What accounts for the emergence of a new interaction pattern? On generative mechanisms, constitutive rules and charging routines

Drawing on the notion of generative mechanisms as constitutive rules, this paper advocates a shift away from the notion of routines as sources of ongoing change and towards a rule-based understanding of routines as institutional facts. While the recent practice turn to routines studies has highlighted sources of endogenous change, this paper adopts a Critical Realist stance to investigate exogenous forces that account for the emergence of a new routine. To this end, the paper endeavours to analyse the passing of new legislation in the criminal justice system of England and Wales. By examining what makes the rules of the game change between the Police and the Crown Prosecution Service, the paper explains an instance of institutionalisation in the making. Theoretical and practical implications are discussed. Our contribution stresses that constitutive rules play a pivotal role for recognising, identifying and labelling organisational routines, thus generating order, stability and patterning.


Introduction
The concept of generative mechanisms is arousing interest in the social sciences (Mingers, 2014). Defined as the processes that generate social events (Mingers, 2014), generative mechanisms can shed a new light on the concept of organisational routines because they can shift the focus away from the notion of routines as sources of ongoing change and towards a rule-based understanding of routines (Geiger and Schröder, 2014). The rule-based perspective on routines is capable of filling important gaps in the literature such as 'how organisational routines emerge' and the 'influence of external shocks and disruptions on rules' (Geiger and Schröder, 2014, p. 186). It can also highlight important analytical differentiations between constitutive rules at the deepest level (the Real), shared interpretations, meanings or understandings (the Actual) and situated performances as 'doings' and 'sayings' (the observable level of the Empirical). As Becker (2005. p. 252) has stated, rules, that is, codified or normative behavioural expectations, 'might be among the underlying mechanisms influencing action and thus giving rise to repeated action patterns'. Indeed, 'it is the (shared) interpretation of rules that ultimately guides the performance of routines' (Geiger and Schröder, 2014, p. 179).
Drawing on the idea that routines functioning as procedural standards are a subset of rules (Ortmann, 2010), we distinguish between constitutive and regulative rules, the former being constitutive of new interaction patterns, the latter determining, regulating and controlling pre-existing interaction patterns (Searle, 1995). While constitutive rules relate to the constitution of meaning, regulative rules refer to the 'sanctioning of modes of social conduct' (Giddens, 1984, p. 18). More than being a difference of kind, the difference between constitutive and regulative rule is a difference of degree. For example, 'constitutive rules not only regulate but rather constitute the very behaviour they regulate, because acting in accordance with a sufficient number of the rules is constitutive of the behaviour in question' (Searle, 2005, p. 9). Furthermore, constitutive rules form a system and represent the background for organisational routines (Reynaud, 2005). By iterating upwards indefinitely, constitutive rules form a structure made up of several interlocked subsets each nested within a larger system. Akin to the assembly rules discussed in the process of organising (Weick, 1979), constitutive rules are the glue that holds organisational routines together.
Since organisational routines are about collectively-shared ideas, they call for an understanding of participants' shared interpretations of procedural standards (Okhuysen and Bechky, 2009). Indeed, we argue that an understanding of participants' shared interpretations of a routine is critical to understanding their interactions. To coordinate their actions, participants have to make sense of their routines ('the ways they do things around there'). But, in this sense-making process, they develop shared attitudes, beliefs and expectations which then serve to shape subsequent interactions in an orderly fashion. Accordingly, participants are bound to develop a reasonable amount of 'implicit agreement' about 'the appropriate meaning of information or events' (Finney and Mitroff, 1986. p 320) so as to reduce conflict and, indirectly, promote order and stability. By generating shared meanings, interpretations or understandings, constitutive rules orient participants on how to make sense of their routines, thus generating shared behavioural dispositions that inform their actions and interactions.
The rule-based perspective on routines is dramatically different from the mainstream view of routines as change. While the latter perspective argues that routines are continuously changing entities where individual agents are the core protagonists of change, the former perspective claims that organisational routines are built on stable rules and, therefore, that routines are sources of order and stability (Schulz, 2008). The routine-as-change perspective focuses on an understanding of routines revolving around double interacts where 'the unit of analysis in organizing is contingent response patterns, patterns in which an action by actor A evokes a specific response by actor B (so far this is an interact), which is then responded to by actor A (this complete sequence is a double interact)' (Weick, 1979, p-89). Since the double interact is the fundamental unit of analysis in the routineas-change perspective (Langley and Tsoukas, 2017), routines may be described as effortful accomplishments undertaken by individual agents. Nevertheless, the routine-as-change perspective raises more questions than answers. If routines are sources of continuous change, how does stability emerge in the first place? If routines are so fluid, how do they stabilise enough for 'entities' to emerge and be labelled as 'the ways we do things around here'? Hence, the impetus for this research is a deeper analysis of the complex phenomenon of how routines emerge (Sargis-Roussel et al., 2017). We define routines emergence as a process aimed at 'bridging' participants' understandings (Sargis-Roussel et al., 2017, pp. 102, 109), thus producing 'new, repetitive, recognizable patterns of concerted, individual actions, when no previous pattern has existed ' (Sargis-Roussel et al., 2017, p. 102). By using a Critical Realist perspective (Porpora, 2015) that takes a middle path between the subjectivist pole of reality as a social construction and the objectivist pole of reality as an objective facticity, we ask the following question: What accounts for the emergence of a new interaction pattern?
