School-to-school collaboration in England: A configurative review of the empirical evidence

In recent years school-to-school collaboration in the English context has been promoted by a myriad of policy initiatives. Many of these initiatives have been directed at structural reforms seeking to facilitate a ’self-improving system’ in which schools support one another to raise standards of teaching and learning and address educational inequality. Yet, at the same time, the English school system remains a deeply marketised and competitive arena while there are debates concerning the extent to which collaboration between schools can meaningfully facilitate educational improvement and equity. Taking these issues as a starting point, this paper reports on findings from a configurative review of the empirical evidence on school-to-school collaboration in England. Drawing on 46 peer-reviewed empirical studies from 2000 onwards, the paper provides insight into the reasons why schools enter into collaborative arrangements and the conditions and factors that can facilitate and hinder such activity, as well as the possible benefits that can result from collaboration between schools. A number of weaknesses within the field are also identified. For example, there is a need for more conceptual and terminological clarity and a stronger theoretical basis for research in this area. We also argue that the field is deficient in respect of critical perspective and interpretation (of collaborative practice). Furthermore, research into school-to-school collaboration is lagging behind policy and practice, presenting a formidable challenge for a system increasingly underpinned by an expectation that schools will work in partnership with one another. Reviews that are collecting empirical data to describe and test predefined concepts can be thought of as using an ‘aggregative’ logic. The primary research and reviews are adding up (aggregating) and averaging empirical observations to make empirical statements (within predefined conceptual positions). In contrast, reviews that are trying to . .. understand the world are interpreting and arranging (configuring) information and are developing concepts. (p. 3)


Introduction
The potential for school-to-school partnership and support to act as a vehicle for educational improvement and equity is well rehearsed in literature (Chapman & Fullan, 2007;Muijs et al., 2010;Muijs et al., 2011) although the extent to which this plays out in practice remains contested (Huxham & Vangen, 2000;Gunter, 2015;Keddie, 2015). In recent years assumptions surrounding the possibilities for school-to-school collaboration have been tested out at scale within the English context as policymakers experiment with structural reforms and directives underpinned by the notion of a self-improving system in which schools will supposedly support themselves and each other to raise standards of teaching and learning and address educational inequality (Greany, 2015). At the same time, the English educational system remains a deeply marketised arena in which schools must compete over pupils, funding and resources in order to survive (Keddie & Lingard, 2015), leading to what Armstrong and Ainscow (2018) describe as an 'unusual cocktail of collaboration and competition ' (p. 4). Within this context, how is it that schools can work together, and what are the necessary conditions for such activity?
The notion of school improvement through partnership and professional dialogue is a powerful and seductive one with which few would argue in principle. Yet for all of the possibilities this approach presents, there are pitfalls and barriers that often hinder or prevent meaningful collaborative activity (Keddie, 2015). Moreover, there are debates surrounding the strength and depth of evidence within this area of the field and uncertainties over whether, how and to what degree collaboration between schools can facilitate educational improvement and equity (Croft, 2015). Nevertheless, the principle of schools working together under formal collaborative arrangements remains central to educational policy in England.
Such issues and concerns provide the starting point for this paper in which we undertake a configurative literature review of the empirical evidence on school-toschool collaboration in England. A distinctive feature of the English education system is the strong emphasis on competition-a feature that would seem problematic in terms of collaboration.
Our aim, then, is to ascertain what we know about this phenomenon and, just as importantly, what we do not. In so doing we provide some insight into the conditions and factors that can facilitate and hinder collaborative activity between schools while also drawing attention to areas of weakness within the field. We then discuss the implications of these gaps in the knowledge base. In particular, we argue that research into school-to-school collaboration is lagging behind policy and practice and that this presents a serious challenge for a system increasingly underpinned by an expectation that schools will work in partnership with one another.
We begin by detailing the context in which school-to-school collaboration has emerged and evolved in the English system. We also consider different theoretical understandings of school (and organisational) collaboration to inform our thinking around the complexity of this concept before outlining our rationale and the research questions we used to frame the review. We then describe the methods employed in gathering, synthesising and reviewing the evidence before presenting our findings and discussing the implications for research, policy and practice.

School-to-school collaboration in England: An overview of context
Over the past 20 years, the school system in England has been subject to seismic shifts across almost every aspect and phase of educational provision (Chapman & Gunter, 2008;Jones, 2016), accompanied by political justifications couched in familiar and broad discourses of educational equity, school improvement, teacher effectiveness and pupil performance (Gunter, 2018). Throughout this period, governments have invested considerable amounts of public funds in national 320 P. W. Armstrong et al. academy trusts (MATs), formalised structural arrangements between two or more academy schools often with shared leadership and governance (Armstrong, 2015).
The two most recent government White Papers in education illustrate this point. In the first, The Importance of Teaching it is stated that: schools working together leads to better results. . . Along with our best schools, we will encourage strong and experienced sponsors to play a leadership role in driving the improvement of the whole school system, including through leading more formal federations and chains (DfE, 2010, p. 60) In the second, Educational Excellence Everywhere, we are informed that: MATs are the only structures which formally bring together leadership, autonomy, funding and accountability across a group of academies in an enduring way, and are the best long term formal arrangement for stronger schools to support the improvement of weaker schools (DfE, 2016, p. 57) Despite a paucity of evidence to support these bold assertions (Greany & Higham, 2018;Hutchings & Francis, 2018), the scale of the academies programme and the pace at which it has developed is noteworthy. As of 2018, of the 21,513 state schools within the English system, around 35% had converted to academy status-including 72% of all secondary schools and 27% of all primary schools (NAO, 2018). Moreover, well over half of all academies (65%) are members of a MAT comprising two or more schools (HoC, 2017).
Furthermore, a number of policy initiatives underpinned by collaboration between schools have emerged in recent years. These include a national network of 'teaching schools' (Matthews & Berwick, 2013)-high-performing schools tasked with training and developing teachers and school leaders within a network of other schools and strategic partners known as a teaching school alliance (Gu et al., 2016a(Gu et al., , 2016b-and other arrangements such as specialist networks or 'hubs' for subjects including maths and English (DfE, 2018). The recent introduction of the 'research schools network'-an initiative funded by the government-backed Education Endowment Foundation (EEF) and Institute for Effective Education (IEE) to support groups of schools that utilise evidence to inform and improve teaching (EEF, 2020)-has added a further dimension to an already complex web of inter-school collaborative activity within and across the English school system.
In addition to these collaborative arrangements there are a growing cadre of national, local and specialist leaders of education, successful school leaders and practitioners that work across the system to support schools that are underperforming (Close & Kendrick, 2019). There are also a number of regional and local networks of schools that have emerged across the English system to facilitate context-specific improvement, some of which have been mapped in the literature (e.g. Gilbert, 2017), although there are doubtless many other examples of informal partnership activity between schools that go undocumented.
In spite of the highly competitive environment in which schools in England must operate, there exists a plethora of collaborative activity within the system and an appetite among teachers and school leaders to work with, learn from and support their fellow professionals (Armstrong, 2015;Greatbatch & Tate, 2019). In undertaking this review, we attempt to make some inroads into understanding the complexity and fluidity of this environment while also drawing attention to the gaps and 'silences' within the literature in this area of the field.

