“ Thanks for helping me ﬁnd my enthusiasm for physics! ” The lasting impacts ‘research in schools’ projects can have on students, teachers, and schools

. Using 6 years of evaluation data we assess the medium-and long-term impacts upon a diverse range of students, teachers, and schools from participating in a programme of protracted university-mentored projects based in cutting-edge space science, astronomy, and particle physics research. After having completed their 6-month-long projects, the 14–18 year-old school students report having substantially increased in conﬁdence relating to relevant scientiﬁc topics and methods as well as having developed numerous skills, outcomes which are corroborated by teachers. There is evidence that the projects 5 helped increase students’ aspirations towards physics, whereas science aspirations (generally high to begin with) were typically maintained or conﬁrmed through their involvement. Longitudinal evaluation 3 years later has revealed that these projects have been lasting experiences for students which they have beneﬁted and drawn upon in their subsequent university education. Data on students’ destinations suggests that their involvement in research projects has made them more likely to undertake physics and STEM degrees than would otherwise be expected. Cases of co-created novel physics research resulting from PRiSE also 10 has seemed to have a powerful effect, not only on the student co-authors but participating students from other schools also. Teachers have also been positively affected through participating, with the programme having inﬂuenced their own knowledge, skills, and pedagogy, as well as having advantageous effects felt across their wider schools. These impacts suggest that similar ‘research in schools’ initiatives may have a role to play in aiding the increased uptake and diversity of physics and/or STEM in higher education as well as meaningfully enhancing the STEM environment within schools.

university campus. Students typically spend 1-2 hours per week throughout working on the project in groups, usually outside 55 of lesson time. Support from the university is provided through workshops, school visits, monthly webinars, printed/multimedia resources, and ad hoc emails as required. The provision within the programme is explored in more detail in M.O. . PRiSE has engaged a much more diverse set of school students and significantly more disadvantaged groups than is typical. Furthermore it has been found that students' success within the programme appears independent of background, which has been attributed by teachers as due to the extraordinary level of support offered (M.O. Archer, 2020). 60 The evaluation of PRiSE's pilot, which ran from 2014-2016 and involved 6 schools, suggested that students' awareness of current scientific research, understanding of the scientific method, and skills were enhanced by the programme and that teachers benefited through reconnecting with their subject at an academic level, being challenged, and being supported in their professional development (M.O. Archer, 2017). The programme has grown significantly since then, having involved 67 London schools by 2020. This paper expands the evaluation of PRiSE's potential impacts.
only completed a PRiSE-wide questionnaire, which not only asked for their observations of impacts upon students but also how involvement in the project has affected their own knowledge, skills, practice, and wider school environments. The questions posed to both students and teachers varied slightly from year-to-year and are found in Appendix A. The PRiSE-wide questionnaires also included feedback on participants' experience of the programme, with this data forming the focus of a separate paper (M.O. . 80 These instruments were chosen in order to collect data from as wide a range of students and teachers as possible as well as respecting the limited time/resources of all involved (both on the school and university sides). All data gathered was anonymous, with students and teachers only indicating their school and which project they were involved with. We have further anonymised the data by using pseudonyms for the schools. More detailed information about the schools involved can be found in M.O. Archer (2020). No protected characteristics (such as gender or race) or sensitive information (such as socio-economic 85 background) were recorded. Our ethics statement on the forms informed participants that the data was being collected for evaluation purposes to determine the programme's impact, and that they could leave any question they felt uncomfortable answering blank (also true of the online form).
For longer term evaluation, students were also asked on a separate paper form at our conference to share their personal email addresses so that we could follow up with them a few years later, in order to explore potential lasting impacts of the 90 programme. As with the main questionnaires, this was presented as optional with an ethics statement and description on how their data would be used clearly presented. It was decided to contact cohorts of PRiSE students 3 years after they started the project for this follow-up, so that students would either be studying at university or at least (in the case of the youngest PRiSE students) making university applications, hence giving us insight into university destinations/plans. The students were emailed and asked to fill out an online form, detailed in Appendix B. The form contained primarily open-ended questions, to enable us 95 to understand PRiSE students' long-term attitudes to the programme, their higher education destinations, and what may have affected these decisions. This sort of contextual data would not be available by simply obtaining destination data from services such as the Higher Education Access Tracker (HEAT, https://heat.ac.uk) or requesting that schools provide it as a condition of their participation. Doing so would have also risked schools declining to participate. However, we do acknowledge that our approach reduces the number of responses that could realistically be collected.