Challenging the assumption that routines are anchored into multiple, divergent and conflicting understandings (Pentland and Feldman, 2008), our purpose is to investigate the emergence of a new interaction pattern (or empirical regularity) stemming from an exogenous change involving the enactment of a new law. Since passing a new law (X) may 'count as' granting new powers (Y) in particular contexts (C) and since routines are about collective agreements, it follows that routines may be the result of a complex and interlocked nexus of generative mechanisms (labelled hereafter as 'count-as' mechanisms or, more simply, as constitutive rules). The form or syntax of constitutive rules is the 'X counts as Y in C' mechanism where 'Y' stands for an institutional fact, 'X' stands for its empirical instantiation and 'C' for the context (Searle, 1995). In addition to objects (e.g., written documents, handbooks, etc.), the 'count-as' mechanism can also be applied to persons, events and/or interactions (Hindriks, 2009). For example, the situated practice consisting of two-way interactions between Investigating Officers and Prosecutors (X) 'counts as' the 'charging routine' (Y) in the new legislative context (C) because participants have collectively accepted or recognised the two-way interactions as their new charging routine. Accordingly, in the X 'counts as' Y mechanism, X terms are not the organisational routine. Neither situated practices observed or experienced in the 'here and now' nor objects (or artefacts) are the organisational routine. Rather, organisational routines are Y terms (or collectively-shared ideas) that, broadly speaking, depict 'the way we do things around here'. As such, routines are crucial labels for identifying the state of affairs specified by the X terms because they presuppose collective agreements whereby participants stipulate objective facts. When enacted, such collectively-shared ideas (or routines) function as social standards, that is, shared or agreed upon procedures that orient or even dictate interdependent practices or behaviours (e.g., requesting a decision, consulting the Duty Prosecutor and then obtaining a charging decision is our standard operating procedure, that is, our charging routine). In this paper we highlight the role that constitutive rules play in the generation of shared meanings and understandings and, therefore, in the emergence of organisational routines. Our contribution stresses that constitutive rules play a pivotal role for recognising, identifying and labelling organisational routines, thus generating order, stability and patterning. The remainder of this paper is organised as follows. After this brief Introduction, the following section contrasts the routine-as-change perspective with the rule-based understanding of routines to investigate alternate lenses for studying organisational routines. The third section outlines our research design, data collection and analysis methods. The fourth section breaks down our data analysis into a sequence of five steps (Wynn and Williams, 2012). The fifth section discusses our findings and the sixth section concludes our work with theoretical and practical implications bearing on the study of organisational routines. Appendix Table A1 summarises the data sources used in the paper.
Background: contrasting the routine-as-change perspective with the rule-based approach to organisational routines In contrast to conventional approaches to organisational routines, which emphasised structure, Feldman and Pentland (2003) set out to bring agency and, therefore, subjectivity back into the picture of routines studies. Informed by Weick's (1979) pioneering work, they shifted the focus from routines as standard operating procedures to routines as dynamic practices undertaken by human agents. Whether analysed with structural or agentic lenses, however, routines are sources of order and stability (Schulz, 2008). Although their enactment may produce new practices and behaviour patterns, routines possess an intrinsic degree of stability that locks action into narrow channels that we recognise as largely the same (Schulz, 2008). Organisational routines are relatively stable and enduring features of organisational life. Yet, when researchers zoom in on them, organisational routines appear as dynamic accomplishments. The incorporation of stability and change in the same phenomenon poses a paradox because each is defined as the opposite of the other (Poole and Van de Ven, 1989).
Drawing on structuration theory, Feldman and Pentland (2003) developed a theory of organisational routines where structure is simultaneously the medium and outcome of agency, thus proposing an insightful way to solve the paradox of the (n)ever changing world. Feldman and Pentland's (2003) ground-breaking article has made a lasting contribution in the field of routines studies because it has laid out the foundation for the routine-as-change approach. As Howard- Grenville and Rerup (2017, p. 323) argue, this approach is consistent with a process (or practice) perspective because 'it sees routines not as entities that encapsulate organizational knowledge, but as emergent (i.e., coming into being only through specific performances) and generative (i.e., capable of producing continuity or change in the actions they spawn)'. More recently, the routine-as-change approach heralded by Feldman and colleagues has sparked a new interest in sociomateriality (Feldman and Orlikowski, 2011). Not only do structure and agency exist in relation to one another. The social (e.g., coordinated practices) and the material (e.g., technologies) too are considered to be inextricably related -'there is no social that is not also material, and no material that is not also social' (Orlikowski, 2007(Orlikowski, , p. 1437). Yet, outstanding issues still mire the routine-as-change perspective. In particular, this approach downplays the historical and institutional context where routines are embedded for the sake of highlighting only those aspects that are within 'the immediate control or even knowledge of the participants' (Mutch, 2016(Mutch, , p. 1184. Furthermore, the routine-as-change perspective cannot easily integrate continuous micro change with frame-breaking, exogenous change (Geiger and Schröder, 2014). Indeed, the routineas-change perspective conceptualises change as an endogenous process of variation-selection-retention where, in their enactment of the routine, individual agents are simultaneously building new interaction patterns that form the substrate for subsequent performances. Hence, according to the routine-as-change perspective, organisational routines are endowed with an evolutionary process leading to the emergence of patterns in variety (Foss et al., 2012;Mariano and Casey, 2016). In addition, the emphasis on variety, plasticity and fluidity has shifted the focus away from cognitive regularities and towards transient and ephemeral understandings of the routine which are in a constant state of flux. If routines are so fluid, how do they emerge in the first place? While routines scholars have made many assertions about the ephemeral status of routines, they have provided very thin arguments as to how that status is achieved and fixed, albeit temporarily (Hardy and Thomas, 2017).
