The Power of the Law vs. the Power of the Strongest? Explaining Successes and Failures of the European Court of Justice in Promoting Member State Compliance with EU Law. Paper prepared for the Ninth Biennial International Conference EUSA, March 31 – April 2, 2005 Austin, Texas Abstract Legalization has become one of the most central concepts in capturing the interaction between international law and politics. However, there is still disagreement whether and under which conditions legalization matters. In the attempt to increase the effectiveness of international law, states have more and more relied on the strengthening of legal dispute settlement proce- dures, e.g. in the WTO. But does adjudication generally result in higher levels of state com- pliance with law beyond the nation-state? Drawing on the European Union as the empirical extreme type for high legalization, this paper explores how and under which conditions judi- cial discourses within international institutions promote state compliance. The empirical analysis shows that judicial discourses before the European Court of Justice result in varying patterns of compliance, reaching from stable compliance, over unstable compliance, to con- tinued non-compliance. None of the prominent compliance theories can sufficiently explain such variation. Emphasizing the importance of institutional learning, this paper offers an al- ternative approach, which accounts for differential patterns of compliance both between and within EU member-states. The explanatory power of the approach is illustrated by two case studies in the field of environmental policy. Author: Diana Panke Institute for Political Science Free University of Berlin Ihnestr.22 14195 Berlin Germany Email: Diana.Panke@urz.uni-heidelberg.de I. The Transformation of Non-Compliance: A Matter of the Power of the Law? With the increase of international institutions in the 20th century, liberal theories on interna- tional co-operation have mushroomed. Remarkably, these theories overwhelmingly assume compliance with international norms to be the rule. States, it is argued, agree to a norm only, when it reflects their interests to an extent, which is somehow proportional to their bargaining power. The existence of an international norm is explained with reference to the state's own interests which prevents non-compliance. This line of reasoning overlooks important sources of non-compliance, such as changing interests or circumstances over time, the uncertainty as condition under which norms are developed, free-riding incentives, lacking resources for compliance, or ambiguities in the norm itself. Empirical research pinpoints that non- compliance is worth studying: Although states' compliance with law beyond the nation-state is generally high, non-compliance occurs frequently (Börzel, 2001, Chayes and Handler- Chayes, 1993, Reinhardt, 2001, Tallberg and Jönsson, 2001). Since compliance is a precondi- tion for the effectiveness of law beyond the nation-state, international institutions provide for various institutional mechanisms for the transformation of non-compliance into compliance. The European Union is characterized by an extraordinary high degree of legalization, going hand in hand with a highly legalized infringement procedure: thereby the legal systems of the EU and its member states are interwoven, the European Commission as a strong institu- tional actor manages compliance monitoring and initiates infringement procedures by its own. Moreover, the adjudication system relies on a strong third party, since the European Court of Justice is a highly independent actor with competencies for issuing rulings backed up with financial penalties. However, unlike modern states, the EU cannot rely on the legitimate use of force as the last resort for restoring compliance. Post agreement interactions are, therefore, ultimately based on bargaining and argumentative strategies for the transformation of non- compliance into compliance. Via bargaining stronger actors rather preserve their substantial interests than weak actors. This decreases the effectiveness of international law to the advan- tage of the power of the strongest. Argumentative strategies, in contrast, allow for transforma- tions of non-compliance into compliance, which are unbiased by power disparities of states. Successful argumentative strategies increase therefore the effectiveness of international law. This paper explores the conditions under which non-compliance can be transformed into two types of compliance (stable and unstable compliance) within the EU's adjudication stage. This stage is of special interest, since cases challenging the effectiveness of interna- tional law most severely are to be found at the very end of post-agreement interactions. The adjudication phase represents least likely cases for further transformations because especially those cases of voluntary non-compliance are carried far, in which a state's substantial interests or strategic preferences are eminently strong. Against this background, it is not surprising that the outcomes of the EU's adjudication phase do not reveal a story of pure success, in spite of the remarkable high degree of legalization (see II). The empirical pattern shows variation re- garding the outcomes stable, unstable, and continued non-compliance between policies and states. The outcomes are defined the following way: Stable compliance is characterized by lasting transformations of non-compliance into compliance. Unstable compliance resembles repeated norm-violations, which reduces the power of the law. The effectiveness of law is not increased at all, when continued non-compliance occurs, in which non-compliance in not only transformed into compliance temporarily. In the light of the EU infringement procedure's constant institutional design, it is puzzling why some cases are transformed into two types of compliance successfully, while other transformations fail. Existing compliance theories, drawing on formal institutional and state-specific variables, cannot solve this puzzle (II). This paper offers a 'compliance as process' perspective (III). The analytical focus is on interactions and contextual variables influencing dynamics of different types of collective and institutional learning, which are ultimately responsible for the outcomes as stable, unstable or continued non-compliance. While argumentative dynamics in adjudication-related interactions allow for reflexive learning and are conducive to stable compliance, bargaining dynamics allow for instrumental learning and are conducive to unstable compliance. This paper offers a range of hypotheses on the prospects for the successful transformation of non- compliance into different types of compliance during the adjudication stage, aiming for the explanation of the empirical puzzle. It proceeds in three steps. In the first step, collective learning processes of governmental actors induced by the judicial discourse are discussed (IV). Adaptations of member states' legal acts to the norm interpretation as developed in the adjudication phase are responsible for the outcomes as stable, unstable or continued non- compliance. Thus, collective learning results of governmental actors have to be transformed into institutional learning, since parliaments and ministerial bureaucracies are important actors for the production of legal acts (V). Since collective and institutional learning processes cannot only be induced top-down by judicial discourses but also bottom-up via strategies of societal actors, hypotheses on the conditions of success of re-framing and shaming strategies are developed in a third step (VI). The theoretical considerations are illustrated by two case studies on the transformational interactions which occurred in Germany with regard to two court cases concerned with environmental directives. II. The Pattern of Transformations in the Adjudication Phase The European Union's infringement proceeding (Art. 226 ECT) combines management, adjudication and enforcement elements (Zangl, 2001). In the management phase, the European Commission interacts with the government of the accused state on a purely bilateral basis (Tallberg, 2002). Only when the informal interactions are not brought to an end by either the conclusion that no norm-violation occurred or by the transformation of non- compliance into compliance, the formal phase is initiated by the Commission in sending a reasoned opinion to the respective state. When non-compliance is not transformed into compliance during the interactions between the national government and the Commission after the reasoned opinion has been sent, the Commission refers the case to the European Court of Justice. Thereby, the adjudication phase is initiated. The adjudication phase starts with a written procedure, in which the European Advocate General and national legal representatives exchange views on facts and legal aspects. The oral procedure foresees two open court settings: a public hearing and final statement of the Advocate General. After the second oral hearing the ECJ issues a binding ruling. Only when states do not comply with the judgment within the foreseen time, the adjudication phase is followed by an enforcement phase according to Article 228 ECT. This procedure is similarly designed as the article 226 procedure, but can end with a second court judgment, in which monetary sanctions are imposed, should non-compliance prevail. Even though the vast majority of cases is solved during the management phase, (Mendrinou, 1996: 4-6, Tallberg, 2002, Tallberg and Jönsson, 2001), for three reasons this paper focuses only on the interactions in the adjudication phase. Firstly, all cases referred to the ECJ have in common that states' substantial interests and/ or strategic preferences are extremely strong, since a consensus or compromise regarding the interpretation of the disputed norm would have emerged during the management phase otherwise. In this sense, all cases transferred to the ECJ are least likely cases for processes of legal adaptation (institutional learning). How can it be explained that there are cases in which institutional learning occurs nevertheless? Secondly, in the wake of the current trend towards increasing legalization of international institutions, it is interesting to explore the contribution of highly legalized institutional designs, such as the arena of the ECJ, to the effectiveness of law beyond the nation-state. Is a highly legalized institutional design a most likely setting for learning, even though substantial interests and/or strategic preferences of states are relatively rigid and therefore least likely cases for further transformations? What insights can the European adjudication system as the empirical extreme type provide for international institutions with lesser degrees of legalization? Thirdly, an interesting empirical pattern can be observed regarding the transformative interactions in the phase between the referral to the ECJ and the pronunciation of a judgment by the European Court of Justice in regard to incorrect transformations of European directives into national law. There are three different types of outcomes of interactions in front of the ECJ: stable compliance, unstable compliance and continued non-compliance (for the operationalization of the dependent variable see Panke, 2004a). Figure 1: The distribution of the dependent variable In the adjudication phase continued non-compliance (no transformations qualifying as compliance), unstable compliance (non-compliance is only incompletely transformed into compliance) and stable compliance (non-compliance has been completely and lastingly transformed into compliance) can occur. The prospects for the transformation of non- compliance into stable compliance, unstable compliance and continued non-compliance differ enormously between states. The two prominent approaches, the enforcement and the management theory cannot explain the pattern. Enforcement approaches (Martin, 1992a, 1992b) would expect that strong states reveal an extraordinary high rate of continued non- compliance (for the deduction of the hypotheses, see Panke, 2004b). Prominent outliers are weak states with high rates of continued non-compliance (such as Portugal and Spain). Contrary to enforcement theories expect management approaches high rates of stable compliance for all states with high political/administrative capacities (c.f. Chayes and Chayes Handler, 1995, Chayes and Handler-Chayes, 1991, 1993). Among the most obvious outliers are Luxembourg and Netherlands. Quite evidently, both theories suffer from high numbers of outlier cases. A second weakness is that management and enforcement theories alike focus on state specific independent variables and can, thus, not explain intra state variation. Thirdly, management and enforcement theories do not cover all three parameter values of the dependent variable. In a theoretical consistent manner, management approaches can only account for stable compliance and continued non-compliance (see at length Panke, 2004a, 2004b). The same holds true for enforcement theories: they capture only continued non- compliance and unstable compliance but cannot account for the outcome stable compliance in a theoretical sound way. III. The Transformation of Non-compliance and Types of Learning Neither management nor enforcement approaches can solve the empirical puzzle. Enforcement approaches overemphasize the power of the strongest, while management approaches overestimate the power of the law. In order to develop a theoretical frame, which provides an adequate account of the empirical puzzle, the simultaneous conceptual coverage of stable, unstable and continued non-compliance is necessary. Because of its action-theoretical foundation constructivism is well suited to develop accounts for stable compliance and continued non-compliance, while rationalism allows capturing unstable compliance and continued non-compliance in a theoretically consistent manner. Rationalist approaches capture the parameter value 'unstable compliance' in a theoretically consistent manner, because unstable compliance resembles instrumental learning. In the wake of external constraints (such as article 228 proceedings), non- compliance can become costly. When the costs for non-compliance are higher than the benefits form non-compliance, states do no longer maintain their strategic preference of non- compliance. Instead a strategic preference change into compliance occurs and states engage in legal adaptations. However, since substantial interests are not altered, states shift back into their strategic preferences of non-compliance again, as soon as external constraints lessen. Public attention or the Commissions supervision declines, when the legal acts passed that do not obviously conflict with the ECJ's norm interpretation. Therefore states do not comprehensively incorporate the ECJ's norm interpretation and define only insufficiently what constitutes norm-reproducing and norm-violating action. Thereby, windows of opportunity for future norm-violations are created. Constructivist approaches, on the other hand, offer accounts for stable compliance, which is characterized by a change of substantial interests. As opposed to strategic preferences is a change in substantial interests conducive