Participants
Data was collected from 153 students (aged 14-18) and 45 teachers across 37 London schools. A breakdown of the number of responses per year and how many schools these responses came from is given in Table 2 where the total number of participants (and their schools) in attendance at our conferences are also indicated. We note that, due to the COVID-19 pandemic, the 2019/20 programme was disrupted so we do not have reliable information on how many students, teachers, and schools would 105 have successfully completed the programme that year. There is no indication that the respondents differed in any substantive way from the wider cohorts participating in the programme.

Analysis
Both qualitative and quantitative approaches were utilised in data analysis, as the open and closed ended questions in the surveys generated different types of data.

110
For all quantitative (numerical) data, uncertainties presented represent standard (i.e. 68%) confidence intervals. For proportions/probabilities these are determined through the Clopper and Pearson (1934) method, a conservative estimate based on the exact expression for the binomial distribution, and therefore represent the expected variance due to counting statistics only. Several statistical hypothesis tests are used with effect sizes and two-tailed p-values being quoted, with a statistically significant result being deemed as p < 0.05. In general we opt to use nonparametric tests as these are more conservative and 115 suffer from fewer assumptions (e.g. normality, interval-scaling) than their parametric equivalents such as t-tests (Hollander and Wolfe, 1999;Gibbons and Chakraborti, 2011). When comparing unpaired samples a Wilcoxon rank-sum test is used, which tests whether one sample is stochastically greater than the other (often interpreted as a difference in medians). The Wilcoxon signed-rank test is used to compare both a single sample to a hypothetical value or data from paired samples to one another.
Both versions test whether differences in the data are symmetric about zero in rank. Finally, for proportions we use a binomial 120 test, an exact test based on the binomial distribution of whether a sample proportion is different from a hypothesized value (Howell, 2007). For ease of reference, further details about the quantitative analyses are incorporated into the relevant sections of the findings.
Thematic analysis (Braun and Clarke, 2006) was used to analyse the textual (qualitative) responses. Instead of using predetermined qualitative codes to categorise the data, our analyses drew on a grounded theory approach (Robson, 2011;Silver-125 man, 2010), letting the themes emerge from the data itself. This process involved the following steps: change across all projects was 0.92 ± 0.04 points, indicated as the black bar in Figure 1 along with interquartile range (grey area), whereas the mean was slightly higher at 1.08 ± 0.06 due to a positive skewness (the uncertainty refers to the standard error in the mean). The overall results show positive changes across all projects to a high level of confidence, as indicated in 165 the figure, with no real variation in results between the different projects or between schools. Therefore, students' confidence in scientific topics and methods seems to have substantially increased as a result of PRiSE and almost all students reported this benefit. This gain in confidence has also been noted in teachers' comments: "They have become more confident in communicating their ideas and realised that they are not too young to do research." (Teacher 1, Hogwarts, SCREAM 2015) 170 "This has been a challenging experience for the students taking part. Students have gained a better appreciation of real science and built confidence." (Teacher 3, Xavier's Institute for Higher Learning, MUSICS 2016)

Skills
From 2016 onwards we asked students (n = 140) to list which skills, if any, they felt they had developed through their PRiSE project. Teachers (n = 40) were also asked to indicate their observations on skills that their students had developed during the 175 programme. We extracted keywords from any prose responses and sifted through the data performing keyword clustering. This latter step involved identifying synonymous skills and relabelling them so there was a consistency of terminology throughout.
This processing resulted in a dataset of 79 unique skills. Students tended to identify on average around 2 distinct skills each whereas teachers typically listed 3, though the responses per person ranged up to 5 and 6 respectively. Figure 2 shows the skills identified as a word cloud, where students and teachers have been given equal weight by normalising their counts by 180 their respective totals. Colours indicate from whom the words originated, showing a large amount of agreement between M e d ia n c h a n g e I n t e r q u a r t il e R a n g e N o C h a n g e Overall Student Confidence After Figure 1. Overall student confidence in relevant scientific topics/methods before and after taking part in PRiSE (n = 127). Data points are coloured by project. Overall changes are also indicated through the interquartile range (grey area) and standard confidence interval in median (black bar). Overall and ranges of z-scores in two-tailed Wilcoxon signed-rank tests are also listed by project.