Conceptualising collaboration
In the editorial for a special issue of School Effectiveness and School Improvement that focused on networking and collaboration for school improvement, Muijs (2010) suggests that empirical research within this area of the field is under-theorised and predominantly evaluation based, having emanated largely from practitioners and researchers with a school improvement focus. The prominence of evaluative research within the knowledge base relating to school-to-school collaboration is a matter to which we return later in this paper. However, the a-theoretical nature of the empirical literature in this field is also a notable feature. Some authors have sought to conceptualise educational collaboration by drawing on theory from other fields such as psychology, sociology and business, where notions of individual and organisational networking are more established. For example, Muijs et al. (2010) highlight the following four theoretical perspectives that can be applied to collaborative activity: constructivist organisational theory, the theory of social capital, 'New Social Movements' theory and Durkheimian network theory.
Constructivist theory posits that organisations are sense-making systems with their own distinct perceptions of reality and as a result are at risk of becoming myopic. This can be mitigated somewhat through inter-organisational networking or collaboration with external partners to facilitate a broader world view. According to Muijs et al. (2010), organisations such as schools that exist in complex and uncertain environments, particularly those in challenging circumstances, are more susceptible to myopia and, therefore, from a constructivist perspective have the most to gain from collaboration.
Social capital theory offers a more functional perspective in emphasising how organisational collaboration can provide a vehicle to utilise and share the resources held by actors within and across a network and thus increase information flows. Burt (1992) refers to this kind of activity within the context of structural gaps in which there exists a dearth of intelligence. In this sense, collaboration is a potentially constructive endeavour for all stakeholders as each brings their own contributions to plugging these gaps in knowledge. As a consequence, the network becomes greater than the sum of its parts. In his thinking around school-led systems, Hargreaves (2011) underscores the importance of social capital; key pillars of which he suggests are reciprocity and trust. He argues that deep partnerships between schools are those in which expertise and intelligence are shared freely, thus enhancing reciprocity and trust between members of the network.
According to Muijs et al. (2010), networks that are formed from the bottom up as a result of shared priorities can be viewed through new social movements theory. This social action perspective acknowledges the transient nature of networking and recognises that while network members may not share the same values they typically share the same goals. Authors such as Townsend (2013) and Hadfield (2005) have suggested the fluidity and complexity of collaborative activity between schools is closely School-to-school collaboration in England 323 aligned to and can be understood through the lens of the new social movements theory.
Durkheimian network theory is underpinned by the notion of moral purpose whereby collaboration is a means of mitigation against organisational anomie that might occur when there is a disassociation between individuals' values and the behaviour they must engage in on behalf of their organisation. This can be a particular issue for schools facing challenging circumstances (Chapman & Muijs, 2014). Lumby and Morrison (2006) offer an alternative perspective by locating collaboration within the context of competition. Drawing on game theory, they suggest that partnership activity requires organisations to move from zero-sum to non-zero-sum games. Such games can be played cooperatively or uncooperatively. In their research with groups of schools working together on the 14-19 Pathfinder project in England, they point to a lack of evidence that schools are working in cooperation 'to meet the needs or wants of all local learners' (p. 337). Rather, they found that schools tended to focus on their own priorities and organisational outcomes. Muijs and Rumyantseva (2014) have also explored notions of competition and cooperation in education, noting how the two might coexist within what they label as 'coopetition'. In doing so, they point to a number of key conditions that are likely to be required for this to occur, including the setting of clear goals and both benefits for and strong trust between partners within the collaboration. They also emphasise forms of leadership that are sensitive to possible tension between partners. Armstrong and Ainscow (2018) have drawn on the notion of coopetition through their research into system leadership and school-to-school support in England and found some support for this theory within their findings.
In a typology of organisational collaboration within education, Chapman (2019) draws together a number of key facets of the aforementioned theoretical perspectives. Building on earlier work by Ainscow and West (2005), he refers to an escalator of collaborative endeavour with deepening levels of partnership activity built on relational trust. The first level is association, which represents traditional hierarchical working interspersed with incidental meetings and with little sharing of knowledge. The second level is cooperation, characterised by short-term activity around specific issues with some knowledge exchange albeit fairly superficial and task-focused. The third level is collaboration, in which the joint activity becomes more sustainable with the emergence of common goals, shared values and the potential to develop new ways of working. The fourth level is collegiality in which longer-term commitments to the partnership emerge underpinned by strategic goals and objectives, common values and a focus on shared knowledge, resource and practice development for the betterment of all partners. Chapman also emphasises the importance of social capital, suggesting that partnership activity is susceptible to failure when this is low. He terms this the dark side of collaboration, in which cooperation is fabricated and contrived as stakeholders tussle for power, influence and status and traditional hierarchies remain. In their research into the self-improving school-led system (SISS), Greany and Higham (2018) identify such conditions across a number of localities in England. Here they draw on governance theory to make sense of the 'complexity and contradictions that underlie the SISS agenda' (p. 26). Drawing on Adler (2001) they consider how hierarchy, markets and networks operate as coordinating (and contradictory) mechanisms within the school system and thus influence collaborative activity between schools. For example, hierarchies facilitate control and coordination through formal structures but can also hinder collaboration and innovation. Markets depend upon price as a means of coordinating supply and demand. This can create the conditions for flexibility but also weaken trust, knowledge mobilization and equity. Networks are underpinned by trust and shared knowledge but are susceptible to complacency or exclusivity when relationships become comfortable.
These theoretical perspectives and understandings are worthy of acknowledgement for the purposes of this review insofar as they facilitate an understanding of the potential motives and rationale for collaboration between schools and also the conditions in which such activity is likely to succeed and the reasons why it might fail.
At this point it is important to assert what we take the notion of school-to-school collaboration to be and how we defined such activity for the purposes of this review. In respect of the theory (and research) in this area of the field, there exist a multitude of different understandings of collaborative activity between schools. Because of this we employed a broad set of parameters to ensure we could capture as much empirical evidence as possible on collaborative activity between schools. We therefore focused on research involving any kind of partnership work between two or more schools where the school is the primary unit of analysis. There are obvious limitations to this broad definition, not least the argument that it is individuals not organisations that collaborate (raising legitimate questions as to whether the school is the most appropriate or useful unit of analysis). Nevertheless, we decided this was a necessary approach in attempting an initial configuration of such a multifaceted field. These issues are discussed in more depth within the final section of the review.