A profoundly different approach to studying the paradox of the (n)ever changing world revolves around a rule-based understanding of routines (Frost and Tischer, 2014;Geiger and Schröder, 2014). The rule-based understanding of routines is premised on a reconceptualisation of structure that is consistent with Critical Realism (Runde et al., 2009). This approach offers novel insights into the study of organisational routines in general and organisational change in particular. First, it includes an explicit theory of morphogenesis (and temporality) that is premised on the assumption that structures are enduring and stable arrangements of social positions that logically pre-exist agency (Archer, 1995). Accordingly, 'any change is observable only in contrast to some stable state' (Poole and Van de Ven, 1989,  p. 564), thus leading to a transformation or reproduction of the pre-existing social structure. Second, being grounded in Critical Realism, the rule-based approach admits the existence of both internal (or necessary) relations, as well as external (or contingent) relations. On the one hand, internal relations are necessary relations where social actors presuppose each other for their existence (e.g., the relation between landlord and tenant). On the other hand, external relations are contingent relationships where each entity can exist without the other. For example, the relation between a Duty Prosecutor and the Case Management System is contingent upon the Duty Prosecutor using this tool. Indeed, it is entirely possible to envisage a Duty Prosecutor performing the function of making charging decisions or providing early advice and guidance to Investigating Officers without using this tool. Thirdly, the rule-based approach brings back the notion of the social actor, thus shifting the focus from patterns of interactions between and among individual agents to interactions between and among social actors performing roles within and between organisations (Kallinikos, 2004). Roles are socially-engineered templates of action (Kallinikos, 2004 ). It is through the enactment of roles that human agents become social actors endowed with collective intentionality (Searle, 1995). This, in turn, fits the notion of routines as collective (or social) phenomena considering that routines are the organisational analogue of individual habits being 'one ontological layer above habits' (Hodgson, 2008, p. 18). Though scholars have argued that approaches anchored within the ontological foundations of Critical Realism bring to the forefront the mechanisms that link action (agency) and institutions (structures) over time (Archer, 1995), very few scholars have focused on constitutive rules, that is, cognitive mechanisms that give rise to shared meanings and understandings. With few exceptions (e.g., Frost and Tischer, 2014), very few scholars have focused on the relationship between rules, routines and situated practices, let alone reappraised the concept of organisational routines on the occasion of exogenous changes occurring at salient even points in time. Table 1 summarises the core insights stemming from both routine-as-change and rule-based approaches and their outstanding issues.
In the remainder of this paper, we endeavour to address these outstanding issues by deploying a rule-based approach to the study of organisational routines using the case study method as our research strategy. In short, we argue that systems of constitutive rules of the type 'X counts as Y in context C' are the generative mechanisms that explain the emergence of new roles and, therefore, new position-practice systems at given moments in time. In other words, constitutive rules (the generative mechanisms) are a powerful engine in the generation of institutional reality in general and organisational routines in particular because they generate a host of deontic consequences (e.g., new rights, duties, obligations, etc.).

Methodology: research design and data collection methods
Our research design is based on an explanatory case study aimed at unravelling the factors presumed to be the logically-necessary conditions of possibility for a new social practice. Several scholars have argued that the case study method is the best approach to uncover the mechanisms and contextual factors that generate the phenomenon of interest (Wynn and Williams, 2012). Though multiple approaches are appropriate for case study designs (Blatter and Haverland, 2012), in this paper we use Lawson's (2013) quasi-experimental logic which can be seen as a special case of contrast explanation (MacKay and McKiernan, 2004). Case studies are particularly fruitful strategies for studying specific events in their real-life contexts when the boundaries between the phenomenon of interest and its context are not clear-cut (Pettigrew, 1990). Drawing on a temporal bracketing strategy (Langley, 1999), we look at the events occurring around November 2003 to explain a period of significant legislative change that is relatively-brief compared with the duration of the path-dependent processes that it triggered. By abstracting away from several idiosyncrasies, we take a snapshot of a significant legislative change in an effort to account for the emergence of a new interaction pattern.
Using counterfactual analysis, we describe the causes leading to the emergence of a new interaction pattern in Includes a theory of morphogenesis (and temporality). Acknowledges both internal (necessary) and external (contingent) relations. Fits the concept of organisational routines as collective phenomena.

Outstanding issues
Organisational routines are in a constant state of flux so much so that it is difficult to establish their identity. Routine-as-change approaches overlook the historical and institutional context and downplay frame-breaking, exogenous change.
Rule-based approaches overlook the cognitive mechanisms that give rise to shared meanings and understandings. a theory-guided fashion (Blatter and Haverland, 2012 (2003), the Duty Prosecutor has acquired the power or ability to generate events (e.g., charging decisions) and associated interactions that would not exist if this legislative act had not been passed. In other words, this legislative act has granted Duty Prosecutors the legitimacy to make early charging decisions, thus transforming the interactions between Police and Crown Prosecution Service (CPS) from one-way interactions to two-way, back-and-forth interactions. Table 2 outlines this counterfactual approach. Case studies enable researchers to move from surface to depth, thus taking context seriously and unpacking important structural conditions for new social practices to be possible (Danermark and Ekstrom, 2002). Accordingly, the phenomenon of interest ('our case') is an instance of institutional transformation in the making, namely the passing of new legislation (i.e., the Criminal Justice Act 2003) which accounted for the emergence of the 'Duty Prosecutor' and the transformation of Police-CPS interactions. We used multiple data sources with the purpose of investigating this phenomenon in its real-life context. Furthermore, we analysed our data using theoretical propositions informed by Critical Realist tenets (Wynn and Williams, 2012). Appendix Table A1 summarises the data sources used in this paper. Retrospective interviews of a select few informants were the primary data collection method because they focus on more distant events and the ensuing patterns of interactions. Nevertheless, these interviews were aptly triangulated with focus groups, observations and secondary data. Interviews were conducted between 2006 and 2014. Interviewees were selected through purposeful sampling at first and through snowball-sampling techniques thereafter. Table 3 provides an overview of the codes extracted from primary and secondary data through an iterative dialogue between theoretical insights and empirical data. Table 3 also provides illustrative quotes for each code.
Several principles have been formulated for conducting Critical Realist case study research in organisation studies in general (O'Mahoney and Vincent, 2014) and management information systems in particular (Wynn and Williams, 2012). Though Critical Realist case studies are very different from one another, they do aim to seek out and clarify the generative mechanisms at work through a process of retroduction (Wynn and Williams, 2012 ). This process consists of 'imagining a model of a mechanism which, if it were real, would account for the phenomenon in question' (Bhaskar, 2014, p. vii). By moving from the observable domain of empirical events, Critical Realists propose hypothetical generative mechanisms that, if they existed, would make the events in question possible. Such mechanisms are subsequently corroborated by means of methodological and theoretical triangulation.