Key
the two groups. All the skills listed are highly relevant to being a scientist, with the most cited being (in descending order) teamwork, research, data analysis, programming and presentation skills. These results remain fairly consistent with those from the pilot and arguably indicate areas where university / research physics differ substantially from the regular school experience.
Therefore, through experiencing and being involved in research-level physics, it seems likely students have gained new, or 185 further developed existing, skills, constituting a positive impact upon them. This has been further expanded upon in the teacher feedback: "They have developed presenting skills, they do get that [at school] but not for academic poster sessions. The unique skills from the project were the exposure to the physics, analysis, independence; it has allowed them to access the world." (Teacher 1, Hogwarts, SCREAM 2015) Students and teachers have been given equal total weight.

Aspirations
To assess whether students' aspirations were affected at the 6-month stage, we first undertook a qualitative analysis in 2018- of underlying codes. The first of these concerned how much the students' felt that they had wanted to study either physics or a STEM subject before even undertaking the project (n = 22), which tended to be raised if they already wanted to pursue this route or if they were not interested in these subjects. The second dimension covers the students' aspirations following the project 200 (n = 13), revealing that students were either now wanting to pursue physics/STEM, were considering these as potential options, or were simply unsure. Finally, the third dimension was an expression of change as a result of their involvement in PRiSE (n = 22), which typically stated that it had confirmed their subject choice, made them more likely to pursue physics/STEM, had not affected them, or (in a small number of cases) had deterred them from continuing with physics/STEM. These themes align with the findings of Dunlop et al. (2019) on how independent research projects can affect students' aspirations. 205 We show the dimensions and codes in Table 4, also giving counts of the number of responses which fall within them (cf. Sandelowski, 2001;Sandelowski et al., 2009;Maxwell, 2010). We note that some students' responses covered more than one of the dimensions, but none spanned all three. Out of the 63 responses to this question, 11 did not fit into any of these three dimensions, instead highlighting aspects of the programme they enjoyed (research, teamwork, real applications of physics, and what physics at university is like) but not explicitly stating their subject aspirations or how they may have been changed by 210 the project. From these counts, it is clear that in both dimensions 2 and 3 the totals indicating positive effects from PRiSE are greater than the neutral or negative responses. However, given that these numbers are rather small and derived from a qualitative coding we do not attempt to make a statistical interpretation about the impact of PRiSE on students' physics/STEM aspirations based on them.  Instead, informed by these promising preliminary results, we implemented in 2020 a quantitative approach to assessing how 215 PRiSE may have affected students' aspirations. In a similar manner and with similar justification to our evaluation of students' confidence, we asked students (n = 35) to assess their likelihood (using a 5-point Likert scale) of continuing with physics and STEM as well as reassessing these from before the project. We shall call this the 'absolute scale', since it pertains to students' absolute likelihood of continuing these subjects, and code it to values of 1-5 as shown in Figure 3a-b. In addition, we asked how working on the project affected their thoughts on physics and STEM as future subject choices using a different 5-point 220 scale, where the wording of the options used were informed by the previous qualitative results in Table 4. We refer to this as the 'relative scale', since it concerns whether the students feel their thoughts changed as a result of PRiSE, and code it to values from -2 to +2, as shown in Figure 3c. One student only answered these questions relating to physics but not for STEM.
