** -.34** -.44** -.30** 8. PWB 79.60 (13.67) - .55** -.49** -.55** -.31** 9. PANAS-P 31.19 (9.51) - -.18** -.36** -.13* 10. PANAS-N 15.12 (6.47) - .67** .49** 11. Anxiety 7.24 (3.92) - .58** 12. PHQ 4.84 (4.48) - Note. N =361 *p < .05 **p < .01 (2-tailed)
IVO = Intrinsic value orientation, ILO =Intrinsic likelihood orientation, SWLS = Satisfaction with Life Scale, PWB = Psychological Well-Being, PANAS-P = PANAS positive affect subscale, PANAS-N = PANAS negative affect subscale, PHQ = PHQ-15 score.
3.5 SEM Analysis
3.5.1 Measurement Model
The results found that all variables loaded adequately on their respective factors (range between .59 and .93) and all factor loadings were significant (p < .001). These results suggest that the observed indicator variables relate sufficiently to the underlying latent constructs, and that residual variance covaries as expected. This provides support for proceeding with the proposed structural model (Cole & Maxwell, 2003).!Unstandardised, standardized and significance values, along with all information about model fit and parameter estimates for all subsequent models can be seen in Appendix C.
To further investigate the measurement model, model fit was examined. Overall, the model was a reasonable fit with acceptable SRMR (.41), CFI (.97) and RMSEA (.09). While RMSEA was found to be relatively high, it is still below .10 representing adequate fit (Browne & Cudeck, 1993). Additionally, given that SRMR and CFI were well within the acceptable range, the two latent factor measurement model was used.
Following the verification of distinct Well-Being and Ill-Being latent constructs, pathways were inserted from the independent and moderating variables to the latent variables and testing of the structural model was performed.
3.5.2 Structural Model
The basic theoretical path model was first tested for IVO and ILO in separate models. These basic models did not provide a good fit to the data based on the fit indices. Therefore, IVO and ILO were put together into a combined model. The Lagrange Multiplier test (also referred to as modification indices; Arbuckle, 2012) was used to find parameters that could be changed to improve model fit. One covariance between residual error terms was added, which was consistent with the theory behind the model. The
complete structural model showed acceptable model fit (SRMR = .05, CFI = .95, RMSEA = .09). The final model can be seen in Figure 5.
3.5.3 SEM Moderation
Age was added as a control based on a preliminary regression analysis finding it a significant predictor of Well-Being. Differences between male and female gender was tested using a moderated moderation model (see section 3.5.5).
To test each hypothesis, a step-wise process to the moderation analysis was taken. In the first step we explored the predictive ability of Grit on Well-Being and Ill-Being (hypothesis one). The model consists of two observed variables and two latent variables; the independent exogenous observed variables are Age and Grit, and the dependent endogenous latent variables are Well-Being and Ill-Being. The model with the standardized path coefficients can be seen in Figure 3. The path coefficients were all statistically significant (p <.001). The model accounts for 29% of Well-Being variance and 19% of Ill-Being variance. The model indicates a direct association of Grit on Well-Being of .54 and also on Ill-Well-Being of -.44. This suggests that increases in Grit can moderately predict increases in Well-Being and decreases in Ill-Being.
Figure 3. Path diagram representing a direct relationship between Grit and
Well-Being, and Grit and Ill-Being.
Note. All parameters are standardized. Abbreviations: SWLS = Satisfaction With Life Scale; PWB = Psychological Well-Being; HADS-Anx = HADS anxiety subscale; PHQ
= PHQ-15 score ; P_pos = PANAS positive affect subscale; P_neg = PANAS
negative affect subscale.