Data analysis: the emergence of the duty prosecutor and the transformation of police-CPS interactions
Constructing mechanisms-based explanations involves analytical movement across three layers of reality: from the Empirical domain, where situations or events are actually observed or experienced, through the Actual domain of events (and non-events), to the Real domain of mechanisms and structures with enduring properties or powers (Mingers, 2004). In this section we analyse primary and secondary data using Wynn and Williams' (2012) five-step approach, but in the following In the absence of legislation, the localised interference of Prosecutors in the early stages of the investigation could become a potential source of conflict between these two social actors because they would challenge the authority of the Police to lead the investigation. Hence, one would expect no systematic two-way interactions between Police and CPS. section (Discussion) we move analytically across these three layers of reality being mindful 'that movement from the Empirical to the Real involves movement along a continuum from an "open" toward a "closed" system' (Gross, 2009, p. 361).
Step 1: Explaining the emergence of duty prosecutors

Drivers of change
Statutory Charging is one of the most significant changes that happened in recent years. The CPS now have the responsibility to determine the charge in all but the most simple and straightforward cases. The specific aims of Statutory Charging are: (1) the elimination at the earliest opportunity of hopeless cases; (2) the production of more robust prosecution cases: and (3) the elimination of unnecessary or unwarranted delays in the period between charge and disposal (Duty Prosecutor).

Full-code tests
Once a file has been received from the Police, the CPS have a duty under the Code for Crown Prosecutors to determine whether there is sufficient evidence to prosecute and secondly whether there is a public interest in doing so. Even if the Police are able to charge the suspect with an offence, the CPS must still review that decision. The review of charges falls into three distinct categories: (1) cases charged by the Police that require a CPS review before going into court; (2) cases where the CPS are asked to make a pre-charge decision; and (3) reviews of files by local Area lawyers on complex cases or cases which require viewing of physical exhibits prior to charge (HMCPSI/HMIC, 2016).

Conflict
There used to be a situation when once the Police had somebody charged they counted that as a detection and it did not matter whether the case had ultimately resulted in a conviction. That situation created conflicts and tensions between us (i.e., Prosecutors) and the Police. But that's gone really. We now all have a target to increase the number of offences brought to justice. And is the same with things like assisting an offender, ineffective trials. All these are cases of joint targets between the Police, CPS and the Courts (District Crown Prosecutor, Focus Group). Institutional context Part of the reason why joint working between the Police and the CPS has not landed before is because you have a very fragmented system with many independent Police forces. The criminal justice system in England and Wales is very complex, involving many different agencies, including Police forces, the CPS, the Criminal Courts, etc. Too often, these organisations have worked in silos rather than working together. This has created a fragmented system with clear institutional boundaries. Agencies were encouraged to pursue individual targets: a focus on volumes rather than outcomes; quantity over quality (Programme Manager, Criminal Justice System Efficiency Programme, Focus Group).

Collective role
The Ministry of Justice, the Home Office and the Attorney General's Office all have a collective role, either directly or indirectly through their executive agencies, in overseeing the effective running of the criminal justice system (Ministry of Justice, 2016). The leaders of all (Police) forces need to reflect individually and collectively on the issues in part 1 of this report so that, where appropriate, policing can be undertaken efficiently and effectively at a national level while preserving local accountability. In so doing, the Police service will retain the confidence, trust and respect of the communities which it serves … Collectively, there is no common understanding across the police service of the numbers of officers required to maintain operational viability or to provide a safe level of policing . A network code is a collective decision-making mechanism for establishing and securing adherence to principles, practices and procedures that need to be common on a regional or national basis, using a system of voting. The United Kingdom has already introduced such mechanisms in several essential public services. The legal technology already exists: it is tried and tested, and it works … In other words, the network code would help to overcome parochialism. Police leadersand, in some cases, leaders of other parts of the criminal justice system tooshould adopt a network code (HMICFRS, 2019).

Power
The pertinent developments can be summarised briefly:  of the CPS to request early advice and guidance with regard to the lines of investigation they need to pursue. The first edition of the Director's Guidance on Charging specifically states that: In order to facilitate efficient and effective early consultations and make charging decisions in prosecution cases, Crown Prosecutors will be deployed as Duty Prosecutors for such hours as shall be agreed locally to provide guidance and make charging decisions. This service will be complemented by a centrally managed out of hours Duty Prosecutor arrangement to ensure a continuous 24-hour service (Macdonald, 2004, p. 3) During these consultations, Duty Prosecutors are expected to apply the Full-Code test which prescribes two conditions to be met, namely (1) a 'realistic prospect' of conviction and (2) the case must be in the 'public interest' for the suspect to be charged. If the Duty Prosecutor has not been given enough evidence during the consultation but believes that the suspect is too dangerous to be set free, the Duty Prosecutor can apply a lower-level test (i.e., the 'threshold test') which presupposes a 'reasonable suspicion' against the suspect. This test, in turn, triggers a process whereby the Investigating Officer is required to gather further evidence by an agreed date to fulfil the Full-Code test requirement for a 'realistic prospect' of conviction. The first edition of the Director's Guidance clarifies this procedure as follows: Where the required evidential material (for a 'realistic prospect' of conviction) is not available, the Duty Prosecutor will assess the case against the 'threshold test' set out below. This should be noted on the Manual of Guidance 3 (MG3) form and a review date for a Full-Code test agreed as part of the action plan (Macdonald, 2004, p. 6) A Duty Prosecutor reiterated this procedure by maintaining that: The Police are the investigators. With the Statutory Charging scheme, they now have to take directions from us (i.e., Prosecutors) in certain categories of cases. The idea is that they are supposed to be coming to us at an increasingly early stage and looking for guidance and we set an action plan and say 'you should go and get this or that evidential material'. (Duty Prosecutor) Whenever the 'threshold test' is applied, a specific date must be agreed between the Duty Prosecutor and the Investigating Officer for a review of the case in accordance with the Full-Code tests (i.e., 'realistic prospect' of conviction based on sufficiency of evidence and 'public interest' tests), these tests setting out the general criteria to regulate the discretion of Duty Prosecutors. Decisions on all charging matters are recorded on an ad-hoc form called the 'MG3' form which contains two sides: the front side is a 'equest for a charging decision' and must be completed by the Investigating Officer, while the rear side, which is finalised by the Duty Prosecutor, includes the charging decision or, alternatively, either the steps to be taken before this decision can be reached or the reason for declaring no further action. Under the new Statutory Charging scheme, two procedures are outlined. While straightforward cases can be dealt with over the telephone with the use of electronic exchanges, the more serious cases require face-to-face consultations. The new and revised Guidance on Charging maintains that: In order to ensure a speedy and responsive charging service, referral arrangements for all but the most serious and complex cases will be to CPS Direct (a team of over 200 Duty Prosecutors based throughout England and Wales). The Police will submit pre-charge reports and key evidence across the electronic exchange (Starmer, 2013, p. 10) While the most serious cases require face-to-face consultations between Investigators and Duty Prosecutors, the less serious cases can be handled with telephone calls to expedite the charging service. As well as making relevant arrangements by telephone, Investigating Officers must send the MG3 forms and crucial evidential material by direct input in their Case Management Systems, secure e-mails or fax.