When considering physics aspirations, there was no clear positive bias towards the subject beforehand (see horizontal distribution in Figure 3a) with a mean value on the absolute scale of 3.23 ± 0.21 (p = 0.264 in a one-sample Wilcoxon signed-rank 225 test against null hypothesis of 3). The vertical distribution shows some shift towards greater values on the absolute scale after the projects, now exhibiting overall positive results (mean of 3.69±0.20, p = 0.006). From the paired data, 40±10% of students increased in likelihood of studying physics on the absolute scale (though no students who were very unlikely before showed any increase) and only one student's likelihood decreased (to a neutral stance), with the mean change being +0.46 ± 0.12. While this indicates only moderate changes in students' absolute physics aspirations, they are statistically significant (p = 8×10 −4 in 230 a Wilcoxon signed-rank test). On the relative scale displayed in the top panel of Figure 3c, however, 69 ± 9% of students report that the projects either made them more likely to continue with physics or confirmed it, with again only one student becaming less likely to pursue physics. The average was +0.89 ± 0.13, greater than zero with high confidence (p = 2 × 10 −5 ). No trends were present by project or school. Since we colour the datapoints in panel a of Figure 3 based on the students' responses on the relative scale (panel c), it is clear that most of the students whose absolute physics aspirations increased attribute this in some 235 way to PRiSE, while around half of those which did not increase on the absolute scale still claim positive influence by PRiSE.
Students' STEM aspirations, unlike physics, were already incredibly high before the projects as shown in Figure 3b with a mean value on the absolute scale of 4.53 ± 0.14. Because of this, a large proportion (67 ± 9%) of students would be unable to increase in value on the absolute scale due to already giving the highest rating. No students decreased in likelihood on the absolute scale and only 6 students out of the 34 increased, though this constituted around half of the students who could possibly 240 increase (gave a 4 or below beforehand). This slight change in the paired data (average +0.18 ± 0.07 across all students) was still statistically significant (p = 0.031). The bottom panel of Figure 3c reveals a bimodal distribution on the relative scale of whether PRiSE affected students' STEM aspirations, with the largest peak at +2 (it confirmed their subject choice) and a smaller one at 0 (no change). No students reported being less likely to pursue STEM due to PRiSE. A similar proportion to with physics of 68 ± 9% indicated PRiSE's likely positive influence on their STEM aspirations, with the mean being 1.15 ± 0.15 245 which is again a clear positive result (p = 1 × 10 −5 ). As before there was no real variation in these results by project or school.
All students who increased on the absolute scale attribute this to PRiSE, whereas a majority who did not change still indicate PRiSE had a positive effect on their STEM aspirations.
Ideally one would benchmark the likelihoods before PRiSE against larger surveys of similarly-aged students' aspirations to test whether PRiSE students were more likely to continue with physics/STEM anyway. Unfortunately, direct comparisons 250 are not possible due to the differing ways the relevant questions have been structured across national surveys. However, such research has shown that STEM degree aspirations amongst all students remains similar to the makeup of STEM vs. non-STEM A-Level subject choices, implying that almost all students studying at least one STEM A-Level likely aspire towards a STEM degree (Hamlyn et al., 2020). Furthermore, science aspirations are highly correlated to 'science capital' and begin to form at an early age (Moote et al., 2020). Therefore, it is not surprising that PRiSE students' likelihood of wanting to continue with 255 STEM was high beforehand.
The follow up qualitative question, asking students to explain how or why their thoughts about subject choices had been affected by the project typically mentioned how their interest, enjoyment, or understanding had been enhanced, e.g. The one student who reported being less likely to pursue physics (but was not affected with regards to STEM) noted that "I already had my mind set on doing STEM subjects at university, but now I am less interested in physics as I have come to see how some areas are more challenging than others and I wouldn't want to specialise in those areas."

