Early childhood education leadership in Finland through the lens of structure and agency

In early childhood education, the concept of distributed leadership has emerged as a key analytical tool for understanding leadership as well as a normative guide for what leadership should be. The concept originates in Peter Gronn’s work, where it is positioned as overcoming the structure-agency debate, which is a foundational question in the study of social reality. While distributed leadership itself has been extensively studied, the problem motivating Gronn’s work—the structure-agency problematique—has rarely been investigated. In an effort to create a deeper understanding of the role of structure and agency in constituting early childhood education leadership, this study examines how these two key dimensions of social reality structure early childhood education center leaders’ understanding of leadership. The data for the study consist of focus group interviews where early childhood education center leaders discuss various aspects of leadership. The data are analyzed in the broad framework of post-structural discourse analysis, using the analytic concept of frame, which reveals the interplay of structure and agency in early childhood education leaders’ understandings of their work. The findings show that early childhood education center leaders’ understanding of leadership is mainly focused on the side of structure and offers few chances for the kind of collective effort hoped for by Gronn.


Introduction
Scholarly and policy discussions in early childhood education (ECE) have recently begun to reflect a broader trend in educational leadership theory by highlighting distributed leadership (DL) as a key analytical tool for understanding leadership as well as a normative guide for what leadership should be (see e.g. Heikka et al., 2012;Heikka and Hujala, 2013). As Heikka et al. (2012: 30; see also Bolden, 2011) point out, the recent interest in DL originates in a seminal paper by Gronn (2000), where he proposes an interpretation of DL as a solution to a more fundamental disagreement about the nature of leadership. He describes leadership theory as polarized around an idea of a superlative individual transforming institutional practices, on the one hand, and an idea of institutional practices and structures swamping the individual on the other (Gronn, 2000: 317). Lurking behind this polarization is an even more fundamental problem of social theory: the structureagency debate, that is, do the actions of agents constitute social structure or does the social structure constitute the agents and their actions? In the context of this study, this translates to: do the actions of leaders constitute institutional structures or do the institutional structures constitute the leaders' actions? This debate has a long history (Archer, 2000;Giddens, 1984;Howarth, 2000) as a foundational question of social ontology, but Gronn (2000: 317-318) argues that if we are to move beyond the simplistic polarization in leadership theory, we need to find a way to accommodate both structure and agency as forces that shape the social world.
Several studies have investigated DL in ECE since Gronn's original suggestion, yet his claim about the polarization of leadership around structure and agency has received relatively little attention. While structure and agency appear to be implicitly present in ECE leadership studies (see e.g. Heikka, 2013;Sims et al., 2018) and have been employed as analytic categories (Colmer 2017;O'Gorman and Hard, 2013), the interplay of these two dimensions has not been investigated. In fact, earlier studies seem to have taken for granted the understanding of structure and agency promoted by Gronn, who sees structures and agents as mutually constituted in an interplay through time (see e.g. Colmer, 2017;Dovemark, 2010: 233;Eteläpelto et al., 2013;O'Gorman and Hard, 2013;Varelas et al., 2015;Woods et al., 2004). Although distinguished theoreticians of the structure-agency debate have defended such a position (Archer, 1982(Archer, , 2000Giddens, 1984; see also Archer, 1995as cited in Gronn, 2000, there are substantial theoretical problems inherent in the view, as can be seen in a related discussion in international relations theory (Doty, 1997). It therefore seems that Gronn's initial assumption about the polarization of leadership around the structure-agency debate merits more explicit attention.
The aim of this study is to examine whether ECE leadership is polarized in terms of structure and agency, that is, whether Gronn's initial assumption about leadership holds in ECE as well. The focus of the study is therefore not DL per se, but rather the assumption about leadership on which Gronn's concept of DL is based on. The following concrete research questions help in realizing my general aim: 1. What kind of phenomenon is leadership in ECE, when viewed through the lens of structure and agency? 2. What kind of subject positions are constructed for ECE leaders in their discussions of leadership? 3. What kind of general discourse of ECE leadership emerges in these subject positions?
