Identifying characteristics that enable academic achievement and that distinguish individuals who are successful from those who are not, setting intellectual capacity aside, remains a worthy pursuit for educational research and practice. One such characteristic is resilience. Resilience is defined by Masten et al. (1990, p. 426) as “the process of, capacity for, or outcome of successful adaptation despite challenging or threatening circumstances,” and by Riley and Masten (2005, p. 13) as “referring to patterns of positive adaptation in the face of adversity.”

On the basis that judgements about risk and adversity and evaluations of competencies and outcomes all relate directly to specific events occurring in specific contexts—in a similar way to self-efficacy beliefs (Bandura, 1997)—the existence and relevance of a unidimensional generalized global resilience construct has been questioned in favor of a multidimensional context-specific approach to resilience (Liddle, 1994; Waxman et al., 2003; Riley and Masten, 2005). As such, academic [sometimes educational] resilience has emerged as a context-specific form of individual psychological resilience and, as argued by Colp and Nordstokke (2014), was created to offer greater assessment and prediction specificity to resilience research. Closely related to individual psychological resilience, which examines the capacity for dealing with challenge and adversity, academic resilience is concerned primarily with the relevance of resilience in educational contexts and is defined as “a capacity to overcome acute and/or chronic adversity that is seen as a major threat to a student’s educational development” (Martin, 2013, p. 488). Discussing academic resilience, Martin and Marsh (2006) note that whilst there are many students who perform poorly and continue to perform poorly, there are a significant number of others who manage to turn around their academic misfortunes, flourishing and thriving despite adversity. An often-cited adversity that affects academic achievement is poverty (Kanevsky et al., 2008), and it is the capacity of some children to overcome the limitations of poverty and to succeed when others do not (Gizir, 2004), that illustrates the existence of individual resilience and underlines its importance as a psychological construct. In an academic context, resilience is characterized by those students that present with the capacity to reverse academic misfortune and failure and succeed while others continue to perform poorly and fail (Martin and Marsh, 2006).

Resilience is—universally—considered a strength or asset, a desirable and advantageous quality, characteristic or process that is likely to impact positively on aspects of an individual’s performance, achievement, health and wellbeing (Bartley et al., 2010). Martin and Marsh (2009), for instance, refer to their approach to academic resilience as an inherently asset-orientated, strength-based and aspirational approach to students’ response to academic adversity and the benefits of academic resilience are demonstrated by studies that report resilience as a significant predictor of coping at university (McLafferty et al., 2012), that report a positive relationship between academic resilience and academic achievement (Fallon, 2010), that suggest the potential to foster increased resilience through interventions (Gardynik, 2008) and that report the positive effects of educational interventions that incorporate aspects of academic resilience (Martin and Marsh, 2008). Martin and Marsh (2006) have suggested that that all students, at some point, will experience poor performance, challenge or pressure. Citing the work of Topham and Moller (2011), along with increasing suicide figures among university students (Office for National Statistics, 2013), Cheng and Catling (2015) suggest that university students have an increased vulnerability to mental illness that implies low resilience in coping with academic stress and change. Considered together, these points help to illustrate the continued relevance of academic resilience and its value as a desirable characteristic in students.

In accepting the value and relevance of resilience, we are faced with the task of capturing its essence in a reliable and valid construct measure. As with many latent psychological constructs, measurement of resilience has inevitably involved psychometric scales. Examples of notable resilience scales include Wagnild and Young’s (1993) Resilience Scale, the Connor-Davidson Resilience Scale (Connor and Davidson, 2003), the Resilience Scale for Adults (Friborg et al., 2003) and the Brief Resilience Scale (Smith et al., 2008). Each of these scales presents respondents with attitudinal statements constructed according to characteristics commonly associated with resilience. These include personal and social competence, acceptance of self and life, self-esteem, action-orientation, adaptability, goal-orientated strategies, problem solving, social support and family coherence, personal structure, sense of humor, endurance, and optimism and relate to the key resilience categories of dispositional attributes, family cohesion, and external supporting systems (Hoge et al., 2007). The emergent factor structure of these scales can provide further insight in to the composite elements of resilience. The Connor-Davidson Resilience Scale, for example, has been reported as yielding a five factor structure: personal competence, high standards and tenacity; trust in one’s instincts, tolerance of negative affect and strengthening effect of stress/stress-related growth; positive acceptance of change and secure personal relationships; personal control; and spiritual influences/spiritual orientation to the future, with personal competence, high standards and tenacity identified as the factor accounting for the largest proportion of variance (Connor and Davidson, 2003; see also Lamond et al., 2009).