3-year stage evaluation
To date we have undertaken long-term evaluation for three cohorts of PRiSE students who had participated in the academic years 2015/16 (cohort 1), 2016/17 (cohort 2), and 2017/18 (cohort 3). At our student conferences 72 students from these three cohorts left contact details with us for this purpose, which were well spread across the different schools involved. Across the 295 three cohorts, the bounce rate was 46 ± 7% (predominantly due to now inactive school email addresses being given) and the emails were opened by 24 PRiSE students (a rate of 62 ± 9% from non-bouncing emails, again well spread across different schools) with 14 filling out the survey though we purposely cannot identify individuals from responses. While this is a relatively small number of responses, longitudinal evaluation is notoriously difficult for under-resourced engagement programmes (e.g.

M.O. Archer et al., Under
Review) and there are still significant and useful results from the data, which we present in this 300 section. The evaluation covers the perceived legacy of PRiSE on these students, as well as the students' higher education destinations and the factors which may have affected these.

Legacy
Most of the 14 PRiSE students who responded were aged 16-17 when they undertook their projects, with two students aged 15-16, and one student each in the ranges 14-15 and 17-18. All of them remembered undertaking a physics research project with [and] observing building schematics to find how much matter muons pass through during travel into [the] building." (Student 12, cohort 3) "I remember getting to experience some more advanced practical physics that was more reminiscent of university "I now study architecture, so the observing building schematics and 3D mathematics were both useful experiences." (Student 12, cohort 3) "The lab work was useful as it gave me an idea of how to work in a uni lab, which is particularly helpful now that I am at uni." (Student 13, cohort 3)

345
Of the two negative responses, both stating "I haven't", one caveated though that "but that's just because I haven't had to do a group project since" (Student 9, cohort 3). Overall, students' responses suggest that their PRiSE projects were lasting and beneficial experiences that they have been able to draw from in their subsequent educational activities and development.

Destinations
All the PRiSE students reported that they were studying at university when the survey was conducted, apart from one who due 350 to their age (they were 14-15 when involved in PRiSE) intended to. We asked the students what subject they were studying at university (or planned to study in the case of the one student) giving the options of physics (or combination including physics), another STEM subject, or a non-STEM degree. The results of this are shown in blue in Figure 4, where we compare these to the degree destinations of physics A-Level students nationally (McWhinnie, 2012) in orange (this data only includes students that went on to higher education). The 7 out of 14 PRiSE students going on to study a physics degree is considerably 355 higher than the national rate of 9.7% (p = 1 × 10 −4 in a binomial test). While the number of students going on to study other STEM subjects (6 out of 14) are consistent with the national statistics, the increased uptake of physics leads to the overall STEM-degree proportion (13 out of 14) also being significantly greater than found nationally (59.3%, p = 0.012). Given the diversity of schools involved (discussed in-depth in M.O. Archer, 2020) and the known barriers to STEM higher education for underrepresented groups (e.g. Campaign for Science and Engineering, 2014; Hamlyn et al., 2020), it is highly unlikely 360 these results can be explained simply by PRiSE schools tending to produce more physics and STEM students anyway. Another consideration may be that PRiSE students were already highly likely to continue their physics education beforehand anyway.
While this did not appear to be the case for the 2019/20 cohort discussed in section 4.1.3, with it being shown that PRiSE led to some increased physics and STEM aspirations, the destinations results here came from different cohorts so this remains a possibility. However, given that PRiSE has been a consistent framework (both in terms of schools targeting and delivery) 365 throughout, it is reasonable to assume that the 2019/20 cohort is representative of the others used here and thus may be used as comparable samples. This then suggests that students' involvement in PRiSE may very well make them more likely to pursue physics and STEM degrees.
The students' reasons behind choosing their degree subjects and what influenced them varied. We note that one student out of the 12 that responded to these questions referenced the research project (which was not prompted in the question)

370
"I did the sounds of space project that you organised a couple of years ago and am now pursuing a physics degree from Cambridge. Thanks for helping me find my enthusiasm for physics!" (Student 4, cohort 1) However, in general the responses were quite brief and did not give much insight into the likely many factors which may have played a role in their subject choices (cf. L. Archer et al., 2013Archer et al., , 2020b. While for cohorts 2 and 3 we added questions that explicitly asked about PRiSE's influence on the students' degree choices, the small number of responses and brief answers 375 simply highlight the need for more in-depth qualitative longitudinal evaluation, though this is challenging to undertake. We note that there is little research into engagement programmes' effectiveness at influencing students' destinations, both correlative and causal, as raised in a recent review (Robinson and Salvestrini, 2020). However, given that PRiSE shows some statistically significant positive effects on students' aspirations, which are known to be highly resilient, at the 6-month stage as well as PRiSE students being more likely to study physics and STEM degrees than would otherwise be expected (further statistically 380 significant effects), the results of this empirical enquiry into PRiSE's potential effect on students' destinations are perhaps promising.