The structure of my paper is as follows. I first outline my understanding of the interplay of structure and agency, based on the concept of 'practice' suggested by Doty (1997: 375-379) and a poststructuralist view of social reality (Howarth, 2000;Laclau and Mouffe, 2001). I then describe the methods of the study as well as the concept of frame (Goffman, 1974), which helps in applying this perspective on structure and agency to an analysis of ECE leaders' perceptions of leadership. Next, I present the results of my analysis, that is, the frames that ECE leaders use in their talk of leadership. I expand on these in the discussion by connecting the frames to their socio-cultural background, which allows identifying the general discourse of leadership that emerges in these framings. Having thereby answered the research questions, I conclude by considering what these answers can tell us about the applicability of Gronn's claim about the polarization of leadership as well as the implications for ECE leadership research and practice.

Structure and agency
As was already noted in the introduction, Gronn's understanding of the structure-agency debate focuses on giving both structure and agency the constitutive power they deserve (Gronn, 2000: 318; see also Archer, 1982Archer, , 2000. In his view, this is achieved by seeing the two dimensions of social reality as interacting through time. According to Gronn (2000: 318; see also Archer, 1982), while an existing social structure is always a precondition for the actions of future cohorts of agents, it also represents the outcome of the purposive, generative (as well as unintended) actions of previous sets of agents and is potentially modifiable by agents yet-to-be-born.
However, introducing time-morphogenesis (Archer, 1982;Doty, 1997: 373)-as a solution to the structure-agency debate leads to contradictions with Gronn's own ontology, where leadership is seen as socially constructed (see Gronn, 2000: 320). Doty (1997) develops two arguments that can be used to show the problems with Gronn's understanding of structure and agency. First, Gronn states that structure is a precondition for the actions of agents. When combined with the autonomous power of the agent, this raises the obvious question of, "if an agent could have acted otherwise, why did he/she not act otherwise?" (Doty, 1997: 373). As Doty (1997: 373) notes, in order to answer this question, one would need to admit to either of the following: (a) the agent did not act otherwise because of the existing structures, in which case the structures would obtain a determining primacy over the agent or (b) the agent did not act otherwise because of his/her own reasons, in which case the agent would obtain a determining primacy over the structures. In both cases, one element becomes primary and the chance of reaching a mutually constitutive relation is lost. This is in direct contradiction with Gronn's (2000: 318) explicit statement that he seeks to overcome the emphasis on either the superlative individual transforming institutional structures or institutional structures that determine the actions of the individual leaders.
Second, Doty (1997: 373-374) discusses whether introducing time into this conundrum solves the contradictions highlighted above. In international relations literature, it had been suggestedas it has by Gronn (2000: 318; see also Archer, 1982) -that understanding structures as both predating and post-dating the actions of agents would solve the problem of mutual constitution. The problem with this solution is that, in any real-life situation, it is impossible to escape the prison of the present. At any given time, either structure or agency has to be accepted as primary. (Doty, 1997: 374.) This, however, raises the ontological question of how to understand the backgroundwhat happens to structure when agency is in the foreground and to agency when structure is in the foreground. To solve this problem, both agency and structure would need to be posited in 'essential' terms-as having an independent existence outside their interplay-leading to problems with mutual constitution along the lines discussed above.
I argue, based on Doty (1997) and post-structuralist social theory (Howarth, 2000;Laclau and Mouffe, 2001), that these problems can be avoided by relying on a third choice: the agent and structure are both constituted by "practices." In ECE, practices constitutive of agency and structure might be, for example, official meetings, everyday discussions or even email exchanges. The concept of "practice" is elaborated by Doty (1997: 375-379) as an alternative constituent of both structure and agency built on two fundamental insights that have been developed in the poststructuralist literature. First, practices always contain a signifying element-practices signify meanings (Doty, 1997: 377). Furthermore, all practices are articulatory in the sense that they establish the meanings they signify (Laclau and Mouffe, 2001: 105, 113). Second, when understood in the poststructuralist framework, there is no necessary link between the signifier and the signified. What the practices signify-their meaning-is not fixed because the relationship between a sign and its meaning is arbitrary, being constructed by differentiation from other signs instead of by representation of an "extralinguistic entity" 1 (Doty, 1997: 377). The articulated significations inherent in the practices are never complete but unstable and subject to possible dislocation (Laclau and Mouffe, 2001: 106-107, 110-111).
