Comparative Planning Research, Learning, and Governance: The Benefits and Limitations of Learning Policy by Comparison

In this article, the authors develop a perspective on the value of, and methodologies for, comparative planning research. Through comparative research, similarities and differences between planning cases and experiences can be disentangled. This opens up possibilities for learning across planning systems, and possibly even the transfer of best planning and policy practices across systems, places, or countries. Learning in governance systems is always constrained; learning in planning systems is further constrained by the characteristics of the wider governance system in which planning is embedded. Moreover, self-transformation of planning systems always takes place, not always driven by intentional learning activities of individuals and organizations, or of the system as a whole. One can strive to increase the reflexivity in planning systems though, so that the system becomes more aware of its own features, driving forces, and modes of self-transformation. This can, in turn, increase the space for intentional learning. One important source of such learning is the comparison of systems at different scales and learning from successes and failures. We place this comparative learning in the context of other forms of learning and argue that there is always space for comparative learning, despite the rigidities that characterize planning and governance. Dialectical learning is presented as the pinnacle of governance learning, into which comparative learning, as well as other forms of learning, feed.


Introduction: A First Mapping of the Field
In this article, we aim to provide a new conceptual frame for considering the possibilities and limitations of comparative planning research as a topic both worthy of investigation by itself and which sheds new light on the bigger theme of policy learning and policy comparison.
If we see comparison in the service of learning, we cannot escape the confluence of ideas and the clashes between approaches that have marked the broad field of investigations on learning in planning, policy, and public administration (cf. Gerlak, Heikkila, Smolinski, Huitema, & Armitage, 2018). Ideas of change and adaptation in planning and governance enter the discussion quickly, and it is here we encounter the rich field of studies on the diffusion, transfer, and travel of policies and ideas (Mukhtarov, 2014). Key concepts for our investigation thus include planning comparison, learning, travel, and change (Bennett & Howlett, 1992).
Governance systems change continuously, and this can be the result of intentional learning or not. If we speak of learning it is clear that some form of change is implied-either in thinking, in organization, or in actionas well as some form of intention and awareness (Gould, 2009). What is learned does not necessarily benefit the organization, on the one hand because it might be simply wrong or irrelevant as compared to the latter's stated goals and preferences, and on the other because private (individual) goals might override the goals of the organization, or those of the governance systems and the associated community. Learning, for example, can be focused on ways to circumvent existing rules about environmental protection (Beunen, 2006;Chapron, Epstein, Trouwborst, & López-Bao, 2017). Learning can also focus on ways to weaken public goods as, for example, reflected in the literature on lobbying (Bouwen & McCown, 2007;Mazey & Richardson, 2006). The literature on social learning, sustainability learning, or organizational learning often tend to ignore this amoral feature of learning.
Within the literature on the travel and transfer of policies and practices, one can distinguish several generations, with early understandings speaking of policy diffusion, policy transfer, and best practices. More recent generations became more context-sensitive and sensitive to the processes of transformation taking place in the travel (Mukhtarov, 2014;Newig, Kochskämper, Challies, & Jager, 2016;Reed et al., 2010). Concepts such as policy travel and policy mobilities tend to signal deeper awareness of such changes and adaptations: Policies change, are reinterpreted, and adapted to the receiving context. These recent bodies of literature tend to assign more ownership and agency to the 'recipient' of the ideas or lessons, and to the learning context. Also, they showcase a diversity in the entities that can travel: not just policies, but also plans, laws, informal institutions, concepts, narratives, images, metaphors, expectations, ideologies, principles of system design, procedures (e.g., of participation), assessment methods, and more. As the poststructuralists already knew (e.g., Bal, 2002;Eco, 1976;Kristeva, 1980), the travel of signs and concepts is always partly metaphorical, as they do not travel without people using them. Yet the people using them and learning through them are also influenced unconsciously by the properties of sign systems and discourses, by discursive shifts, and discursive migrations. Traveling and learning always combine.