to stable outcomes of transformational processes, regardless of further changes in external constraints. The transformation of continued non-compliance at the beginning of the adjudication phase into stable compliance during the judicial discourse can be explained by processes of reflexive learning. During interactions in front of the ECJ new ideas can be communicated. Thereby participants might learn that their ideas and their interpretations of the content and/or scope of the disputed norm, underlying their original substantial interests, are less true, rightful or appropriate than the ideas communicated during the judicial discourse. As a result the original ideas can be substituted by new ideas in a process of reflexive learning. Such reflexive learning processes can culminate in altered substantial interests being now in accordance with the norm interpretations as developed consensually during the interactions in front of the ECJ. Such substantial interest changes remain stable regardless of changes in external constraints. Constructivist approaches are, therefore, best suited to conceptualize the parameter value 'stable compliance'. The third parameter value, 'continued non-compliance' resembles the null hypotheses for instrumental and reflexive learning. Continued non-compliance occurs in the absence of both: instrumental and reflexive learning. Based on these considerations the theoretical frame underlying the 'compliance as process' perspective must fulfill two tasks: (1) combining rationalist and constructivist elements in a meta-theoretically consistent manner and (2) providing hypotheses on contextual conditions, favoring characteristics in social interactions, which are either more adequately grasped by rationalist or by constructivist approaches (IV- VI). Starting with the first task, the following question is crucial: Under which conditions can the different types of learning be expected? Interactions are essential for both reflexive and instrumental learning, because interactions accelerate learning by increasing the flow of ideas. However, the flow of ideas alone is not sufficient for the deduction of ideal scopes of rationalist and constructivist theories because it cannot account for the type of learning that might occur. In order to differentiate whether communicated ideas are conducive for reflexive or for instrumental learning, a systemic perspective, avoiding the predominance of one-sided action theoretical assumptions, on interactions is necessary (see at length Panke, 2002). A system is characterized by two necessary conditions. These are "(a) a set of units or elements is interconnected so that changes in some elements or their relations produce changes in other parts of the system, and (b) the entire system exhibits properties and behaviors that are different from those of the parts" (Jervis, 1997: 6). A system of interaction is composed of the totality of all speech acts, which were expressed by the participating actors, as the units of the system, during the interactions (starting after the referral to the ECJ and ending with either a withdrawal or a court judgment). In every system, structures can evolve. In systems of interactions, structures are the dominant pattern of speech acts, which influence the dynamics of ideational change (unconscious reflexive or instrumental learning). According to the systemic approach, collective learning, as learning processes of the participants in a system of interaction, is a systemic effect of interactions. Collective learning (as opposed to individual learning by only a single actor) occurs only under specific conditions, conditions that constitute the two possible structures that can evolve within systems of interaction: arguing and bargaining. Structures of interaction are defined by certain relationships between structure and content of the dominant pattern of speech acts (Panke, 2002). Both elements structure and content of speech acts are elaborated in turn (for an overview see table 1). The structure of speech acts can take two different forms. It can either be an argument or a speech act of bargaining. An argument links a proposition to reasons related to the intersubjective world. A speech act of bargaining is characterized by a demand, a concession or a rejection, which can additionally be linked with a threat or reasons that are related to the subjective world. However, a dominant pattern of speech acts is not sufficient to bring processes of collective learning about. Collective learning, as learning processes of the participants in a system of interaction, requires meaningful communication. Communication is not meaningful when actors cannot relate to each other and talk cross-purposes. Meaningful communication presupposes that all participants share standards of how to evaluate the content of speech acts. Meaningful communication is characterized by the possibility that B (as well as the other participants) understands the content of the speech act of A, evaluates the quality of communicated ideas and replies to A in a manner that allows A (and also the other participants) to reply meaningful again. In interactions that are based on the mutual exchange of meaningful speech acts, results (compromises or consensus) can be achieved incrementally, to which all participants can agree (without voting or authoritative decision). Hence, collective learning can only take place when communication is meaningful. For meaningful interaction to evolve, it is necessary to have a consensus among the actors of how the content (not the intention!) of speech acts is to be understood. Only when this precondition is fulfilled, meaningful communication is possible. In order to initiate processes of collective reflexive or instrumental learning of the participants of interactions, the contents of the speech acts must therefore fulfill certain criteria. Which criteria for the quality of the content of speech acts can be defined in the abstract? The possibility for processes of reflexive collective learning to take place, presupposes two elements. The necessary condition is that arguments are the dominant pattern of speech acts. The sufficient condition is that standards for the evaluation of the quality of ideas are shared among the actors. Such standards refer to what constitutes truth (causal ideas), rightness (normative ideas) or appropriateness (ideas on values) in a given context to a particular point in time (Habermas, 1995b). When both conditions are fulfilled, I refer to this pattern of meaningful communication as 'arguing as a structure of interaction'. Only when arguing as the structure of interaction has emerged, it is likely that argumentative speech acts lead the participants to question the ideas, which underlie their own substantial interests without having been consciously prepared or motivated before. When ideational change occurs, a change of substantive preferences is possible, when the ideas underlying the original substantial interests are affected by the ideational change (reflexive collective learning). Processes of reflexive collective learning can result in a consensus as the result of interactions. There is a second pattern of meaningful communication, to which I refer as 'bargaining as the structure of interaction'. For bargaining as a structure of interaction to evolve, it is not only required that acts of bargaining constitute the predominant pattern of speech acts, but also that actors share a standard for the evaluation of credibility. The standard of credibility has two components, incorporating a subjective and an intersubjective part. The intersubjective standard for the evaluation of a bargaining speech act refers to the bargaining power of an actor. Bargaining power is a complex social construct, which does not only entail formal vetoes but also such elements as the preference intensity and the alternatives of action. In regard to the subjective part, it is necessary that a positive attribution of a particular actor's reputation is undertaken by the other actors. Otherwise a threat, demand or concession is not meaningful, because the other actors cannot rely on its realization. Besides bargaining acts are the predominant pattern of speech acts, it is necessary that actors share a common conception of bargaining power and a common attribution of the reputation of each actor for bargaining as the structure of interaction to evolve. Within bargaining as the structure of interaction, instrumental collective learning about the distribution and nature of external constraints (such as the costs imposed by threats) is likely and can result in a compromise. Table 1 Two structures in systems of interaction Structure 'arguing' Structure 'bargaining' Pattern of dominant speech acts Arguments (reasons related to the intersubjective world) Bargaining acts (demands, threats, concessions, reasons related to the subjective world) Shared standards for the evaluation of the content of speech acts Common standard for truth or rightness or appropriateness Common conception of what constitutes bargaining power AND shared attitude on the reputation of the speakers Systemic effect: possible influence of communicated ideas on the majority of actors Reflexive collective learning (conducive to stable compliance regardless of changes in external constraints) Instrumental collective learning (conducive to unstable compliance when changes in external constraints occur) Results of interactions Consensus Compromise The systemic approach on interactions has the advantage of accounting for the coexistence of argumentative speech acts and speech acts of bargaining. This meta-theoretical frame and its concept of two structures of interaction is a heuristic yardstick with which the potential impact of ideas from reflexive to instrumental learning can be grasped. Since the systemic approach abstracts from logics of action, the gap between rationalist and constructivist theories with similar substantial foci can be bridged neutrally. This requires that ideal scopes of both approaches are examined with recourse to the contextual existence of the preconditions for the evolution or maintenance of any of the structures of interaction. Contextual conditions provided by the judicial discourse are discussed next. Principally, actors are always free in choosing between arguments and bargaining acts as two types of speech acts. However, the power of the law, as opposed to the prevalence of the preferences of the stronger, can only be strengthened during interactions in front of the ECJ, when argumentative speech acts are successful. Only when collective reflexive learning takes place, actors alter their substantial interests according to the developed consensual norm interpretation, which allows for stable compliance. When, on the other hand, bargaining dynamics evolve the power of the stronger could prevail over the power of the law in the longer run. This, however, would require that states accomplish their substantial interests against the ECJ's norm interpretation. Comparing the bargaining power, the ECJ possesses the threat of an article 228 proceeding, while states potential threats are severely reduced. Given the power disparities (among other reasons, see next paragraph), it is unlikely that bargaining as a structure evolves, collective instrumental learning occurs and a compromise is achieved. IV. The Top-Down Impact of Judicial Discourses IV.1 The ECJ as Platform for Argumentative Speech Acts The ECJ serves as an institutionalized arena that is conducive to the development of arguing as the structure of interaction to the disadvantage of bargaining in various ways. (1) The existence of a third adjudicating party alone does not negate differences in bargaining power of the states as the constitutional actors in international institutions. When third parties are not highly independent of the states in regard to their composition, tenure, payment, and terms of recall, they might anticipate ex-post sanctions (in pre-agreement interactions) in rulings favoring strong states (similar Abbott et al., 2000: 419; Keohane, Moravcsik and Slaughter, 2000: 460). A highly independent third party reduces the success of a state's bargaining strategies, since potential threats decline. Institutional designs with low degrees of legalization (third parties are not independent and rulings are non binding), allow for threats and ex-post sanctions by states (such as the dismissal of individual judges). In such settings, third parties anticipate more likely the power of and power differences between states in the contents of their rulings. Thereby they strengthen the power of the strongest. If, however, the degree of legalization is high, the resources required for bargaining are severely restricted and prevent states from extensive bargaining strategies. Hence, adjudication mechanisms, as installed in the infringement proceeding of the EU, reduce the possibility that a common standard for the evaluation of bargaining power prevails. (2) Within adjudication processes only arguments related to the world of the treaty and to the intersubjective world of truth are considered as legitimate speech acts (Alexy, 1983, Onuf, 1989). In adjudicational settings, speech acts of bargaining are not considered as contextual appropriate and are, therefore, not used in front of the ECJ (interview ECJ #2). (3) The EU's infringement procedure combines management and adjudication mechanisms. Before the adjudication phase is started through the European Commission's referral of the case to the ECJ, political aspects of potential norm violations are in the centre of the debate (Tallberg, 2002, Tallberg and Jönsson, 2001). Judicial aspects become increasingly important with the ECJ referral. Additionally heuristics for the interpretation of the content and scope of norms (wording, historical, teleological and systematic interpretations) are institutionalized. This allows for the reduction of subjectivism (Fiss, 1982, Gulmann, 1980). Moreover, each heuristic introduces an additional yardstick, on which the quality of arguments can be evaluated. IV.2 What Makes an Argument Convincing? IV.2.1 Different Standards of Reference. Because of the highly legalized design, interactions in front of the ECJ favor speech acts of arguing over bargaining. Nevertheless, the argumentative communication of ideas alone provides no yardsticks for evaluating whether an argument is convincing: Not every argument is per se good and thus not per se suited to persuade an actor. Therefore, the crucial question is: What characterizes a good argument? Or put differently: Which ideas are likely to change actors' substantial interests? Judicial positivism claims that there is always one correct and authentic interpretation of a norm, which judicial reasoning has to uncover during adjudicational interactions. However, the 'true' interpretation of a norm is not just out there. Firstly, objectivism is prevented by the value laden character of judicial reasoning itself, stemming from the hierarchy of heuristics and the necessity of recurring to other values and norms of the legal document in systematic and teleological interpretations (Brest, 1982, Dewey, 1924, Fiss, 1982, Habermas, 1998, Klare, 1998, Rasmussen, 1986, Wheeler Cook, 1927: 24). Secondly, subjectivity is