Rationale and research focus
Given the centrality of school-to-school collaboration to national educational policy and practice in England in recent years (Armstrong, 2015;Brown & Flood, 2019) this represents an opportune moment to reflect on the empirical evidence from this area of the field so as to acknowledge and make sense of what we know about schools working with other schools. Widespread academisation, coupled with the growth of the MAT model as the preferred structural arrangement for schools, has positioned England's education system at the forefront of such activity. There are few comparable contexts in which formalised school-to-school collaboration is so central to educational policy (Earl & Katz, 2007;Salokangas & Ainscow, 2017). However, the pace of this structural reform has left research lagging behind policy and practice. In collating what we know about school-to-school collaboration, this review will also throw light on the gaps in our knowledge in this area to identify where the field needs to focus its attention in respect of research priorities and also as a means of informing policy and practice.
As far as we are aware, though there exist a small number of commissioned reviews looking at school partnerships and collaboration (e.g. Arnold, 2006;Bell et al., 2006;Atkinson et al., 2007;Armstrong, 2015;Greatbatch & Tate, 2019), this paper represents the first scholarly attempt to review the empirical literature relating to collaboration between schools in England. We perceive this review as a starting point rather School-to-school collaboration in England 325 than a definitive mapping of this area of the field and the quality of the research within it. Indeed, as we will discuss, the complexities of school-to-school collaboration and the many different forms it can take make such an exercise unrealistic within the margins of an academic journal article.
Our broad aim is therefore to start a meaningful dialogue as to what the empirical scholarly evidence can tell us about school-to-school collaboration and, just as importantly, what it cannot. We then consider some implications for future research agendas and policy decisions within this area. Bearing this in mind, we devised the following research questions which were used to frame our review of the literature: 1. What are the prominent drivers for school-to-school collaboration in England? 2. What factors are likely to facilitate and/or hinder school-to-school collaboration? 3. What are the consequences of school-to-school collaboration for teaching and learning and student outcomes?
The overarching rationale for this review is to understand what is known about school-to-school collaboration in England and the factors that influence such activity. In this sense we follow Petticrew (2015) in his assertion that to grasp what is purposeful (research question 3), one must first understand what is happening and why (research questions 1 and 2).

Method
In reviewing the literature, we began by searching two of the most comprehensive educational research and social science databases, the Education Resource Information Centre (ERIC) and Scopus. Different combinations of relevant terms were used as keyword searches, including 'school' or 'academy ' + 'partnership' or 'collaboration' or 'cluster' or 'network' or 'trust' or 'federation' or 'chain'. Other terms that were likely to produce relevant results were also employed (e.g. 'teaching school', 'professional learning community', 'networked learning community', 'city challenge'.) In addition, searches were limited to research undertaken in England from 2000 onwards. This parameter was applied due to the increased emphasis on and interest in school-to-school collaboration within the educational policy, practice and research communities around the start of the millennium (Muijs et al., 2011). Additional, supplementary searches were then undertaken through Google Scholar and the reference lists of key texts within this area of the field identified by the three reviewers to source additional literature that had not already been located.
Having identified approximately 300 texts as part of these initial searches we then applied further parameters to narrow the focus of the review. First, we read through the abstracts of these returns retaining only scholarly literature (i.e. peer reviewed journal articles, books and book chapters) that contained empirical research in which some form of school-to-school collaboration could be identified within the research. More specifically, as we have previously highlighted, this meant including research involving any kind of partnership work between two or more schools where the school is the primary unit of analysis. We also rejected texts focusing on school partnerships with other agencies. This was an important step to enable us to generate a rigorous and 326 P. W. Armstrong et al. robust platform on which to raise questions, base arguments, draw conclusions and address our research aims. It was also necessary given the volume of think, opinion and advocacy pieces circulating within the non-academic literature in this area of the field (e.g. Hill, 2010;Hargreaves, 2010Hargreaves, , 2011Gilbert, 2017).
In applying these parameters we recognise that a number of research reports have been excluded from this review. 1 Many of these reports comprise evaluations of government-funded initiatives underpinned by or including an element of school-toschool collaboration (Armstrong, 2015). A smaller number report on government commissioned research (e.g. DfE, National College) or studies funded by research councils. While these sources fell outside our peer-reviewed parameter, we found that a number of the academic articles we reviewed had drawn upon data generated through these studies. Excluding these reports from the review therefore reduced the possibility of source duplication. As a result, a final total of 46 texts were accepted for review.
As Gorard et al. (2019) observe, typically with systematic reviews, quality assessments and the trustworthiness of findings should be judged on research design, scale, missing data, quality and relevance of measurements, fidelity, validity and so on. For the following reasons, we did not make judgements of this kind on the research selected for this study. First, collaborative activity between schools is complex and multifaceted, typically encompassing a wide range of different practices, resulting in a lack of clarity and consistency within the empirical research regarding such activity (including what and who it involves, its purposes and how it is being defined). Secondly, the field is dominated by small-scale exploratory research, which often fails to detail fully the design, sample, instruments used or potential issues of validity and reliability. Correspondingly, we take the view that the field is not yet ready for a systematic review of the kind that would provide an accurate quality assessment of the research (indeed, such an exercise would likely conclude that evidence of impact is poor or weak). Hence, we took the decision to adopt a configurative or organisational approach to the review, the rationale for which we discuss in the following section.

Analysis: A configurative approach
The notion of collaboration in education is under-defined (Muijs, 2010). Moreover, the literature reflect the complexity and multi-faceted nature of school-to-school collaboration in practice. This poses a challenge for a review of the literature in this area of the field, particularly the traditional aggregative approach to reviewing, the purpose of which is to provide a summary of the findings from similar studies of phenomena that are more clearly defined and understood (Levinsson & Prøitz, 2017). Such an approach is useful for addressing issues of effectiveness, impact and improvement (i.e. what works) but less so when synthesising complex bodies of research to consider what is happening and why.
Instead we adopted a configurative approach to the review in which the synthesis is primarily concerned with organising (configuring) findings from the literature to address the guiding research question(s). As Gough et al. (2012)

explain:
School-to-school collaboration in England 327 Reviews that are collecting empirical data to describe and test predefined concepts can be thought of as using an 'aggregative' logic. The primary research and reviews are adding up (aggregating) and averaging empirical observations to make empirical statements (within predefined conceptual positions). In contrast, reviews that are trying to . . . understand the world are interpreting and arranging (configuring) information and are developing concepts. (p. 3) As such, configurative approaches have been promoted as suitable for providing an oversight of complex bodies of research (Levinsson & Prøitz, 2017) such as those we focus on in this paper.
In exploring the complex nature of collaboration between schools we follow Petticrew (2015) who asserts that: 'asking the simple question "does it work?" about highly complex social change processes, where evidence is often sparse and heterogenous, is often meaningless and usually unanswerable' (p. 2). Schools work with, learn from and support other schools in a multitude of ways for a whole range of reasons with different motives and over varying periods of time . As Higham and Yeomans (2010) remind us: Partnership is a process, not an event and is therefore inherently unstable and dynamic and subject to changes over time. These changes may be as a result of shifts in policy or in the configuration and use of levers and drivers. Change may also be triggered at the local level, perhaps as a result of changes in key staff, for example, where new principals or head teachers may wish to change the ways in which their institutions engage collaboratively. (p. 397) The empirical research into school-to-school collaboration reflects this complexity and therefore lends itself to a configurative method of review. Moreover, configurative reviews typically pose the kind of open questions that are addressed with qualitative data and iterative methods of exploration concerning experience and meaning (Gough et al., 2012).
Aside from a small number of notable exceptions, the research in this area of the field is predominantly qualitative in design. This is likely to be a consequence of the nebulous and intangible nature of collaboration, a concept that has proved difficult to 'measure'. As such, the research in this area has tended to focus more on the nature, process and key features of partnership activity rather than the outcomes (Hayes & Lynch, 2013). It is also important to point out that in contrast to aggregative approaches that generally set out to be exhaustive in scope, configurative reviews tend to follow what Levinsson and Prøitz (2017) refer to as 'an inductive logic that arranges the findings of different studies in a way that offers a meaningful picture of what the research presents' (p. 213). What we present, therefore, is less a systematic mapping of the field but rather a narrative that provides a broad overview of schoolto-school collaboration in England. The configurative approach we have adopted is suitable for these means allowing us to develop an understanding of the key themes within the field and to lay the foundations for further work that might: (a) generate theory and test out both existing and new theoretical interpretations of collaborative activity between schools (hopefully leading to more fine-grained definitions and categorisations of the different forms that such activity can take); (b) design research that can throw more light on the kind of impact that collaborative activity between schools 328 P. W. Armstrong et al.
can have on student outcomes; and (c) provide a more nuanced and accurate map of the research and theory in this area of the field both within the English context and further afield. We suggest how this might be achieved in our recommendations section.
Adopting this approach, the findings have been organised (configured) around key themes that recur within the literature and that help to address our guiding research questions. Following the findings section, we return to the research questions to discuss whether and the extent to which this has been achieved.

