Profile of PSI attendees
Racial/ethnic demographics. Figure 2a shows the self-identified racial and ethnic demographic distribution of students who attended at least one PSI session for a PSI-supported course in a given semester, compared to the demographics of all students enrolled in PSI-supported courses. PSI attendees strongly resembled the overall population of students enrolled in PSI-supported courses. However, PSI attendees were slightly more likely to identify as Black/African American or as one of the racial/ethnic groups in the Other category, and slightly less likely to identify as White or Asian, compared to the overall student population. This is also reflected in Table 2, which shows the percentage of students who were enrolled in a PSI-supported course and attended at least one PSI session, separated according to students’ self-identified race/ethnicity. The fact that students identifying with HU groups were slightly more likely to attend PSI is in agreement with other reports (Kudish et al., 2016; McGee, 2005; Moore & LeDee, 2006; Peterfreund et al., 2008). The racial/ethnic demographic distribution for PSI attendees remained very similar when analyzed according to frequency of attendance (Fig. 2b), indicating that frequency of attendance was similar for students identifying with all races/ethnicities.
HS GPA. Table 3 shows the mean HS GPAs for PSI attendees and for all students enrolled in PSI-supported courses, separated by students’ self-identified race/ethnicity. For students who attended PSI sessions, the mean HS GPAs in Table 3 are also broken down as a function of PSI attendance (1–2, 3–5, 6–9, or 10 + sessions). Again, it is important to note that HS GPA is not necessarily being used here as an indicator of student motivation or ability, but instead may be reflective of the different high school environments (funding, teacher/student ratio, available coursework, etc.) experienced by each student, and thus reflects how prepared they were to transition to the college environment. HS GPA is also commonly used as a tool to identify self-selection bias for academic support services, i.e., the idea that only strong students attend SI services (Dawson et al., 2014).
A one-way ANOVA revealed significant effects of race/ethnicity, F (4, 1605) = 37.48, p < 0.0001, on mean HS GPA. Students identifying as Black/African American, Hispanic/Latino, and with one of the races/ethnicities in the Other category had significantly lower mean HS GPAs compared to their peers who identified as White or Asian. This indicates that these students were at the greatest disadvantage when entering college. It is interesting to note that while students identifying with HU groups had the lowest average HS GPAs, they also had the highest rates of PSI attendance (Fig. 2). It is encouraging that the students most in need of academic support were also the most likely to seek it out.
Statistical analysis showed no correlation between HS GPA and PSI attendance for students identifying as Asian, Black/African American, White, or a race/ethnicity in the Other category. Furthermore, no significant difference existed between the mean HS GPA of PSI attendees vs. the overall average for all students enrolled in these courses within each racial/ethnic group. These results suggest that for students identifying with these racial/ethnic groups, college preparedness was not a good predictor of who would attend PSI, or if they did attend, who would attend more frequently. Any differences in final course grades can thus be attributed (at least to some extent) to the effect of attending PSI sessions.
On the other hand, a significant correlation between HS GPA and PSI attendance did exist for students identifying as Hispanic/Latino (r(349) = 0.16, p < 0.01); students identifying as Hispanic/Latino who attended PSI 3 + times per semester had slightly higher (0.15–0.18) mean HS GPAs compared to that of all students identifying as Hispanic/Latino who were enrolled in PSI-supported courses. Based on this, it appears that students identifying as Hispanic/Latino who attended PSI at least 3 times may have been slightly better prepared for college coursework compared to those who attended infrequently or not at all. This indicates that self-selection bias may have been a factor for these students attending PSI. This point will be discussed further in the next section.
Final course grades as a function of PSI attendance and self-identified race/ethnicity
Final course grades were examined for PSI attendees as a function of both PSI attendance and students’ identified race/ethnicity; these results are shown in Fig. 3a–d. Also shown for comparison are the final course grades for all studentsFootnote 3 identifying with each racial/ethnic group who were enrolled in PSI-supported courses. The “Other” category was omitted in this and subsequent analyses due to an insufficiently large sample size for this group, and because this group contained students identifying with a variety of racial/ethnic backgrounds. Students attending 1–2 PSI sessions in a given semester exhibited very similar final course grade distributions and DFW rates compared to those of all enrolled students, regardless of students’ self-identified race/ethnicity. PSI attendees saw increasing AB rates and decreasing DFW rates with increasing PSI attendance; these trends are more clearly seen in Fig. 4a and b. Averaging the data in Fig. 4a and b across all students, there is an average increase in AB rates of 29.0 percentage points, and an average decrease in DFW rates of 26.1 percentage points, for students who attended 10 + vs. 1–2 PSI sessions.