introduced during the construction of the situation and the selection of relevant 'facts' (Brest, 1982, Dewey, 1924, Fiss, 1982, Habermas, 1998, Klare, 1998, Rasmussen, 1986, Wheeler Cook, 1927). Hence, contrary to claims of judicial positivism, judicial interpretation is a process of social construction (Brest, 1982, Dewey, 1924, Fiss, 1982, Habermas, 1998, Klare, 1998, Rasmussen, 1986, Wheeler Cook, 1927). But how and under what conditions can consensual norm interpretations emerge during judicial discourses and when do they fail? In general, with truth, rightness and appropriateness, there are three possible standards to assess the quality of arguments (Habermas, 1995b). The standard of truth encompasses epistemological and methodological principles and also ontological elements. Argumentative interactions, in which the quality of arguments can be measured based on standards of truth, are conducive to collective reflexive learning, when the actors share expertise on the subject matter. However, truth-related reasoning becomes meaningless, when there is no consensus of whether an effect reproduces or reinforces the norm, proper to its content and scope. Thus, a necessary but not sufficient condition for reflexive collective learning based on the argumentative communication of causal ideas is the existence of a consensus on the purpose of a norm among the actors. Norms are expression of a common interest of the norm- producing actors (Habermas, 1992, 1995a) and the quality of normative ideas is measured by the extent to which they express the purpose of a norm, as the standard for rightfulness (Habermas, 1995b: 42). For incorrect transpositions of European directives, however, content and/or scope of the respective norm is most likely to be disputed. During interactions in front of the ECJ reflexive collective learning is in so far aggravated, as it is unlikely that a shared standard for rightness exists, on which the quality of normative ideas can be equally assessed. Nevertheless, actors are not trapped in the dilemma that reflexive collective learning can only occur when there is consensus of a norm's purpose, while the very fact that the case has been carried on to the adjudication phase indicates that such a consensus is not existent. The judicial discourse offers an expedient, since it aims at the clarification of the content and scope of a disputed norm and thus of the standard of rightness itself. In order to identify and clarify a norm's content and scope, judicial heuristics are applied. With wording, historical, systematic and teleological interpretations, there are four different heuristics for interpretations. These instruments of judicial reasoning can serve as additional yardsticks to measure the quality of arguments. Thus, the application of judicial heuristics provides opportunities for reflexive collective learning (which, in turn, is conducive to stable compliance), opportunities which are absent in the interactions prior to the initiation of the adjudication phase. IV.2.2 Heuristics of Judicial Interpretation Under which conditions are states likely to experience processes of reflexive collective learning during judicial discourses? The ECJ has often been criticized for its dominant pattern of pro-Commission and thus pro-integration rulings (Garrett, 1992, 1995, Garrett and Weingast, 1993; see also Rasmussen, 1986). Prominent cases, such as the Costa case, in which the ECJ developed the doctrine of supremacy of European law, reveal that ECJ rulings can have an enormous impact on further dynamics of European Integration. In such rulings, the ECJ goes beyond clarifying the status quo of European Integration, but strengthens the supranational character of the project. Many authors argue that such far reaching interpretations (Rechtsfortbildung) are only possible, since the ECJ uses foremost the teleological heuristic (Rasmussen, 1986: 149, 173, 180, 264; Snyder, 1993: 40; Gulmann, 1980: 189, 199). Teleological interpretations allow for readings of norms in the light of the preamble, in which the aim of further integration is explicitly stated. Despite the tendency of the ECJ to side with the Commission in its rulings, the teleological method opens broad windows of persuasion being conducive to collective reflexive learning. This is because abstract, consensual aims can serve as yardsticks for the evaluation of arguments regarding norm interpretations and allow, if applied, for a high number of possible, judicially 'correct' norm interpretations. An empirical survey of judgments regarding employment and environmental directives in my data set reveals that the ECJ does not apply the broad teleological heuristic at all. Instead, it extensively relied on wording, sometimes on directive-immanent teleological and almost never on historical methods of judicial reasoning. Unlike the application of the broad teleological heuristic (referring to further market integration in general) and to a lesser extent systematical method of interpretation (which the ECJ does not use at all), directive- immanent teleological, wording and historical methods provide relatively small rooms for the argumentative development of consensual norm-interpretations. Hence, the prospects for reflexive collective learning are relatively restricted – compared to the broader teleological method. The lowest likelihood for reflexive learning provides the historical method of interpretation. In using this heuristic, the ECJ defines scope and content of directives through reference to the will of the norm-creators. However, states and not the ECJ were prominently involved in European policy-making. Arguments related to 'the original will' of the norm creators are likely to be evaluated as 'not right', when the norm creators themselves have diverging memories of the 'original will'. This prevents the evolution of arguing as the structure of interaction and thus reflexive collective learning. The wording heuristic is not as restrictive as the historical method. Its room for consensual norm interpretations is on a medium scale, while the directive-immanent teleological method allows for a slightly broader range of possible interpretations. Hence, chances for reflexive collective learning increase, when the directive-immanent teleological method is applied (either alone or supported by the wording heuristic). However, the possible space for consensual interpretations is heavily influenced by policy characteristics. Policy characteristics influence a norm's scope of interpretation and are very likely to have an impact on the prospects for reflexive collective learning during the judicial discourse. The degree of precision is often defined in regard to whether a norm qualifies as hard or soft law (Abbott and Snidal, 2000, Cini, 2000). Yet within the EU's infringement procedure only (primary and secondary) hard law is at stake. This paper focuses exclusively on the incorrect transposition of directives into national law. Compared with regulations, directives are less precise in regard to their content (aims and procedural instruments). Nevertheless, directives vary in the degree of clarity to which contents (aims and instruments) and applicatory scopes of norms are defined (scope of interpretation, see appendix 1). A broad scope of interpretation is characterized by the extensive use of ambiguous concepts and definitions regarding purpose, instruments and the applicatory scope. A narrow scope of interpretation, on the other hand, relies on clear definitions, does not introduce new and unspecified concepts and distinctions. The broader interpretational scopes are the stronger might norm interpretations diverge. In the extreme, interpretations of a norm with a broad interpretational scope can compromise the norm's purpose according to a narrow interpretation. Judicial heuristics such as the wording heuristic that leave only a very limited scope for consensual interpretations are suited to solve interpretational conflicts and allow for processes of collective reflexive learning, if the norm at hand has a narrow scope of interpretation. Broad interpretational scopes of norms, on the other hand, are better dealt with instruments (such as the teleological heuristic) that open wide windows for consensual interpretations. H1: The more narrow a norm's scope of interpretation is, the higher is the likelihood that processes of collective reflexive learning take place (which are conducive to stable compliance), when the wording heuristic is applied. The broader the scope of interpretation, the lower is the likelihood for reflexive collective learning, when wording heuristics are applied. H2: When broad teleological heuristics are applied, the prospects for reflexive collective learning increase (which are conducive to stable compliance), when the norm at hand is strongly vague and imprecise (reveals a broad interpretational scope). For illustration purposes, two environmental directives (the drinking water directive and the environmental impact assessment directive) for which the ECJ issued rulings against Germany are contrasted. The used methodology is qualitative in character. I disentangle the causal mechanisms (resting on different types of learning) through process-tracing. Thereby I examine the relevant actors' strategic preferences and/or substantial interests and their respective changes based on the examination of the contextual prerequisites for instrumental and/or reflexive learning (see the hypotheses). However, since this would bring about the danger of circular reasoning, I additionally examine directly through the analysis of speeches, other primary sources and especially through interviews whether and which type of learning occurred. According to hypotheses one, reflexive collective learning (of governmental actors, the Commission and the ECJ judges themselves) is more likely the narrower the interpretational scope of a directive is, when the wording heuristic is applied. While the drinking water directive is characterized by a narrow scope of interpretation, the EIA directive's interpretational scope is very broad. The drinking water directive (80/778) very precisely sets out a series of quality standards for water intended for human consumption and prescribes in detail the procedural instruments for their measurement (see at length annex one). In the drinking water case (237/90), the Commission accused Germany for not correctly transposing the directive into national law. The first reprehension on the interpretation of 'states of emergency' as an exceptional clause to the quality aims was already solved by the day of the judgment, since Germany adopted a decree, which no longer considered events such as thunderstorms as qualifying for exceptions according to the directive. The second criticism of German law by the Commission was on the communicational requirements of departures. While article 9 I of the directive demands a wholesome communication of failed parameters, German law only required communications of some parameters, if the cause for the failure was geogen in nature. The advocate general and the ECJ applied the wording heuristic. The latter concluded in its 1992 judgment (237/90) that the German transposition is inadequately, since article 9I requires complete reports. The second case is on the environmental impact assessment directive (85/337). This directive prescribes a mandatory procedure for the evaluation of projects' impacts on the environment. Its aim is the prevention of environmental damages, since only when a projects' impact on the environment is not regarded as significant concerning all media, it may be approved. The EIA directive is very imprecise as regards the definition of aims and procedures combined with a complex definition of its applicatory scope (see further appendix one). Among other elements, the Commission criticized the German legal transposition as treating the directive's annex two (in which classes and projects which should be subject to the EIAs are listed) as facultative rather than obligatory and defining parameters too vaguely. As result the German legal transposition reduces the directive's scope enormously and negatively affects its preventive aim. Hypotheses one would lead us to expect, that reflexive collective learning (which is conducive to stable compliance) during the judicial discourse is more likely in the case of the drinking water directive than regarding the environmental impact assessment directive. A glance on the dependent variables reveals that the outcome of the drinking water case is coded as stable compliance, while the EIA case is an instance of continued non-compliance (reasoned opinion based on article 228 after the 226 court judgment). This correlation, however, does not in itself indicate that learning dynamics induced (or not induced) by the judicial discourse are responsible for the variation in the dependent variable. So far it seems as if strong substantial interests were pointing towards non-compliance in the drinking water case in the first place. A staff member of the BGA (a high authority on health matters; Bundesgesundheitsamt) commission formerly working on drinking water exclaimed that "incorrect and incomplete transpositions are better options for all cases (such as the drinking water case), in which European directives do not correspondent with German thoroughness. Since the drinking water directive lacks detailed descriptions of the evaluation methods, it opens windows for lax application and derogates the high German standards through the back door" (interview UBA #2). "German authorities justified the delayed and incomplete transposition of the drinking water directive with the high quality of drinking water, which Germany had enjoyed since more than 100 years" (Börzel, 2003: 82). The first German legal implementation (for which the health ministry had the lead responsibility) transposed the health related elements of the directive (Kromarek 1987: 43), while broader environmental aspects (e.g. the complete communication of transgressed parameters) were left out. In their reaction the advocate general and the ECJ pointed towards the wording of article 9 I of the drinking water directive and its incompatibility with the German reading (Paragraph 4(3) of the TrinkwV, and Section III of Annex 4 to that regulation). Because of the stringent application of the wording heuristic, the German distinction between health and environmental related aspects was not discussed at all. Hence, the German delegation had no opportunity of defending the German way of the drinking water directive's legal transposition based on broader normative grounds. Based on the wording heuristic, the arguments on the 'correct' interpretation of the directive of the advocate general were superior to the German ones (interview organized interests #12). Hence, the judicial discourse eliminated competing readings of the narrow interpretational scope directive on drinking water. In the EIA case it is very unlikely that a consensual norm interpretation emerged during the judicial discourse. Since German environmental law is based on a media specific quality aims approach (Knill, 2001: 144; Wurzel, 2004: 103) while the EIA directive resembles a cross-media approach and additionally relies on procedural regulation, the misfit on