The reported factor structure for the Connor-Davidson Resilience Scale is not fully consistent (Campbell-Sills and Stein, 2007; Lamond et al., 2009; Green et al., 2014), and with the hardiness characteristics of commitment, control and challenge also reported as features of resilience, Hoge et al. (2007) notes that despite the development of a number of scales purporting to measure resilience, there exists little consensus regarding which of these scales best captures and quantifies the construct of resilience. Lamond et al. (2009) also notes the lack of consensus on the construct definition of resilience and Hoge et al. (2007) go on to suggest that this lack of consensus is indicative of an inherent difficulty in defining the ‘notion’ of resilience; the measurement of academic resilience does not transcend this ‘inherent difficulty.’

Whilst interest in the field continues (e.g., Ricketts et al., 2015; Edwards et al., 2016), research specifically focussing on academic resilience is limited, and advances in terms of defining the construct and its associated predictive factors has been slow (Martin, 2002; Martin et al., 2010). Such advances, it is argued, are a necessary precondition for subsequent advances in construct measurement and this may, in part, explain the lack of available standardized measures of academic resilience (Cassidy, 2015). Although reference to standardized context-specific measures are rare, one popular measure—often cited in the related literature—used to measure academic resilience was presented by Martin and Marsh (2006) in their study examining educational correlates of resilience in high school children. Comprising just six items, this academic resilience subscale asks students to rate their ability to deal with setbacks, challenge, adversity and pressure in an academic setting. Individual items refer specifically to mental toughness, study stress, bouncing back from a poor mark, dealing with schoolwork pressures, confidence and dealing with such setbacks as bad marks and negative feedback. Although a brief attitudinal scale developed for use with school children, the Martin and Marsh scale arguably remains the prevalent measure of academic resilience currently available.

Questioning the validity of resilience measures that capture only state characteristics or positive attitudes or mood, and referring to a definition of resilience that includes reference to a response to some specific event or situation, Hoge et al. (2007) argue that resilience should in fact be measured by observing individuals during a stressful experience and assessing how well they return to normal functioning (i.e., capacity for ‘bouncing back’), “a true resilience scale measures an individual’s reaction to an experimental stress paradigm.” (Hoge et al., 2007, p. 147). Friedland (2005) also emphasizes the importance of behavioral responses (in addition to attitudinal measures) in the measurement of resilience, yet the majority of studies examining academic resilience employ generalized attitudinal response scales, mainly with samples of school children. Thus, there is an apparent lack of suitably developed standardized construct measures available to investigate academic resilience, particularly in samples of university students (Khalaf, 2014).

Waxman et al. (2003) have described resilience as referring to factors and processes that limit negative behaviors associated with stress and result in adaptive outcomes in the presence of adversity, while Morales (2008), citing McGubbin (2001), notes the discussion of resilience that considers whether it should be characterized in terms of an outcome (e.g., academic success, good grades) or process (e.g., protective factors such as a strong work ethic that helps mitigate risk and adversity). The present study offers an alternative process-based measure of academic resilience, focusing on adaptive and non-adaptive cognitive-affective and behavioral responses to academic adversity. Recognizing the need for significant adversity and adequate adaptability in the face of such adversity (Riley and Masten, 2005; Hoge et al., 2007), the 30 item Academic Resilience Scale (ARS-30) developed in the present study measures the responses of university students to a hypothetical, but authentic, academic adversity case vignette. The vignette was developed to portray adversity in an educational context, allowing students to respond in an adaptive or non-adaptive manner, thus providing a measure of academic resilience that is based on responses to a specific instance of academic adversity in a similar way to other resilience scales developed to capture resilience responses to specific life events (e.g., Hardy et al., 2004).

Self-regulated learning (SRL) has been conceptualized as the way in which learners control their thoughts, feelings and actions in order to achieve academically (Zimmerman and Schunk, 2001) and self-efficacy as “people’s judgments of their capabilities to organize and execute courses of action required to attain designated types of performances” (Bandura, 1997, p. 391). Both concepts are salient features of the literature related to academic resilience. Martin and Marsh (2006) for example refer closely to these concepts when proposing their 5-C model of academic resilience: confidence (self-efficacy), commitment (persistence), coordination (planning), control (how hard work and effective strategies impact achievement) and composure (low anxiety), and Newman (2002) and Sautelle et al. (2015) reported an association between high self-regulation (including adaptive help-seeking) and resilience. As such, the items comprising the ARS-30 were selected to reflect the conceptual areas of self-efficacy and self-regulation together with the range of attributes, characteristics and factors commonly associated with resilience.