These introductory notions on learning and travel provide the background for the perspective on learning modes presented in the next section. Each mode involves, to an extent, a combination of travelling ideas and learning of ideas, and each is acknowledged in diverse bodies of literature. In the subsequent section, we focus on learning through comparison, particularly in spatial planning. This is an exercise in theory building, meaning that we enter into a dialogue with established literature without identifying with one existing theory in particular.

Learning Modes
We distinguish four modes of learning for governance and planning systems: 1. Learning from the past; 2. Learning from other places (comparative learning); 3. Learning from experts and expert knowledge; 4. Learning through dialectic engagement (discussion).

Learning from the Past
Both organizations and governance systems can learn from experience (Dunlop & Radaelli, 2013;Newig et al., 2016). That experience, though, is never unproblematically accessible, as each system remembers and forgets in its own manner and links what is remembered in different ways to procedures of assessment and adaptation of decision-making (McCann & Ward, 2015;Stein, Michel, Glasze, & Pütz, 2017). We do argue, however, that reflexivity can be cultivated, so as to enhance deeper awareness of a system's past and how it affects the present system (Edmondson & Moingeon, 1998;Gherardi, Cozza, & Poggio, 2018;Voß, Bauknecht, & Kemp, 2006). This, in turn, can inspire modifications of decision-making, i.e., learning (Golden, 1992). Learning from the past in governance can entail many things. Increased awareness of how things actually worked in the past can shed a new light on successes and failures, on assets seen and unseen, on patterns of inclusion and exclusion, on plans which remained a paper reality, on mechanisms which helped the implementation of plans, or on big promising concepts which gained little traction in the community. The learning can focus on a particular project which succeeded or failed, on expertise now forgotten, but also on the mechanisms of learning, adaptation, and change that were present. A park one is now proud of was maybe perceived as a failure before, as an authoritarian imposition, or a by-product of a political crisis and a reparation attempt. Each of these observations can trigger its own responses. Maybe it is decided that the growing city needs more institutionalized ways to enable such projects. It is also possible that the park is now seen as a major asset and part of an identity one can build on, thus allowing it to guide future developments. Several theories can offer guidance regarding the self-analysis of organizations and governance systems, but we highlight here the potential of Evolutionary Governance Theory (EGT) as a theory focused on the way the past of governance systems shapes both its present functioning and its transformation options (Beunen, Van Assche, & Duineveld, 2015;Van Assche, Beunen, & Duineveld, 2014). For EGT, each governance system has a unique path, a unique co-evolution of actors and institutions, and of power and knowledge. Each governance path is marked by its own pattern of dependencies, or rigidities in its evolution, which enable and constrain future transformation options. Each governance system thus has unique capacities for learning and adaptation.

Learning from Other Places
EGT, and many recent perspectives on management, organization, and governance, explore the possibilities for learning from other places. Yet, it is important to emphasize that such learning cannot rely consistently and solely on the identification of 'best' or 'worst' practices, for the rather simple reason that what works in one context does not necessarily work in a different one. Just as what worked in the past does not necessarily work in the present anymore (Sheldrick, Evans, & Schliwa, 2017;Stein et al., 2017). Context, then, has a double meaning: the internal context of the governance system, which never remains identical to itself, and the external context, the community itself, its values, expectations, resources, narratives, and power relations (Czarniawska, 2001).
Why something was possible for a given system in a given context is only accessible as a narrative reconstruction that reflects the narrative world of the observing system (Downing, 2005). Switching to indicators for the assessment of success does not alter this; the indicators function on silent assumptions within often latent narratives (Apaza, 2009). Presentations of success and failure are also performances, with strategic aims, never merely descriptions (Van Assche, Beunen, & Duineveld, 2012). Furthermore, interpretations of success and failure are always shaped by a particular understanding of the situation, of both the internal and external contexts (Bunnell, 2015;Dunlop, 2017;McFarlane, 2010).
When attempting to learn from other systems, the features and contexts of the learning (observing) system, and that of the observed system, play a role. One can say that a higher degree of reflexivity in the observing system (see previous section) will help discern what can be learned from the observed system, i.e., to identify what could fit into the receiving context. As to the observed system, there the limits of observation of other systems always apply (Cilliers, 2001), and what can be observed will be interpreted through the categories of the learning system (Seidl, 2005). Intermediaries, such as experts telling about other systems, can help clarifying why something worked in a particular context, but the intermediaries also add a layer of interpretation, and often a particular interest (Sultana, 2011).