polity (organizational restructuring), policy (quality aims) and politics (procedural regulations) dimensions between the EIA directive and the German environmental law is very high (see further Liefferink and Jordan, 2004: 37). Substantial interests against compliance existed especially within the states' ministerial bureaucracy and also (but to a slightly lesser extend) in the federal ministerial bureaucracy (interview state ministry #2). The sentiments against the EIA directive are rooted in the belief that German environmental law (based on single media, emission standards, and BAT) is superior to European environmental law approaches in general and especially in the widely shared perception that the EIA is unnecessary since German environmental law already had sectoral assessment procedures (interview state ministry #2). Against this background, the major challenge for the development of a consensual norm interpretation during the judicial discourse was not the EIA's high misfit, but the norm's broad interpretational scope. While the drinking water directive has a relatively narrow interpretational scope, the EIA directive is lowly precise and highly complex and allows, thus, for strongly diverging readings. The broad interpretational scope of the EIA directive opened the window for a German interpretation of the scope of application of the EIA directive as very narrow (reflecting the states (and federal) ministerial bureaucracies' conception of the EIA directive as unnecessary in the wake of already existing German laws on sectoral approvals). The narrow German interpretation compromised the purpose of the EIA directive as interpreted by the Commission and the advocate general. Based on the directive's exceptional clause (article four), the German federal state considered the application areas out of annex two as voluntary. Accordingly, annex two 'classes of projects' were excluded from the legal transposition. On the contrary, the Commission and the advocate general held that exceptions must be restricted as far as possible and that annex two is obligatory and not an 'annex á la carte' (see the statement of the advocate general). While the German reading is based on a wording interpretation of article 4 II (allowing for exceptions), the Commission's position rests on a directive immanent systemic reading: a comparison between the structures of annex one and two reveals that annex two is as obligatory as annex one (see the statement of the advocate general). Given these discrepancies, the challenges of the judicial discourse are high, since the parties implicitly dissent on the interpretational heuristic to be applied. The doctrinal hierarchy of heuristics as developed by the ECJ usually solves interpretational conflicts through the application of the wording heuristic (interview COM #1). Regarding the EIA directive, however, a wording interpretation of article 4 II (exceptions) would sacrifice the directive's aim according to a wording interpretation of article three (aim) and according to a broader interpretation (e.g. teleological). Not only the advocate general but also the ECJ in its 1998 judgment (case 301/95) applied a directive immanent teleological interpretation (see statement of the advocate general) which is compatible with the systemic but incompatible with the wording interpretation. Hence, the Commission succeeded while Germany lost the case. Broader heuristics (such as the directive immanent teleological heuristic) are generally better suited for the development of consensual norm interpretations, the broader the interpretational scope of the norm at hand is, because interpretations going beyond a narrow reading are not a priori excluded. Nevertheless, it is doubtful whether processes of collective reflexive learning took place during the oral and written procedure in front of the ECJ. As regards the hierarchy of norms, there is no logical way to deduct a hierarchy of judicial heuristics. Rather the hierarchy of judicial means (serving as additional standards on which the quality of ideas can be commonly evaluated) is set apodictically. When the advocate general and the ECJ depart from the institutionalized hierarchy of standards (and start of with the directive immanent teleological instead with the wording heuristic), it might well be the case that in effect, a common standard is lacking in such interactions. This, in turn, prevents the development of arguing as the structure of interactions and processes of collective reflexive learning. Since the advocate general used the systemic interpretational method in reaction to the German wording approach, it seems plausible that no collective reflexive learning occurred, because this would require that a judicial heuristic is shared (allowing to serve as a standard on which the quality of arguments can be commonly measured) in the first place. An established hierarchy of standards, on the other hand, implies that a second order heuristic can be applied, when the first order instrument fails to provide results. Applying the wording heuristic in the EIA case could lead to two different assessments, depending on which article (three or four) is given priority. It is, thus, not surprising that the ECJ applied the directive-immanent teleological heuristic instead of the wording method. Regarding the shift towards a second order tool, the prospects for reflexive collective learning during the announcement of the judgment strongly depend on whether the ECJ made explicit why the wording cannot be applied in the case – which, however, is not explained in the judgment at all (see ECJ judgment). From the empirical insights gained so far, it is much more likely that reflexive collective learning took place in the drinking water case than in the EIA case. For two reasons, reflexive collective learning processes induced by the judicial discourse do not automatically translate into the outcome of stable compliance (while the absence of collective reflexive learning during judicial discourses leads not deterministically to the outcome 'continued non- compliance'). Firstly, for the outcomes as stable compliance, unstable compliance or continued non-compliance to occur, national legal acts (laws or decrees) must be changed. Therefore, it is not sufficient that governmental actors undergo processes of collective learning. Rather, collective learning of governmental actors has to be transformed into institutional learning (ministerial bureaucracy and/ or legislative actors). Secondly, learning dynamics (collective and institutional) can not only be induced top down (by judicial discourses in case of governmental actors and in by the government in case of the ministerial bureaucracy and the parliament), but also bottom-up via strategies of societal actors. V. Institutional Learning - The importance of horizontal and vertical fragmentation Interactions in front of the ECJ are purely judicial in character. As a result, experts often agree on the norm interpretation (interview ECJ #2). There are also cases in which substantial differences between the actors persist (interview ECJ #2). In both cases, governmental actors have to communicate the essentials of the judicial discourses to the ministerial bureaucracy and/or to the parliament, in order to transform collective into institutional learning (responsible for the outcome of legal acts). Against this background, a member of the Commission stated that 'the major battle field is within the states and not in front of the ECJ' (interview COM #1). The debate on arguing and bargaining suffers from an underspecified concept of agency (Checkel, 2000, 2001a, 2001b, 2002). In implicitly conceptualizing states as unitary actors, a distinction between individual, collective and institutional learning is lacking. Institutional learning, as the learning processes of all actors who are prominently involved in the 'production' of legal acts (e.g. federal laws, decrees, state laws), however, is crucial for the outcomes of judicial discourses. The major questions, therefore, are: How are judicial discourses successfully transmitted into the domestic arena? Which instruments can governmental actors apply for transferring collective into institutional learning? Under which conditions can the transformation of collective into institutional learning be expected? The veto player approach points towards the importance of formal decision-making rules and competences as well as the distribution of substantial interests for the production of legal acts (Tsebelis, 1990, 2002). In its simplest reading it states that decision-making (law production) is the more difficult, the more actors are involved. Since governmental actors involved in the judicial discourses cannot apply hierarchical steering mechanisms (authoritative decision- making) for the adaptation of national legal acts, they must rely on argumentative and bargaining means. This is the more important, since primary (governmental actors) as well as secondary addressees (parliamentary actors, ministerial bureaucracies) of judicial discourses overwhelmingly perceive ECJ judgments as being ambivalent and requiring further interpretation before national legislation can be adapted accordingly . Which interpretation of judgments will make the day? Under which condition are governmental actors (primary addressees) able to transform their collective learning results into institutional learning (outcomes of legal acts)? The following veto-player related hypothesis can serve as a useful staring point for considerations on the prospects of success, for the transformation of the governmental's collective learning results into institutional learning (outcomes of legal acts: either laws or decrees). The higher the horizontal and vertical fragmentation the more likely is a high diversity of existing relevant substantial interests and strategic preferences, the more are further transformations of the results of governmental collective learning processes into institutional learning (and outcomes) be expected. The lower the horizontal and vertical fragmentation is, the less diverse are interests and preferences among the actors, who are relevant for the production of legal acts, the less likely it is that collective learning results of governmental actors are further transformed by parliamentary or ministerial bureaucratic actors. The distribution of vertical and horizontal competencies is influenced by state-specific and policy variables. The state specific variables (e.g. federalism) are constant in the German case and therefore not discussed any further. While the EIA directive was framed as a matter of environmental policy, the Germany Ministry of the Environment took the lead responsibility for the legal implementation of the directive as regards the state-level. Environmental matters do not fall under the exclusive competencies in the German constitution but belongs to the shared competencies. On the federal and on the state level the EIA directive was dealt with in a highly fragmented manner. Not only resembles the EIA directive the cross cutting policy character of environmental policy, it is also based on a cross-media approach (see appendix one). Since German environmental law is organized sectoral (media specific), the EIA affects a high number of actors. Hence, the vertical as well as the horizontal fragmentation is relatively high as regards the Environmental Impact Assessment Directive. The drinking water directive is older than the EIA directive and was dealt with primarily by the German Ministry for Health, because no single environmental ministry existed in 1980 and the quality of drinking water was framed as a health policy. Unlike the EIA, drinking water is less cross- cutting in its policy character and resembles a single media approach. The horizontal fragmentation as well as the vertical fragmentation was less in case of the drinking water directive than for the EIA directive. On the federal level, the drinking water directive was mainly opposed by the BGA (health high authority) (interview UBA #3). Another opponent was the water supplying industry, which lobbied especially the conservative parties on the federal and the state level (interview federal ministry #4). The BGA is not a 'real' veto player, since it cannot block decisions but merely provides scientific expertise to the federal ministries. The BGA lost influence over the health ministry's take on drinking water during the ongoing judicial discourse (not the least because the health ministry was no longer (if ever) convinced of the BGA's interpretation of 'states of emergency'). Accordingly, the German federal drinking water regulations were chanced in 1990. At the time of the judgment the remaining problem was the lack of German legal acts regulating the report requirements regarding all failures to meet the 64 quality parameters. Unfortunately I do not yet know which strategies were applied for transforming collective into institutional learning and which legal acts were adopted after the ECJ judgment on the federal level. I expect that institutional reflexive learning occurred on the federal level , which according to the German constitution was crucial for the adaptation of legal acts according to the ECJ ruling. The EIA directive was dealt with in a highly fragmented manner on the federal level and involved a broader range of institutional actors as compared to the water case. The federal elections of September 27th 1998 led to a change in government, since the conservative/liberal coalition was succeeded by a social-democratic/green coalition on October 27th 1998. Since the change in government occurred shortly after the date of the ECJ judgment (October 22nd 1998), the following argumentation proceeds in two steps. Firstly, I focus on the efforts of the conservative government to bring institutional learning about, during the ongoing judicial discourse (September 20th 1995 - October 22nd 1998) and directly after the judgment. Secondly, the emphasis is on the preferences and interests of the social- democratic/green government, its position to the ECJ judgment and the strategies of bringing institutional learning about. The state and federal ministerial bureaucracy in the environmental field strongly opposed the EIA directive and possessed the potential of blocking the production of decrees and initiatives for laws (interview state ministry #5). Since the primary addressees (CDU/FDP governmental actors) did most likely not undergo processes of reflexive collective learning during the judicial discourse, they applied no argumentative strategies in order to bring the ministerial bureaucracy in line with compliance before the government passes on to the social-democrats/green coalition. During the ongoing adjudication procedure, the possibility of a penalty based on article 228 played a minor role, since it was regarded as remote future (interview state ministry #3). This indicates that Merkel's ministry abstained from strategies designed to induce instrumental institutional learning. The ECJ ruling against Germany (10/22/1998), was in between the federal elections (09/27/98) and the adjuration of the new government (10/271998). For the first time in German history, the green party came into government and one of its members, Jürgen Trittin, became the