In the second step, we explored the predictive ability of Age, Grit, IVO and ILO on Well-Being and Ill-Being (hypothesis two). This model consists of four observed variables and two latent variables: the independent exogenous variables are Age, Grit, IVO and ILO. The dependent endogenous variables are Well-Being and Ill-Being. The model with the standardized path coefficients can be seen in Figure 4. The path
coefficients were statistically significant (p < .01) apart from IVO to Ill-Being (p > .05). The model accounts for 49% of the variance scores of Well-Being and 28% of the variance of Ill-Being. The model indicates a direct association of Grit on wellbeing of .42 and also on Ill-Being of -.35. This suggests that increases in Grit can moderately predict
increases in Well-Being and decreases in Ill-Being. The model also indicates an
association of IVO on Well-Being of .20. This suggests that an increase in focus towards valuing intrinsic goals higher is predictive of greater Well-Being. Finally, the model shows an association of ILO on Well-Being of .34 and on Ill-Being of -.39. This suggests that an increase in likelihood of achieving intrinsic goals over extrinsic goals is predictive of a moderate increase in Well-Being and decrease in Ill-Being.
Figure 4. Path diagram representing a direct relationship between Grit and
Well-Being, Grit and Ill-Well-Being, IVO and Well-Well-Being, IVO and Ill-Well-Being, ILO and Well-being, ILO and Ill-Being.
Note. All parameters are standardized. Abbreviations: SWLS = Satisfaction With Life Scale; PWB = Psychological Well-Being; HADS-Anx = HADS anxiety subscale; PHQ
= PHQ-15 score; P_pos = PANAS positive affect subscale; P_neg = PANAS
negative affect subscale; IVO = Intrinsic value orientation; ILO = Intrinsic likelihood orientation.
The third and final step was performed to investigate the moderating effect of aspiration importance and likelihood on the relationship between Grit and Well-being and Ill-Being (hypothesis three and four). Figure 5 represents the final model with the
standardized path coefficients. The path coefficients were statistically significant (p < .05) apart from IVO to Being, the IVO interaction variable with both Well-Being and Ill-Being, and the ILO interaction variable with Ill-Being. The final model accounts for 50% of Well-Being and 28% of the variance of Ill-Being. The model indicates a direct
association of Grit on Well-Being of .42 and also on Ill-Being of -.35. It indicates an association of IVO on Well-Being of .20. Finally, ILO shows a significant moderating interaction effect between Grit and Well-Being of -.17, suggesting that an increase in likelihood of obtaining intrinsic goals, dampens the association between Grit and Well-being. In other words, the importance of Grit in predicting Well-Being is greater when an individual’s likelihood of obtaining intrinsic goals is lower (and likelihood of obtaining extrinsic goals is higher).
Figure 5. Path diagram representing a direct relationship between Grit and
Well-Being, Grit and Ill-Well-Being, IVO and Well-Well-Being, IVO and Ill-Well-Being, ILO and Well-being, ILO and Being, IVO moderation effect on Well-Being, IVO moderation effect on Ill-Being, ILO moderation effect on Well-Being and ILO moderation effect on Ill-Being.
Note. All parameters are standardized. Abbreviations: SWLS = Satisfaction With Life Scale; PWB = Psychological Well-Being; HADS-Anx = HADS anxiety subscale; PHQ
= PHQ-15 score; P_pos = PANAS positive affect subscale; P_neg = PANAS
negative affect subscale; IVO = intrinsic value orientation; ILO = intrinsic likelihood orientation.
3.5.4 SEM Group Moderation
To provide another way of checking the data for moderation effects, a comparison was conducted between the upper and lower quartiles of IVO and ILO using the step 1 model (Figure 3).
Before testing structural group moderation, the measurement model was
investigated for configural and metric invariance. The IVO group comparison was found to have acceptable configural invariance (SRMR = .05, CFI = .95 and RMSEA = .09) as well as metric invariance between the unconstrained and fully-constrained model (χ2 (6) = 9.26, p = .16). This suggests that there are no significant differences between the
indicator pathways to the latent factors between the IVO groups. The ILO group model was found to have a Heywood case (Brown, 2006) where the standardized estimate was over 1. This required two indicator regression weights to be constrained to be equal to reduce error (see Appendix D). Once this was done the model was found to have poor but acceptable configural invariance (SRMR = .085, CFI = .92 and RMSEA = .09), however was not found to have metric invariance (χ2 (3) = 17.22, p < .001). This suggests that there are systematic differences in the indicator pathways to the latent factors depending on difference ILO group. By looking at the data, it appears that this was due to differences in the loadings of PWB on Well-Being. However, partial invariance can be assumed given that only one indicator was found to be different (Vandenberg & Lance, 2000).