Defined in this way, "practice" differs in an important respect from the concept of activity Gronn (2000: 326-329) sees as a bridge between agency and structure. Gronn's concept is grounded in socio-cultural activity theory. Crucially, as noted by Eteläpelto et al. (2013: 55-56), the strand of the socio-cultural framework Gronn uses posits activity as a closed system of meaning-in poststructuralist language, the relationship between the signifier and the signified is a fixed one. Thus, it cannot offer space for agency and consequently needs to take agency as a power external to the activity itself, reproducing the problems with the mutual constitution of structure and agency discussed above. By contrast, "practice," in its openness with respect to its meaning, can constitute both agents and structure without the determinism inherent in the model used by Gronn (Doty, 1997).
This distinction and, consequently, the way "practices" constitute structure and agency, are centered on the notion of "subject position" (Howarth, 2000: 108-109;Laclau and Mouffe, 2001: 114-122). "Subject positions" can be understood as ways of positioning speakers within discursive structures (Howarth, 2000: 108). For example, in the context of ECE, the leader could be seen as someone who knows better, as someone who can give orders or as someone who can give you advice in difficult situations, and so on. Such positioning is made available and sustained by practices. However, since the meaning of a practice is not fixed, the subject positions themselves are unstable (Howarth, 2000: 109). The instability inherent in practices makes possible a process of "dislocation," where the existing subject positions cannot confer stable identities to social actors. As a result, social actors are compelled to "create and identify with" new discourses 2 . (Howarth, 2000: 121) Thus, it is the inherent instability and malleability of the significations that confer meaning to practices that constitute the participating actors as agents.
This further implies that, although both structure and agency are constituted by practices and thus lack any ontological independency, it is nevertheless necessary to analytically separate the two (cf. Archer, 1982;Eteläpelto et al., 2013: 62;Gronn, 2000: 318). To do this, I employ Emirbayer and Mische's work on agency. They (Emirbayer and Mische, 1998: 971) distinguish between the iterative dimension of agency, which reproduces past courses of action, and the projective dimension of agency, which imagines alternative futures. It is certainly clear that Emirbayer and Mische's work is not based on a post-structuralist framework. However, their iterative and projective dimensions can be interpreted as analytic categories that highlight the relative stability inherent in existing subject positions (iterative) and the openness inherent in dislocation (projective). Emirbayer and Mische's work gives a concrete form for the analytical distinction between agency and structure: the analytic category of structure represents the stability inherent in practices while the category of agency represents their malleability (Archer, 1982;Howarth, 2000;Laclau and Mouffe, 2001). Their categories thus provide an operational definition of structure and agency, which I discuss in more detail in the section "analysis procedures" below.
In this study, I interpret ECE leadership as a "practice" and use ECE leaders' discussions of leadership to map out subject positions and possible dislocations constituted in the "practice" of ECE leadership. More specifically, I examine what kind of meanings frame (Goffman, 1977) ECE leaders' understanding of institutional structures. This allows an analysis of leadership in terms of structure-the stable identities conferred to the leaders-as well as agency-possible dislocations that call into question existing subject positions. ECE leaders are people that are involved in the practice of leadership on a day-to-day basis. This suggests that the meanings they employ when discussing leadership are constituted primarily by the practice of ECE leadership. It is therefore possible to claim that ECE leaders' understanding of structure and agency is indicative of what kind of structure-agency relations are constituted by the practice of ECE leadership.

Methodology
I employ post-structural discourse analysis as a methodological framework (Howarth, 2000;Laclau and Mouffe, 2001). Discourse analysis is a methodological approach that, perhaps more than most, is defined by its ontological and epistemological claims (Gee, 2010: 11;Jokinen et al., 2016;Mills, 2004). Although many of the ontological principles have already been explored in the discussion of structure and agency, I repeat here, mainly for the sake of clarity, the key aspects of the post-structuralist understanding (Howarth, 2000;Laclau and Mouffe, 2001). Laclau and Mouffe (2001: 105) "call articulation any practice establishing a relation among elements such that their identity is modified as a result of the articulatory practice." Furthermore, for Laclau and Mouffe (2001: 105), discourse is "the structured totality resulting from the articulatory practice," while differential positions are moments when "they appear articulated within a discourse" and elements when they are "not discursively articulated." It is important that a distinction be maintained between discourse as a result of articulation and the discursive field as a general background (Howarth, 2000:103). Elements are members of the discursive field but not moments of a discourse. Any particular discourse is articulated as a totality but is always vulnerable to the exteriority of the discursive field, which, by contrast, is not a closed totality.

