Summary The idea of a 'growth mindset', or the belief - UCL (PDF)

Doctorate in Educational and Child Psychology Trudy Kearney Case Study 1: An Evidence-Based Practice Review Report Theme: School Based Interventions for Learning Are school-based interventions that promote a growth mindset effective in raising pupil attainment? Summary The idea of a ‘growth mindset’, or the belief that intelligence is something we can change, is becoming increasingly popular in education due to research by Carol Dweck. Dweck’s findings suggest that having a growth mindset is associated with better academic achievement, and that it is possible for students to change their mindset to this more adaptive way-of-thinking. Research has begun to combine these two aspects and evaluate whether interventions to change a student’s mindset from fixed to growth can have an impact on their academic attainment. A systematic review of the literature was carried out to establish whether these interventions are effective in schools. Five studies were included and evaluated in relation to their methodological quality, methodological relevance and topic relevance to the review question. The review found some evidence to suggest that interventions that promote a growth mindset are effective at raising pupil attainment; however, this was predominantly true for either pupils identified as at risk of educational disadvantage, or pupils who held a fixed mindset prior to the intervention. Further research areas are identified based on limitations of the design of studies and gaps in participant demographics that could increase the generalisability of findings. Introduction What is a ‘growth mindset’?  Doctorate in Educational and Child Psychology Trudy Kearney 1 The phrase ‘growth mindset’ developed from Carol Dweck’s research into self- theories and motivation. Dweck (2000) argues that how people view intelligence can be classified into two main types. The first is referred to as a ‘fixed mindset’ or entity theory, where intelligence is viewed as something that is inherent and fixed, and little can be done to change it. The opposite is termed a ‘growth mindset’ or incremental theory, where a person believes that intelligence is changeable and something that can develop through effort and perseverance. Growth Mindset Interventions As the aim of the intervention is to promote a new way of thinking, there are many different ways in which this has been implemented. Previous research has involved a teaching session, followed by a short activity where a student writes to another student outlining the theory they have learnt (Paunesku et al., 2015). Interventions similar to this have ranged from 25 minutes (Yeager et al., 2014) to almost one academic year (Good, Aronson & Inzlicht, 2003). There is at least one commercially available intervention based on implicit theories of intelligence, Brainology®, which was developed by Dweck and her team. Brainology® is an online programme that teaches children to see intelligence as malleable using material based on neuroscience and what happens to the brain when we learn. The programme consists of approximately 2.5 hours of online sessions, which can be combined with up to 10 hours of classroom-based activities. There are also free resources available online, such as Mindset Kit (https://www.mindsetkit.org/), which provide structured sessions for teachers, parents or senior management teams to implement with children and staff.  Doctorate in Educational and Child Psychology Trudy Kearney 2 Psychological Theory The theory behind growth mindsets originally stemmed from research into how children respond to failure. Diener and Dweck (1978; 1980) identified two main types of responses: one termed ‘helpless’ and the other ‘mastery-oriented’. Diener and Dweck (1978; 1980) found that children with a helpless response viewed the failure as out of their control and unchangeable, whereas those who were mastery-oriented persisted despite the failure and focussed on mastering a task. Dweck and Leggett (1988) extended this further by linking the helpless and mastery-oriented responses to the goals that a child pursued, and how their implicit beliefs can influence the type of goal they prefer. Dweck and Leggett (1988) suggested that a child who responds to failure in a helpless manner is likely to prefer performance goals (where the aim is to gain judgement about their current competence), and hold an entity theory of intelligence. On the other hand, a mastery-oriented child would be more likely to pursue learning goals (where the aim is to increase competence) and hold an incremental theory of intelligence. Research has supported this theory and found that implicit beliefs about intelligence predicted a pupil’s response to failure and goals pursued (Robins & Pals, 2002), and, importantly, predicted their grades (Blackwell, Trzesniewski & Dweck, 2007), with incremental theorists scoring higher than entity theorists. Considering educational psychology practice, if it is possible to change a child’s implicit belief to that of a growth