Step 2: explicating the structural context In the second step, one needs to explicate the structure and context where events occur by identifying 'components of social and physical structure, contextual environment, along with relationships among them' (Wynn and Williams, 2012, p. 796). Structure is a tricky concept in Critical Realist studies, but essentially it is used to point to a 'set of internally-related objects' (Danermark and Ekstrom, 2002, p. 47) whether these objects (or entities) are social (e.g., actors/roles) or material (e.g., technological artefacts). The structural entities under investigation are relational in the sense that they are what they are by virtue of the relations they enter into. Thus, a Duty Prosecutor exists only in relation to an Investigating Officer. Nevertheless, this is an asymmetrical necessary relation because Investigating Officers did exist even in the absence of Duty Prosecutors and, therefore, they do not presuppose Duty Prosecutors for their existence. Since Police Officers and Prosecutors are measured against different targets, a situation of reciprocal interdependence has arisen between these two social actors (Hoekzema, 2020). A Prosecutor specifically claimed: We have a mismatch because the Police disposals are charges, whereas Prosecutors are measured against It is all about 'sanction detection' which does not include a case which is 'NFAed'. But that is not what we (i.e., Prosecutors) are measured on. We are measured on how many of the cases that are then charged result in a conviction. Until that mismatch is resolved there will be a conflict (Borough Prosecutor) The whole point of having Duty Prosecutors was to solve this conflict because the function of Duty Prosecutors is to charge cases with a 'realistic prospect' of conviction or, at a minimum, a 'reasonable suspicion'. Accordingly, both Prosecutors and Investigating Officers had to agree common objectives to solve this conflict. In the new charging context, Prosecutors and Investigating Officers agreed that their common goal was to gather good evidential material. They reasoned that Duty Prosecutors need Investigating Officers gathering good evidential material to be measured favourably but, symmetrically, Investigating Officers themselves have a vested interest in gathering good evidential material to increase the volume of charges and, eventually, successful convictions. Hence, the new charging routine created a much needed 'truce' between Police and Prosecutors (Nelson and Winter, 1982). Another informant confirmed this insight: There is no conflict. There used to be a conflict because of separate targets, but over the years they have been brought together. The most important one is what the CPS call 'offences brought to justice' and the Police call 'sanction detection', but it is essentially the same thing. That's successful prosecutions basically. We both have the same target for that, but there may not be the same percentages for each (District Crown Prosecutor, Focus Group) The internal (or necessary) relations between social actors coupled with the interlocked system of powers (and liabilities) create the structural context where routines are enacted. For example, the charging routine takes place within a legislative context where Duty Prosecutors have the power of providing early advice and guidance to Investigators including, where appropriate, the making of early charging decisions (and symmetrically, Investigators have the duty of requesting early charging decisions or advice with regard to specific lines of investigation).
Step 3: retroducing the emergence of charging routines In the third step one needs to ask 'what makes the phenomenon of interest possible?' (Wynn and Williams, 2012, p. 800). To address this question, one needs to deploy the principle of retroduction whereby one proposes hypothetical mechanisms that, if they existed, would explain the events under investigation (Mingers, 2004). Or, to put it another way, by way of thought experiments and counterfactual thinking (MacKay and McKiernan, 2004), one needs to move from a description and analysis of concrete phenomena to the 'basic conditions for these phenomena to be what they are' (Danermark and Ekstrom, 2002, p. 80). Since social interactions in general and patterns of interaction in particular presuppose pre-existing structural arrangements (Archer, 1995), to arrive at 'what is basically characteristic and constitutive of these structures' (Danermark and Ekstrom, 2002, p. 96) we draw on Searle's notion of constitutive rules (Searle, 1995). Searle attempts to reconstruct the system of social positions, the norms and rules, as well as the socially (and culturally) acquired dispositions structuring particular activities by looking at the structural features of organised societies.
Though social structures do not exist independently of agents' interpretations of what they are doing, within the domain of social reality there is a particular class of events that require collective agreement for their existence. Searle labels such events as institutional facts because they exist only by human agreement. Therefore, institutional facts are ontologically intersubjective because they only exist in relation to human beings. But knowledge about them is an objective phenomenon. Thus, they are epistemologically objective because, for example, the fact that a person is a Duty Prosecutor is not a matter of opinion, but it can be ascertained as true or false in an objective fashion. Hence, institutional facts are epistemologically objective, even though human attitudes, expectations and beliefs are part of their mode of existence (Iannacci and Hatzaras, 2012).