In the High IVO model, Grit and Age accounts for 24% of Well-Being variance and 17% of Ill-Being variance. The model indicates a direct association of Grit on Well-Being of .51 and also on Ill-Well-Being of -.42. In the low IVO model, Grit and Age accounts for 37% of Well-Being variance and 20% of Ill-Being variance. This model indicates a direct association of Grit on Well-Being of .62 and on Ill-Being of -.45.
Critical ratios for differences (z-scores) were created to compare the results of the two models, by dividing the difference between two path estimates by an estimate of the standard error of the difference. This provides a method to identify significant differences on each path of interest, and compare scores between groups. This is a statistically robust alternative to the chi-squared difference test (Byrne 2010), however is easier to calculate and results in less human error (Gaskin, 2012). The critical ratio is significant at the p = 0.01 level if it is greater than 2.58, and at the p = 0.05 level if greater than 1.96 (Field, 2009; Hopwood, 2007). The analysis found a significant difference in the relationship of Grit and Well-Being between the Low and high IVO groups (CR difference = 2.33, p < .01). By looking at the regression weights, it is possible to
conclude that this score represents a decrease in the prediction of Well-Being in the high IVO group. This suggests that with greater importance placed on intrinsic goals, the relationship between Grit and Well-Being decreases1.
In the High ILO model, Grit and Age accounts for 13% of Well-Being variance and 16% of Ill-Being variance. The model indicates a direct association of Grit on Well-Being of .35 and also on Ill-Well-Being of -.41. In the low IVO model, Grit and Age accounts for 38% of Well-Being variance and 13% of Ill-Being variance. This model indicates a direct association of Grit on Well-Being of .61 and on Ill-Being of -.37.
The analysis found that the prediction of Well-Being from Grit changed
depending on the ILO score (CR difference = 3.58, p < .01). This suggests that having a greater likelihood of achieving intrinsic goals, leads to a decrease in the relationship between Grit and Well-Being. This result was the same if using the extra constraints required in the configural invariance test (see Appendix D).
No significant differences in the relationship between Grit and either Well-Being or Ill-Being were found when comparing high IVO and ILO and Low IVO and ILO.
3.5.5 Gender Moderation
Given the significant differences found between gender in the demographic analyses, a moderation based on gender (male and female) was also performed on the step 3 model.
The measurement model was first tested for configural and metric invariance. The model was found to have acceptable configural invariance (SRMR = .056, CFI = .97 and RMSEA = .07) as well as metric invariance between the unconstrained and fully-constrained model (χ2 (6) =10.49, p = .11). This suggests that there are no significant differences in the indicator pathways to the latent factors between males and females.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
1 This group moderation effect was no longer significant when investigating a USA only
In the Male structural model, Age, Grit, IVO, ILO and the interaction variables account for 47% of Well-Being variance and 32% of Ill-Being variance. The path coefficients were all statistically significant (p < .05) apart from IVO-Ill-Being and the interaction variables. The model indicates a direct association of Grit on Well-Being of .40 and also on Ill-Being of -.38. The model indicates an association of IVO on Well-Being of .31. Additionally, it shows an association between ILO and Well-Well-Being of .21 and with Ill-Being of -.28.
In the Female structural model, Age, Grit, IVO, ILO and the interaction variables account for 60% of Well-Being variance and 31% of Ill-Being variance. The path coefficients were all statistically significant (p < .05) apart from IVO to Well-Being and Ill-Being and the interaction variables to Ill-Being. The model indicates a direct association of Grit on Well-Being of .45 and also on Ill-Being of -.30. The model indicates an association of IVO on Well-Being. Additionally, it shows an association between ILO and Well-Being of .51 and with Ill-Being of -.49. Lastly, there is significant moderation effect of IVO on the relationship between Grit and Well-Being of .21 and a significant moderation effect of ILO on the relationship between Grit and Well-Being of -.23.