environmental minister. As regards the states, the new federal governmental actors in the lead resort did not explain or justify the court judgment any further. Instead of persuading the states argumentatively, the federal level passed the judgment with the note 'please pay attention' ('Bitte um Beachtung') to the states' resorts (interview state ministry #2). They also did not point towards potential future costs of a 228 judgment as the legal adaptation proceeded very slowly after the 226 judgment (interview state ministry #2, #3). Hence, the new primary addressees stipulated no instrumental institutional learning. There are two interpretations of the new government's inaction stipulating reflexive or instrumental institutional learning. Firstly, it could be the case that substantial interests or strategic preferences of the new environmental ministry were pointing towards non-compliance (after the reading of the text of the ECJ's judgment). While it is likely that processes of collective learning were not induced by the judgment (see IV), it seems at odds that a green environmental minister should oppose the EIA directive. A second possibility is that the environmental ministry did not undergo processes of reflexive collective learning after reading the judgment, but the substantial interests/ and or strategic preferences were already in favor of compliance. Since the green party pressed for a comprehensive restructuring of environmental law (Umweltgesetzbuch) in the coalition agreement, which – as the EIA – should integrate all media under a single umbrella, it seems to be plausible that substantial interest were pointing towards compliance with the EIA and, in turn, with the court judgment. The legal adaptation, however, was severely delayed on the federal level. Plans for the comprehensive restructuring of the German environmental law (Umweltgesetzbuch), which should incorporate the EIA as an integral part, failed finally in the beginning of 2000 (interview state ministry #3). This was due to a shift of substantial discussions towards legal aspects, which brought about the insight that the vertical distribution of competencies prohibited a comprehensive reform of the environmental law. Instead a constitutional reform (article 75) would be necessary for the realization of the project. From the end of 1999 onwards, the environmental ministry realized the vertical fragmentation as an insurmountable barrier and pushed for legal adaptations according to the ECJ judgment via a single federal law ('Artikelgesetz'). The second chamber (Bundesrat) discussed the UVPG proposal at the end of 2000, which was not passed before September 2001. Even than, however, the EIA was not correctly and completely transposed, since some states delayed legal adaptations (of the minor aspects, they were responsible for). However, the responsible federal environmental resort did not even informally point towards the possibility of penalties according to article 228 procedures in order to fasten legal adaptations. One of the reasons was the 'federal state's remorse' based on the fact that the federal law, the UVPG (from 1990, changed in September 2001), shifted all remaining (and difficult!) tasks from the federal to the state level, so that the states are the ones who must bear all major costs (interview state ministry #2). So far, it seems very likely that reflexive collective learning occurred in the drinking water case and could have been transformed into institutional reflexive learning on the federal level. A change of the government substituted for top-down induced processes of reflexive collective learning in the EIA case. Because of the vertical fragmentation of competencies, the new government's substantial interests were transformed into institutional learning very late. These conclusions remain preliminary, since the strategies of societal actors are not yet discussed albeit they can influence dynamics of institutional learning towards both: instrumental or reflexive learning. VI. Judicial Discourses and the Differential Empowerment of Societal Actors There is disagreement on whether the transparency of settings influences the likelihood that arguments matter. Approaches taking communicative logic of action as naturally dominant argue that transparency increases the impact of arguments, because the public serves as a third standard, allowing for the triadic structure of arguing ({Saretzki, 1996 #171}). On the contrary, when the strategic logic of action is primary, it is argued that 'in camera settings' free actors from the public pressure of interest representation (Checkel, 2001b: 563). According to this argumentation, in camera settings allow for processes of persuasion, because public constraints on changes of substantial interests can be side-stepped by pretending bargaining dynamics and restraints. The discrepancy of both assessments is grounded in the weight put on either the communicative or on the strategic logic of action, and cannot be solved on this level, because the logics are mutually exclusive since they are based on different conceptions of rationality (Panke, 2002). The systemic approach on interaction arrives at a different assessment of the impact of transparency. Such an approach does not rely on one-sided action theoretical assumptions because it makes a distinction between speech acts and logics of actions presuming that neither reflexive nor instrumental learning requires a conscious motivation of actors. Transparency would only promote reflexive or instrumental learning, if it had an impact on the selection of speech acts or on the standards to which actors can refer. The selection of speech acts is not influenced by the degree of transparency. While in some settings, arguing might be enforced through publicity of interactions, in others, the audience might expect the representatives to pursue given preferences via bargaining. Also, transparency is not in itself conducive to the reference of either one of the standards for the evaluation of ideas. It does not influence the likelihood for bargaining power to be equally assessed, credibility is attributed, or actors share standards for what constitutes true, rightful or appropriate ideas. Hence, transparency has no influence on the likelihood of the evolution of either arguing or bargaining as structures of interaction (Panke, 2002). In public as well as in in-camera settings, actors' speech acts can refer to argumentative or bargaining standards alike. Nevertheless, public attention might influence the directions and dynamics of learning of governmental and parliamentary actors and thereby the outcomes of transformations of non-compliance during the interactions in front of the ECJ. The public cannot be reduced to a passive audience, which state actors have to convince through argumentative means. Instead, actors belonging to the public sphere can have interests in certain results of the adjudication phase, and try, thus, to exert influence on national decision-makers. As it will be discussed next, the judicial discourse empowers some societal actors over others. Not only does it influence the resources of actors, but also the conditions for the success of some strategies. In general, there are three types of resources, which can be used for various strategies aiming at influencing governmental and parliamentary learning dynamics. These are channels of access to constitutional actors, bargaining power (potential threats and credibility) and ideas/ information. The importance of the resources varies with the strategies actors pursue. Among the strategies between which the actors can choose, there are strategies, which operate according to rationalist (shaming, i.e. internal reputation costs, and peer pressure, i.e. external reputation costs) or constructivist (arguing and (re)framing) mechanisms (see further Schimmelfennig, 1997, 2001; Ulbert, 1997; Deitelhoff, 2003; ). Strategies operating according to a rationalist mechanism are based on the communication of external constrains (explicit or implicit threats), inducing costs on the executive/ legislative actors, when they refuse to alter their strategic preferences. While peer pressure is based on the imposition of costs, stemming from losses in a state's external reputation, shaming strategies are related to the threat of declining domestic reputation of governmental/ parliamentary actors. On the contrary, strategies of framing and arguing are based on a constructivist mechanism and focus on the far reaching impact of ideas: on reflexive learning. A survey among German federal and state members of parliament and ministerial bureaucracy, which I conducted in December 2004, revealed that court judgments empower those societal actors, whose substantive interests are in line with compliance (see also Kahler, 2000). These actors can firstly draw on the legitimacy and authority of the ECJ and secondly refer to the shrinking shadow of sanctions, in order to strengthen their claims. On the contrary, actors opposing compliance can no longer refer to ideas of appropriateness. Also it becomes increasingly difficult, to emphasize domestic costs of compliance and reinforce a state's strategic preference for non-compliance. During ongoing judicial discourses, resource distributions for both types of strategies – persuasive and pressuring – are altered in favor of pro-compliance actors. Thus, depending on the strategies societal actors pursue, processes of reflexive or instrumental institutional learning favorable to the transformation of non-compliance can be reinforced. Neither pure arguing nor pure bargaining strategies of societal actors can be expected to be very successful. The judicial discourse could empower societal actors in providing them with new ideas, strengthening their substantive preferences. However, it is unlikely that societal actors can persuade governmental and parliamentary actors with the communication of these ideas, if the latter did not learn in the judicial discourse, when these ideas have been brought up in the first place. Rather, societal actors can be expected to influence decision- makers with argumentative means, when they engage in re-framing activities. Under which conditions can we expect success of re-framing activities? Under which conditions are new frames launched successfully? There are two options for re-framing activities. (1) Either a highly technical norm-debate is reframed in terms of values, or (2) a highly value-laden norm debate is re-framed in technical, truth-related terms. (1) The value- laden re-framing (politicization) is the more successful the less complex the norm at hand is. This is because a low complexity allows for a rather inclusive debate after re-framing activities (which is important for institutional learning). Such re-framing activities are, however, only conducive to reflexive collective learning, when the value to which the frame refers is deeply institutionalized. (2) A de-politicization frame tries to re-frame a highly value- laden discourse into more technical, truth-related terms. It is the more successful the less deeply institutionalized the values are, since this affects the openness for de-politicization and contributed to a rather inclusive debate. De-politicized re-framing is conducive to arguing as the structure of interaction to emerge and, in turn, to collective reflexive learning. This is because the standard of truthfulness is the standard, most often and most likely shared in interactions among experts (scientists, ministries, legislatures' committees, organized interests, issue-specific public) (see Panke 2002). Re-framing activities, allowing for reflexive collective learning, are the more successful, (1)the higher the complexity of a norm is and the more deeply the value to which the new frame refers is institutionalized (for politicization frames) or (2)the less deeply the values under the 'old frame' are institutionalized (for de-politicization frames). Such new frames provide grounds for reflexive learning of state actors and are conducive to stable compliance. Successful bargaining strategies would presuppose that a concept of bargaining power is shared among societal actors and the national decision-makers. Societal actors do not possess per se potential threats, but must first create them, through shaming strategies. Strategies of shaming aim for an indirect influence on the primary and secondary addressees. They mobilize the general public, who in turn can be used as an external cost-imposing constraint preventing continued non-compliance. The first step of a shaming strategy is the adoption of a frame, which highlights the inappropriateness of a state's non-compliance. This is the easier, the higher the goodness of fit between judicial arguments and domestically institutionalized ideas (the domestic lifeworld) (for resonance arguments see further Checkel 2000, 2002 and Ulbert, 1997). When such a normative frame is applied successfully and resonates within the public discourse, the public becomes mobilized against further non-compliance. This can be used in a second step by societal actors, in order to point towards the domestic reputational costs for the governmental and parliamentary actors, arising from future non-compliance. The cost-benefit calculations of decision-makers are more strongly affected by the threat of reputational losses, when the particular policy is of high relevance for the profile of the governing parties or when elections are coming up soon. Under these conditions, shaming strategies are probably successful and instrumental institutional learning is likely to take place. The higher the goodness of fit of a European directive to institutionalized values in the member state, the higher are the prospects of success of societal actors' shaming strategies, which, in turn, are favorable to instrumental learning and, thus, conducive to unstable compliance. Before the drinking water case was referred to the court, there was a strong domestic opposition combined with almost lacking societal pressure for compliance. Water providers opposed the drinking water directive's parameter going beyond the already existent German health-related regulations; because the end-of-pipe approach on which the directive rests would let them bear the major costs (interview UBA #3). Hence, those actors had strong strategic preferences privileging non-compliance over compliance. Before the case was referred to the ECJ pro-compliance actors were largely passive, since environmental organizations supported incremental adaptation (through cooperation with the responsible authorities) instead of mobilizing the public (Börzel, 2003: 83). After the ECJ referral and the judgment, certain arguments against the legal transposition of the drinking water directive can no longer be made within a frame of appropriateness. This favored pro-compliance advocates to the disadvantage of non-compliant proponents. In the drinking water case, the court ruling changed the strategic environment even as far, as leading to cooperation between the water providers (as formerly pressuring for non-compliance) and environmental organizations: opposing the end of pipe approach, water providers and environmental organizations lobbied for limitations of fertilizers (Börzel, 2003: 83), as the problems