The analytic concept of "frame"
Under the general umbrella of discourse analysis, a broad range of approaches to the actual analysis can be discerned (Jokinen et al., 2016). In this study, the analytic concept of "frame" (Goffman, 1977: 11; see also Jokinen et al., 2016: 283) is used to identify different subject positions (see Howarth, 2000: 101-102) occupied by the leaders with respect to institutional structure. "Frames" are interpretive perspectives that organize and structure experience (Goffman, 1974: 13). Framing is primarily about making choices of what aspects of reality are brought to focus-different frames emphasize different aspects of social reality. This understanding is very much in line with the notion of "subject position," although Howarth (2000: 3) considers frames to belong to a different strand of discourse analysis. I use 'frames' as a lens that opens a view from within the practice constitutive of the subject positions (cf. Howarth, 2000: 3). Although the choice of framing is dependent on individual actors, the frames themselves are not the property of individuals. Rather, "frames" are experiential manifestations of the meanings inherent in "practices" and individuals are often not aware they are "choosing" to use a frame 3 . (Goffman, 1974). To put this in the language of Laclau and Mouffe, frames are moments of a discourse as experienced by the ECE leaders. Interpreted in this manner, frames offer a lens that reveals the subject positions and dislocation the leaders experience to be open to them in the practices that constitute leadership.

Data
This study is part of a larger research project under way in the University of Helsinki, called "Discourses of leadership in the diverse field of early childhood education." Data for the project was collected through thematized focus group interviews where ECE leaders and practitioners discussed their views on leadership. The focus group discussions were facilitated by the researchers involved in the project, with the help of a set of guiding questions. The guiding questions (see appendix 1) were decided in a discussion among the researchers participating in the project and were based on earlier ECE research on leadership. The data for the analysis presented in this paper was obtained from leaders' focus group discussions in three Finnish municipalities. ECE in Finland (see Kumpulainen, 2018) is organized around daycare units. One unit might include two or three daycare centers. In this context, "leader" means the person in charge of a daycare unit. "Practitioners," by contrast, are responsible for the work with children. A leader is thus part of middle management: in the organizational structure they supervise the practitioners but are answerable to 'upper management' where many of the strategic and financial decisions are made.

Participants
Five, four and two leaders were present in the three focus group discussions, respectively. Each participant held a position as a leader of a day care unit at the time of the focus group discussions. The participants were recruited via email. Some of them responded to an open invitation sent to all ECE center leaders in each municipality, while others were contacted personally by the researchers to ensure a sufficient number of participants. There was limited interest in participating in the study and all the interested leaders were accepted as participants. The leaders came from diverse backgrounds and their career lengths varied, yet they cannot be considered a representative sample of Finnish ECE leaders. In particular, since participating in the study was the result of active action on the part of the participants, they are likely to have above average interest in professional development and leadership theory. To counter the possibility that the analysis might only be able to reveal the opinions of these specific leaders, the results of the analysis were carefully contextualized with reference to existing literature regarding Finnish ECE (see the section "cultural constitution of frames" below).

Analysis procedures
First, from the initial data source, two data sets were created, one for the iterative dimension and another for the projective dimension (Emirbayer and Mische, 1998). All instances where the center leaders discussed the use of existing patterns of thought or action were collected into one data set representing the analytic dimension of structure and all instances where they discussed alternative possibilities into another, representing the analytic dimension of agency. In cases where a passage of text displayed both categories, it was included in both data sets since, in any given passage of text, several frames can be present simultaneously (Puroila, 2002: 36). For the sake of clarity, I refer to these two data sets as the iterative and the projective data set, respectively.