mindset, there are clear benefits that this could have for pupil motivation and academic achievement. Interventions that promote a growth mindset could be implemented universally across schools to encourage this way of thinking in all pupils. Considering the nature of the intervention, educational  Doctorate in Educational and Child Psychology Trudy Kearney 3 psychologists would be in the ideal position to train and support schools with its implementation. Rationale for Review The increased focus on accountability and the need to apply evidence-based practice in schools (DfE & DH, 2015) means that school staff need to have an understanding of the research behind interventions and a clear rationale for their implementation with specific pupils. Hattie (2008) also suggests that it is no longer enough to rely on ‘what works’ but that educators need to be able to compare different interventions to know which is likely to be most effective and why. There is a need, therefore, for clear unbiased reviews of school-based interventions that enable comparisons to be made between them. The effect of growth mindset interventions on academic attainment has been positively reviewed in literature focussing on wider topics, such as psychosocial interventions in general (Spitzer & Aronson, 2015; Yeager & Walton, 2011), but as yet there has been no systematic review into the effectiveness of these interventions on pupil attainment. A systematic review was considered necessary for the following reasons. Firstly, while Yeager and Walton (2011) approached their search systematically, only specific journals were searched and many studies may have been missed. Secondly, the authors of the reviews had also written at least one of the studies included in the review. A systematic review would remove any subjectivity that could occur due to a conflict of interest. Thirdly, past reviews have included studies completed with adult students from colleges and Universities (e.g. Aronson, Fried & Good, 2002) and the findings, therefore, are less relevant to school-age children.  Doctorate in Educational and Child Psychology Trudy Kearney 4 The current review aims to establish whether interventions that stem from this psychological theory are having the positive impact on children’s attainment as hypothesised in the literature. Due to the wide variation in how these interventions can be implemented, and a dearth of research into newly-produced manualised interventions, the review will include any school-based intervention that promotes a growth mindset, regardless of how and by whom it was implemented. Review Question Are school-based interventions that promote a growth mindset effective in raising pupil attainment? Critical Review of the Evidence Base A comprehensive literature search of the following electronic databases was carried out between 18th December 2015 and 2nd January 2016:  PsycINFO  PsycEXTRA  PsycARTICLES  ERIC (EBSCO)  ERIC (ProQuest)  Web of Science Search terms were created to find studies that met the following conditions:  The research design is experimental and consisted of an intervention condition.  The intervention was based on Carol Dweck’s theory of growth mindset.  Doctorate in Educational and Child Psychology Trudy Kearney 5  The intervention took place in a school and the participants were school-age children. The exact search terms used are presented below in Table 1. Table 1 Search Terms Entered into Databases. 1 2 3 implicit intelligen* OR malleable intelligen* OR growth mindset OR fixed mindset OR entity theory OR incremental theory AND intervention* OR experiment* AND school* OR child* OR educat* OR pupil* OR student* OR Note. The asterix means that any word that contains all letters before the asterix will be included. For example, intelligen* would encompass intelligent and intelligence. If available, a filter was applied so that only peer-reviewed studies were included in the search. The search yielded 732 studies in total, 318 of which were found to be duplicates. The research listed on the Brainology® website (http://www.mindsetworks.com/) was also screened on 2nd January 2016 for any additional studies that could be included; however, only duplicates were found. This meant 414 studies were title screened adhering to the inclusion and exclusion criteria outlined in Table 2. A further 334 studies were removed during title screening, leaving 80 requiring abstract screening applying the same criteria. 70 studies were removed during abstract screening, leaving 10 that would have been appropriate for full-text screening. Despite all abstracts being available in English, three of these studies, however, were not published in English and were, therefore, removed from the review as full-text screening was not possible.  