Findings
In configuring the literature we identified the following broad themes that link to and inform our overarching research focus (framed by the three guiding research questions). It is important to acknowledge that these themes are by no means mutually exclusive and overlap in a multitude of ways.
(i) Drivers for collaboration Ainscow et al. (2006) identify three prominent drivers for schools to engage in collaborative activity with other schools. First, schools may choose to do so voluntarily because of a common interest or priority. Secondly, schools may be incentivised to collaborate typically around a centrally driven initiative that is underpinned by or includes a strong element of partnership activity. There have been several examples of such initiatives within the English context over the last 25 years often accompanied with a short-term financial incentive for participating schools (Armstrong, 2015). Thirdly, in some cases schools are forced to enter into collaborative arrangements by central government. This tends to be the case when a school is deemed to be performing poorly (often following an inspection) and is partnered with a higher-performing school or group of schools such as a MAT. The third driver is somewhat questionable in the sense that forcing a school to join a particular governance structure does not necessarily mean that school will engage in collaborative activity with other schools within that structure. We would argue that it is more useful to consider governance structures (such as federations and MATs) as factors that make school-to-school collaboration more likely). This leaves two main drivers for school-to-school collaboration.
In respect of the first driver, voluntary collaboration, a number of studies report on small schools, typically in rural locations, entering into partnership arrangements as a means of sharing resources and taking advantage of economies of scale through shared contracts and resources (Muijs, 2008;Chapman et al., 2010) and opportunities for joint professional and curriculum development (Ainscow et al., 2006;Hayes & Lynch, 2013). Other reasons why schools may choose to work together voluntarily relate to reduced or limited capacity at local authority level and an associated paucity of adequate services and provision (Ainscow et al., 2006;Coldron et al., 2014). Such drivers for school-led collaboration can be understood as representing social capital approaches to partnership working (Hargreaves, 2011) in which schools are entering into such arrangements to utilise and share resources. Equally, this kind of activity could also be characterised as a social movement in which the collaboration is formed School-to-school collaboration in England 329 from the bottom up in response to the identification of common priorities .
The second driver, incentivised collaboration, very much underpins the many centrally driven school improvement initiatives in England since the turn of the millennium involving schools working with and supporting other schools. Much of the literature in this particular area of the field is comprised of research reports evaluating such initiatives on behalf of the funder (typically the Department for Education) although some of this work has made its way into the scholarly domain. For instance, Smith (2015) undertook retrospective research into the Beacon Schools initiative that ran between 1998 and 2004 and aimed to address improvement through school diversity, collaboration and partnership. Following analysis of variations in academic and social data from 322 of the secondary schools involved in this programme, she concludes that there is little evidence that the Beacon Schools initiative provided any advantage for the schools involved in respect of improvements to student outcomes. Evans et al. (2005) draw on research findings from the Diversity Pathfinders initiative, an area-based project established in 2001 in which different regions across England were identified as pathfinders to demonstrate the benefits of collaboration between schools and invited to bid for funding to participate in this initiative. In the following year a similar initiative aimed at supporting reform within secondary and further education was established. The 14-19 Pathfinders initiative also comprised regional collaboration between schools and included colleges and other training providers. Again, groups were encouraged to submit proposals for projects that developed and enabled innovative collaborative activity as a means of facilitating and sharing good practice within the 14-19 phase of education. Successful bids were awarded financial support for their projects. Other empirical studies reflecting on the process, impact and implications of this initiative include Lumby and Morrison (2006), Higham and Yeomans (2010) and Hayes and Lynch (2013).
Established in 2002, the Networked Learning Communities (NLC) programme, promoted by the National College for School Leadership (NCSL), had similar characteristics whereby volunteer networks of schools were invited to submit proposals for networked learning activity that promoted student learning and professional development among staff. Successful proposals received three years funding and access to a team of specialist support and administrative staff to facilitate their projects and collaborative activity. A number of studies have drawn on research into NLCs (e.g. Earl & Katz, 2007;Sammons et al., 2007;Katz & Earl, 2010).
The School Business Manager (SBM) Demonstration Project Programme (SBMDPP) adopted a slightly different approach in which the schools involved engaged in collaborative activity around organisational, financial and resource management. Launched in 2008 by the New Labour government through the NCSL, this programme was again centrally funded and again provided participating groups of schools with a financial incentive to work together under shared business management arrangements to foster expertise in this area (thus easing headteacher workload) and improve financial and organisational efficiency. Data from this initiative is reported within a number of scholarly sources (e.g. Woods et al., 2012Woods et al., , 2013.
The City Challenges were established in 2003 with a focus on improving schools in London and later expanded to include schools in the Midlands and Greater Manchester. This government-funded initiative was characterised by a model of school-to-school collaboration that partnered struggling schools with those with a record of high performance as a means of support and both professional and school development. Access to funding and additional resources for participating schools was a key pillar of this programme. Within the literature, Ainscow (2015), Ainscow et al. (2006Ainscow et al. ( , 2012 and Brighouse and Fullick (2007) provide useful and first-hand accounts of and perspectives and reflections on the process, outcomes and legacy of the City Challenges. However, there is a surprising paucity of peer-reviewed scholarship that looks at data from this initiative.
The partnering of high-performing schools with their lower-performing counterparts, as a means of support and improvement for the latter, is a key feature of many centrally funded initiatives. For example, the Specialist School Achievement Programme (SSAP) employed a strategy in which a low-performing specialist school was matched with a high-performing school with the same specialism and given additional funding and resources for capacity to facilitate a one-year focused improvement programme (Allen, 2007). Teaching Schools represent the most recent centrally driven attempt to financially incentivise schools to collaborate, in this respect through leading teacher training and professional development across networks of partner schools. Eligible schools (criteria include those regarded as at least 'good' in their most recent national inspection grade) receive financial support annually for three years upon being awarded teaching school status. This initiative remains in its infancy (the first cohort of teaching schools were designated in 2011) and as such features sporadically within the empirical scholarly literature (see Greany, 2015;Dowling, 2016).
It is worth highlighting that funding alone is not necessarily the only motive for schools to enter into collaborative arrangements. For example, Higham and Yeomans (2010) suggest many of the partnership projects they observed 'were often based upon pre-existing partnership activity, which was then shaped in various ways by the funding and accountability requirements of the Pathfinder . . . initiative' (p. 390). In other words, groups of schools may utilise a centrally driven initiative and any attached funding to support, continue and develop collaborative activity with which they are already involved.
According to Lumby and Morrison (2006), centrally driven or top-down initiatives such as these, particularly those that partner high-and low-performing schools, do not create the conditions for equitable outcomes. Drawing on game theory they suggest this can lead to a situation in which schools move from a zero sum to non-zero sum and start to prioritise their own interests over that of the partnership or collaboration. The governance theory perspective utilised by Greany and Higham (2018) is also useful here. They suggest that top-down initiatives or drivers for collaborative structures often serve to benefit existing hierarchies within the school system whereby the most advantaged schools gain more resources and influence to the detriment of the less advantaged schools (ii) Factors that can facilitate school-to-school collaboration There are a number of commonalities within the literature with regards to the conditions that can potentially support or facilitate purposeful collaboration between schools. Chief among these are themes relating to leadership (including coordination, School-to-school collaboration in England 331 shared responsibility and capacity building) and relational factors (including trust and clear communication). Contextual features such as a history and pre-existing culture of collaboration also appear frequently.
For instance, reporting on case study research with groups of schools working together in both formal and informal collaborations, Chapman et al. (2010) identified robust yet fluid structures that maintained beyond the turnover of key personnel; approaches to leadership that are open to collaboration as a means of improvement; and self-governance underpinned by trust and shared values as key characteristics of the most effective collaborations in their sample. Similarly, in their research into federations of schools in disadvantaged areas, Chapman and Allen (2006) underscore a number of key conditions for partnerships to flourish, including: a strong and clear focus on teaching and learning; dispersed leadership responsibility that builds capacity across the network; and a shared commitment to professional development at all levels. Earl and Katz (2007) distinguish between two types of leadership within effective school networks. They refer to formal leadership as the work of the headteacher, typically involving motivation and encouragement, setting and monitoring the agenda and building capacity, and distributed leadership which signifies the extent to which staff members and other stakeholders are involved and engaged with the networked activity. This echoes Higham and Yeomans (2010) who refer to the depth of engagement with the collaborative activity in their research that 'determined the extent to which a partnership was principally a talking shop for senior managers or had an impact upon the day-to-day practices of member institutions' (p. 389). A number of other studies identify leadership, in respect of establishing clear goals, coordination and the sharing of responsibility, as central to collaborative practice between schools (see, for example, Ainscow et al., 2006;Allen, 2007;Chapman & Hadfield, 2010;Hayes & Lynch, 2013;Chapman & Muijs, 2013;Howland, 2014;Muijs, 2015a;Armstrong & Ainscow, 2018;Brown et al., in press).
Relational factors are also a common feature within the literature on school-toschool collaboration. For instance, drawing on 49 interviews with headteachers, school leaders and other senior educational stakeholders, Coleman (2012) identifies trust as a key feature of school-to-school partnership whereby the participants engaging in the deepest and most mature collaboration also had the highest perceived levels of trustworthiness. In his research with networks of small schools Muijs (2015a) suggests 'partnerships were more successful where they focused on a small number of key goals, established trust between the schools, used a phased approach to change, and developed a mutually beneficial relationship' (p. 578).
Similarly, research into networked school leadership and governance identifies trust between headteachers as the bedrock of successful partnership activity (Chapman et al., 2010). Varga-Atkins et al. (2010) talk about the creation of trusting environments where individual views are respected, ideas can be tested and risks taken as crucial conditions for partnership activity between schools. In research with primary school networks, Muijs (2015aMuijs ( , 2015b lists trust and personal relationships between schools as a key facilitating factor for partnership work. Looking more closely at interleader relationships, Hadfield (2007) draws on a large mixed methods study of over 100 school networks to explore the dynamic between leaders within such collaborative groups. Findings point to tensions that can arise between middle leaders within networks because of differential rates of individual agentic development and coupled with a shift in their identification with the network and its aims over time. Hadfield suggests there is work to be done to further explore the kind of leadership models that are applicable to and appropriate for networked activity. Other studies that emphasise the importance of relational factors include Day and Hadfield (2004), Allen (2007) (2018) and Brown et al. (in press). Relational factors relating to trust and reciprocity between partners align with social capital theories of collaboration (e.g. Burt, 1992;Hargreaves, 2011) and offer some explanation as to why such factors are often seen as key to successful and sustainable collaboration.
Contextual factors, such as history, culture and geography, feature prominently with the literature as enablers to collaboration. A history of partnership is commonly cited as a key facilitator to the success of inter-school collaborative activity. If there is a pre-existing culture of shared practice within a network, then mature professional relationships and high levels of trust (or social capital) are likely to have been established between partners, thus forming robust foundations for any future collaborative activity. This is borne out across a number of studies (e.g. West, 2010;Hayes & Lynch, 2013;Armstrong & Ainscow, 2018).
Similarly, contextual sensitivity was an important feature of the City Challenges, in which an understanding of the local conditions that characterised individual schools and their surrounding communities was central to the collaborative structures and support mechanisms established as part of this initiative (Ainscow, 2012). Geography is also influential. Research suggests that urban contexts, in which there are higher numbers of schools in closer proximity to one another, often make for more collaborative favourable conditions. For instance, in such contexts it is more straightforward for staff or students to travel between sites to engage in joint practice and learning than in more sparsely populated, rural locales where there are less schools, greater distances between them and fewer transport options (Muijs, 2015b;Armstrong & Ainscow, 2018).
That is not say partnership between schools cannot happen in rural contexts. Indeed, in such areas the need to collaborate is often more acute due to fewer resources and less staffing capacity. For instance, Woods et al. (2012) report on groups of small schools in rural contexts sharing business management expertise and provision as a more efficient alternative to employing their own school business manager. An option many could ill afford to take. The pooling of resources to achieve economies of scale is also reported in research undertaken by Howland (2014) among a group of 10 schools working together. Based on her findings, she also concludes that there is a greater likelihood of successful collaboration where the schools concerned have a common set of vision and values, and shared history, geography and demographics. Again, new social movements theory  affords a useful lens for interpreting this kind of collaboration in which schools develop partnerships and engage in collaborative working as a result of shared educational or financial objectives, priorities and/or needs.
Finally, we would add governance structures to the list of factors that can facilitate collaborative activity between schools. This is not attended to explicitly within the literature. However, the number of research studies we reviewed that involve structures School-to-school collaboration in England 333 such as federations, Teaching Schools and MATs suggest such structures, together with other facilitating factors we have highlighted, might make school-to-school collaboration more likely.