A principal feature of resilience is the capacity to bounce back [from adversity], to recover and restore previous, pre-adversity, level normal functioning (Smith et al., 2008). In combining the academic adversity vignette with the presentation of associated adaptive and non-adaptive cognitive-affective and behavioral responses, it is suggested that the ARS-30 encompasses, to some degree, both the stressful experience and the quantification of the capacity to return to normal functioning posited as necessary components of a true measure of [academic] resilience (Friedland, 2005; Hoge et al., 2007). It is argued that the adversity vignette represents the critical incident from which it is necessary to recover and the degree to which adaptive responses are selected [over non-adaptive responses] confers capacity for ‘bounce back.’ The aim of the present study is to assess, evaluate and report the psychometric properties of the ARS-30 and consider its potential as a valid and reliable construct measure of academic resilience in university students.

The sample consisted of 532 British undergraduate university students (mean age 22.4 years, SD = 6.2). The main analysis was conducted using a sub-group of 321 participants who completed the original vignette version of the ARS-30 (Table 1); the remaining participants (n = 211) completed the alternative vignette version of the ARS-30 that was used to assess discriminant validity [see Section The Academic Resilience Scale-30 (ARS-30)]. Participants were randomly allocated to sub-groups. The nature of the study—questionnaire development—required that a self-report questionnaire-based design with correlational and between-subjects components was employed. Academic resilience and academic self-efficacy measures were completed during a single data collection point at which time participants’ gender, age and year of study data were also recorded. Though females were overrepresented in the sample, introducing potential bias in to the data, this imbalance has been reported as representative of typical undergraduate intakes in a number of disciplines including psychology, education, subjects allied to medicine, social science, creative arts and design, veterinary science, languages and law (Bourne, 2014; Hillman and Robinson, 2016) and reflects the growing trend for females to outnumber males on two thirds of university courses (Universities and Colleges Admissions Service [UCAS], 2016) and in five sixths of higher education institutions (Hillman and Robinson, 2016).

Once informed consent had been obtained participants completed the GASE and ARS-30 together with a demographics questionnaire recording age, gender, and year of study. A sub-group of the sample (n = 211) completed the ASR-30 following exposure to the alternative vignette modified to depict academic adversity experienced by a fellow student instead of being personally experienced by the participant [see Section The Academic Resilience Scale-30 (ARS-30)]. The remainder of the sample (n = 321) completed the ARS-30 on the basis of the original—personal adversity—vignette [see Section The Academic Resilience Scale-30 (ARS-30)]. Participants were randomly assigned to the sub-groups exposed to either the alternative or original vignette. Data collection was anonymous in order to improve the validity of responses.

The study was carried out in accordance with the recommendations of both the British Psychological Society Code of Ethics and Conduct and the Research, Innovation and Academic Engagement Ethical Approval Panel, University of Salford with written informed consent from all participants in accordance with the Declaration of Helsinki.

Table 2 presents mean scores and standard deviations for each of the 30 items of the ARS-30, along with the global academic resilience score, based on the responses of a sample of undergraduate students (n = 321). For all items, a higher score (range 1–5) indicates greater agreement with the statement. The global ARS-30 score represents the summation of responses to the 30 individual items, with a higher global score (theoretical range 30–150) reflecting greater academic resilience.

The scree plot presented in Figure 1 indicates that there are three meaningful factors for extraction, including factor 3 located at the point of inflection (Cattell, 1966; Field, 2013). Initial retention of three factors is supported by the total variance of 42.4% accounted for by the three factors: 27, 9.1, and 5.5%, respectively. Given that the sample size exceeds 200 the scree plot is considered a reliable basis for factor selection (Stevens, 2002; Field, 2013). Equally, retention of the three factors meets Kaiser’s (1960) criterion for retaining factors with eigenvalues greater than 1(factor 1 = 8.359, factor 2 = 2.716, factor 3 = 1.644), although Kaiser’s criteria was not the primary criterion for selection given the suggestion that this can overestimate the number of factors to be retained (Field, 2013). Sampling adequacy was verified by KMO = 0.897 (Kaiser, 1970; Hutcheson and Sofroniou, 1999, as cited in Field, 2013). Bartlett’s test of sphericity (χ² = 3457.39, df = 435, p < 0.001) and determinant of R-matrix > 0.00001 indicate that inter-variable correlations are suitable for factor analysis (Field, 2013).

Table 3 shows factor loadings after maximum likelihood oblique (promax) rotation, with loading of 0.3 and above in bold (Field, 2013). Item clustering suggests that factor 1 (items 1, 2, 3, 4, 5, 8, 9, 10, 11, 13, 15, 16, 17, and 30) represents perseverance, factor 2 (items 18, 20, 21, 22, 24, 25, 26, 27, and 29) reflecting and adaptive help-seeking and factor 3 (items 6, 7, 12, 14, 19, 23, and 28) negative affect and emotional response.