While all students saw improved grades with increasing PSI attendance, marked differences in AB and DFW rates as a function of students’ self-identified race/ethnicity can be seen in Fig. 4a and b. Students identifying as Asian consistently had the highest AB rates, followed by students identifying as White, Hispanic/Latino, and Black/African American, respectively. The reverse trend is seen for DFW rates, except for students attending 10 + sessions. For this group, the DFW rates of students identifying as Asian and White converged (2.9% and 2.1%, respectively), as did the DFW rates for students identifying as Black/African American and Hispanic/Latino (14.8% and 16.3%, respectively). A significant gap exists in the DFW rates of students identifying with HU vs. non-HU groups. Notably, students identifying as Black/African American typically had the lowest AB rates and highest DFW rates. However, when comparing students who attended 1–2 PSI sessions vs. 10 + sessions, this group showed the largest increase in AB rates (from 28.7 to 60.5%) and the largest decrease in DFW rates (from 47.1 to 14.8%). Statistical analysis of the impact of PSI participation on students identifying with different racial/ethnic groups is discussed in more detail below.
Mean final course GPA was also investigated as an overall indicator of successful course completion and used for further statistical analysis. Figure 5 shows the mean final course GPA for PSI attendees as a function of both students’ self-identified race/ethnicity and number of PSI sessions attended. A two-way ANOVA revealed significant effects of both students’ self-identified race/ethnicity (F (3, 1773) = 30.18, p < 0.0001) and PSI attendance (F (3, 1773) = 35.25, p < 0.0001) on final course GPA, though no interaction was found between these two factors. This indicates that attending PSI sessions had a similar effect on overall course GPA for students identifying with all racial/ethnic groups. This was confirmed using a moderator test, which also showed that there was no significant interaction between PSI attendance and students’ self-identified race/ethnicity on course GPA (b = 0.006, SE = 0.023, p = 0.794, CI = (− 0.039, 0.051)). Thus, the effect of attendance on GPA does not depend on students’ self-identified race/ethnicity (i.e., there was no significant moderation effect). However, the moderator test found that the main effect of attendance was significant as it influenced the course GPA (b = 0.300, SE = 0.067, p < 0.001, CI = (0.170, 0.430)), confirming the results from the two-way ANOVA.
Analyzed separately via one-way ANOVAs, students identifying with the four racial/ethnic groups analyzed here all showed a statistically significant increase in final course GPA with increasing PSI attendance. Although there was no statistical difference in the overall effect of attending PSI for students identifying with different racial/ethnic groups in the two-way ANOVA, comparing the results of the one-way ANOVAs, the effect size was largest for students identifying as Black/African American (F (3, 675) = 18.79, p < 0.0001). These students saw an average increase of 1.03 GPA units (i.e., one letter grade) for students attending 10+ vs. 1–2 sessions. Students identifying as Asian (increase of 0.81 GPA units, F (3, 261) = 8.26, p < 0.0001), Hispanic/Latino (increase of 0.90 GPA units, F (3, 397) = 8.04, p < 0.0001), and White (increase of 0.85 GPA units, F (3, 440) = 7.02, p < 0.0001) all had similar gains to each other.
Further analysis also revealed differences in the threshold level of PSI required to see a significant improvement in final course GPA. Tukey post hoc analyses showed that while students identifying as Black/African American and Asian exhibited significant increases in course GPA when attending just 3–5 sessions (p < 0.05), students identifying as Hispanic/Latino and White required at least 6+ sessions to see a significant effect (p < 0.05). At the same time, students identifying as Asian appeared to experience a ceiling effect, seeing no additional benefit from participating in 10 + sessions as compared to 6–9 sessions. Students identifying with all other racial/ethnic groups continued to see increasing benefit with increasing PSI attendance. Overall, as students identifying as Black/African American saw improved final course grades from attending as few as 3–5 PSI sessions, continued to see an increasing benefit with increasing PSI attendance, and exhibited the largest overall increase in their final course grade when attending 10+ vs. 1–2 sessions, they appear to have gained the largest benefits from participating in PSI.