Of particular interest here is the analysis of modes of self-transformation, meaning here the mechanisms by which the governance system can induce change in and by itself (Van Assche et al., 2014). Self-transformation can rely on learning, but not necessarily so; it can also emerge from internal interactions, both strategic and routinely (Luhmann, 1995). If in comparative learning the focus is merely on the content of a policy, the outcome, or the form of a procedure towards a policy goal or outcome, without paying attention to the modes of self-transformation, then it is not clear at all whether an attractive model from elsewhere could be emulated with the existing governance configuration as starting point, nor if it would have the same effects on its evolution.
Self-observation and observation of others thus have to be sharpened simultaneously to enable learning through comparison (Alvesson, Lee Ashcraft, & Thomas, 2008;Luhmann, 1995). Importantly, these observations cannot be restricted to features of the system, tools, and to substantive choices made (e.g., a particular policy, plan, or law), but should also include how the system changes and can change itself. It requires observation and analysis of the (internal) dynamics of the systems. Benefiting from comparative learning needs an understanding of how both observer and observed change, and of how forms of learning are implicated therein. Reflexivity is key to both improving observation and selfobservation, but reflexivity itself has limitations, stemming from the fact that a system cannot observe itself entirely, and furthermore from the overburdening and slowing down of the observing system through accumulation of complexity-reflexivity as transaction costs and as impediment to action (Alvesson et al., 2008).
All this retains the possibility of learning from other places and from other governance systems. We highlight four mechanisms that make it easier to learn.
First of all, in western societies and a globalizing world, many assumptions, expectations, and features of governance are shared. We live in a functionally differentiated world, a globalized economy, most people in some form of democracy, and many ideas on basic values are similar (cf. Brans & Rossbach, 1997;Luhmann, 1995). Many narratives on success in governance are persuasive in many places, because people have at least overlapping expectations on good governance, its functioning, and its results (Bunnell, 2015;. Second, and consequently, governance and planning systems are open systems and therefore-through their interaction with other ideas, values, shared institutional (e.g., legal) frameworks, etc.-share characteristics with other systems, although differences and particular unique aspects remain (Allen, 1998;Buijs, Eshuis, & Byrne, 2009). Therefore, it might not be necessary to completely understand why something works, or doesn't work, as long as the analysis of conditions across places is sufficiently similar: It seems to work there, and we don't see any real difference in relevant conditions (Spicer, Alvesson, & Kärreman, 2009). Third, experimenting might be possible, accepted, and worthwhile (McCann & Ward, 2015). If we are not talking about a major investment in financial or political capital, or about a major overhaul of the governance system, and if the system can reproduce itself during the experiment, there might be no need for existential pondering (Van Assche & Hornidge, 2015). Fourth, it might be possible to change the context enough to make a proposed solution work (cf. Gerrits & Verweij, 2018). A new policy picked up in one place might not fit the learning system in its current state, but it might be possible to embark on a larger self-transformation in which the desirable policy might fit (Sheldrick et al., 2017). For such larger transformations, the observed system might also give inspiration, but not necessarily so; that idea could also come from self-analysis or from other places.
Different reasons for and forms of comparison might succeed each other. A North American city might look at Copenhagen and its success in promoting cycling, heritage preservation, green space development, and innovation in conjunction. It might first see the reason for success in the combination of policies, tries it out, and fails. Then, spurred by internal experts and active citizens, the city might hire a Danish consultant with extensive local knowledge. The consultant comes over, studies the American city, and proposes a different combination of policies, emphasizing innovation and downplaying heritage, while forgetting the whole biking part. Local politicians might now be fully awake, organize a visit to Copenhagen, as well as participatory visioning sessions, where the emphasis on innovation is picked up, but now in tandem with a new, slow traffic network, inspired by but not copied from Scandinavia.