Co-production of research
Co-producing publishable physics research between researchers and school students is not an explicit aim of PRiSE, unlike for instance the more researcher-driven ORBYTS programme (Sousa-Silva et al., 2018). The rationale behind this is explored fur-385 ther in M.O. . Nonetheless, in a few instances genuinely novel preliminary results have come from students' independently motivated work on the MUSICS project (the other PRiSE projects are unlikely to result in publishable physics research). The first of these originated from the 2016/17 cohort, where a group from a girls' school in an area of particularly high deprivation discovered a series of decreasing-pitch "whistle" sounds which lasted several seconds (corresponding to 5 days in reality). Through collaboration with the students and further investigation by professional scientists, it was discovered 390 that these unexpected sounds corresponded to the natural oscillations of Earth's magnetic field lines following a solar storm.
Such wave events had been deemed rare previously but, due to the accessibility of exploring the sonified data, were found to in fact be quite common thanks to the students' discovery. The work was presented at several international scientific conferences and eventually published in the journal 'Space Weather' with the students and their teacher listed as co-authors (M.O. Archer et al., 2018). Information about these developments resulting from the students' project work was continually passed on via 395 their teacher, who in turn responded with the students' comments: "It was a very rewarding experience which allowed us an insight into the research conducted at university level.
This helped us to develop crucial skills needed in the next years of our studies. It was truly amazing to hear how significant the event we found was and that it will be forming the basis of a proper scientific paper." "Being a part of the university's research and the subsequent paper published is truly an amazing opportunity. It 400 was really interesting to find such a significant event and we gained so much experience and developed many skills during our research that will be useful in our university careers." The publication garnered widespread media attention, for example featuring on BBC Radio 4's 'Inside Science'. Unfortunately we were unable to find out how the news of the publication had been shared across the school involved and affected other students' thoughts about physics. However, through publicising the result across all schools involved in PRiSE via teachers, it 405 appears to have had a powerful effect on PRiSE students at other schools: "Hearing that other kids at other schools have actually produced a paper, it just gives you hope that it's actually something I can do." (Student, Summer Heights High, MUSICS, BBC Radio 4 interview, Oct 2018), potentially highlighting through demonstration by their peers that (research-level) physics is something which is accessible to 'people like me', thereby breaking down known barriers to participation in physics and science generally.

Quantitative analysis
Based on the areas of impact on teachers and schools emerging from the qualitative data (from 2015-2018), from 2019 onwards we sought to quantitatively assess how prevalent they might be. Teachers (n = 23) were asked to identify for each of the 8 495 themes whether they felt that they (or their school) had been affected by the project in that area, using the closed options of: "I have", "I will eventually", "I have not", and "Unsure". This scale was chosen over a 5-point Likert due to an expected low level of responses. We exclude any blank or unsure answers (which were rare) and divide the remaining 172 responses across the 8 themes into negatives ("I have not") and positives, with the latter being subdivided into planned ("I will") and definite ("I have") impacts. largely reliable and likely did not fall prey to unreflective responses. Therefore, it appears that the identified areas of impact upon teachers and schools as a result of PRiSE may indeed be quite widespread.

Conclusions
We have investigated the medium-and long-term impacts on students, teachers, and schools who have participated in a 6- Medium-term impacts on the participating 14-18 year-old school students were assessed after they had completed their 525 6-month-long projects. Students' confidence in relevant scientific topics and methods seems to have substantially increased as a result of PRiSE, with nearly all students reporting this benefit. Furthermore, through experiencing and being involved in research-level physics, students report having gained new, or further developed existing, skills. Both of these impacts upon students have been corroborated by teachers' observations. While the students involved with PRiSE were fairly committed to STEM in general beforehand, our data suggest that they had no clear bias in aspirations towards the subject of physics in 530 particular. Following the programme it appears that students' attitudes towards pursuing STEM were typically maintained or confirmed through their involvement and physics aspirations seem to have been moderately enhanced. We find no evidence that these impacts varied by the different projects or schools. These results should be deemed successful, as a drop-off in STEM aspirations is often seen at this age (Davenport et al., 2020), with these issues being particularly pertinent in physics (L. Archer et al., 2020a). Thus at this stage of a student's educational journey they are likely to require interventions that sustain and 535 support their science identity which, in turn, has an influence on their educational choices (L. .
Longitudinal evaluation has also been performed for three cohorts of PRiSE students 3 years after they commenced their projects. While a relatively small sample, the evidence suggests that these projects have been highly memorable and beneficial experiences that students have been able to draw upon in their later educational activities and development. The data on PRiSE students' degree destinations show increased uptake of physics and STEM at degree-level than would typically be expected, 540 suggesting that their involvement in the research projects has helped transform their aspirations into destinations -a key aim of the programme. Further in-depth qualitative research, such as interviews or focus groups, could provide richer and more reflective information on how students' thoughts and feelings about their association with physics and STEM may have been affected by participating in PRiSE, given the nuance and multiple factors at play with students' aspirations in general, which are difficult to capture and interpret with questionnaire data alone (L. Archer et al., 2013Archer et al., , 2020b.