Searle claims that 'constitutive rules' play a fundamental role in our understanding of institutional facts. He draws a contrast between regulative and constitutive rules, the former regulating pre-existing activities, the latter creating the very possibility of the activities they regulate. Constitutive rules take the form 'X counts as Y in context C'. Searle refers to this form as the 'count-as' mechanism (Searle, 2010, p. 7). A constitutive rule entails the assignment of a function (Y) to an object, person and/or interaction (X) in a specific context (C), even though the object, person and/or interaction in question does not have the intrinsic properties to perform the function under investigation. For example, being a Duty Prosecutor (X) 'counts as' having the power of making early charging decisions (Y) even though there is nothing intrinsic in a public official to entrust her with such powers. Again, our secondary data has clarified this idea as follows: The Code for Crown Prosecutors (the Code) is the public document that sets out the framework for prosecution decision-making. Crown prosecutors have the Director of Public Prosecutions' powers to determine cases delegated to them, but must exercise them in accordance with the Code and its two-stage test the evidential and the public interest stages. Cases should only proceed if, firstly, there is sufficient evidence to provide a realistic prospect of conviction and, secondly, if the prosecution is required in the public interest (HMCPSI, 2012) Searle (2010) argues that constitutive rules come into interlocked systems and that such rules are the foundation of social institutions. Searle claims that the assignment of the function (Y) to an object, person and/or interaction (X) is performed by virtue of a collective agreement, that is, a collective intention, desire or belief shared within an actual group of people. Indeed, it is thanks to such a collective agreement that power relationships exist because formal or informal rights, duties and obligations presuppose their collective endorsement or enforcement (Elder-Vass, 2015). By way of counterfactual thinking (MacKay and McKiernan, 2004), we claim that constitutive rules are the logically-necessary processes for the emergence of routines. They serve as generative mechanisms that make the emergence of a new (formal) position (i.e., Duty Prosecutor) and, therefore, a new social standard or expectation about Police-CPS interactions possible (i.e., the two-way interactions between Investigating Officers and Duty Prosecutors). In the absence of collective acceptance or recognition, a person (X) would not 'count as' a Duty Prosecutor (Y) in a particular context (C). For example, a self-proclaimed Duty Prosecutor would not 'count as' such because of the lack of legitimacy (Suchman, 1995). Therefore, Police Officers (as a collective) would neither endorse nor enforce the decisions made by a self-proclaimed Duty Prosecutor.
Step 4: corroborating the counterfactual explanation In the fourth step one needs to 'overcome the tentativeness of inferences derived through retroduction by attempting to validate the existence of the proposed mechanisms' through empirical corroboration (Wynn and Williams, 2012, p. 801). Blatter and Haverland (2012) discuss three criteria to make counterfactual analysis more compelling, namely clarity, historical consistency and theoretical consistency. Though focusing on the logic of pattern matching, Wynn and Williams (2012) offer similar criteria because they argue that the postulated mechanisms must survive empirical tests and that longitudinal research (i.e., historical consistency) is particularly useful within the Critical Realist paradigm.
It is clear that, in the absence of the Criminal Justice Act (2003), there would be no Duty Prosecutor. Hence, the relation between this Act (a social object) and the Duty Prosecutor (a social actor) is internal (or symmetrically necessary) because they presuppose each other for their existence. Furthermore, this counterfactual requires minimal re-writing of historical contingencies because the English system of criminal justice would still be an adversarial system even in the absence of the Criminal Justice Act (2003). In this hypothetical scenario, the CPS would still be leading the prosecution against the Defence even though the Police would be responsible for charging the suspect. Lastly, the evidence is consistent with a Searlian account of constitutive rules because in the absence of the Criminal Justice Act (2003) there would be no institutional fact. Clearly, the Criminal Justice Act (2003) is an act of Parliament and, as such, it is collectively accepted and recognised as a formal piece of legislation. Accordingly, the absence of legislation would entail the lack of collective acceptance and recognition and, indirectly, the lack of a generative mechanism whereby Duty Prosecutors come into being. Again, our secondary data confirms this insight: The Criminal Justice Act 2003 was implemented on November 16th 2003. It puts the power of the CPS to direct the charge on a statutory footing … Aligned to this statutory power was the Director's Guidance on Charging. (HMCPSI/HMIC, 2015)The 'count-as' mechanism we have suggested has better explanatory power than alternative theories of routines Pentland and Feldman, 2005;Pentland and Feldman, 2008, 20102011) because it accounts for the generative mechanism 'metaphor' rather than conflating it with the 'generative system'. In other words, the generative mechanism is neither the situated practice nor the disposition to enact a particular pattern of action. Rather, a procedure inscribed in an artefact (X) 'counts as' the collectively-agreed expectation or standard (Y) in a particular context (C) because participants have collectively agreed that it stands for 'the way they do things around there'. Therefore, routines participants 'will evolve a set of dispositions that are sensitive to the (constitutive) rule structure' (Searle, 1995, p. 145), that is, to the 'X count as Y' mechanism.
Step 5: triangulating methods and theories In the fifth and final step, researchers deploy multiple data sources and theories to validate their findings. Admittedly, above we have used multiple data sources to corroborate our findings (specifically, retrospective interviews, focus groups and documentary evidence as outlined in Appendix Table A1). However, while the use of counterfactual thinking enables the exploration of rival explanations, we used only one theory (i.e., the Searlian theory of institutional facts) rather than multiple theories. Again, it is easy to see how alternative but complementary theories fit our account. For example, Archer's (1995) notion of analytical dualism that separates structure and agency resonates with the idea that institutional facts in general and routines in particular occur within pre-existing institutional structures that can be decoupled from the practices and activities enacted by individual agents (or actors). Likewise, one can separate social and technical systems from each other, thus formulating stronger analytical explanations where systems of constitutive rules generate the possibility of institutional facts (Iannacci, 2014). Table 4 summarises these Critical Realist case study principles.