caused by nitrates and pesticides for the water quality became obvious after 1990. Although environmental organizations were more active after the ECJ issued the ruling against Germany than in earlier phases of the infringement procedure, their influence was relatively limited. After the court judgment, the media attention was relatively low. Combined with possible credibility problems, arising form the ‚alliance' between environmental groups and the administration before the case was referred to the ECJ, the conditions for success of societal shaming strategies became additionally difficult – even though the goodness of fit between the drinking water directive and the German environmental and health law was relatively high (e.g. setting of quality parameters, single media approach, predominance of substantial aims instead of procedural regulation). Societal actors did not conduct re-framing strategies, which might be due to the difficulties of launching a politicization frame, emphasizing environmental protection . While there was severe domestic societal opposition in the drinking water case at the beginning of the infringement procedure, the EIA case did not raise as many sentiments by societal actors. The ministerial bureaucracy rather opposed the EIA because of the organizational adaptations the directive required. Even after the court judgment in the article 226 procedure, environmental groups did not intensify their activities greatly. Pro- compliance societal actors rather emerged and organized themselves ad hoc, in reaction to single applicational problems. Since even in most of these cases, the "resources of societal actors are too restricted to fight potential article 234 cases to the end" (interview organized interests #6), they cannot serve as a push factor for the legal adaptation via broader shaming or re-framing strategies. Even if the resources of societal actors would have been sufficient, successful strategies would have been very unlikely. Shaming strategies would be difficult because of the low goodness of fit between the EIA and the German environmental law. Politicizing re-framing strategies would most likely have also failed, because the EIA is an extraordinary complex norm. Not only was the number of pro-compliance societal actors higher in the drinking water case than in the EIA case. Also were the pro-compliance societal actors in the drinking water case better organized, had more resources, and possessed better channels of access to decision-makers and the national media. In the EIA case the media attention was very low (interview organized interests #6), which – combined with the high horizontal fragmentation – created a difficult environment to influence dynamics of legal adaptation. Moreover, the EIA directive is very complex and not easy to deal with. Since it requires extraordinary high expertise, 'the EIA directive is not an easy instrument for societal actors' (interview state ministry #2). This combination of factors prevented societal actors to conduct successful shaming or re-framing strategies, transforming the absence of institutional learning (leading to continued non-compliance) into either reflexive institutional learning (successful re-framing strategies) and stable compliance or into instrumental institutional learning (successful shaming strategies) and unstable compliance. VII. The Conditional Power of the Law – Explaining Successes and Failures of Adjudicational Interactions In the wake of the current trend towards increasing legalization of world politics, it is important to explore how adjudication phases contribute to the transformation of non- compliance. The legalization literature teaches us that highly legalized adjudicational settings provide institutional environments, which are very conducive to further transformations of non-compliance into compliance. At the same time, however, are all cases carried to the adjudicational stage instances of rigid preferences and interests and, in this sense, least likely cases for further transformations of non-compliance into compliance, since settlements would have been achieved in earlier stages otherwise. Within the European Union's infringement procedure (article 226), an interesting variation of transformational prospects between policy fields and between states can be observed regarding incorrectly transposed directives into national law, which cannot be explained by existing theories. This paper tries to solve this empirical puzzle by presenting an alternative theoretical approach, which is illustrated by two case studies (both on environmental directives in Germany). The theoretical argument proceeds in three steps: Firstly, the top-down influences of judicial discourses on collective learning dynamics are discussed. The most important hypotheses in this regard distinguishes between the interpretational scope of directives and between different judicial heuristics for interpretation and links them towards the likelihood for reflexive collective learning and the absence of top- down induced learning. While the EIA is a relatively imprecise and highly complex directive, the drinking water directive has a narrow interpretational scope. Since the wording heuristic is suited as an additional yardstick for the evaluation of arguments, if the norm at hand has a narrow scope of interpretation, collective reflexive learning was expected for the drinking water case. The judicial discourse in the EIA case did not bring about collective learning, since the EIA directive was dealt with by different judicial heuristics. This eventually led to the lack of a common standard for the evaluation of arguments during the adjudicational phase. The government changed in the same month as the court ruling. It brought a green environmental minister into office, whose substantial interests were most likely already in line with compliance, substituting for the absence of reflexive collective learning processes during the judicial discourse. Secondly, collective learning must be transformed into institutional learning in order to construct national legal acts according to norm interpretations as developed in the adjudication stage. It was hypothesized that horizontal and the vertical fragmentation are important parameters for the transformation of collective into institutional learning. In the EIA case, in which no collective learning took place during the judicial discourse, no institutional learning can be expected before the ECJ issues a ruling. The ruling itself was not suited to bring reflexive collective learning about, since it was based on uncommon interpretational heuristics. Because of the change in government few days after the ruling was issued, top-down induced reflexive learning was no longer necessary: the new environmental minister's substantial interests were already in line with compliance. However, because of the vertical fragmentation, the competencies for a comprehensive restructuring (with the EIA as an integral part) of the German environmental law were lacking on the federal level. This delayed the transformation of the primary addressees' substantial interest into institutional learning and, in turn, legal adaptations on both, the federal and the state level, so that the Commission initiated an article 228 procedure. In the drinking water case, the horizontal fragmentation was considerable, but the major opposing actor (the BGA) had no veto powers and could have been brought into line with compliance by the German health ministry. Thirdly, collective learning cannot only be induced top-down via the judicial discourse, but also bottom-up through strategies of societal actors. While reflexive institutional learning (leading to stable compliance) would have succeeded in the drinking water case without strategies of societal actors, societal actors could potentially have made a difference in the EIA case. If shaming strategies were conducted successfully, instrumental institutional learning would have occurred (conducive to unstable compliance), while successful re-framing strategies would have brought about reflexive institutional learning (leading to stable compliance). However, societal actors were not very active in the EIA case – not at last because the directive itself is very complex and difficult to handle. Hence, the outcome of the EIA case 'continued non-compliance' was due to the slowness of institutional learning processes – which were not fastened by bottom-up strategies of societal actors. In the drinking water case, societal actors were largely inactive. Nevertheless did reflexive learning occur during the judicial discourse and was transformed into reflexive institutional learning, due to path of appropriateness set forth by the judicial discourse. This paper is part of a PhD project, for which eighth case studies (Germany and UK in regard to environmental and employment directives with narrow and broad interpretational scopes) are planned. The empirical cases presented in this paper are first insights and serve, therefore, for illustrative purposes rather than for rigid hypotheses testing. Appendix 1 – The interpretational scope of two environmental directives The directive number 80/778 refers to the quality of water that is intended for human consumption. Overall the interpretational scope of this directive is narrow, since its precision is high and the complexity is low. The drinking water directive aims at promoting a common quality standard for water, which is intended for human consumption. Its aim is very concrete, since the quality standards are defined in a numerical manner in annex one, to which article seven refers. The quality standards serve as the lower bottom line, with which the water quality in the member states must comply. There are only three exceptions (article nine and ten), which are relatively clear defined. Firstly, member states may diverge from the quality standards in order to take account of 'situations arising from the nature and structure of the ground in the area from which the supply in question emanates'. Since neither nature nor structure of the ground is further specified in the directive, this dilutes the precision and provides interpretational focal points for narrow and broad readings. Secondly, exceptional meteorological conditions constitute reasons for derogation. Thirdly, member states may violate quality standards in events of emergencies. All exceptions are restricted by article nine paragraph three and by information-related requirements. The substantial limit set out in paragraph three of article nine prohibits any exceptions related to toxical or microbiological factors. Furthermore, no derogations are allowed, which would pose a hazard to public health. The procedural provisions oblige the member states to inform the European Commission of all exceptions being made. This allows for oversee activities and prevents the unequal interpretations regarding applications. In sum, the substantive aim of the drinking water directive is highly precise. The drinking water directive encompasses not only quality standards but also instrumental provisions of how these standards have to be monitored by the member states. Procedures for granting water quality are laid down in article twelve and the annexes two and three. Annex two prescribes in detail how the member states shall conduct their monitoring activities. It establishes four different types of control analyses, defines the parameters and control activities which must be examined during the different types of control, and the frequency of the controls in a clear and unequivocal manner. The precision is also very high for the scientific methods of analysis, which shall be adopted for the member states' monitoring activities as specified in annex III. None of the procedural annexes encompasses procedural exceptions. At first sight, the applicational scope of the directive 80/778 seems to be highly complex regarding the definition of the applicatory scope. This is because drinking water is defined as water that is intended for human consumption. The reference to intentions always introduces a high interpretational scope. However, article two defines more explicit what is meant by drinking water, since it entails an enumerative list of what 'human consumption' includes. Within this list, one criterion is ambiguous: the wholesomeness of the foodstuff. In regard to this area of application, the interpretational scope is relatively broad. Lacking an intersubjective definition of what wholesomeness means, the directive introduces a procedural component (article six), according to which the member states are obliged to report the information on their negative definitions to the Commission on a regularly basis. This allows the European Commission to oversee the member state definitions and application of this criterion. The interpretational scope of the drinking water directive is additionally increased, since its applicational scope of is restricted by article four. Natural mineral waters and medical waters are excluded from the directive according to this provision. This is insofar problematical as the definition of medical water is left to the competent national authorities. Summed up, the substantial and procedural elements of the directive number 1980/778 are very precise. Clear definitions, well defined concepts and the low number of cross-references provide only for a narrow window of divergent interpretations. While the substantive and especially the procedural provisions are highly precise, the scope of application is ambivalent in regard to three elements: the wholesomeness of the foodstuff, natural mineral waters and medical waters. However, natural mineral water is defined in directive 1980/777. In taking the mineral water directive into account, the complexity of scope of application of the drinking water directive declines. Only medical water is not further defined in other directives. This, in turn, leads to an overall complexity on a low to medium level. Summed up, there are only a few substantial and scope-related points, based on which divergent interpretations of the drinking water directive can be developed. Even broad interpretations of the exceptions and ambiguities cannot sacrifice the aim and the scope of the directive, if one would interpret it in a restricted manner. The directive number 1985/337 concerns the environmental impact assessment of private and public projects. It is a directive with an extraordinary high interpretational scope in regard to its scope of application: possible narrow and possible broad interpretations diverge strongly. The environmental impact assessment directive aims at preventing environmental damages caused by public and private projects. Therefore all effects that private and public projects exert on the environment shall be assessed by the member states' authorities before consent is given. Article three specifies the preventive aims in defining the elements on which projects can have an impact by an enumerative list, which encompasses all media and even includes cultural heritage and material assets. Overall, the substantive component of the environmental assessment directive is relatively precise: in regard to the definition of the EIA's substantive aim judicial figures are only rarely used and