Learning from Experts and Expert Knowledge
Governance systems can learn from experts, either internal or external. External experts include both academics and consultants. Consultants often have a financial motive, which might inspire copy/paste attitudes, to save time and to sell branded solutions. Consultants might also have a rich experience observing other organizations and governance systems and could therefore have a trust that academics lack, a trust rooted in perceptions of 'real world' testing and of efficiency, a trust sometimes necessary to trigger change (Fincham, 1999). Academics might have more time and creative freedom, yet might lack the experience, networks, and prestige to cause change. They are therefore often not recognized as the potential bringers of messages that management could not bring themselves. Internal experts can bring insider knowledge to the table but might not have the freedom to think and speak, and might also uncritically identify with the existing system, its problem definitions, grounding narratives, etc. (Fischer, 1990).
What enters the learning system and can spark understanding and change thus hinges not only on the content of what is offered, the manner in which it is offered, and the features of the learning system, but also on the roles assigned to different actors (Dunlop & Radaelli, 2013;Gould, 2009;Newig et al., 2016). Different people with the same message will be welcomed differently, and the acceptance of the message will be contingent upon a series of factors. Some of those have been highlighted in the first section of this article, and others include the positionality (the roles taken or assigned) of those introducing the expert knowledge supposed to bring change and induce learning (Alvesson et al., 2008;Spicer et al., 2009).
One can also understand the process of inserting expert knowledge towards governance learning as a series of translations and confrontations taking place in and between networks or systems (Sultana, 2011;Van Assche, Beunen, Holm, & Lo, 2013). The role of knowledge brokers as mediators or connectors has been highlighted (Hering, 2016;Reed, Stringer, Fazey, Evely, & Kruijsen, 2014). Governance systems often include a variety of experts, channels for external expertise to enter, and several centers and scales of decision-making. It is easy to see then how the effect of new knowledge on the learning system (i.e., the learning itself) is the result of a highly complex interplay and competition between governance actors (preferring a particular learning and direction) and between the knowledge brokers themselves (either preferring a particular policy direction or marketing of particular expertise; Hoppe, 2009).
Picking up the example of learning from Copenhagen in America, a local planning expert might have whispered an advice very similar to what the Danish expert said, with little impact. A different Danish expert might have been better informed about the North American city but lacked the prestige and networks of the one hired. Meanwhile, in the American city, infrastructure experts might have sidelined planners and landscape architects for a long time, so at first, Copenhagen was dismissed as too dense and difficult for car traffic. Later, the technical specifications of the bike lanes were scrutinized for car safety implications, while ignoring the context of the bike network and the linkages with open spaces and transit. The participatory visioning sessions might bring up calls for a different expertise, not perceived as present in either Copenhagen or the American city, associated with a forgotten indigenous heritage.

Dialectical Learning
In the literature on policy learning, there is a peculiar tendency to omit or forget the kind of learning that has been central to Enlightenment ideas of learning-the kind of learning that is in fact central to late modernist discourses on participatory governance, communicative planning, deliberative policymaking, etc. We speak here of dialectical learning: the production of new insights through discussion and deliberation. Indeed, if we take Habermas and others seriously, then we cannot present their view on deliberation as simply adding up preferences, and then grinding them up in a process of calculation supposedly producing a conception of the common good (Hillier, 2003). For Habermas-for the institutionalists interested in deliberative governance-deliberation entails discussion, a testing of alternatives or, at least, the creation of new knowledge out of the confrontation and combination of existing ideas (Flyvbjerg, 1998;Tewdwr-Jones & Allmendinger, 1998). It does make sense, therefore, to speak of dialectics.
Holding a belief in dialectical learning does not position oneself in the modernist tradition of policy and planning, as it does not necessarily assume an objective and universal truth, nor the idea that the best, most persuasive argument, is the most rational one. Dialectical learning for a governance system does have to be more than the mere construction of new insights or arguments; it has to entail an effect of those arguments (Fischer, 2009;Kennedy, 2016). Whether an adaptive response to the new insight happens, hinges, again, on a variety of parameters. The literature, in our view, does not fully elucidate those conditions, as there is usually an a priori normative; an embrace of a particular idea of rationality, form of governance, or procedure (Hillier, 2002;Voß et al., 2006).