In categorizing the data according to the iterative and projective dimensions, a few clues were used to ascertain the correct category for a passage of text. These clues were based on the features Emirbayer and Mische (1998) ascribe to the elements. For the iterative category, past tenses of verbs used to describe common practices were a clear clue, since it implies the practices have existed for a while and are "ongoing," so to speak. In addition, mentioning upper levels of management as the source of a given practice was taken as a clue for including the passage of text in the iterative category. Although a practice defined by upper management might be new for the leader in question, implementing a pre-defined course of action is clearly a case of reproducing an already existing institutional practice rather than imagining an alternative future (Emirbayer and Mische, 1998: 971). Thus, for example, "we also have proper general familiarization procedures defined by the upper management for new permanent employees" (Leader 1) 4 was categorized as iterative, since reference is made to upper management as the source of the procedure being implemented.
In the above example, the present tense "we have" is used to describe something as already existing. However, since there is no future tense in the Finnish language, no corresponding clue could be used for the projective element. Consequently, the presence of the conditional modus was used to indicate instances of imagining alternatives. The following excerpt does not explicitly refer to the future: "if you could sort of stay one step ahead and not have a feeling of barely keeping your head above the water" (Leader 1). However, the use of the conditional modus in the expression "if you could" clearly makes this an imaginary, "as if" situation, justifying its categorization as projective. In addition to the conditional modus, explicit reference to change was considered an indication of imagining alternative futures.
After the initial categorization of the data into two sets based on the concepts elaborated by Emirbayer and Mische (1998), the next step in the analysis was to identify the different frames used to talk about each of the elements. Based on earlier works employing frame analysis in ECE (Puroila, 2002: 37;Paananen et al., 2018), it was necessary to establish guiding questions for my analysis. Since the aim of this study is to inquire into how ECE leaders position themselves with respect to institutional practices, the guiding questions should highlight the possibilities the leaders see in either reproducing or transforming existing practices. Based on this, I limit myself to one analytical question: What purpose do the existing practices, or the imagining of alternatives, serve, according to the leaders?

Iterative frames
Three frames were identified within the iterative element: (a) enabling the enactment of the core function, (b) creating a shared understanding by keeping everyone informed and (c) ensuring work well-being and social justice. I discuss each in detail below.
Enabling the performance of the core function. A key purpose for upholding existing institutional practices is to make sure the core function of the institution-providing education, care and instruction for children-can be performed as well as possible. The following excerpt is illuminating: I think that leadership is about enabling, the structures and resources need to be in order...substance management is what is always talked about like do we have time for that and how can we do it but I feel that it is possible with good structures like the ones that have been constructed and created for years in [town one]. (Leader 1) The leader does not actively participate in enacting the core function of the institution. Instead, she creates the conditions for others to do so. It is also important to note that, although the leader uses the word "create," the existing practices are something that have been created during a time span of many years. In addition, the use of the passive voice makes "no-one" the creator of these structures-they are constituted in the practice of ECE. When it comes to the management of substance, it is only the good structures and their smooth operation that makes it possible. Thus, the leaders are in line with Emirbayer and Mische (1998: 971) in seeing that the iterative activation of the familiar institutional practices sustains the very core of the institution.
Creating a shared understanding. The second frame for repeating habitual actions is connected to knowledge. "These structures, for example weekly meetings, how you organize those, you should have written agendas, who is responsible, who writes the memo and you have to read the memos to know what has been decided" (leader 3), "for knowledge transfer we also have established structures so that you send quick ones with an SMS and use the email for less urgent matters" (leader 1). The existing practices ensure that knowledge gets transferred to everyone involved. The structures also guard against people getting different information: "the system [of familiarizing new employees] should be the same for all of us" (leader 2). Structures not only make sure knowledge reaches all the relevant people. Habitual courses of action also embody knowledge that makes everyday action easier: "we know who to ask what and who is responsible for what" (leader 2). Without clear inscriptions for what to do in a given situation, everyday decision-making would be chaotic rather than safe and comfortable.
Ensuring work well-being and social justice. In addition to the perspectives of core function and knowledge, the leaders also framed the iterative element of their agency from the point of view of work well-being.