only a few concepts and distinctions are introduced. The instrumental provisions of the EIA directive are very extensive (article four to ten) and out weight the substantial part. According article four, member states can establish threshold values for the decision, of whether certain projects (according to annex two, which enumerates projects subject to article four, paragraph two) shall be subject to the EIA. In a narrow definition, absolute threshold values are in accordance with the EIA. This would mean that the EIA can be omitted, when one of the threshold criteria (nature, size, and location) is not met. According to a broad definition, on the other hand, threshold values are conceptualized as being of relative character. As long as at least on of the three threshold criteria (nature, size, and location) is reached, an environmental impact assessment must take place. This ambivalence reduces the precision of the procedural provisions enormously, since narrow readings of the possibilities in setting threshold values possibly sacrifice the aim of the directive. The remaining procedural provisions are of higher precision. Article five (in combination with annex three) enumerates all information, a developer has to provide (and the member states must request). The other articles are related to the access of information and the decision of the competent national authorities in regard to the result of the EIA of a project. As article five, they are relatively precise, since abstract and ambivalent definitions are avoided. There are only a few context specific exceptions regarding the distribution of information (e.g. article ten), which introduce possibilities for diverging interpretations. Only the extensive use of cross-references decreases the precision of the procedural rules. Although most procedural provisions are relatively precise, the overall precision is on a low t level, due to the prominence and low precision of article four. The scope of application of the directive 1985/337 as defined in articles one and four allows for an extraordinary wide range of diverging interpretations. The EIA introduces a range of new concepts and distinctions, some of which are only vaguely defined. A number of the core concepts, namely 'project', 'developer', and 'development consent', are specified in article one. However, these definitions are not sufficiently concrete. Especially the definition of 'project' is ambiguous, since it encompasses two residual categories. The first part of the definition refers to projects as 'the execution of construction works or other installations or schemes'. Installations and schemes are not further specified and serve thus, as a residual category for constructional affairs. Ambiguity in regard to the scope of application is introduced in the second part of the definition, according to which 'project' includes 'other interventions in the natural surroundings and landscape including the extraction of mineral resources'. The content of this residual category can be interpreted in different manners. According to the first (broad) reading, 'other interventions' includes all human interference into nature. A rather narrow reading would suggest that only those interferences are meant, which somehow relate to the extraction or production of natural resources, because the second and the first definitional part refer to certain types of economic conduct. Moreover, which actions qualify as 'intervention'? Article two of the environmental impact assessment directive provides neither a qualifiable nor a quantifiable criterion for an intervention. Not all private and public projects in terms of article one paragraph two are subject to an environmental impact assessment. Paragraph four and five constitute two exceptions, namely defense-related projects and projects adopted by a specific national legislative act. These exceptions do not increase the complexity, since they are clearly defined and do not introduce unspecified judicial figures. The scope is further specified in article two. Paragraph one states that all project are subjects to assessment when it is likely that they have significant effects on the environment by virtue or their nature, size and location. While nature, size and location are concepts on which intersubjective consensus is likely, the term 'significant effects' allows for a wide range of possible interpretations. Even more problematical is paragraph three of article two. Divergent interpretations of this paragraph are possible. Does it pose a general or a conditional opportunity for the member states to set out the application of the directive? What constitutes an 'exceptional case'? The projects subject to an environmental impact assessment are further specified in article four, that refers to annexes one and two. In both annexes classes of projects are listed. While annex I is clearly obligatory, article four, paragraph two (in combination with annex II) can be interpreted broadly or rather narrow. The wording states "Projects of the classes listed in Annex II shall be made subject to an assessment, in accordance with articles 5 to 10, where Member States consider that their characteristics so require. To this end Member States may inter alia specify certain types of projects as being subject to an assessment or may establish the criteria and/or thresholds necessary to determine which of the projects of the classes listed in Annex II are to be subject to an assessment in accordance with Articles 5 to 10". In a broad reading, the member states' scope of discretion is based upon article two, paragraph two. According to this line of interpretation, the member states can exclude selected projects out of annex two. Annex two would, thus, become an annex á la carte. The narrow interpretation (which restricts the member states' area of discretion), on the other hand, is based on a comparison between annex one and two. In this line of interpretation the similarities in the definition of 'projects' and 'classes of projects' in annexes one and two suggest that the projects in annex two are as obligatory as the projects in annex one. This ambivalence can have far reaching impacts in regard to the applicatory scope of directive 85/337. It strongly increases the complexity of the EIA-directive's scope of application. Taken all three elements together, the interpretational scope of the EIA directive is very broad. This is due to an extraordinary high complexity combined with a low level of precision regarding the procedural provisions (but a relatively high precision of the less important substantial aims). Literature: Abbott, K.W., Keohane, R.O., Moravcsik, A., Slaughter, A.M. and Snidal, D. (2000) 'The Concept of Legalization'. International Organization, 54 Vol., No. (3), pp. 401-19. Abbott, K.W. and Snidal, D. (2000) 'Hard and Soft Law in International Governance'. International Organization, 54 Vol., No. (3), pp. 421-56. Alexy, R. (1983) Theorie der juristischen Argumentation. Die Theorie des rationalen Diskurses als Theorie der juristischen Begründung (Frankfurt: Suhrkamp). Börzel, T.A. (2001) 'Non-Compliance in the European Union: Pathology or Statistical Artefact?' Journal of European Public Policy, 8 Vol., No. (5), pp. 803-24. Börzel, T.A. (2003) Environmental Leaders and Laggards in the European Union. Why There is (Not) a Southern Problem (London: Ashgate). Brest, P. (1982) 'Interpretation and Interest'. Stanford Law Review, 34 Vol., No., pp. 765-74. Chayes, A. and Chayes Handler, A. (1995) The New Sovereignty. Compliance with International Regulatory Agreements (Cambridge, MASS: Harvard University Press). Chayes, A. and Handler-Chayes, A. (1991) 'Compliance Without Enforcement: State Behavior Under Regulatory Treaties'. Negotiation Journal, 7 Vol., No. (2), pp. 311- 30. Chayes, A. and Handler-Chayes, A. (1993) 'On Compliance'. International Organization, 47 Vol., No. (2), pp. 175-205. Checkel, J.T. (2000) 'Bridging the Rational-Choice/ Constructivist Gap? Theorizing Social Interaction in European Institutions'. ARENA Working Papers, WP 00/11 Vol., No., pp. 1-14. Checkel, J.T. (2001a) Taking Deliberation Seriously In. Checkel, J.T. (2001b) 'Why Comply? Social Learning and European Identity Change'. International Organization, 55 Vol., No. (3), pp. 553-88. Checkel, J.T. (2002) 'Persuasion in International Institutions'. ARENA Working Papers, WP 02/14 Vol., No., pp. 1-25. Cini, M. (2000) 'From Soft Law to Hard Law?: Discretion and Rule-Making in the Commission's State Aid Regime'. EUI Working Papers, 2000/35 Vol., No., pp. 1-31. Deitelhoff, N. (2003) Taking States into Persuasion. Norm-Entrepreneurs and the Manipulative Nature of Communicative Settings in Norm-Construction In http://www.ibdiskurse.de/ipc/papers/deitelhoff.doc (Frankfurt/Main. Dewey, J. (1924) 'Logical Method and the Law'. Cornell Law Quarterly, 10 Vol., No. (17). Fiss, O. (1982) 'Objectivity and Interpretation'. Stanford Law Review, 34 Vol., No., pp. 739- 63. Garrett, G. (1992) 'International Cooperation and Institutional Choice: The European Community's Internal Market'. International Organization, 46 Vol., No. (2), pp. 533- 60. Garrett, G. (1995) 'The Politics of Legal Integration in the European Union'. International Organization, 49 Vol., No. (1), pp. 171-81. Garrett, G. and Weingast, B.R. (1993) 'Ideas, Interests and Institutions: Constructing the European Community's Internal Market.' In R.O. Keohane (ed.), Ideas and Foreign Policy. Beliefs, Institutions and Political Change (Ithaca / London: Cornell University Press), pp. 173-206. Gulmann, C. (1980) 'Methods of Interpretation of the Court of Justice'. Scandinavian Studies in LawVol., No., pp. 198-204. Habermas, J. (1992) 'Handlungen. Sprechakte, Sprachlich Vermittelte Interaktion und Lebenswelt.' In J. Habermas (ed.), Nachmetaphysisches Denken - Philosophische Aufsätze (Frankfurt am Main: Suhrkamp), pp. 63-104. Habermas, J. (1995a) Theorie des Kommunikativen Handelns (Frankfurt am Main: Suhrkamp). Habermas, J. (1995b) Theorie des Kommunikativen Handelns. Band 1 Handlungsrationalität und gesellschaftliche Rationalisierung (Frankfurt am Main: Suhrkamp). Habermas, J. (1998) Faktizität und Geltung. Beiträge zur Diskurstheorie des Rechts und des demokratischen Rechtsstaats (Frankfurt: Suhrkamp Verlag). Holzinger, K. (2001) 'Kommunikationsmodi und Handlungstypen in den Internationalen Beziehungen. Anmerkungen zu einigen irreführenden Dichotomien'. Zeitschrift für Internationale Beziehungen, 8 Vol., No. (2), pp. 243-86. Jervis, R. (1997) System Effects. Complexity in Political and Social Life (Princeton: Princeton University Press). Kahler, M. (2000) 'Conclusion: The Causes and Consequences of Legalization'. International Organization, 54 Vol., No. (3), pp. 661-83. Keohane, R.O., Moravcsik, A. and Slaughter, A.M. (2000) 'Legalized Dispute Resolution: Interstate and Transnational'. International Organization, 54 Vol., No. (3), pp. 457-88. Klare, K.E. (1998) 'Legal Culture and Transformative Constitutionalism'. South African Journal on Human Rights, 14 Vol., No. (1), pp. 146-88. Knill, C. (2001) The Europeanisation of National Administrations. Patterns of Institutional Change and Persistence (Cambridge: Cambridge University Press). Liefferink, D. and Jordan, A. (2004) 'Measuring Europeanization of Policy Convergence. National Baseline Conditions.' In D. Liefferink (ed.), Environmental Policy in Europe. The Europeanization of National Environmental Policy (London: Routledge), pp. 32- 46. Martin, L.L. (1992a) Coercive Cooperation - Explaining Multilateral Economic Sanctions (Princeton, New Jersey: Princeton University Press). Martin, L.L. (1992b) 'Interests, Power, and Multilateralism'. International Organization, 44 Vol., No. (4), pp. 765-92. Mendrinou, M. (1996) 'Non-Compliance and the European Commission's Role in Integration'. Journal of European Public Policy, 3 Vol., No. (1), pp. 1-22. Müller, H. (1994) 'Internationale Beziehungen als Kommunikatives Handeln. Zur Kritik der utilitaristischen Handlungstheorien'. Zeitschrift für Internationale Beziehungen, 1 Vol., No. (1), pp. 15-44. Müller, H. (2002) Arguing, Bargaining and all that. Reflections on the Relationship of Communicative Action and Rationalist Theory in Analysing International Negotiations In (Frankfurt/ Main. Onuf, N.G. (1989) World of Our Making: Rules and Rule in Social Theory and International Relations (Columbia: University of South Carolina Press). Panke, D. (2002) Argumentieren und Verhandeln im Europäischen Verhandlungssystem und die Frage nach der Einflussreichweite der Kommission In http://www.mzes.uni- mannheim.de/lvs/texte/panke_ma.pdf (Mannheim. Panke, D. (2004a) Institutional Learning in the Shadow of Judicial Discourses. Argumentative Strategies and Compliance with European Norms In Paper prepared for the ECPR Fifth Pan-European Conference on International Relations, Section 1 "Legalization and World Politics", Panel 4 "Dimensions of Legalization: Delegation (Judicialization)" (The Hague, Sept. 9-11, 2004. Panke, D. (2004b) Learning during Judicial Discourses In Manuscript (Heidelberg. Payne, R.A. (2001) 'Persuasion, Frames and Norm Construction'. European Journal of International Relations, 7 Vol., No. (1), pp. 37-61. Rasmussen, H. (1986) On Law and Policy in the European Court of Justice. A Comparative Study in Judicial Policymaking (Dordrecht, Boston, Lancester: Martinus Nijhoff Publishers). Rein, M. and Schön, D. (1993) 'Reframing Political Discourse.' In J. Forester (ed.), The Argumentative Turn in Policy Analysis and Planning (Durham: Duke University Press), Reinhardt, E. (2001) 'Adjudication without Enforcement in GATT Disputes'. Journal of Conflict Resolution, 45 Vol., No. (2), pp. 174-95. Risse, T. (1999) 'International Norms and Domestic Change: Arguing and Communicative Behavior in the Human Rights Area'. Politics and Society, 27 Vol., No. (4), pp. 529- 59. Risse, T. (2000) '"Let's Argue!": Communicative Action in World Politics'. International Organization, 54 Vol., No. (1), pp. 1-39. Risse, T. (2002) Konstruktivismus, Rationalismus und die Theorie Internationaler Beziehungen - Warum empirisch nichts so heiß gegessen wird, wie es theoretisch gekocht wurde In M. Zürn (ed.), Forschungsstand und Perspektiven der Internationalen Beziehungen in Deutschland. Risse, T. (2003) 'Konstruktivismus, Rationalismus und die Theorie Internationaler Beziehungen - Warum empirisch nichts so heiß gegessen wird, wie es theoretisch gekocht wurde.' In M. Zürn (ed.), Forschungsstand und Perspektiven der Internationalen Beziehungen in Deutschland (Baden-Baden: Nomos Verlagsgesellschaft), pp. 99-132. Schimmelfennig, F. (1997) 'Rhetorisches Handeln in der Internationalen Politik'. Zeitschrift für Internationale Beziehungen, 4 Vol., No. (2), pp. 219-54. Schimmelfennig, F. (2001) 'The Community Trap: Liberal Norms, Rhetorical Action, and the Eastern Enlargement of the European Union'. International Organization, 55 Vol., No. (1), pp. 47-80. Smith, M.E. (2000) 'Conforming to Europe: The Domestic Impact of the EU Foreign Policy Co-operation'. Journal of European Public Policy, 7 Vol., No. (4), pp. 613-31. Snyder, F. (1993) 'The Effectiveness of European Community Law: Institutions, Processes, Tools and Techniques'. Modern Law Review, 56 Vol., No. (1), pp. 19-54. Tallberg, J. (2002) 'Paths to Compliance: Enforcement, Management, and the