As noted above, the two-way ANOVA also revealed significant effects of students’ self-identified race/ethnicity on final course GPA. In other words, students identifying with different races/ethnicities had significantly different final course GPAs. Students identifying as Black/African American had the lowest overall course GPA (mean course GPA of 1.98 across all attendance groups), followed by students identifying as Hispanic/Latino. Students identifying as Asian consistently had the highest overall course GPA, with a mean course GPA of 2.88 across all attendance groups. These results are consistent with the trends in AB rates and DFW rates seen in Fig. 4, and are seen across all attendance levels; even with frequent PSI attendance (10+ sessions/semester), students identifying as Black/African American and Hispanic/Latino had lower final course GPAs compared to their peers who identified as White and Asian.
One interesting point is that although students identifying as Hispanic/Latino who attended 3 + PSI sessions had similar HS GPAs compared to students identifying as White and Asian (Table 3), they received lower grades and final course GPAs for all levels of PSI attendance (Figs. 3, 4, 5). In fact, while the performance of students identifying as White and Asian essentially converged when attending 10 + PSI sessions, students identifying as Hispanic/Latino in this attendance group instead performed more similarly to students identifying as Black/African American, despite being seemingly better prepared. This removes some concern regarding self-selection bias for PSI attendees identifying as Hispanic/Latino, as self-selection bias should result in higher final course grades than expected, not lower. Moreover, students identifying as Hispanic/Latino who attended 10 + sessions per semester had better course outcomes compared to students who only attended 3–5 sessions, although they had similar HS GPAs. Together this suggests that although PSI attendees identifying as Hispanic/Latino may have been slightly better prepared for college coursework in general (compared to students identifying as Hispanic/Latino who did not attend PSI), they still directly benefitted from attending PSI sessions. Indeed, the fact that they underperformed compared to PSI attendees identifying as White and Asian, despite having similar HS GPAs, indicates that there may be additional factors negatively affecting the performance of students identifying with HU groups that are not considered here.
In summary, PSI attendance was positively correlated with improved course outcomes for students identifying with all racial/ethnic groups. The degree of benefit increased with the number of PSI sessions attended. Frequent PSI attendance resulted in drastically reduced DFW rates, decreasing from an average of 35% for students attending 1–2 sessions to 10% for students attending 10 + sessions (Fig. 4b). Average AB rates simultaneously increased from 46 to 75% (Fig. 4a), and overall final course GPAs increased by roughly one letter grade (Fig. 5). At the same time, the consistent disparity in final grades between students identifying with different racial/ethnic groups indicates that while PSI is a promising method to support student success, it is not able to fully compensate for the fact that students identifying with HU groups enter GGC at an academic disadvantage compared to their non-HU peers, and/or experience additional educational barriers once they are in college.
Interaction between college preparedness, PSI attendance, and race/ethnicity
In our previous work, we demonstrated that PSI is particularly beneficial for underprepared students, as determined by their HS GPA (Achat-Mendes et al., 2020). Since students identifying with different races/ethnicities enter GGC with different average HS GPAs, we first sought to confirm these findings, so that we could further investigate how students identifying with different racial/ethnic groups benefitted from PSI. A two-way ANOVA revealed significant effects of both HS GPA (F (2, 1442) = 72.07; p < 0.0001) and PSI attendance (F (3, 1442) = 18.40; p < 0.0001) on final course GPA for PSI attendees. In addition, the ANOVA revealed a significant interaction between these two factors (F (6, 1442) = 2.20; p < 0.05). The interaction between HS GPA and PSI attendance was confirmed via a second moderator test, in which we examined whether the relationship between the PSI attendance of students and their final course GPA was influenced by their high school GPA (i.e., preparedness). The model showed there was a significant interaction between PSI attendance and HS GPA on final course GPA (b = − 0.135, SE = 0.047, p = 0.004, CI = (− 0.227, − 0.043)). These analyses indicate that while students with lower HS GPAs received lower final course grades in their PSI-supported STEM courses compared to students with higher HS GPAs, they also received a greater benefit from frequent PSI participation. These trends may in part explain why students identifying as Black/African American, who entered GGC with the lowest average HS GPA, experienced the largest benefit from PSI participation. However, it is also possible that PSI offered additional benefits to students identifying with HU groups related to a sense of belonging and peer mentorship.