The group moderation analysis found no significant difference in predictive ability of IVs on Well-Being or Ill-Being.
!
3.6. Additional Analysis
While not specifically hypothesized about, it is possible that Grit may be moderating the relationship between aspiration type and mental health. Therefore upper and lower quartile groups were made for Grit and a comparison in the predictive ability of IVO and ILO on the mental health constructs was investigated. However, no significant
differences in the relationship between aspiration scores and Well-Being or Ill-Being were found when comparing the two models (p > .05).
Finally, a one-way Multivariate Analysis of Variance (MANOVA) was run to determine the interaction effect of Grit and ILO on mean Well-Being and Ill-Being scores. Four groups were created based on participants’ Grit and ILO Scores: Low Grit and Low ILO (N = 26), High Grit and Low ILO (N = 15), Low Grit and High ILO (N = 20), High Grit and High ILO (N = 37). A high score represented than an individual was in the top quartile for that variable. Similarly, a low score represented that an individual was in the bottom quartile for that variable. Well-Being and Ill-Being scores were based on the standardized result created during PCA. There was a statistically significant difference between the groups on the combined dependent variables, F(6, 186) = 17.16, p < .001; Wilks' Λ = .41; partial η2 = .36. One-way Analysis of Variance
(ANOVA) were inspected and found that there was a significant difference between the four groups on Well-Being scores, F(3, 94) = 30.34, p < .001; partial η2 = .49.
Additionally, there was a significant difference in Ill-Being scores, F(3, 94) = 15.82, p < .001; partial η2 = .34. Overall, these analyses suggest that there are systematic differences
in Well-Being and Ill-Being depending on levels of Grit and ILO. To understand the direction of these differences, two sets of post-hoc analyses were run.
To investigate group differences in Well-Being four Benjamini-Hochberg False Discovery Rate (Benjamini & Hochberg, 1995) corrected t-tests were run.
Firstly, differences in Well-Being within individuals scoring high in Grit were compared between the top and bottom quartiles of ILO. This allowed investigation of whether Gritty individuals who are likely to obtain more intrinsic goals relative to extrinsic goals, differed in their level of Well-Being compared to individuals who are more likely to obtain extrinsic goals relative to intrinsic goals. An independent samples t-test revealed significantly greater Well-Being in Gritty individuals in the high ILO group
(M = 0.88, SD = 0.91) compared to the low ILO group (M = 0.02, SD = 0.62), t(50) = -3.93, p < .001, d = 1.11. This suggests that Gritty individuals with a focus on intrinsic goals have the greatest level of Well-Being.
Secondly, a similar comparison was done within scoring low on Grit. Greater Well-Being was found in individuals with a focus on intrinsic goals (M =!0.53, SD = 0.93) compared to extrinsic goals (M = -1.18, SD = 1.08), t(44) = -5.66, p < .001, d = 1.70. This suggests that non-Gritty individuals with a focus on intrinsic goals have the greatest level of Well-Being.
Thirdly, a comparison was made within the top quartile of ILO between individuals who reported having high and low Grit. No significant differences between these groups were found (p > .05), suggesting that for individuals who are likely to achieve a greater balance of intrinsic goals, Well-Being does not differ despite different levels of Grit.
Fourthly, individuals within the bottom quartile of ILO were compared based on whether they had high or low Grit. Greater Well-Being was found in individuals with high levels of Grit (M =!0.02, SD = 0.62) compared to low levels of Grit (M = -1.18, SD = 1.08), t(39) = 3.93, p < .001. This suggests that if an individual is likely to achieve more extrinsic goals, then an increase in Grit leads to a substantial increase in Well-Being
Figure 6. Well-Being comparisons between high and low Grit and likelihood
orientations of intrinsic and extrinsic goals. ***p < .001.