You need to think and prioritize whether I just answer myself and tell them [upper management] that our unit has already handled this issue --our upper management is very pro-development and we have a lot of different projects and so on, so in the field they feel it's already too much...you have to find a reasonable limit for how much you can try to implement. (Leader 1) Here the leader describes a difficult situation where she is forced to choose between implementing a new practice defined by upper management and choosing not to implement it because the staff might become overworked. In reflecting on the decision, she frames the habitual courses of action as providing better opportunities for work well-being. Through work well-being, this frame obtains a dimension of social justice. "If a group is responsible for security matters and they make sure the updates are in order and information and training is available --you have to trust them and give them [the responsibility]" (Leader 1). Transforming the institutional practices is seen as potentially harmful since it might promote a feeling of exclusion, whereas sustaining a course of action agreed upon earlier is seen to provide a sense of inclusion and worth for the practitioners.

Projective frames
Within the data set thematized as projective, two frames were identified: (a) the role of the leader and (b) what changes? Identifying the second frame required considerable interpretation on my part and it is not comparable to the other frames. Instead of being an answer to the question guiding my analysis (What is the purpose of reproducing/transforming the existing institutional practices?), this frame is characterized by the lack of a clear answer. Because of this, I interpreted it as an example of "dislocation." The role of the leader. Throughout the projective data set, the leaders framed their talk of imagining alternatives from the point of view of the tasks constituting the leaders' job description. The following excerpts illustrate this frame nicely: It would be interesting to try out what I have spoken about [town 4] where they have had a shared leadership model in pairs like we have in upper management with one leader being responsible for personnel and finance while the other is responsible for customers and substance. (Leader 1) It would be possible to focus on your own responsibilities and there would be no need to try to be in all the places at once. (Leader 2) The leaders hope that the alternative future they imagine would clarify and simplify their job descriptions, thus allowing them to focus on the tasks they feel most at home with: "Clearly one aspect is closer to me...the other person could focus on the substance while the other could focus on personnel and finance matters" (Leader 1). Another, related dimension was the resources the leaders operate with: "The structures and a bit more staff and things would be fine" (leader 4). It is interesting to note that the alternative future expressed here is not imagined into being from nothingness-leader 1 refers to another city that has piloted the model discussed here and it seems plausible that the two leaders who know each other have discussed the model before. Thus, at least in this case, the imagined alternative is significantly influenced by what existing alternatives the person is aware of. In fact, the leaders do not imagine the alternative per se, rather, they consider what they see presently in their work and what they have seen elsewhere and compare the two. This shows how the agency of the leaders is not only constituted by the practice they are currently part of but also by practices they have participated in in the past 5 .
What change? Whereas the previous frame was present in much of the discussion categorized into the projective data set, this frame emerged only at one very specific point in one of the focus group interviews. In the focus group discussions, the facilitators (researchers involved in the project) used open questions to guide the group's discussion (see appendix 1). One of these guiding questions was meant to direct the discussion towards critical perspectives by asking the leaders what they would like to change in leadership/management of ECE. In one of the focus group discussions, the following exchange occurred at this point: In the second line of the third utterance (Leader 2), the "role of the leader" frame makes an appearance. I will leave this aside for the moment because the other parts of the exchange are more interesting. The long pause after the facilitator's question already implies what is stated explicitly by leader one: this question has not been considered before. This is surprising to say the least, since the idea of developing organizational culture, institutional practices and pedagogical content is central to the role of the leader in Finnish ECE (see OPH, 2018). Since it does not seem likely that the leaders would not consider themselves as part of the ECE center that they are developing, an alternative explanation is required.
I argue that the crucial issue is the wording of the question: "what would you like to change in ECE leadership?." This leads to considering how the leaders understand change. The following, rather long excerpt from the same session helps to delve deeper into this issue: Leader 1: We have so many things we need to know and we get so much information from upper management that we need to implement Leader 2: And often in a hurry Leader 1: Yes and from below they expect less development work...so we are kind of in between like from the above they expect us to implement this develop that and we have to choose what is reasonable Leader 2: The new national curriculum and such that have required work and still require work . . . a lot of challenging groups of children.