Participation and deliberation come in many forms, and not every form is conducive to actual dialectic learning (Fischer, 2003;Reed, 2008). The aforementioned openings for external expert knowledge, and positions for internal experts, can both encourage and limit dialectic learning. From a systems perspective, one can argue that cultivating diversity within and between organizations is of the essence (Alvesson et al., 2008;Ashmos, Duchon, & McDaniel, 2000;Seidl, 2005). In order to spark dialectic learning, one has to start with truly different perspectives on the state of affairs. Often, governance systems-in the name of efficiency or of shared values, identity, or consensus, or of supposed agreement on 'best practices'-start from a situation of minimizing difference. Such institutionalization of an a priori agreement does not enhance sharp observation, nor does it stimulate the construction of different perspectives, which can then, in discussion, lead to new ideas that could become shared and trigger organizational learning. Since Machiavelli, we know that conflict can be productive too, that dialectical learning can be a matter of quiet deliberation, and of strong differences in interest and interpretation. Discursive production can be the result of both polarization and the attempts at later reconciliation or cooperation (Bennett & Howlett, 1992;Hillier, 2002;Van Assche et al., 2014).
Dialectic learning can take place in different settings, and in the learning from Copenhagen example, it happened in city council, in administration, at the visioning sessions with locals, when hosting the Danish guest, and when visiting Denmark. It was helped by the diversity of perspectives brought in and the diverse forms of comparison in the lengthy process.

Comparative Learning in Planning Systems
If we understand planning broadly as the coordination of policies and practices affecting spatial organization, then it is clear that planning is spatial governance (Van Assche, Beunen, Duineveld, & de Jong, 2013). We can understand planning systems as the set of actors (individuals, groups, organizations) and institutions (plans, laws, policies, informalities) which make up the configuration structuring spatial governance. Planning systems are always embedded in larger governance systems that represent special needs for learning, but also coming with particular obstacles for it (Nadin & Stead, 2008). The generation and sharing of knowledge across planning systems and from research to practice is at the core of what planning researchers do (Silva, Healey, Harris, & van den Broeck, 2015).
Clearly, spatial governance is imbued with cultural values; each culture has different ideals and acceptable modes of organization of space. Planning is linked to ideas of the good community, and to the pursuit of both collective and individual goods. This means that planning is likely to be a site of policy integration and at the same time an arena where different interests compete for greater influence on spatial organization. Moreover, planning is supposed to provide both flexibility (adapting to new public and private interests and goals) and stability (protecting property and reasonable expectations of transaction rules), which further entrenches a planning system in a locale and makes simple import of foreign practices unlikely to be successful (Beunen, Patterson, & Van Assche, 2017).
Any observer of the American planning perspectives of 'smart growth' or 'new urbanism'-each assuming that their recipe can be metabolized anywhere-can tell us that the American reception of their recipe is not an easy digestion, but instead a rejection or tough struggle, in an environment where property rights politics has shifted to the right (Platt, 2004). These two American approaches to 'good planning' also show how very different discourses on planning affect their implementation or non-implementation, including discourses rejecting spatial planning as such. Learning, then, becomes an unlikely event, as discussion and openness are either suppressed, or take the form of a debate where winning rather than dialectic learning is the goal.
Parallel to our observations on the limiting and enabling conditions for comparative learning in governance generally, we also observe that comparative learning does happen. Indeed, the idea of planning itself spread from town to town, before higher level administrations enabled it, and to an extent imposed it (Scott, 1998). Let us not repeat here our observations on governance and comparative learning, but instead specify what they could mean in the case of planning and how, despite their particular evolutionary rigidities, planning systems are not immune to comparative learning. We can speak here of planning systems that learn through direct comparison, and of the academic (outsider) benefits of planning comparisons, some of them with implications for comparative learning within planning systems.