(iii) Factors that can hinder school-to-school collaboration
The most frequently cited potential barriers to the initiation, efficacy and sustainability of collaboration between schools include threats to school autonomy (and perceived power imbalances), capacity (including funding and resources), workload and a marketised national policy context that fosters and actively encourages competition.
In their research with school federations, Chapman et al. (2010) report tensions arising between schools that enter into collaborative activity for reasons of protecting or enhancing their power or influence with the aim of the collaboration to acquire resource. This can lead to weak collaborations and create conflict between stakeholders. The authors found this to be more prevalent in informal collaborations such as those without shared leadership or governance, suggesting that formalised arrangements may be more robust. Reporting on research into school-to-school support, Muijs (2015b) differentiates between issues within the supported school (such as resistance to outside assistance or internal conflicts between staff members) and issues within the supporting school (such as staff overload and pressures on capacity) as a result of the support they are providing. Such barriers might be interpreted through a Durkheimian lens whereby individual priorities and values do not necessarily align with those of the school or wider collaborative group (i.e. anomie).
In their research into school-to-school support and challenge in disadvantaged communities, Ainscow and Howes (2007), underscore the importance of a thirdparty coordinating function (in this case the local authority) to oversee, broker and orchestrate collaborative activity thereby easing the pressure on individual schools to undertake such work. This aligns with the constructivist perspective adopted by Muijs et al. (2010) to think about schools facing challenging circumstances. Because of their circumstances, such schools are at risk of myopia and therefore in a position to benefit from the support of a third-party broker to facilitate a wider view of their situation and smooth the inter-school collaborative support process.
As we have already noted, many of the centrally or externally driven collaborative initiatives have additional funding and resourcing attached (at least in the short term) to support schools with capacity issues associated with additional labour required to undertake and manage the partnership activity. However, a number of authors conclude that problems can often arise once this additional (initial) support ceases. For example, in their research into the 14-19 Pathfinder initiative, Hayes and Lynch (2013) distinguish between partnerships formed in response 'to government demands for collaboration' (enacted) and those formed in response to 'locally identified interests/needs' (community). They conclude that enacted partnerships are likely to be 'less effective' than community partnerships and also 'superficial' and at risk of subsiding once funding ends or policy changes (p. 425), a view supported by Woods et al. (2013) and their research into groups of schools sharing business management expertise.
This speaks to a broader issue highlighted by a number of authors that so long as competition and parental choice continue to be the dominant drivers for educational 334 P. W. Armstrong et al. policy in England, purposeful and authentic collaboration between schools will be a challenge (e.g. Lumby & Morrison, 2006;Ainscow & Howes, 2007;Townsend, 2013;Woods et al., 2013;Smith, 2015;Armstrong & Ainscow, 2018) This view is summarised by Keddie (2015) who draws attention to the 'the difficulties of creating socially responsive and responsible collaboratives in the current "heterarchical" and market-oriented policy environment' in England (p. 1). Indeed, such a competitive environment acts as a counter to the development and cohesiveness of social capital (trust, reciprocity) that appears central to purposeful collaboration (Hargreaves, 2011).
Systemic barriers to collaboration are also identified by Ehren and Godfrey (2017) in a case study of a MAT. They argue that external accountability in relation to networks can be problematic if network partners are ill at ease with being held to account for something they do not have complete control over. At the same time vertical, oneway, top-down forms of accountability are not supportive of the creation of inter-organisational networks that are agile and flexible enough to effect change: this, they argue, is because single-member accountability can prevent network development and motivate a structure of strong internal hierarchical control around the framework used to hold individual members to account. Small-scale case study research undertaken by Keddie (2016) highlights similar tensions associated with school-to-school collaboration. Her data from interviews with six primary school leaders discussing the academies programme reveals concerns about school collaboration within a competitive system and the potential loss of individual school autonomy.
More broadly, Chapman and Ainscow (2019), drawing on their experiences of using research to change education systems, including the role that collaboration plays in this process, identify three factors that can hinder organisational partnership activity (including collaboration between schools). These are social factors (including pre-existing relationships between partners that encourage mutual support), political factors (attitudes and priorities of key partners) and cultural factors (local traditions and expectations of partners as to what is possible).
(iv) School-to-school collaboration: What difference does it make?
As we have highlighted, one of the main criticisms levelled at school-to-school collaboration is the lack of tangible evidence as to the difference it makes to teaching and learning. In particular, critics have emphasised a paucity of evidence for the positive impact and influence of such activity on student progress and outcomes (e.g. Croft, 2015). Our findings certainly suggest a degree of inconclusiveness in relation to this. Indeed, we would argue it is premature to be drawing conclusions relating to 'impact' given the vagaries within the literature surrounding the exact nature of school-toschool collaboration, the different forms it takes and who it involves. Bearing this caveat in mind, we have somewhat tentatively configured the findings according to three areas of potential impact: student outcomes, teaching and professional development.
a Student outcomes There have been very few studies that have set out to explicitly or exclusively focus on the direct influence of school-to-school collaboration on student outcomes.