The key phrase here is "we are kind of in between like from the above they expect us to implement this and develop that [and from below they expect not to develop]." On the one hand, change is expected by upper management. On the other hand, there are expectations from the practitioners of knowing what to do and how to handle the difficult issues of curriculum planning as well as pedagogical challenges with the children. The lack of a stable subject position that emerges in response to the facilitator's question of change can be seen as a point of dislocation, a point where the significance of leadership is in a state of flux. The various expectations placed on the leaders-the guidance with the challenges of pedagogy from 'below' and demands for development from 'above'-displace any stable meaning that would give direction to the leadership practice. Moreover, as these conflicting significations are placed on the leaders from the outside, they come to perceive the idea of changing institutional practices in an essentially reactive way. This in turn leads to a crucial narrowing of their agency: the chance of transcending the given, made possible by the point of dislocation and central to many theories of agency (see e.g. Matusov et al., 2016: 435), is lost.

Discussion
The analytic concept of frame is meant to reveal how subjective experience of social situations is organized (Goffman, 1977: 10-11). However, as noted earlier, the frames are not constructions created by the individual. Rather, they are moments of discourse which remains largely outside conscious reflection by individual agents but nevertheless organizes their experience (Goffman, 1977: 21). Consequently, the first task for the discussion here is to show how the frames identified in the data connect with the cultural context of Finnish ECE. After this, I continue the discussion begun at the end of the previous section by considering what kind of general discourse of leadership emerges in the interplay of the interpretive frames.

The cultural constitution of frames
All the frames identified in my analysis are familiar perspectives for anyone involved in ECE leadership theory or practice. In the early studies conducted in Finland around the turn of the millennium by Hujala and Nivala (see e.g. Hujala et al., 1998;Hujala and Nivala, 2002), it was already revealed that ECE leadership is a "multi-dimensional phenomenon, with various administrative and educational foci" (Fonsén et al., 2019). The first projective frame identified in my analysis connects with this defining feature of the profession. The leaders, undoubtedly familiar with at least some of these ideas, which are well known in Finland, frame their experience of leadership from the point of view of the tasks included in the job description. Crucially, this frame emerged in the projective data set, which implies that the leaders see this as something that requires change. Moreover, the leaders introduce as a basis for their suggestion of shared leadership a distinction between day-to-day management issues and content leadership which reflects the earlier finding of leadership as balanced between the twin tasks of substance management and administrative duties (Hujala et al., 2016). Analogously, the first iterative frame connects with the concept of core function, which has been central to Finnish ECE leadership (Fonsén et al., 2019).
The second iterative frame-the perspective of knowledge-does not have a similar straightforward connection with the central themes of ECE leadership research in Finland. However, in the context of Finnish ECE, we can find an influential idea that might make this frame possible. I am referring to the concept of participation, which has become central to Finnish ECE in recent years (see e.g. Kangas et al., 2015a;Kangas et al., 2016;Turja and Vuorisalo, 2017). In discussions about pedagogy of early childhood, the term has developed a noticeably different meaning than the one it has in everyday English. The concept of participation has appropriated some of the features of the concept of agency and is considered to involve a strong element of active engagement and influence (Turja and Vuorisalo, 2017). The roots of this transformation are in the UN convention on the rights of the child, where participation in decision making about their own affairs is promoted as a child's right. In participating in decision making about themselves, children gain access to knowledge that has previously been considered to belong only to adults. Thus, a concept central in defining what is considered good in early childhood pedagogy has its roots in the idea of giving access to knowledge. It is possible that this connection creates the discursive structure which leads the leaders to use access to knowledge to frame their views of leadership.
Finally, the third iterative frame-work well-being and social justice-connects with key features of Finnish ECE in that social justice has been identified as a defining feature of specifically Finnish ECE (see e.g. Paananen, 2017). To conclude, the frames connect with important perspectives found in the professional and wider cultural context of Finnish ECE where the leaders operate.
2. Discourse of leadership resulting from the interplay of the frames I left the second projective frame out of the discussion above on purpose. Since this frame was constructed out of the negation of the analytic question guiding my analysis, it holds a special position with regards to the other frames. It reveals that the leaders do not see a clear purpose in changing the practice of leadership. Based on my analysis regarding the idea of change in the data, it seems that this is due to a fundamental ambiguity in the expectations placed on the leaders from different directions. The leaders do not see this point of dislocation as offering a space, for example, for actively pursuing an interpretation of leadership as oriented towards substance management. Rather, it represents a difficulty that needs to be navigated.