Doctorate in Educational and Child Psychology Trudy Kearney 6 Table 2 Inclusion and Exclusion Criteria Inclusion Criteria Exclusion Criteria Rationale Type of publication Peer-reviewed research only Non peer-reviewed research Increased credibility and rigor of research process Type of study Study included an intervention Study contained no intervention Research question refers to intervention effectiveness Psychological theory of intervention Intervention refers to implicit theories and a growth mindset Any type of intervention that does not directly refer to implicit theories or a growth mindset Research question focussing on interventions stemming from this specific psychological theory only Purpose of intervention Intervention aims to promote a growth mindset in participants Intervention aims to do anything other than promote a growth mindset in participants Research question relates to intervention that promote a growth mindset Outcome measures At least one outcome measure of academic attainment No outcome measure of academic attainment Research question relates to raising pupil attainment Language Article published in English Article published in any language other than English Inability to accurately translate from original language Setting Intervention took place in a school Intervention took place in a setting other than a school Research question concerns effectiveness of school-based interventions Location Research took place in an Organisation for Economic Co- operation and Development Research took place in a non-OECD country Research more able to generalise to UK schools due to social, financial and educational  Doctorate in Educational and Child Psychology Trudy Kearney 7 Inclusion Criteria Exclusion Criteria Rationale (OECD) country similarities. Participants Participants of school age (aged 4-17 at beginning of academic year in which intervention took place) Participants not of school age (below 4 or above 18 at beginning of academic year in which intervention took place) Research question concerns effectiveness of interventions for school-age children Date Published before 2nd January 2016 Published after 2nd January 2016 Final search date before analysis conducted Seven studies were full-text screened using the inclusion and exclusion criteria. Three studies were removed at full-text screening and these are listed alongside the reason for exclusion in Appendix A. This left four articles suitable for inclusion in the review. It is important to note that only Study 2 from Blackwell et al. (2007) was included, and Yeager et al. (2014) included two studies that met the inclusion criteria (Study 2 and Study 3) and these were evaluated separately. This meant that five studies were included from four articles. A flow diagram of the systematic search process is shown in Figure 1. The final five studies are listed in Table 3 and summarised in Appendix B. The studies were evaluated using the Weight of Evidence (WoE) Framework proposed by Harden and Gough (2012). Studies were first evaluated for WoE A, Methodological Quality. An adapted coding protocol for group-based designs from Kratochwill (2003) was used to evaluate the studies on a range of criteria and compare them using scores for seven main areas. The adaptions made to the coding  Doctorate in Educational and Child Psychology Trudy Kearney 8 protocol are listed in Appendix C, and an example coding protocol can be found in Appendix D. The studies were then evaluated for WoE B, Methodological Relevance to the review question, and WoE C, Topic Relevance to the review question. Scores for each WoE were then averaged to provide an overall score, WoE D. Further details and the scores for each WoE are listed in Appendix E. Figure 1 A Flow Diagram of the Literature Search Table 3 List of Studies Included in the Review Full Reference Blackwell, L. S., Trzesniewski, K. H., & Dweck, C. S. (2007). Implicit theories of intelligence predict achievement across an adolescent transition: A longitudinal study and an intervention. Child development, 78(1), 246-263. - Study 2 Database/website search results n = 414 Abstract screened n = 80 Full text screened using inclusion/exclusion criteria n = 7 Studies included in review n = 5 (5 studies from 4 articles) Removed via abstract screen n = 70 Unavailable in English language n = 3 Removed via title screen n = 334 Removed via full-text screen using inclusion/exclusion criteria n = 3  Doctorate in Educational and Child Psychology Trudy Kearney 9 Full Reference Good, C., Aronson, J., & Inzlicht, M. (2003). Improving adolescents' standardized test performance: An intervention to reduce the effects of stereotype threat. Journal of Applied Developmental Psychology, 24(6), 645-662. Paunesku, D., Walton, G. M., Romero, C., Smith, E. N., Yeager, D. S., & Dweck, C. S. (2015). Mind-set interventions are a scalable treatment for academic underachievement. Psychological Science, 26(6), 784–793. Yeager, D. S., Johnson, R., Spitzer, B. J., Trzesniewski, K. H., Powers, J., & Dweck, C. S. (2014). The far-reaching effects of believing people can change: Implicit theories of personality shape stress, health, and achievement during adolescence. Journal of Personality and Social Psychology, 106(6), 867-884. - Study 2 - Study 3 Participants and Setting The review included data from 2,051 students from the USA, ranging from 7th to 12th grade. Each