The inter factor correlations presented in Table 4 indicate medium to large positive correlations between the three factors. Increased perseverance was associated with increased reflecting and adaptive help-seeking (r = 0.71) and increased avoidance of negative affective and emotional response (r = 0.45); increased reflecting and adaptive help-seeking was associated with increased avoidance of negative affective and emotional response (r = 0.39).

Table 5 presents mean scores and standard deviations for each of the three factors of the ARS-30, perseverance, reflecting and adaptive help-seeking, and negative affect and emotional response based on the responses from a sample of undergraduate students (n = 321). For each factor, the factor score represents the summation of responses to the individual items loading highest on that factor (Table 3), with higher scores reflecting more adaptive responses for each factor.

Item-scale analysis is presented in Table 6 for the ARS-30. Cronbach’s α of 0.90 indicated high internal consistency reliability for the global scale (i.e., summation of the 30 items). All item-total correlations were above 0.3 (Field, 2013) with the exception of items 1 (0.14) and 14 (0.12); as deletion of these items does not increase the overall Cronbach’s α it is suggested that all items are contributing positively to the scale’s reliability, supporting the case for retaining these items (Field, 2013). Cronbach’s α was also acceptable for each of the three retained factors: factor 1 α = 0.83; factor 2 α = 0.78; and factor 3 α = 0.80. Item-total correlations ranged between 0.41 and 0.63 for factor 1(with the exception of item 1 = 0.11), between 0.37 and 0.65 for factor 2 (with the exception of item 29 = 0.15), and between 0.45 and 0.65 for factor 3. ‘Alpha if item deleted’ results indicate that deleting items 1 and 29 would increase—marginally—the reliability of factors 1 (by 0.017) and 2 (by 0.031) respectively, potentially raising questions regarding the retention of these items, at least in instances where factor-based subscales are to be utilized in the application of the ARS-30.

Higher global academic resilience scores were associated with increased academic self-efficacy (r = 0.49, N = 319, p < 0.01) and increased age (r = 0.20, N = 317, p < 0.01; Table 7). That separate independent analyses of the factors did not result in higher correlations with academic self-efficacy than analysis of the global score may indicate greater utility of the ARS-30 as a unidimensional measure, with less emphasis on the scale’s multidimensional properties unless these there is clear focus on these in the scale’s application (Sánchez-López and Dresch, 2008). Additionally, robust correlations between factors (reported in Table 4) indicate the existence of a shared psychological variable common across factors, so that factor scores can be meaningfully combined to form a global academic resilience score (Furr, 2011).

Although there were mean differences in academic resilience scores between male and female students and between first, second and third year students (Table 8), these differences did not reach statistical significance (p > 0.05). Whilst unequal sample sizes, as seen here, may increase the potential risk of errors, the risk is reduced provided, as is the case here, the homogeneity of variance assumption is met. Nevertheless, inferential findings related to gender and year of study inter-group comparisons should be interpreted with caution.

When the ARS-30 was completed in response to an alternative form of the original adversity vignette, varied to describe adversity experienced by a fellow student [see Section The Academic Resilience Scale-30 (ARS-30)], significant differences in mean global ARS-30 scores emerged with large effect size (Cohen, 1988) (t = 11.27, df = 525, p < 0.001, d = 0.98), providing evidence supporting the scale’s discriminant validity (Table 9). A comparison of mean GASE scores across the two vignette groups (Original Group M = 145.78 [SD = 19.3], Alternative Group M = 146.37 [SD = 19.4] did not reveal a significant difference (t = 0.341, df = 529, p > 0.05)), indicating that differences in ARS-30 scores resulted from manipulation of the vignette and not group differences in academic self-efficacy.