School-to-school collaboration in England 335
Rather, where student outcomes are discussed they tend to be considered as one of a number of different areas that school-to-school collaboration might impact upon. At best, the findings are mixed. Sammons et al. (2007) analysed national assessment and examination data of primary and secondary schools involved in the NLC programme between 2003 and 2005 to compare their student results with the national average over the same period. Their findings indicated that NLC school data generally aligned with the broader national trend. However, while they found no evidence that NLC primary schools had improved more rapidly or narrowed the attainment gap in relation to national results during this time, they did find some indication of improvements to English results at key stage 3 (pupils aged 11-14) against the national average (though not for science or maths).
In their research into federation membership and student outcomes, Muijs (2013, 2014) present a slightly clearer and more positive picture. They compared three different types of federation (n = 66 schools in total) with matched samples of comparator schools. By controlling for student characteristics such as prior attainment, FSM entitlement, SEN provision, ethnicity, gender and English as an additional language (EAL) they were able to estimate the independent effect of the federation on attainment. While they found no differences in student outcomes in the cross-phase federation sample, in academy federations and particularly performance federations, students outperform their peers in matched non-federated schools from the second and third year the school is part of the federation (respectively). Of the different types, the performance federations (where high-performing schools partner those in the bottom tier) appeared to have the most influence on student outcomes.
A re-analysis of this dataset (coupled with additional telephone interview data) revealed a small yet significant influence of school networks on student outcomes among groups of schools in rural contexts (Muijs, 2015b).
Reporting on outcomes of the City Challenge initiative, Ainscow (2012) highlights a particular strand of activity entitled 'families of schools'. Here schools across the region were grouped together based on their similar contextual features (e.g. prior pupil attainment, socio-economic characteristics, English as an additional language etc.). Such groupings allowed schools serving comparable populations to work together without the conflicting issue of direct competition. While engagement was inconsistent, some (though not all) of these families reported marked improvements to student attainment among particular cohorts of young people as a result of their involvement in and commitment to the collaborative activity.
b Teaching There appears to be a higher degree of confidence within the literature as to the influence of school-to-school collaboration on teachers and teaching. This reflects the predominantly qualitative nature of the research in this area of the field in which practitioner and school leader testimonies, while open to reasonable questions of validity, remain a key source of evidence.
Reporting on research with networks of schools, Ainscow et al. (2006) highlight increased support for problem solving and curriculum development with schools coconstructing joint courses that they would not have had the resources to develop 336 P. W. Armstrong et al. individually. Reflecting on small-scale case study research with a network of providers involved in the Beacon School programme, Bullock and Muschamp (2004) report on the positive influence of collaborative action research projects between teachers from different schools. Notably, teachers reported an improved understanding of their own pedagogy and pupil learning. Drawing on work with Specialist Schools partnerships and federations of schools, Chapman and Allen (2006) suggest school-to-school collaboration can facilitate improvements in school climate and staff development opportunities.
West (2010) discusses findings from research with six groups of schools in urban contexts working in partnership. He reports increased opportunities for staff development and shared professional dialogue among teachers as a result of the collaborative activity. Similarly, Day and Hadfield (2004) draw on research with networks of primary schools working together on collaborative action research projects as a means of school improvement. In follow-up case study analysis to his earlier work (Chapman & Muijs, 2013, Muijs (2015a) explored collaborative activity within performance federations. His findings throw light on a process of intensive intervention characterised by developing approaches to teaching and learning and building practitioner capacity and capital. Chapman (2008) reflects on case study research with four schools situated in disadvantaged contexts and working in three networks. His findings reveal that all three networks to which the case study schools belong had prioritised improvements to teaching and learning as a core feature of their partnership work. This resulted in the sharing of best practice and teaching strategies aimed at pupils within the particular contexts served by these schools.
Brown (2017) reports on a study of Research Learning Communities (RLC), a collaborative arrangement involving groups of teachers (typically 8-12) from across a network of 4-6 schools working together to engage in research as a means of enhancing their own practice and that of their peers. Initial findings from this suggest the RLC model has the potential to improve teaching and learning through teachers becoming more research conscious and learning how to purposefully utilise research and evidence in their practice.
c Professional development Professional development appears to be fertile ground for school-to-school collaboration, with a number of studies reporting favourable outcomes in relation to this theme. For example, drawing on multiple methods research with networks of schools in Liverpool, Varga-Atkins et al. (2010) found that where practitioners benefitted from professional development in these networks, the quality of such practice was a key factor. Similarly, in their research with school federations, Muijs (2013, 2014) found that federating provided more opportunities for professional development between schools whereby shared learning was recognized by teachers as a more powerful medium that was more likely to directly influence practice compared to traditional approaches such as external training courses. In case study research with one Teaching School Alliance, Dowling (2016) suggests a lack of agency among teachers for their own professional development is a barrier to professional learning within school networks (such as Teaching School Alliances). He argues there is work School-to-school collaboration in England 337 to be done by school leaders to raise awareness of the opportunities for professional development within their networks to enable such potential to be fulfilled.
Drawing on their work with school networks, Hadfield (2010a, 2010b) cite similar findings whereby the increased demands of leading a network necessitate a requirement to build leadership capacity and distribute leadership more widely across the school. As such, staff members with little previous leadership experience are handed opportunities to develop their skills and knowledge in this area. Comparable findings are reported elsewhere within the literatures (e.g. Chapman et al., 2010;Chapman & Allen, 2006).
More recent research has focused attention on system leadership and the growing number of school leaders and practitioners working across more than one institution. Close and Kendrick (2019) note that there are now several thousand 'system leaders' working as consultants on aspects of school-to-school collaborative support in England. At the same time there has been no systematic consideration of the long-term professional development of these individuals and the skills and dispositions they require to operate across the system. Following case study research with a sample of system leaders within a Teaching School Alliance, the authors identify an emerging professional development need among this cohort of the school workforce that includes the acquisition of consultancy skills and a greater understanding of the wider political context of the English educational system. A similar study by Boylan (2018) employed a case study design to explore the work of teacher system leaders who were engaged in practice beyond their own school. His findings point to a professional development gap that suggests teacher system leaders require support to develop their adaptive leadership skills. He characterises this as learning to be responsive to needs at the local and system level including brokering, mobilising and forming new networks.
Rempe-Gillen (2018) reports on a small-scale yet novel study that explored the experiences of two primary teachers who engaged in a cross-school, cross-phase collaborative initiative with counterparts from the secondary school sector over a period of one year. Findings highlight opportunities for professional growth, classroom experimentation and the development of subject knowledge for the primary teachers involved. Table 1 shows the 46 studies that were selected for this review. Table 2 details the evaluative reports excluded from this review. The research reported within these sources is underpinned by or involves a significant element of school-to-school collaboration. We have listed these as a reference for those who are interested.