In spatial planning systems, comparison can enter through various modes: The experts in planning them-selves (academic and otherwise) are steeped in comparative learning (Silva et al., 2015) and the professional associations they are members of tend to reinforce this thinking. External planning consultants tend to sell their familiarity with many other cases, especially with ideas that have 'worked.' Planning academics can do the same, but often retain more space for the application of new ideas; ideas which might, in turn, be derived from comparison of cases (Hillier, 2002;Kennedy, 2016). The situation is different, though, for many other brands of academics at work in planning. For engineers, ecologists, hydrologists, and others, the knowledge base is often more based on deduction and modeling rather than on inductive comparison (Van Assche & Hornidge, 2015). Their inclusion in the planning system influences comparative learning as it brings in other criteria of evaluation.
Participatory planning discourses expect higher democratic legitimacy by including more actors directly, while arguing that this also makes planning more efficient (avoiding conflicts later) and adaptive through the inclusion of local knowledge (Hillier, 2002;McFarlane, 2012). Where private commercial parties, either consultants or developers, take on a big role in planning, their experience comes to weigh in, with comparative learning more likely to be introduced through the experts (consultants and experts working for developers). Broadly speaking, one can notice a growing complexity of spatial planning and growing tensions between dialectical learning and expert learning (Fischer, 2009). Where systems attempt to become more participatory, and when this is taken seriously, new opportunities for dialectical learning-and from there, comparative learningmight arise.
Participatory planning is thus expected to solve the issue of knowledge integration in complex governance systems, while fixing the other problems mentioned. Knowledge integration was supposedly already covered by the diversity of experts working for the high modernist state, but that idea of state got in trouble for practical and ideological reasons decades ago (Scott, 1998). The issue of knowledge integration brings us to the issue of policy integration. Indeed, the two are linked, and the way policies are integrated in a spatial plan affects both the way knowledge is weighed and how it is integrated (Van Assche & Hornidge, 2015). This process always creates losers and winners (less and more influential knowledges) and it further shapes how the system can transform (cf. Alvesson et al., 2008). For example, if a spatial plan is structured around water as a first ordering principle, and engineering and hydrological knowledge underpins this first ordering structure, then this cognitive and spatial frame will determine which spatial changes are possible and which are not, and it will influence what other knowledge could induce a system change; as for instance in the Dutch Room for the River Program of the 1990s (Zevenbergen, Rijke, van Herk, Ludy, & Ashley, 2013). For this reason, some have argued for flexible policy integration, as coordination of knowledges rather than a cemented form of comprehensive planning with predefined spots for particular knowledges (Van Assche & Djanibekov, 2012).
Limits to participation and to the flexibility of policy and knowledge integration exist and this brings us back to the specific difficulties for comparative learning in spatial planning. First, the previously existing form of policy integration exerts pressure (Candel & Biesbroek, 2016). Some choices made are not easy to alter, even if one wanted it. Second, the planning system is expected to create stability and predictability as well as adaptivity. Comparative learning can be used here by both proponents of stability and flexibility, by referring to more stable or more adaptive systems, as part of their argument. If strong property rights are the focus of planning, and coordination and collective goal setting move to the background, comparative learning is still possible, as many institutional designs are still possible. These observations reiterate, however, that planning never operates in a vacuum, and that literally every premise of a planning system can be questioned if broader governance configurations shift. That is, spatial planning remains an arena; it is never only a factory or laboratory (Bunnell, 2015;Tewdwr-Jones & Allmendinger, 1998).
This reveals a third limitation of the desirable feature of flexible policy integration: Some forms of knowledge and some forms of policy integration are deeply entrenched because of deeply entrenched discourses, values, and narratives-either in the governance system itself (e.g., city administration is an engineer's domain) or of the broader community (e.g., we are a farming community; Czarniawska, 2001;Scott, 1998; Van Assche, Gruezmacher, & Deacon, in press). We see here again the importance of the embedding of planning in governance and governance in communities, for the enabling or limiting of learning.