of the studies drew samples from a single school, except for Paunesku et al. (2015), which sampled 13 schools. Demographics, such as gender, ethnicity and low socio-economic status (SES, as identified through free or reduced price school lunch) were reported for most schools and participants came from a range of backgrounds. The percentage of males and females was reported for all studies (for Yeager et al., 2014, see supplementary document) and was roughly equal. The percentage of pupils identified as low SES ranged from 0 to 90% (both in Paunesku et al., 2015). The ethnic majority of schools varied between studies: predominantly Hispanic/Latino (Good et al., 2003; Yeager et al., 2014, Study 3), African American (Blackwell, et al., 2007), Asian (school 1 in Paunesku et al., 2015) and White (multiple schools in Paunesku et al., 2015). Clear information on demographics was incorporated into a study’s score for WoE C. Pupils came from either public, charter or private schools and this is reflected in  Doctorate in Educational and Child Psychology Trudy Kearney 10 scores for WoE A and WoE C, with studies that took place in a private school, or where information was unknown, given less weight than those that took place in state-funded schools. The majority of children who would require intervention attend state-funded schools (Kratochwill, 2003) and, therefore, private schools can be considered less representative of the target population. Several of the studies targeted students identified as ‘at-risk’. In both Yeager et al. (2014) Study 2 and 3, participants were taking a maths class that was considered lower ability than the class taken by most ninth grade pupils. In Blackwell et al. (2007), pupil’s maths grades were at the 35th percentile nationally prior to the intervention, meaning they were relatively low-achieving. This is incorporated into scores for WoE C. Attrition rates were low across studies (less than 20%, as described by Kratochwill, 2003), although a common feature was a lack of information on whether the mortality rate was equal across conditions. This was taken into account when scoring WoE A Comparison Group. Design and Control Groups All studies included a control group. For both Yeager et al. (2014) studies and Blackwell et al. (2007), the control group received a weaker form of the growth mindset intervention. In Blackwell et al. (2007), half the sessions received were the same for both the control and intervention group, with alternative sessions covering memory instead of a growth mindset. In the Yeager et al. (2014) studies, all participants received the same initial session. The experimental group then completed an activity aimed at promoting an incremental theory of personality, whereas the control group activity promoted an incremental theory of athletic ability.  Doctorate in Educational and Child Psychology Trudy Kearney 11 In Good et al. (2003), the control groups received alternative interventions and one group received a combination of both an incremental and alternative intervention. In Paunesku et al. (2015), the control group received materials about a different aspect of neuroscience and other groups received either an alternative intervention or a combination of both the growth mindset and alternative intervention. The differences in control group between studies is factored into scores for WoE A Comparison Group and WoE B. All studies randomly assigned participants to each group. Most studies included an intervention delivered at an individual level so this allowed for easy randomisation of participants. Blackwell et al. (2007) randomly assigned pre-existing classes to each condition, but pupils had been randomly allocated to these classes in the first place. Intervention The method and length of intervention varied between studies. The intervention in Good et al. (2003) consisted of two 90 minutes sessions, followed by weekly email communication with a mentor for two school semesters. Participants created a website for hypothetical students based on the information they were taught on the incremental theory of intelligence. Both Yeager et al. (2014) studies and Paunesku et al. (2015) adopted similar approaches. Participants first read material that promoted an incremental theory and then completed a written activity explaining this information to a hypothetical pupil. This activity lasted 25 minutes in both Yeager et al. (2014) studies and 45 minutes in Paunesku et al. (2015). The activity was completed on a computer in both Yeager et al. (2014) Study 3 and Paunesku et al. (2015). For the former of these, this was so that information could be translated into Spanish if necessary. Blackwell et al. (2007) delivered eight weekly 25 minute  Doctorate in Educational and Child Psychology Trudy Kearney 12 sessions, four of which included reading, activities and discussions relating to an incremental theory of intelligence. The studies that involved a weaker form of the intervention as a control group (Blackwell et al., 2007; Yeager et al., 2014, Study 2 and 3) allow the effectiveness