A further consideration regarding the general discourse of leadership is the first projective frame. The purpose these leaders see in transforming existing practices is essentially technical, not substantial. What they propose should be changed is the distribution of labor. Nothing is said of what the leader should aim to do in their work, what is irrelevant in the job description or what might be missing. Thus, this frame (the role of the leader) is qualitatively different than the three iterative frames. Instead of proposing what should be done, it proposes who should do it. This makes the leaders' perception of their agency inherently instrumental-a means to a pre-defined end (Matusov et al., 2016; see also Woodrow and Busch, 2008).
The contours of a discourse about leadership begin to take shape. The leaders see the key elements of leadership, such as support for the core function of the institution, knowledge management and upholding social justice in the workplace, as something to be achieved through reproducing existing practices-through stabilizing the social reality. That is, through keeping things the way they are, good leadership becomes possible. Transforming existing practices is understood mainly in terms of achieving the same things more effectively or more comfortably while the openness of meaning present in moments of dislocation is perceived as a threat instead of a possibility. Thus, the discourse about leadership positions the leaders as instruments of upholding existing values and procedures and ensuring their efficient operation.

Conclusion
The conclusion seems clear: when viewed through the lens of structure and agency, ECE leaders understand their leadership primarily in terms of structure. Thus, in the context of ECE practice, understandings of leadership are not polarized in the way Gronn describes. This conclusion reflects Doty's (1997: 379) theoretical observation that, in the open field of articulatory practices, any fixed discursive structure is an effect of power. Since the position of ECE center leader is one of institutional power, it is no surprise that leaders' understanding of their work is dominated by creating and maintaining discursive order. This conclusion needs to be qualified with the limitations of the present study. Most importantly, there was a limited number of participants in the focus group discussions, which might result in the analysis showing a stronger emphasis on the individual opinions of these leaders rather than shared views. Taking this qualification in to account, the conclusion has implications for the future of the wider phenomenon of DL, both as a normative model for organizing leadership and as an analytic lens for studying leadership (see Bolden, 2011).
Taken as an analytic lens, of considerable interest would be a comparative study where a similar analysis with the one presented here would be conducted regarding the framings of agents holding different positions in institutional hierarchy. Such an analysis would reveal to what extent the significations inherent in the practice of leadership are distributed in an ECE institution. The present study also has the severe limitation of not being able to analyze how the different framings get realized in everyday practice. Thus, a careful analysis of everyday practices, such as pedagogical meetings, would be helpful in extending the scope of the present analysis. Such an analysis might, for example, reveal how existing structures constrain and enable present agentive actions (see Archer, 1982). In a similar vein, it would be potentially very informative to apply the analytical concept of the 'imaginary' in a manner similar to Paananen (2017). She constructed imaginaries of the purpose of ECE based on political discourse and analyzed which of these faded into the background in practice, thus revealing what type of vision of ECE survives in the actual practice in the ECE centers. By analogy, considering which and whose frames survive when decisions are made could reveal informal workings of power in ECE practice. This could help in constructing a more nuanced understanding of the relation between DL and institutional power-a perspective that has not received sufficient attention in the research on DL so far (Bolden, 2011: 261).
My conclusions also have implications with respect to leadership practice in the context of ECE. The conclusions of my analysis show that in ECE the problem for leadership is not one of polarization. Rather, the practice of leadership seems to lean heavily on the side of structure. In light of this, ECE leadership might benefit more from a dose of the superlative-individualtransforming-institutional-practices type of leadership instead of finding a middle way in DL (cf. Gronn, 2000). This could be realized, for example, in leadership training, by a stronger focus on argumentative skills. In addition, it is important to note that, paradoxically, understanding the organizational structure one is part of might help in developing an agentic identity for the leaders (Close and Raynor, 2010). Also, more flexible managerial practices between ECE centers and upper management could help in providing the space for the leaders to view their work as agentive, which shows the importance of noting the constraining and enabling influence of existing structures on agency (Archer, 1982). In the present condition, distribution of leadership risks operating as a method of managerial control (cf. Kangas et al., 2015b). If the main purpose of leadership is understood as the stabilization of the meanings inherent in existing practices, distributing it would only create a stronger culture of conservation instead of the community of learning and innovation hoped for both at the national (OPH, 2018) and the international level (Colmer et al., 2014).