Exploratory factor analysis was conducted to investigate the factor structure of the ARS-30. Three factors emerged: factor 1, interpreted as perseverance; factor 2, interpreted as reflecting and adaptive-help-seeking; and factor 3, interpreted as negative affect and emotional response. The emerging factors accounted for a total of 42.4% of variance in academic resilience scores and resemble factors previously reported in studies focussing on the measurement of resilience (e.g., Connor and Davidson, 2003; Hoge et al., 2007; Lamond et al., 2009) as well as reflecting aspects of self-regulation and self-efficacy. The most important factor was perseverance, accounting for 27% of variance. This was followed by reflecting and adaptive-help-seeking, accounting for 9.1% of variance, and finally negative affect and emotional response accounting for 5.5% of variance. Some authors have suggested that if the largest emerging factor accounts for three times the variance of that of the subsequent factor, the construct measure can be considered unidimensional (Gorsuch, 1983). Equally, robust correlations between the dimensions, as reported here for the ARS-30, mean that dimension scores can be combined to represent a meaningful unitary global academic resilience score (Furr, 2011). However, as the emerging factors reflect previously identified and meaningful aspects of resilience (e.g., Wagnild and Young, 1993; Connor and Davidson, 2003; Martin and Marsh, 2006), and because the primary purpose of the ARS-30 is to facilitate interventions aimed at building academic resilience (Cassidy, 2015), it is suggested that there remains significant value, dependent up on the intended application, in utilizing the multidimensionality of the scale. A similar approach to dimensionality has been suggested for other psychometric instruments such as the Multidimensionality Self-Esteem Inventory (O’Brien and Epstein, 1988) which Furr (2011) report as assessing several correlated dimensions of self-esteem which can be both scored separately or combined to form a total social self-esteem score.

Factor 1, perseverance, includes items featuring hard work and trying, not giving up, sticking to plans and goals, accepting and utilizing feedback, imaginative problem solving and treating adversity as an opportunity to meet challenges and improve as central themes. There are clear parallels between this factor and factors previously identified, including perseverance (involving persistence despite adversity, willingness to continue to struggle and to practice self-discipline, Wagnild and Young, 1993), personal control and tenacity (Connor and Davidson, 2003), commitment and control (i.e., persistence, hard work and effective strategies, Martin and Marsh, 2006), and personal control and goal orientation (Lamond et al., 2009). Items loading on factor 2, reflecting and adaptive-help-seeking, features themes including reflecting on strengths and weakness, altering approaches to study, seeking help, support and encouragement, monitoring effort and achievements and administering rewards and punishments. As with factor 1, there are evident parallels between factor 2 and previously reported factors including self-reliance (belief in one’s capabilities and recognizing personal strengths and limitations) reported by Wagnild and Young (1993), adaptability reported by Lamond et al. (2009) and adaptive help-seeking reported by Newman (2002). Finally, factor 3, negative affect and emotional response features themes including anxiety, catastrophising, avoiding negative emotional responses, optimism and hopelessness and is similar to acceptance of negative affect reported by Connor and Davidson (2003) and Lamond et al. (2009), composure (low anxiety) reported by Martin and Marsh (2006) and meaningfulness (the belief that one has purpose in life and something to live for) reported by Wagnild and Young (1993).

Each of the emerging factors represents common features evident in existing research studies investigating resilience, with clear similarities and overlaps with concepts and constructs identified as relevant in previous studies of general and context-specific resilience. Thus, the emerging factor structure, and the degree to which it relates to accepted theoretical definitions and relevant constituents of resilience, supports the construct validity of the ASR-30 and the notion of academic resilience as a context-specific—multidimensional—resilience construct.

Item analysis presented convincing evidence for the internal consistency reliability of the scale, with the reported Cronbach’s alpha of 0.90 exceeding levels normally considered acceptable (Cronbach, 1990; Field, 2013). Equally acceptable alphas between 0.78 and 0.83 were reported for factor level reliability analysis. Low item-total correlations did raise doubts regarding the functioning of three items (items, 1, 14, and 29). However, as deletion of these items did not raise the overall reliability of the global scale, and did so only marginally at factor level, they were retained on the basis that all items contributed positively to the internal reliability of the scale (Field, 2013).

Previous studies have reported significant associations between resilience and theoretically relevant constructs including self-efficacy (Hamill, 2003; Martin and Marsh, 2006, 2008). The significant positive correlation between ARS-30 scores and academic self-efficacy (r = 0.49) reported in the present study serves to demonstrate the concurrent validity of the scale. The discriminant validity of the scale was supported by significant mean differences and large effect size (d = 0.98) in ASR-30 responses to two independent versions of the academic adversity vignette (p < 0.001), which was not explained by group differences in academic self-efficacy (p > 0.05). Findings from previous studies examining the relationship between resilience and age, gender and experience have been mixed (e.g., Martin and Marsh, 2006, 2008; Allan et al., 2014; Khalaf, 2014). As such, the weak but significant negative correlation between ARS-30 scores and age (r = 0.20), along with small but non-significant (p > 0.05) gender and experience differences in mean ARS-30 scores reported in the present study offer no clearly interpretable additional evidence regarding the validity of the ARS-30. That scale items were selected to reflect generally accepted definitions, theoretical understandings and factors, constructs and attributes commonly associated with resilience is presented as evidence of the Scale’s content validity (Wagnild and Young, 1993).