To investigate this further, the effect of PSI attendance on final course GPA was again analyzed as a function of students’ self-identified race/ethnicity, but this time students were separated into three groups, as shown in Fig. 6: underprepared (HS GPA < 2.5, Fig. 6a), moderately prepared (HS GPA 2.5–3.5, Fig. 6b), and well prepared (HS GPA > 3.5, Fig. 6c) for college-level coursework. Two-way ANOVAs revealed significant effects of both PSI attendance and of race/ethnicity on final course GPA for groups with HS GPA < 2.5 [attendance, F (3, 291) = 8.13; p < 0.0001 and race/ethnicity, F (3, 291) = 6.24; p < 0.001] and HS GPA 2.5–3.5 [attendance, F (3, 813) = 13.84; p < 0.0001 and race/ethnicity, F (3, 813) = 7.58; p < 0.0001]. No significant effects of either were found for students with HS GPA > 3.5, and no significant interaction was found between race/ethnicity and PSI attendance in the two-way ANOVAs for any HS GPA group.
First, we can examine the overall relationships between HS GPA, PSI attendance, and final course GPA shown in Fig. 6. Averaging the data across students identifying with all racial/ethnic groups, underprepared students saw an average increase of 1.25 GPA units when attending 10 + vs. 1–2 PSI sessions. Moderately prepared students saw an average increase of 0.85 GPA units when attending 10 + vs. 1–2 PSI sessions. Well prepared students did not see an impact even from frequent PSI participation; however, this may simply be because this group already performed very well, with a mean final course GPA of 3.22 across all attendance groups. These results are consistent with previous reports (Achat-Mendes et al., 2020; Dancer et al., 2015; Yue et al., 2018). Thus, the current study adds support to previous research indicating that structured peer-led study methods particularly benefit underprepared students. The fact that the least prepared students saw the largest benefit from PSI also strengthens the conclusion that the improvement in course outcomes was a direct result of attending PSI sessions and not due to PSI attendees being inherently stronger students.
Figure 6 also allows us to more closely examine the relationships between HS GPA, PSI attendance, and final course GPA, this time exploring differences between students identifying with different racial/ethnic groups. For underprepared and moderately prepared students, students identifying with HU groups had significantly lower mean course GPAs compared to their similarly prepared peers for all levels of PSI attendance, as shown in Fig. 6a and b. Within each HS GPA group, students identifying with all racial/ethnic groups experienced a similar increase in their final course GPA with increasing PSI attendance. The result is that there was a consistent gap in the mean final course GPAs for under- and moderately prepared students identifying with HU vs. non-HU groups. This gap was largest for the underprepared students (Fig. 6a), and did not diminish even with high levels of PSI participation. Promisingly, there were no significant differences in final course GPAs for well prepared students identifying with different races/ethnicities (Fig. 6c). This was also the only group for which PSI participation did not have a significant impact on final course GPA for any group. This shows that if students enter college well prepared, they will perform equally well, regardless of students’ self-identified race/ethnicity. Unfortunately, underprepared students were much more likely to identify with HU groups, and in particular as Black/African American, compared to moderately or well prepared students, as seen in Fig. 7. This again indicates that students identifying with HU groups were more likely to enter college at an academic disadvantage. As a whole, these results are concerning, as they indicate that academic inequities at the K-12 level resulted in even greater disparities in student success rates at the college level for underprepared and moderately prepared students. Since students identifying with HU groups were more likely to experience these educational inequities, they were also struggling more at the college level.