!
A second set of False Discovery Rate corrected t-tests were run to investigate group differences in Ill-Being.
Firstly, differences in Ill-Being within individuals scoring high in Grit were compared between the top and bottom quartiles of ILO. An independent sample’s t-test showed significantly lower Ill-Being in Gritty individuals in the high ILO group (M = -0.78, SD = 0.56) compared to the low ILO group (M = -0.05, SD = 0.93), t(18.27) = 2.85, p = .011, d = -0.95 (Equal variances not assumed). This suggests that Gritty individuals with a focus on intrinsic goals have a lower level of Ill-Being.
Secondly, no significant difference in Ill-Being was found in individuals scoring low in Grit between high ILO (M =!0.12, SD = 0.96) and low ILO (M = 0.76, SD = 1.15), t(44) = 2.01, p = .051, d = -.0.60 This suggests that Ill-Being does not significantly
change depending on the likelihood of individuals obtaining intrinsic or extrinsic goals when those individuals have low Grit.
Thirdly, within individuals scoring high in ILO, there was a significant difference in Ill-Being depending on being high in Grit (M = 0.12, SD = 0.97) or low (M = -0.78, SD = 0.56), t(26.13) = 3.85, p = .001, d = 1.14 (Equal variance not assumed). This indicates that for Gritty individuals, those who are likely to achieve a greater balance of intrinsic goals have decreased levels of Ill-Being.
Fourthly, within individuals scoring low on ILO, there was a significant
difference in Ill-Being between high Grit (M = -0.05, SD =!0.93) and low Grit (M =!0.76, SD = 1.15), t(39) = 2.319, p = .026, d = -0.77 These data illustrate that Ill-Being is significantly greater in participants who have low Grit and are more likely to achieve extrinsic goals.
Figure 7. Ill-Being comparisons between high and low Grit and likelihood orientations
of intrinsic and extrinsic goals. **p < .01 *p < .05
! !
4. Discussion
This study aimed to investigate whether the ratio of intrinsic to extrinsic goals aspired to, and the likelihood of achieving these goals, moderates the relationship between Grit and Well-Being and Ill-Being. Four hypotheses were made based on this aim.
Firstly, it was hypothesised that Grit would be positively predictive of Well-Being and negatively predictive of Ill-Being. The results support this hypothesis. Grit was found to be a significant positive predictor of Well-Being. This follows previous evidence that Grit can predict positive emotions and PWB (Vainio & Daukantaite, 2015; Singh & Jha, 2008; Hill et al., 2014). The current study also shows that Grit is able to predict a higher-order factor of Well-Being. While this study did not investigate the mechanisms
specifically, the pattern supports the idea of Grit acting similarly to the growth
motivation proposed in Organismic Valuing Process (Roger, 1951), moving individuals towards a state of greater PWB. In addition, this study supports the theorization that Gritty individuals are more likely to show positive emotions. It was suggested this may be due to a pre-disposition due to associated personality traits, that building goals was more likely if one had a positive emotional base, and/or that Grit helps to maintain happiness (Hill et al., 2014).
Grit was showed a significantly negative prediction of Ill-Being. This follows previous evidence that high Grit can predict decreases in fear, sadness, anxiety, burnout, depression and suicide ideation (Singh & Jha, 2008; Sheridan, et al., 2015; Kleiman et al., 2013; Salles et al., 2014). The results of this study show that Gritty individuals are less likely to suffer from mental and physical discomfort. One interpretation of this finding is that the increase in mental Well-Being predicted by Grit acts to buffer against
environmental stimuli that lead to Ill-Being. Indeed a number of studies have shown that SWB can buffer against future psychopathology (Suldo & Huebner, 2004; Kelly, Hills, Huebner & McQuillin, 2012). A second alternative is that Grit acts to actively decrease mental health problems and physical ailments. This is feasible as evidence shows that the meaning of life that Grit provides, can act to decrease suicidal ideation (Kleiman et al., 2013). A third alternative is that Grit is associated with a personality profile that is less likely to report problems. Previous evidence has shown that individuals high in Grit often show lower levels of neuroticism, which is a personality trait predicting feelings such as anxiety, anger, moodiness, guilt, and depressed mood (Costa & McCrae, 1980; Duckworth and Quinn, 2009). It is possible therefore, that the decrease in Ill-Being could be due to Gritty individuals not complaining as much as others with similar symptomology. A fourth alternative is that increased Ill-Being leads to differences in Gritty behaviour. Previous research has shown that psychopathology leads to decreases
in persistence and cognitive functioning (Vansteenkiste, Lens, & Deci, 2006). Therefore, Grit may not have an effect on Ill-Being, but increases or decreases as a result of
changing cognitive functioning. Again, this study did not investigate the specific mechanisms that Grit acts on mental Ill-Being, but the pattern of results follow that expected pattern based on this previous research.