Discussion
Our discussion of the findings is organised around the three guiding research questions that have framed this review and to which we now return. 1. What are the prominent drivers for school-to-school collaboration in England?
The majority of the empirical literature we reviewed report on government-led or top-down initiatives that are underpinned by or include a significant element of 338 P. W. Armstrong et al.     school-to-school collaboration. Such initiatives typically include a financial incentive for schools to participate (though this tends to be limited to a single payment of a predesignated amount or a series of payments that reduce over time and end at a specified date). This leads us to conclude that incentivised collaboration remains the most prominent driver for this kind of activity. While monetary enticements can be effective in the initiation and early phases of initiatives, question marks remain over longer-term sustainability and motives for engagement. Reporting on the government-sponsored SBM Demonstration Project Programme involving collaborative models of school business management, Woods et al. (2013) make the point that: encouraging schools to engage in competitive bidding processes and other entrepreneurial activity to increase their funding were giving rise to adverse consequences . . . the government funds awarded to successful projects, rather than the espoused purpose of forming partnerships, were often a prime incentive for submitting a project proposal. (p. 762) Indeed, Hayes and Lynch (2013) suggest collaborative activity would be more likely to endure were schools incentivised to engage in such activity through other means: Direct funding can incentivise and facilitate collaboration, but there is little evidence from our study that partnership working will be sustained once funding has been removed, unless a simple model of partnership working is adopted or unless other, but equally strong, policy levers are utilised, such as performance indicator tables and inspections that place a value on collaboration. (p. 444) This, of course, raises more fundamental questions surrounding a policy environment in England in which schools are primarily in competition with one another and held to account through their individual pupil and school level data and inspection grades rather than their partnership activity with other schools.
Our findings also point to evidence of voluntary collaboration, in which schools have chosen to work with and support one another through the identification of a common interest or need. In some cases, such arrangements have utilised government-backed initiatives to build upon and bolster this existing partnership activity. This is encouraging and suggestive of a willingness and appetite within the system for schools to work together despite the competitive conditions in which they operate (Armstrong & Ainscow, 2018). This will need to be harnessed more effectively, if the growing trend toward academisation and MATs continues and policymakers are to fulfil their somewhat lofty ambitions of creating a truly self-improving school-led system.
Recent findings as to the impact and efficacy of academies, MATs and teaching schools suggest the long-standing and dominant forces of marketisation and hierarchy remain formidable barriers to collaborative activity (between schools) that promotes equity, inclusion and improvement (Greany & Higham, 2018;Hutchings & Francis, 2018).
2 What factors are likely to facilitate and/or hinder school-to-school collaboration? 342 P. W. Armstrong et al.
Our findings point to a number of factors that can support the conditions in which purposeful collaboration between schools is likely to occur. For example, the role of leaders and leadership in coordinating and managing collaborative activity, ensuring responsibility for such activity is dispersed and shared among stakeholders and to build capacity to support collaborative efforts. Relational factors also feature strongly, with many studies highlighting the importance of trust and clear communication between partners (i.e. the development of social capital). Linked to this are contextual factors such as a history and pre-existing culture of collaboration between partner schools.
The review also highlights a number of barriers to the initiation, efficacy and sustainability of collaboration between schools. Chief among these are threats to autonomy and perceived power imbalances between schools, issues surrounding additional capacity, resource (financial and human) and workload associated with collaborative activity and difficulties in establishing shared objectives and common goals between partners. The marketised education system that encourages and rewards competition is also a major obstacle for collaborative practice between schools. As Keddie (2014) contends: When schools are concerned with economic imperativessuch as competing with each other for their 'market share' of students and generating incomegenuine collaboration is undermined and a focus on students and learning is sidelined. (p. 3) Of course, none of this is particularly surprising. These are logical and well-reported factors that are likely to influence purposeful collaboration between schools. This does, however, underscore the peculiarities and paradoxical nature (Greany & Higham, 2018) of the educational policy context in England. The policy discourse is awash with aspirational statements about creating a self-improving system led by schools working together and supporting one another to collectively improve the educational attainment and life chances of the young people and communities they serve. In reality the school system is a marketised space in which schools are held accountable for their own outcomes and therefore encouraged to be introspective. Such conditions, as the empirical evidence suggests, are not conducive to purposeful collaborative practice. Reflecting on his work with groups of schools involved in the Greater Manchester Challenge, Ainscow (2015) offers an alternative perspective: In essence, it builds on the idea that within schools and the communities they serve there are untapped resources that can be mobilised in order to transform schools from places that do well for many children and young people so that they can do well for all of them. It also shows how an engagement with evidence of various forms can act as a catalyst for such developments, not least by making the familiar unfamiliar. (p. 14) In this sense, incentivising schools to collaborate is not sufficient. Rather, schools need to be carefully and contextually matched so that they are able to provide mutual challenge and critical friendship informed by evidence as to their strengths and weaknesses. Uncomfortable as this may be, it moves the collaboration beyond cosy friendships to a potentially more powerful and purposeful space.
Evidence as to how collaboration between schools might directly impact upon student outcomes is limited. Moreover, in the very few studies in which attempts have been made to explore this area the findings are mixed. That there have been so few attempts to study empirically the influence of school-to-school collaboration on pupil outcomes is perhaps suggestive of two things. First, that the complexities of what the collaborative activity involves and exactly how and why this might influence pupil outcomes have proven too big a barrier for researchers to overcome. Secondly, that we need a more accurate definition of what school-to-school collaboration involves and the different forms it takes before thinking about the kind of impact such activity might have.
Considering the consequences for teachers and teaching practice, the literature provides more clarity. Many studies report improvements in areas such as staff professional development and career opportunities and the sharing of practice and innovation together with reductions and realignments in headteacher workload and organisational and financial efficiency as a result of school-to-school collaboration. More specifically, the evidence suggests teachers are often beneficiaries of collaboration between schools, with practitioners reporting an increased motivation to engage in professional dialogue with their colleagues, knowledge mobilisation and a general shift towards more learning-oriented and enquiry-based cultures.
Even though there remains uncertainty and a paucity of evidence regarding how and why collaboration between schools might promote improved student outcomes, we contend this to be a narrow view of such activity. Taking a broader perspective, our review suggests teachers, school leaders and other educational stakeholders all gain from working collaboratively with colleagues outside of their institutions and that this is (indirectly) to the advantage of the educational experiences and outcomes of the young people within their schools and classrooms.