PSI’s impact on students identifying with HU groups
Overall, it does not appear that the PSI program at GGC particularly benefitted any students based on their self-identified race/ethnicity alone. Instead, our data extend upon our previous work which showed that PSI addressed differences in college preparedness based on prior academic experiences. The result is that GGC students identifying with some racial/ethnic groups—in particular, students identifying as Black/African American—received a larger benefit from PSI because they arrived at GGC underserved and at an academic disadvantage compared to their peers. These results are important given the previous disagreement in the literature on whether SI is particularly beneficial for students identifying with HU racial/ethnic groups. Additional studies investigating the dual impact of both students’ prior academic experiences and students’ self-identified race/ethnicity at institutions of varying types and demographics may shed additional light on this. For example, future studies might perform a similar investigation at a predominantly white institution to determine if the student demographic makeup influences these results, or at a research-based institution with large class sizes to determine if class format has an effect. In both of these examples, the social benefits of a peer-led academic support program may be more prominent compared to at GGC, where students regularly interact with students and faculty identifying with HU groups in small classroom settings.
These results also reveal a troubling pattern that students identifying with HU groups appear to have additional barriers in both their high school and college education, as evidenced by the different HS GPAs and college STEM course outcomes for these students. While PSI does not equalize final course outcomes between similarly prepared students identifying with HU and non-HU groups, it is still a promising academic support model. PSI disproportionately benefitted less prepared students, and thus helped to close achievement gaps related to college preparedness/prior academic experience. Since students identifying with HU groups had the highest rate of PSI participation (Fig. 2) and made up a majority of the underprepared and moderately prepared groups (Fig. 7), these were also the students who received the greatest benefit from participating in PSI. We propose that this model can be used in conjunction with additional student success support services to more fully address inequities in educational outcomes for students identifying with different racial/ethnic groups, as explained in more detail below.
Looking towards equitable outcomes
The data shown here give some interesting insights into the factors affecting student performance in introductory STEM courses at GGC. Our findings confirm that preparedness plays a major role in student success; only well prepared students showed a passing mean course GPA for students identifying with all racial/ethnic groups. PSI attendance had no significant effect on final course GPA for this group, suggesting that these students would have been successful even without this academic support service. This group also showed no achievement gap in final course grades for students identifying as Hispanic/Latino and Black/African American. These results indicate that if students identifying with HU groups have strong academic experiences before they enter GGC, they are just as likely to succeed as their peers. Unfortunately, as shown in Fig. 7, students identifying as Black/African American are not well-represented in this group. This disparity must be addressed at the K-12 level in order to put these students on equal footing with their peers. Until that time, however, it is necessary for us to address this inequity as best we can at the college level.
Further study is needed to identify other factors that might affect college performance for students identifying with HU groups attending GGC, and interventions that might assist them. These interventions may or may not be part of the PSI program. For example, the PSI program has recently incorporated academic mindset interventions into PSI sessions, which address growth mindset, sense of belonging, and value of coursework for PSI attendees. Jordt et al. (2017) found that values affirmation interventions in an introductory biology course reduced disparities in course outcomes between students identifying with different racial/ethnic groups, and Fink et al. (2018) showed that incorporating growth mindset interventions in a general chemistry course essentially eliminated them when controlling for prior academic achievement. Incorporating academic mindset activities into PSI may therefore prove particularly beneficial to students identifying with HU groups. Future analysis of the effectiveness of the PSI program will include determining the dual impact of PSI and academic mindset interventions on underprepared students and students identifying with HU groups.