Discussion
Spurred by recent calls to investigate the emergence of routines in the field of management studies (Sargis-Roussel et al., 2017), we have asked at the outset of this paper 'what accounts for the emergence of a new interaction pattern?' We are now able to address this question. Though new interaction patterns may be the outcome of an endogenous process of variationselection-retention (Feldman and Pentland, 2003), there are also instances where new interaction patterns may emerge from exogenous forces when 'new legislation is passed or the authorities change the rules of a game' (Searle, 1995, p. 125). Though the latter are special cases of revolutionary (or episodic) change, they reveal a process of construction of institutional reality that is underpinned by an interlocked system of generative mechanisms (or constitutive rules) of the type 'X counts as Y in context C' (Searle, 1995).
Arguing for the operation of legislative forces to account for the emergence of new interaction patterns is beneficial because it enables routines scholars to combine endogenous and exogenous models of change at once (Van de Ven and Poole, 1995), thus developing a more nuanced explanation of routines transformation. Though routines normally change over time through a process of endogenous change, participants may acquire new powers at a salient event point in time when new legislation is enacted. For example, passing a new legislative act in 2003 (X) 'counted as' enacting the Criminal Justice Act 2003 (Y). But, within the context of this Criminal Justice Act 2003 (C), a Crown Prosecutor (X) 'counts as' the Duty Prosecutor (Y) and, therefore, acquires new charging powers (and symmetrically, the Investigating Officer is subjected to specific liabilities because the Investigator has the obligation of requesting an early charging decision). In addition, it is likely that, over time, the individuals filling these roles will 'evolve a set of dispositions that are sensitive to the (constitutive) rule structure' (Searle, 1995, p. 145), thus developing mutually-coherent action dispositions that account for the emergence of patterns in variety.
Accordingly, the emergence of new interaction patterns presupposes a great deal of change that encompasses both teleological processes (e.g., the Association of Chief Police Officers' and the Director of Public Prosecutions' purposeful endorsement or enforcement of the Criminal Justice Act 2003) and evolutionary dynamics (e.g., gradual evolution of background capacities, propensities and dispositions that conform to the constitutive rule structure). Furthermore, the new piece of legislation under investigation lays out the 'core' or 'basic' sub-procedures of organisational routines because it provides participants with shared understandings that 'enable the coordination of joint activities' (Dionysiou and Tsoukas, 2013, p, 196). In other words, the new law provides 'participants with compatible interpretations about what is happening and reciprocal expectations for what is likely to happen next or what actions are appropriate' (Dionysiou and Tsoukas, 2013, p, 196). For example, witnessing or experiencing a two-way interaction flow between Investigating Officers and Duty Prosecutors in the domain of the Empirical presupposes a large set of 'core' or 'basic' events in the domain of the Actual. Indeed, the Investigating Officer must file a 'request for a charging decision' (i.e., requesting a charge). This request, in turn, will prompt a "consultation' process between Investigating Officer and Duty Prosecutor based on faceto-face or more remote interactions (i.e., face-to-face meeting or remote consultation). The consultation, in turn, will enable the Duty Prosecutor to have a more thorough picture of the suspect and, therefore, make a decision regardless of whether the Duty Prosecutor agrees to charge the suspect, ask for further evidence or declare no further action (i.e., charging, asking for further evidence or declaring no further action). In addition, these events presuppose the existence of structural relations between humans (e.g., Police-CPS relations), artefacts (e.g., front-end interfaces and back-end databases and data structures) and humans and artefacts (e.g., the Police and CPS interact by means of direct inputs in their Case Management Systems, email attachments or, more simply, face-to-face interactions).
We submit that anchoring the rule-based approach within the ontological foundation of Critical Realism is useful because it removes several blind spots. First, it helps us conceive of routines as collectively-accepted and recognised ideas that inform expectations about new practices. As such, routines (or collectively-shared ideas) are standing possibilities that endure even when they are not enacted in the domain of the Empirical. Hence, the routine for charging someone with a crime endures even in the absence of an alleged crime. Second, routines are placeholders for regular procedures (i.e., request for a charging decision then consultation and then charging) even when such cognitive regularities are unrealised due to intervening or countervailing factors (Bhaskar, 2008). For example, the 'Full-Code test decision' may be unrealised because of countervailing factors whereby, due to the lack of evidential material, the Duty Prosecutor has not been able to charge the suspect in accordance with the Full-Code test (which prescribes a 'realistic prospect' of conviction). Regardless of this countervailing factor, the routine for charging someone with a crime will be enacted in the normal way through the application of the 'threshold test' prescribing a 'reasonable suspicion' even though it may not produce the 'Full-Code test charging' decision because there is no 'realistic prospect' of conviction (i.e., no full evidence). Third, constitutive rules have iterative properties that account for the layered and stratified nature of organised activities. Since in Searle's (1995, p. 125) theory of institutional facts, 'the "Y"' term of one level can be the "X"' term or the "C"' term of the next level up or even higher levels', it follows that being a Duty Prosecutor (X) 'counts as' having the power (Y) of providing early legal advice or guidance with respect to investigative decisions. But providing this advice based on a 'reasonable suspicion' against the suspect (new X at a higher level of abstraction) 'counts as' enacting the 'threshold test' (Y). But enacting the 'threshold test' (new X at a higher level of abstraction) 'counts as' having the power (Y) of leading the investigation, and so on. Likewise, making a charging decision based on a 'realistic prospect' of conviction (X) 'counts as' enacting the Full-Code test (Y), but enacting the Full-Code test charging decision (new X) 'counts as' prosecuting a suspect in the Courts (new Y), and so on. Figure 1 gives a sequence of snapshots of the layered and stratified ontology of these organised activities at each salient event point in time.
Ultimately, once organisational routines scholars go behind the surface appearance of organised activities, they can discover that there is a relatively-simple logical structure (i.e., the 'X counts as Y in C' mechanism) even though the manifestation of situated practices is extremely diverse. This logical structure can iterate upwards (or downwards depending on the choice of metaphor) indefinitely, thus creating a system of interlocked constitutive rules that explains the logical structure of organisational routines.