of certain components of the intervention to be tested to some extent; however, all studies failed to adopt a design that allowed for all components of the intervention to be evaluated individually. For example, in Good et al. (2003), it is unknown whether the two initial sessions added any value to the intervention, or whether weekly correspondence for only one semester would have resulted in the same effect. This common flaw is reflected in WoE A Identifiable Components, for which all studies scored 0. In terms of intervention fidelity, several of the interventions (Paunesku et al., 2015; Yeager et al., 2014, Study 2 and 3) were delivered through a computer programme or used predesigned materials, and thus can be considered manualised. Both Yeager et al. (2014) studies also used two coders to check whether participants had given a sufficient answer for the written activity. In Blackwell et al. (2007), the sessions were delivered by undergraduate assistants who received prereading and weekly training sessions prior to delivering each session and met for a debriefing afterwards. Participants were also tested on their knowledge of the intervention content after it had finished, and the intervention group received higher scores than the control. The mentors used in Good et al. (2003) received training on how to teach pupils the required message; however, little else is specified about how the intervention was monitored. As a result of the manualisation and use of coders, both Yeager et al. (2014) studies  Doctorate in Educational and Child Psychology Trudy Kearney 13 and Paunesku et al. (2015) scored higher on WoE A Implementation Fidelity than Blackwell et al. (2007) and Good et al. (2003). Manualised interventions provide a description of the procedure for implementing the intervention. This allows for more accurate replication of the intervention and, therefore, fidelity of implementation will be higher than an intervention not accompanied by descriptions of this type. Measurement Most studies used school grade as the attainment measure. Both Paunesku et al. (2015) and the Yeager et al. (2014) studies used a Grade Point Average (GPA) for all subjects considered ‘core’, which is factored into their WoE C score. Blackwell et al. (2007) only looked at mathematics grades, and Good et al. (2003) used end-of- year mathematics and reading scores assessed using the Texas Assessment of Academic Skills (TAAS), a standardised achievement test. Only Yeager et al. (2014) Study 2 reported a reliability coefficient, which resulted in the higher WoE A Measurement score. All studies except Good et al. (2003) collected attainment scores at multiple time- points, enabling a comparison of pre and post intervention scores. Good et al. (2003) only collected grades post-intervention, and compared these to the control group. Most studies collected grades at one time-point post-intervention and, therefore, failed to include any follow-up. The exception to this is Yeager et al. (2014) Studies 2 and 3, which included grades for multiple time-points post-intervention; however, a separate follow-up was not included in the analysis. The lack of systematic follow-up in any study resulted in all receiving a score of 0 for WoE A Follow-up Assessment. Outcomes and Effect Sizes  Doctorate in Educational and Child Psychology Trudy Kearney 14 Blackwell et al. (2007) and Yeager et al. (2014) Study 2 were the only studies that found statistically significant improvements for their whole sample. It is important to note, however, that participants in both these studies were identified as relatively low-achieving for their age. The effect of intervention on grades for Blackwell et al. (2007) and Yeager et al. (2014) Study 2 was small to medium, and small respectively (Cohen, 1988). Good et al. (2003) found a medium effect on reading scores for all pupils who received the growth mindset intervention; for maths scores, however, a statistically significant effect was only found for females, which was large. Paunesku et al. (2015) found small effects only for those pupils identified as at risk of dropping out of school, which was measured by a low GPA or failure of at least one core subject. The statistical analysis, however, collapsed all interventions into one condition and this is reflected in the weighting criteria for WoE B and C. Yeager et al. (2014) Study 3 found a large effect on GPA but only for pupils who held an entity theory of intelligence prior to the intervention. Blackwell et al. (2007) tested whether pupils who originally held an entity theory benefitted more from the intervention; however, their results were not statistically significant (p < .10) and they suggest that this is due to a small sample size. Further information on effect sizes for statistically significant results can be found in Table 4. Table 4 Effect Sizes and Descriptors for Statistically Significant Findings Study Measure Comparison Effect Size Descriptora WoE D Descriptor Blackwell et al. (2007) Study 2 Maths grade Intervention vs control d = 0.45b Small to medium Medium