545
The impacts upon students reported in this paper only relate to those who completed the 6-month programme. However, as should be expected for any extended programme, there is some drop-off in participation with PRiSE. This has been explored in more detail in M.O. Archer (2020), demonstrating that students' success with PRiSE appears to be independent of background and thus not clearly patterned by societal biases present within the field (e.g. Campaign for Science and Engineering, 2014).
Currently though we have no data on what impact the programme has on those students who drop out. While no teachers 550 have communicated any negative effects on students who do not continue, with some highlighting informally their students' attitudes towards the projects, this requires further formal investigation. Such work is required to ensure that no negative effects are being felt by these students and potentially discover what positives, if any, may result from even partial participation.
Furthermore, we have no evidence that the PRiSE approach would be effective for students who are generally uninterested or unengaged with STEM. Indeed, it seems unlikely that such students would want to persist with an extended and challenging 555 extra-curricular physics programme. Young people's aspirations towards science begin to form at an early age (L. Archer et al., 2013Archer et al., , 2020b and therefore interventions throughout their educational journey need to align with their needs and wants, from initial inspiration and positive associations, to informing on career-focused aspects, and finally sustaining those built aspirations (Davenport et al., 2020). PRiSE only aims to address that final part of the chain, since no single programme can fit all stages.
Evaluation of the impacts on teachers and schools has identified several themes. By collaborating on PRiSE, teachers can 560 gain new physics knowledge, become more confident in discussing research, and integrate aspects of the research projects into their regular lessons. Teachers also report developing various technical skills, gaining confidence in mentoring, and reassessing their preconceptions of students' potential. While all these positive changes to teachers' practice will likely be felt across their wider schools, there is more direct evidence of the school environment being affected such as through students' project work being championed, the profile of physics or science being raised, and a university-school relationship being established 565 with significant repeated buy-in from schools over several years. These impacts appear to be fairly widespread across the teachers and schools involved in PRiSE. We note that these results share many similarities to those reported by Rushton and Reiss (2019)  resources and interventions to support teachers' and schools' participation. The similar impacts thus highlight that, with the right support, teachers and schools from a variety of contexts can benefit from 'research in schools' projects. Further research could investigate the validity of the participating teachers' remarks of the impact on the schools' environments, for instance through in-depth interviews or focus groups with other teachers in the schools.
field of 'research in schools' initiatives. They suggest that with more similarly designed and supported programmes at other institutions, we may be able to start to address a key part of the chain of the wider issue of uptake and diversity not just in physics but potentially STEM also. We stress, however, that multi-faceted approaches from a variety of different stakeholders and organisations are required to implement real change on this entire issue, but 'research in schools' may be able to form one piece of the puzzle.

580
Appendix A: 6-month stage evaluation questions Here we list the questions considered within this paper posed in the PRiSE-wide questionnaires at the 6-month stage evaluation.
We detail the phrasing used, how participants could respond, and which years the question was asked. Follow-on questions are indicated by indentation and a down-right arrow ( ). The following questions were posed to students:

Question Response type Year(s)
In what way has this project affected you Open Text 2016-2020 What skills, if any, has the project helped you develop Open Keywords 2016-2020 How has doing the project affected your thoughts about future subject choices / careers Open Text 2018-2019 Before working on the project, how likely were you to continue with the following (Physics/STEM) in the future

5-point Likert 2020
After working on the project, how likely are you now to continue with the following (Physics/STEM) in the future

5-point Likert 2020
How has working on the project affected your thoughts about these future subject choices (Physics/STEM)

5-point Likert 2020
Please explain how or why? Open Text 2020 The questions asked of teachers were: The following questions were asked of students in the 3-year stage evaluation via an online form. Data availability. Data supporting the findings of this study that is not already contained within the article or derived from listed public domain resources are available on request from the corresponding author. This data is not publicly available due to ethical restrictions based on the nature of this work.