Future directions for the field
One of the distinctive features of the English school system is the competitive and market-driven policy environment within which it exists and which poses a formidable barrier to collaborative activity between schools. Based on the findings from this review, our strong contention is that empirical research into school-to-school collaboration is lagging behind policy and that this presents a challenge for a system increasingly underpinned by an expectation that schools will work in partnership with one another. As such, we now outline what we consider to be the areas that most require further scholarly attention in order to move the field forward.

Types of collaborative activity between schools
Although efforts to chart the different types of school provider within the English system have been made (see Courtney, 2015), we found no evidence of any attempts to capture the landscape of collaborative activity between schools and the range of structures and forms this takes. With continued increase of MATs within the system together with a range of other collaborative governance structures (e.g. Teaching Schools, Research Schools), there is work to be done to start to map this complex 344 P. W. Armstrong et al.
web of inter-school activity and practice. Furthermore, it would be useful to understand more about whether and the extent to which such structures do indeed lead to more collaborative activity between the member schools.
Collaboration between schools is by no means confined to the English context (see Pont et al., 2008aPont et al., , 2008b and it is likely that there are useful lessons to be learnt from other contexts. This would justify a similar exercise in mapping the international knowledge base on school-to-school collaboration.
Moreover, the multifaceted nature of collaborative activity between schools is reflected in the broad range of research in this area. It might therefore be prudent to consider fragmenting this field of study into smaller sub-fields that focus on specific aspects of collaboration between schools and the myriad of different practices and areas of educational provision that are currently enveloped within this broader terminology. This would facilitate a more nuanced understanding of this kind of activity and allow for more accurate judgements to be made about the quality and usefulness of research in this area.

Terminological and conceptual clarity
There currently exists a wide range of terms used to describe inter-organisational practice between schools with various definitions and a great deal of overlap and contradiction. If the field is to work towards a more accurate mapping of the landscape of school-to-school collaboration, it will be important to develop terminological clarity and consistency. This will facilitate a more nuanced understanding of the practical and conceptual features of such practice. For example, it might be argued that the school (as the organisation) is not the most useful unit of analysis when thinking about school-to-school collaboration. The growing body of work into professional learning networks in education (see Brown & Poortman, 2018), in which practitioners are typically viewed as the collaborative node, offers some useful insights in this respect.

Inter-organisational dynamics
Much of the literature on school-to-school collaboration focuses on the somewhat superficial aspects of function and process. There is less insight into the actualities of collaborative practice such as the brokering, development and nurturing of relationships and the organisational changes that materialise when schools collaborate. For example, recent research with a research learning community within a network of schools has offered new insights into leadership practices across schools working in partnership (see Brown et al., in press). More insights are needed in this space.

Theory
As we alluded to at the start of this review, in the few instances where authors have utilised theory to think about collaboration between schools this has created potential for a more nuanced understanding of the challenges and opportunities that emerge through such activity. In addition to supplementing the knowledge in this area of the School-to-school collaboration in England 345 field, there are practical applications to theory development whereby conceptual thinking can help to 'understand the complexity that underlies collaborative situations and convey it in a way that seems real to those involved, even though they could not have explicated it themselves' (Huxham, 2003, p. 419).
Based on our findings, we contend that despite these possibilities empirical studies in this area of the field remain lacking in theoretical substance and conceptual clarity. More work is therefore required to combine and test out (new and existing) conceptual models with collaborative school activity and practice. Combining the theoretical with the empirical will help to develop new insights into the dynamics of and motives for such activity and also inform practice and policy in this area.

Critical perspectives
Linked to the previous point, given much of the research we reviewed is functional and descriptive in nature, there is little in the way of critical insight or perspective into the realities of collaborative activity between schools. Bucking this trend, recent research into the school-led system in England suggests traditional hierarchies and power structures are being reinforced and even strengthened by the growing number of MATs (Greany & Higham, 2018). There is more work to be done to problematize partnership activity between schools and to throw light on the 'dark side of collaboration' (Chapman, 2019).

Educational progress and outcomes
Acknowledging the complexities of isolating the direct influence of school-to-school collaboration on student learning, there remains a dearth of evidence within this area. Although we contend that this forms a narrow view of such activity (and acknowledging the definitional issues already discussed), given the high stakes accountability structures under which they must operate, it remains important for schools to be able to demonstrate that working with other schools will improve educational outcomes. There is, therefore, a need for more research in this area building on that which has already been undertaken (see Chapman & Muijs, 2013.