The second hypotheses, suggested that placing more importance on intrinsic goals and being more likely to achieve intrinsic goals, would predict greater Well-Being and decreased Ill-Being. Partial support was found for this hypothesis. The likelihood reported to achieve intrinsic goals over extrinsic goals (ILO) was positively predictive of Well-Being and negatively predictive of Ill-Being. The importance reported of achieving intrinsic or extrinsic goals (IVO) was a positive predictor of Well-Being. However, it was not a significant predictor of Ill-Being. Nonetheless, these results show a similar pattern to previous research showing that intrinsic goals can predict greater mental SWB and PWB, and extrinsic goals can predict mental health problems (Kasser and Ryan, 1996; Niemiec et al., 2008). This is theorized to be due to intrinsic aspirations helping fulfill the basic psychological needs of autonomy, mastery and positive social relationships with others (Deci & Ryan, 2000). Whereas a focus on extrinsic goals thwarts an individual obtaining basic psychological needs, which can lead to greater mental health problems and functional impairments.
In the third hypothesis, it was conjectured that the positive relationship between Grit and Well-Being would be moderated by the importance and likelihood individuals place on achieving intrinsic over extrinsic goals. More specifically, that the positive relationship between Grit and Well-Being will be stronger if individuals report placing more importance on intrinsic goals and report themselves to be more likely to achieve intrinsic goals. This hypothesis was partially supported. ILO was found to show a moderating effect between Grit and Well-Being. This effect was also found in the group
analysis comparing individuals who have the highest ILO to those with the lowest. However, the direction of this effect was unexpected. The results show that an increase in the likelihood of obtaining intrinsic goals decreases the relationship between Grit and Well-Being. Therefore, it seems that when individuals are likely to achieve intrinsic goals, the predictive ability of Grit on Well-Being is not as strong.
While IVO was not found to be a significant moderator of the relationship between Grit and Well-Being in the interaction analysis, the group comparison found a significant difference in moderation between high and low IVO score. Similar to the ILO results, individuals who placed a high level of importance on obtaining intrinsic goals, compared to those reporting high levels of importance placed on extrinsic goals, were found to show a significantly decreased relationship between Grit and Well-Being. This again suggests that when individuals place more importance on intrinsic goals over extrinsic goals, the predictive ability of Grit on Well-Being weakens. Therefore, when individuals focus on extrinsic goals, the benefits to Well-Being from being Gritty increase, even despite the goals not being in line with increasing basic psychological needs.
Similarly, the fourth hypothesis, suggested that the relationship between Grit and Ill-Being will be moderated by the importance and likelihood individuals place on
achieving intrinsic over extrinsic goals. Specifically, the negative relationship between Grit and Ill-Being will be stronger if individuals report placing more importance on intrinsic goals and report themselves to be more likely to achieve intrinsic goals. This hypothesis was not supported. No evidence of a moderating effect of either importance or likelihood of obtaining intrinsic or extrinsic goals was found on the relationship between Grit and Ill-Being.
The main proposed interpretation of the moderation results are that an increase in focus towards intrinsic goals can indeed lead to greater mental Well-Being via the