Planning thus comes with ambitions of policy and knowledge integration, it is marked by its own history and the history and culture of the communities it operates in, it has to balance flexibility and stability and it functions as an arena where many processes of value creation (linked to activities which need a space) are decided upon. These features of planning systems shape the possibilities and limits of learning, and of comparative learning. Grasping the features quickly leads to questioning of formulaic solutions. It inspires doubts about easy recognition and import of 'best practices' independent of unique forms of policy and knowledge integration, power relations in the planning system arena, and the double embedding referred to. Non-learning and learning the wrong thing occur very often, spurred by the desire for easy solutions, for technocratic certainty, and for political glory and economic efficiency (Dunlop, 2017).
In the following section, we elaborate the discussion of comparative learning in and for spatial planning and we consider a few methods that could be of use in the arena of spatial planning.

Methods of Comparison
First, when attempting comparative learning, one has to decide what to compare: • The design of the whole planning system, as shaping learning modes? • The adaptive capacity of the planning system and its modes of transformation, as proxy for learning? • Forms of knowledge, of knowledge integration, or policy integration in the planning system? • The way the planning system is rooted in governance, in cultures, emphasizing embedding as shaping learning?
Second, there is the question of what the goal of the comparison is: • Understanding a feature of the observed system? • Answering a more general question?
• Helping a different system struggling with a particular planning issue?
Third, there is the selection of systems to compare between. This could entail: • A matter of sample size and composition (mostly in quantitative approaches); • A comparison between places and only secondarily their planning systems; • A targeted comparison with one successful area (e.g., Silicon Valley) towards emulation; • A comparison of the embeddings of planning in governance systems; • Comparing forms of competition between planning and other policy domains.
For comparative learning it is advisable to disentangle the specific feature from context: Does it work here because of a particular context, or may it work generally, across contexts? Answering this question can lead us to the analysis of various contexts in the observed systems, to understanding how they affect the observed feature, and to the careful observation of the feature itself in various systems: Is it actually the same with the same benefits? A third option here is to combine a mapping of planning and governance systems at the same time, which can then elucidate the linkages. A fourth approach is to focus on the shared contexts, which may make sharing of solutions easier (cf. Nadin & Stead, 2008). A promising method of comparison for the analysis of spatial planning and governance systems is Qualitative Comparative Analysis (QCA; Gerrits & Verweij, 2018;Verweij, 2017;Verweij & Trell, 2019). QCA is particularly useful for generating explanations about how and why planning processes or systems perform the way they do, or produce certain outcomes, taking into account explicitly the perceived complexity of the contextual en-vironments encountered (Gerrits & Verweij, 2018). The method is geared towards the comparison of systemswith cases as complex entities that consist of multiple aspects or features (cf. Byrne, 2005Byrne, , 2009. Cases are arranged in a 'truth table' that lists all the logically possible combinations of aspects (i.e., configurations), and the length of which is dependent on the number of case aspects or features considered (see Schneider & Wagemann, 2012). By pairwise comparing those configurations that show similar outcomes, and that differ in only one of the aspects, that aspect in which they differ can be eliminated as-in QCA-terminology-a necessary or sufficient 'condition' or 'cause' for explaining the outcome. As such, the method allows to disentangle (contextual) features that have explanatory value in specific cases from features that work across cases, i.e., that are context-unspecific (Verweij & Trell, 2019). Although QCA can be conceptualized to be able to trace the trajectory or development of planning or governance systems over time (Byrne, 2005(Byrne, , 2009, it is actually not designed to do so (Gerrits & Verweij, 2018), and other methods may be better suited to that kind of purpose.
The methods of path and context mapping, derived from EGT, can prove useful to understand the evolution of planning systems more in detail, as well as their embedding in governance (Van Assche et al., 2019). Application of these methods can have the benefit of combining several of the above-mentioned options. Time constraints can be an issue, and comparing governance paths still requires decisions: Which features of the paths do I want to compare and why? Which scales are relevant? Which periods? An additional benefit is that transformation mechanisms can be made visible. Each compared system has different features which can be explained through its context, but also through its mode of transformation. Even if a context might be shared, the transformation mechanisms might not be. When comparing systems, the grasp of transformation potential looks paramount, so mapping of features without understanding existing capacities can be pointless. Moreover, as we pointed out, learning becomes implicated in itself: In order to trigger comparative learning, one has to map out existing modes of learning in the system.