Conclusion
In their insightful commentary on the notion of mindfulness and routines (i.e., mindlessness), Weick and Sutcliffe (2006, p. 523) raise two provocative questions: 'Are we on the verge of a rout for notions of mindfulness and routine given the expanded meanings of these concepts? In trying to link two concepts (i.e., mindfulness and mindlessness), have we dissolved the concepts we started with? ' Feldman's (2016) recent endeavour to define routines as 'performing' and 'patterning' brings to mind Weick and Sutcliffe's (2006) admonition that a retreat from a classic concept may indeed turn into a rout. Are we on the verge of a rout for the notions of performative (or agentic) and ostensive (or structural) aspects considering Feldman's (2016) rhetorical move away from these concepts? Might engaging the terms 'performing' and 'patterning' suffice to capture the constitutive aspects of organisational routines? If routines consist of 'doings and sayings' (Feldman, 2016, pp. 23, 30, 32, 33, 39), are these 'doings and sayings' enacted in a vacuum or are they only intelligible within the context of a social institution or structure? If the core ontological question is 'how do we do patterning?' (Feldman, 2016, p. 39), one wonders whether Feldman's suggestion to focus on observable and consequential actions leads to a flat ontology that denies the stratified nature of organised activities (Seidl and Whittington, 2014).
Rarely are routines performed in the absence of pre-existing institutions. They presuppose an institutional context where they are slotted in. As Mutch (2016Mutch ( , p. 1171 has recently claimed, performative accounts of routines 'tear routines out of their broader context.' Nevertheless, Mutch's advice to bring context back in the study of organisational routines has been widely ignored in the mainstream study of organisational routines. It seems that routines scholars have been mostly preoccupied with countering conventional definitions of routines for the sake of highlighting their fluidity, variety and plasticity. Not only do mainstream accounts of organisational routines bracket out the broader context where routines are performed. They also downplay the maturity levels of organisational routines. If the routine is a mature routine (e.g., traffic-light crossing, penalty-awarding in football games, etc.), there is little need for a collective agreement to spur coordination between and among participants. However, if the routine is a new routine (e.g., charging routine, monitoring live content on social media platforms, fighting the Coronavirus pandemic, etc.) then participants must achieve a collective agreement about 'the way they do things around there' for coordination to emerge. We submit that mainstream accounts of organisational routines apply primarily to mature routines because they focus on multiple and even conflicting understandings of the routine. Yet, these accounts have little application in the domain of new routines triggered by exogenous forces such as legislative changes, new technologies or natural disasters. In other words, in the case of a mature routine all participants have experience of it so much so that multiple and even divergent and conflicting understandings are possible without hindering coordination. Nevertheless, new routines presuppose a common and shared model of understanding for group interaction to take place. Again, this argument sheds a new light on recent findings especially in the context of routines transfer (D'Adderio, 2014(D'Adderio, , p. 1347italics in original). Since routines transfer is not simply about the reproduction of organisational routines 'but also about the effortful recreation of a routine in and through a different context', it follows that the regeneration of routines across spatio-temporal contexts revolves around shared understandings which ensure stability in the face of conflicting demands especially in the early stages of routines transfer. This paper has policy implications too. If constitutive rules are not merely mechanisms for labelling the state of affairs specified by the X terms, but they generate further consequences (e.g., obligations, liabilities, powers, rights, etc.), it is essential that such rules are inscribed within social artefacts (e.g., laws, policies, mission statements, codes of conduct, etc.). More specifically, practitioners must endeavour to make institutional facts (or Y terms) widely understood. In our case, the actors engaged in the new charging routine relied upon a publicly-available Statutory Charging scheme which was piloted beforehand. As well as this scheme, they also used the Data Standards Catalogue to store data standards. As a result, we urge practitioners engaged in the routine to disseminate both social (i.e., the routine) and technical standards (i.e., data standards) as widely as possible.

Limitations and future research
Admittedly, the Critical Realist approach we advocate is not without its limitations. First, given our focus on generative mechanisms explaining the emergence of a new interaction pattern in 2003, we did not look at events occurring before 2003 in a longitudinal fashion. Nevertheless, a perusal of the empirical data before 2003 suggests that the two-way interaction pattern is the outcome of a process of drift of one-way interactions between Police and CPS (Ortmann, 2010). Second, one should not draw statistical generalisations from this research because we relied on a single case study. Nevertheless, one could draw analytic generalisations that demonstrate the operation of the constitutive-rule mechanism and its interlocked system. Third, in our attempt to portray the logical structure of organisational routines, we have downplayed the role of agents' reflexivity, human imagination and self-monitoring. Furthermore, from a cultural perspective, the role of situated interactions was described too briefly. For example, the deployment of inexperienced Junior Duty Prosecutors could prompt a contingent and somehow different adoption of the Statutory Charging scheme across Police-CPS locations (as some areas may not enact this scheme despite its collective endorsement). Rather than a single, overarching constitutive-rule mechanism, future research could investigate the structural-agenticand-cultural mechanisms at work (Archer, 2015). Likewise, our conceptualisation of constitutive rules as the basic building blocks of routines may narrow the conceptual domain of organisational routines to institutional facts (or cognitive regularities/shared understandings), thus simplifying the notion of organisational routines instead of burdening it with too many theoretical aspects (Foss, 2009). It may also help to re-energise the 'rules' or 'procedures' research stream and stimulate new studies aimed at further exploring 'the relations between rules and routines' (Frost and Tischer, 2014, p-205). These limitations notwithstanding, this paper shows that constitutive rules of the type "X counts as Y in context C' are a powerful engine in the generation of institutional reality in general and organisational routines in particular.
We need further research to explore the emergence of new routines. By highlighting the dynamic nature of organisational routines, mainstream scholars have focused on their variety, plasticity and fluidity at the expense of the underlying mechanisms that generate order, stability and patterning.