It is also possible that separate interventions could be combined with PSI to further improve course outcomes for students identifying with HU groups. Previous studies have shown that holistic programs show great promise in reducing achievement gaps in STEM. For example, Toven-Lindsey et al. (2015) reported on the Program for Excellence in Education and Research in the Sciences (PEERS) at the University of California, Los Angeles, which combined academic and career seminars, holistic academic counseling, research seminars, and collaborative-learning workshops. Although PEERS enrollment was not limited to students identifying with HU groups, it was targeted towards students underperforming in STEM, including students identifying with HU groups, students identifying as female, and students from lower socioeconomic backgrounds. PEERS participation significantly increased student success in science and math courses, and PEERS students on average took more science classes and had higher retention in science majors after two years. A 2020 report on the Operation STEM (OpSTEM) program at Cleveland State University described similar results. This program utilized two levels of intervention: SI alone, and a comprehensive version (OpSTEM Scholars) in which students attended a 2-week summer bridge program before their first semester, had mandatory SI in their math classes in their first year, and received advising and research opportunities throughout their undergraduate career. While all students benefitted from SI sessions, students identifying with HU groups saw much greater gains when enrolled in the comprehensive OpSTEM Scholars program. In contrast, students identifying with non-HU groups had similar outcomes whether they participated in SI alone or the comprehensive OpSTEM Scholars Program (Van Sickle et al., 2020). A 2016 study by Lane attempted to explain the efficacy of such comprehensive programs through focus groups and interviews with 50 students enrolled in a Comprehensive STEM Program (CSP) at a large mid-western university. The CSP included a summer bridge program, biweekly advising, peer-led recitation sessions, a first-year seminar, clustered residential assignments, and peer mentoring, among other program components. Lane concluded that the impact of such comprehensive programs results from a combination of four basic components: proactive care, holistic support, community building, and catalysts for STEM identity development (Lane, 2016). At GGC, PSI is part of a larger systems model designed to reform STEM education through the use of HIPs, CUREs, and PSI (Achat-Mendes et al., 2020; Awong-Taylor et al., 2016, 2018). All of these may contribute to community building and STEM identity development, but additional components of the program would be needed to fulfill all four of the areas identified by Lane (2016). Together, these studies all indicate that a program such as PSI can be made even more powerful when combined with other impactful practices such as summer bridge programs, targeted mentoring/advising, and additional career and research seminars for at-risk students.
Limitations of study
This study used HS GPA as an indicator of college preparedness in order to draw conclusions regarding the impact of PSI on different student populations. Using HS GPA, college entry exam scores, or some other academic measure to verify the effect of SI on final course grades is a common practice (Dawson et al., 2014). However, as some studies have noted, measures of previous academic achievement do not necessarily account for current motivation levels of students, and higher motivation levels can be expected to result in higher final course grades (Dawson et al., 2014; McCarthy et al, 1997). We do acknowledge that student motivation may be an additional contributing factor to improved course outcomes for PSI attendees and plan to investigate this effect in the future. Current PSI participants are surveyed regarding academic mindset, which includes questions related to motivation. However, this survey was added in Spring 2020, when all PSI sessions were converted to a virtual format due to the COVID-19 pandemic, and therefore the data are not directly comparable to that presented here. McCarthy et al. (1997) additionally argue that college and high school learning environments are not identical, and success in the former does not necessarily guarantee success in the latter. They suggest using performance in common non-SI college courses as a second controlling variable. The range of courses investigated in this study, as well as the varying English and mathematics courses that serve as pre-requisites, unfortunately made that impractical for this analysis.
A further limitation regarding the use of HS GPA as a measure of college preparedness is based on the fact that HS GPA may vary considerably depending on both the academic rigor of the high school and the specific classes that a student took. Allensworth and Clark (2020) recently investigated this effect by examining how well both HS GPA and ACT scores predicted 6-year college graduation rates for students who had graduated from Chicago Public Schools over a 4-year period. They found that HS GPA was overall a strong and consistent predictor of college readiness, and performed much better than ACT scores. However, they also saw that students with the same HS GPA from different high schools could have fairly different college graduation rates. This agrees with a 2010 study by Fletcher and Tienda, which examined GPA and 4-year graduation data from four Texas public universities. They found that students identifying as Black/African American and White, and students identifying as Hispanic/Latino and White, had significantly different GPAs and graduation rates even when controlling for HS GPA. However, these gaps largely disappeared or were even reversed when comparing students from the same high school (Fletcher & Tienda, 2010). Together these reports indicate that while HS GPA is a good general predictor of college success, it may be that in this study, students within each HS GPA group shown in Figs. 6 and 7 are not exactly matched according to previous academic experiences and achievement, if they attended different high schools. Future studies could examine the data for students from the same high school to see if different course outcomes for students identifying with different racial/ethnic groups persist, and how PSI affects students from different high schools.
Despite these limitations, the strong evidence that HS GPA is a good predictor of college readiness, combined with our own data showing the correlation between HS GPA and final course GPA, lends confidence to our assertion that PSI attendees had similar levels of academic preparedness compared to all students enrolled in PSI-supported courses. While PSI participation may not be the sole reason for the higher final course grades of PSI attendees, the evidence here strongly suggests that students directly benefitted from attending PSI sessions, and less prepared students saw a greater benefit than well prepared students.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.