If the goal of the comparison is ambitious, bricolage and nesting of methods is highly recommended. It is unlikely that one method of comparison can tease out all the information needed to answer the research question and certainly to link knowledge to action. That means that broad methods such as path mapping can contain, or combine with, other methods, such as QCA, but also traditional methods of data collection and analysis, such as: participatory observation, interviews, survey's, descriptive statistics, process tracing, document analysis (including discourse analysis), cartographic analyses, or focus groups (either per case or when comparing; Sheldrick et al., 2017;Verweij & Trell, 2019;Wood, 2016).
We can make a distinction between comparative research in and for planning systems. If we consider com-parative learning within planning, or governance, then one can consider a variety of participatory methods, where comparisons can be included in a more structured manner (beyond the places the planners or council members have visited). One can think here of participatory visioning, charrettes, competitions, public debates, or participatory design, where either comparable cases are spelled out, or brought up during the activity (Innes & Booher, 2010).
Such participatory methods of comparison can also cultivate reflexivity in governance, which in turn can increase chances of discerning features of other systems which might translate well. Methods to encourage reflexivity enhance the conditions in which methods for comparison can be applied, as they encourage learning in its different forms, and the productive combination of forms of learning (Alvesson et al., 2008;Seidl, 2005). We referred to the relevant literature for such methods, and refer to our earlier remarks, yet can highlight here the importance of maintaining difference and discussion in governance, creating access to governance, and avoiding mixing logics (de-differentiation), rigid hierarchies, excessive policy integration and, a common issue, elimination of critical thinking by bureaucratic routines.

Conclusion
Comparing planning systems is as old as planning itself, and comparative learning is part and parcel of any governance process (Friedmann, 1987;Silva et al., 2015). We analyzed planning as spatial governance, as always embedded in governance systems. We situated comparative learning within a set of other forms of learning, which can entangle and enrich each other: learning from one's own past, expert learning, and dialectical learning.
Reflexivity is an important concept in understanding the possibilities of comparative learning and the possibilities of learning as such. Indeed, self-understanding and analysis of the own governance path makes other forms of learning potentially more productive, as it enhances the understanding of what would happen to knowledge in a learning system and how that knowledge could transform it. For comparative learning in governance it is easy to grasp that the observing system needs to be very well aware of its own features, transformation modes, and goals of the comparison. For comparative learning for governance the same applies, even if the observer is less implicated in the process.
Comparison which aims at using practices or ideas from other systems encounters special obstacles in spatial planning. Those are related to the function of spatial planning as a site of policy integration, its function of balancing flexibility and stability, and its deep roots in the communities whose space it organizes. Indeed, space is the expression of shared values, of cultures, as much as it is the expression of competition between values and narratives. What is possible in planning hinges on context, history, and contingent events. This applies to learn-ing and comparative learning, whereby governance and community are relevant contexts.
Planning is always an arena of deliberation and knowledge production. The methods of comparison have to avoid either assuming that planning systems are technical systems which can be rationally optimized, or that the methods of comparison themselves are neutral. Further, the observer, either inside or outside a planning system, can clarify for his/herself what the goal, the scope, the duration and cost, and the detail of the comparison is. This can then lead to a choice for certain methods, or assemblage of methods of comparison.
What any comparative planning analysis has to grapple with is the question to what extent the features of the observed system and its successes are a product of context, of a particular fit between system and context, of contingent events, or of specific performances of success. In other words: comparative learning has to transcend technicalities to achieve real results; it has to trigger dialectical learning.
Comparative learning does not take place in isolation from other forms of learning and the linkages between the forms of learning can be managed in order to optimize the effect. The point is not to take something and adapt it, but to figure out something about other places and systems, about oneself, to learn from theory, and to bring this together in discussions which produce novel insights, which enable dialectical learning. Cultivating reflexivity helps to link the different forms of learning. This fits the image of planning (and governance) as an ongoing and open-ended conversation. Comparison ideally becomes creative comparison, and improving learning capacity involves combining the learning modes in a more conscious manner.