European Union'. International Organization, 56 Vol., No. (3), pp. 609-43. Tallberg, J. and Jönsson, C. (2001) Compliance Bargaining in the European Union In ECSA International Conference (Madison, Wisconsin. Tsebelis, G. (1990) Nested Games: Rational Choice in Comparative Politics (Berkeley: University of California Press). Tsebelis, G. (2002) Veto Players: How Political Institutions Work (Princeton NJ: Princeton University Press). Ulbert, C. (1997) Die Konstruktion von Umwelt. Der Einfluss von Ideen, Institutionen und Kultur auf (inter-)nationale Klimapolitik in den USA und der Bundesrepublik Deutschland (Baden-Baden: Nomos). Wheeler Cook, W. (1927) 'Scientific Method and the Law'. American Bar Association Journal, 13 Vol., No. (305-306). Wurzel, R.K.W. (2004) 'Germany. From Environmental Leadership to Partical Mismatch.' In D. Lieferink (ed.), Environmental Policy in Europe. The Europeanization of National Environmental Policy (New York: Routledge), pp. 99-117. Zangl, B. (2001) 'Bringing Courts Back In: Normdurchsetzung im GATT, in der WTO und der EG'. Schweizerische Zeitschrift für Politikwissenschaft, 7 Vol., No. (2), pp. 49-79. Zukin, C. and Snyder, R. (1984) 'Passive Learning: When the Media Environment is the Message'. Public Opinion Quarterly, 48 Vol., No. (3), pp. 629-38. The emphasis on processes of collective and institutional learning has the advantage that unstable compliance (as repeated norm-violations) can be captured in addition to stable compliance and continued non-compliance. For several reasons (e.g. multiple actors in the implementation stage, specific character of regulations), I exclusively focus on the legal transposition of European directives into national law. Additionally, continued non-compliance can occur when not all actors, relevant for the production of legal acts after an ECJ judgment, undergo processes of reflexive or instrumental (for the distinction between collective and institutional learning see IV, V, and VI). Speech acts and logics of action can be distinguished analytically (see also Holzinger, 2001, Müller, 2002, Risse, 2002; Schimmelfennig, 1997, 2001). Whereas actors behaving according to the logic of communicative action can only use arguments, actors behaving according to the strategic logic of action or the logic of appropriateness can potentially use both types of speech acts, since the selection of the speech acts is subject to the type of rationality. It is, for example, strategically rational for an actor to use an argument instead of a speech act of bargaining, in order to pursue her preferences, when her bargaining power is perceived as too low and the changes for influence are expected to be higher through arguing. Hence, from the fact that actors use arguments it cannot be deduced the actors follow the logic of communicative action and are themselves consciously motivated to become persuaded. This analytical distinction fits well with the interactionist approach, since this approach links prospects for learning to the dominant pattern of speech acts in combination with systemic preconditions and not to logics of actions. From the criterion of a dominant pattern of speech acts, which are characterized through certain relationships between structure and content of speech acts, for a structure of interaction to exist, it follows that structures of interactions cannot exist at the same time. However, structural changes can occur during interactions. Both concepts, the subjective and the intersubjective world, are social constructions. Whereas all actors are affected more or less equally by the intersubjective world, the subjective world refers to the internal conditions (domestic constellations such as positions and influence of organized interests) with which an actor is confronted. His own subjective world affects an actor more intensively than other participants of interactions (who themselves face their own subjective worlds). Since reflexive learning is an unintentional process (see also Checkel, 2001a, 2001b, Zukin and Snyder, 1984: 629-630), it is also possible that short cuts lead some of the actors to accept an argument as true, right or appropriate, even though common standards are lacking, because they attribute authority to the speaker. However, short cuts do not contribute to the establishment or maintenance of any of the structures of interaction, because it is unlikely that all actors undertake similar short cuts simultaneously. It allows for but does not result in stable compliance, since collective reflexive learning has to be transformed into institutional learning, for stable compliance to be achieved (for the intervening variables see part VI). The extensive discussion of theoretical accounts for bargaining dynamics and processes of instrumental learning during interactions in front of the ECJ would be beyond the scope of this paper. With the shadow of financial sanctions (article 228) and the shadow of external reputational losses, there are two sources that might increase the cost-imposing constraints during the adjudication phase. As argued elsewhere, both explanations suffer from several theoretical shortcomings and are not supported by empirical evidence (Panke, 2004b). The institutional context is thus, conducive to reflexive collective learning. However, all states acting within the context of the ECJ have strong substantial interests and/or strong strategic preferences pointing towards non- compliance, since otherwise the case would have been already settled in earlier stages of the infringement proceeding. In this sense, ECJ referrals are least likely cases for further reflexive institutional learning. The hierarchy of heuristics has been developed in the ECJ's daily practise (interview COM #1). Systemic interpretations put a disputed norm into the context of the whole legal document. Norm interpretations in the context of unchallenged primary or secondary law provide insights on scope and content of the disputed norm, since negative exclusions and positive contributions allow for concretizing disputed norms. This method was applied in only four cases. In all those cases, the outcome was continued non-compliance. Moreover, sometimes non-accused states join in the judicial discourse in order to support and help the accused state (institution of 'Streithelfer'). When also those states doubt the correctness of the reconstruction of the original will, reflexive collective learning is additionally unlikely. The overall number of cases carried to the ECJ between 1978 and 1999 for cases of incorrectly transposed European directives into national law is 148. Hence, quantitative methods are not applied. Instead I rely on a qualitative research design, allowing for process tracing. According to the most similar systems design, I select for the policy fields environment and social policy directives with broad and narrow interpretational scopes for two countries (UK and Germany) in order to get variation on the independent variable 'scope of interpretation' (while the other independent variable 'interpretational heuristic' is constant) and on the intervening variables (such as the horizontal and vertical fragmentation) while controlling for alternative variables such as administrative capacity and economic power. So far I have not concluded the eight case studies. Especially the process tracing part on the causal mechanisms is work in progress. Therefore I use the empirical insights; I have gained so far, only for illustrative purposes. Most importantly, in a wide series of interviews the relevant actors are asked what they think of the directive, how they perceived and reacted to the judgment, and what laws (content and scope) they would prefer to regulate the subject matter and how legal adaptation should be conducted. Also it is examined which reasons they give for their conduct. If instrumental learning occurred, an actor most likely points towards the inadequacy of a directive shedding negative light on it, reacts to the judgment with rejection, while the actor (as opposed to an actor who did not learn at all) simultaneously prefers a loose patchwork-adaptation, and highlights external cost- imposing constraints, requiring fast legal adaptations. When reflexive learning occurred, actors give detailed reasons for why the directive should be incorporated into national law comprehensively, based on the usefulness or appropriateness of the directive itself. This corresponds to a press release of the European Commission, in which they referred to the German failures of compliance with the drinking water directive: "According to reports in the German media, the German government intends to avail of this so as to continue to only partially respect Directive 80/778/EEC. It is for these reasons that the Commission has decided to apply to the Court." (Reference IP/90/14, date: 10/01/1990). As a rhetorical reaction did the BGA argue that extensive readings of ‚states of emergency' (as including e.g. thunderstorms) would benefit the affected inhabitants, since they could be informed of the reduced water quality ( interview UBA # 2). The directive's aim was, in fact, almost compromised by this reading. Also strategic preferences were pointing towards non-compliance with the drinking water directive in the first place (how cost-benefit calculations changed after the ECJ judgment see part on the strategies of societal actors). High misfit on at least on of the three dimensions is a variable, constant across all eight case studies. Without a high misfit, adaptational requirements would probably be low and won't provoke serious domestic opposition against the complete legal transposition of a directive. Thus, the concept of ‚misfit' cannot explain variations in the outcome of the adjudication phase. Sometimes, national advocates professionally represent and fight for the national position, but are not convinced by the goodness of their own judicial arguments (interview COM#1). Since especially the earlier debate strongly drew on action theories, the focus on individual learning is not surprising. However, when individuals undergo processes of reflexive or instrumental learning during interactions, they have to transform their altered strategic preferences or substantial interests into institutional learning, for policy change (as the original dependent variable, see Müller, 1994) to occur. This is the result of a representative survey that I distributed among actors in the federal and states' parliaments as well as in the federal and states' ministerial bureaucracy of Germany. One indicator for reflexive institutional learning on the federal level is that no patchwork-style legal adaptation was undertaken for the transposition of the ECJ's ruling (which would have introduced uncertainties and ambiguities in the German legal water related acts, opening wide windows for repeated norm violations). The main actors in the health and environmental ministry (federal resorts responsible for water issues) managed to construct legal acts that were weaved relative comprehensively into the body of German legal acts (interview state ministry #4). Slight changes in the content of the drinking water regulation form 1995 were due to restricted resources for a complete revision of the water related law in 1992 and the horizontal fragmentation that lead to overlapping competencies between the health and the environmental ministry (interview state ministry #4). The analytical distinction between speech acts and logics of action is an important progress, on which Risse's concept of argumentative self-entrapment is built (Risse, 1999, 2000 Risse, 2003). According to his argumentation, the public forces actors to use argumentative speech acts, regardless of the underlying logic of action. Changes of positions occur not because the actors are intrinsically motivated to become persuaded, but rather because they become caught by their own arguments, which cannot be recalled in public without a loss of reputation. Thus, the public as a third standard brings about an argumentative dynamic (Risse 1999; 2003). However, this line of reasoning presupposes that the public appreciates arguing of their representatives more than bargaining. This implicit assumption is not generally valid, because in some situations the public might expect their representatives to push through the ‚national interest', or the preferences of organized interests through bargaining. Additionally, the concept of argumentative self-entrapment is not based on unitary assumptions about the level of strategic rationality. On the one hand, it presupposes perfectly strategic actors, calculating their reputational costs. At the same time, however, it is implicitly assumed that the actors are hardly rational regarding the selection of their speech acts, since they would otherwise anticipate the argumentative trap and eventually avoid the use of arguments at all. Because of these shortcomings, it can not generally be upheld that transparency allows the public to operate as a third standard which automatically favors arguing. One could argue that the politicization of issues favors the use of rightness and appropriateness as standards for the evaluation of ideas, to the disadvantage of truth. This line of argumentation would lead to the hypothesis that public settings favor the development of bargaining as the structure of interaction. This argument, however, requires two additional assumptions. Actors must act in a strategically rational manner, since they wouldn't be sensitive to concerns of the electorate otherwise. Since politicization increases public attention, it would additionally be necessary that the public always expects their representatives fighting for their interests with bargaining strategies. While this might be true in some settings, it is certainly not valid for all cases, especially not when the interested and mobilized public opposes governmental action-plans. Within this paper, peer pressure strategies are not discussed at length. Their application is unlikely, because of (1) the institutionalization of European norms of secrecy (Smith, 2000: 615), (2) the institution of other member states as 'Streithelfer' in front of the ECJ and (3) all states are themselves confronted with ECJ referrals. When the state is willing to comply, ECJ rulings are supportive to the realization of new preferences for compliance. Because of their binding character, a convicted state can refer to moral, normative or factual obligations arising from the ruling and is, in turn, strengthened against domestic opposition (Abbott and Snidal, 2000: 454). Without introducing assumptions on primary and secondary logics of action (which would lead to a rationalist or a constructivist bias (Panke, 2002) and thus to a bias regarding the outcome), it is not possible to determine in the abstract, which strategies actors employ. Therefore, I analyze contextual variables regarding their influence the prospects of success of different societal strategies. Framing is the process of selecting, organizing, interpreting, and making sense of a complex reality (Rein and Schön, 1993: 146). Through framing actors try to transfer a specific construction of a situation into an institutional arena, by highlighting ideational parallels between their conceptions and already shared ideas (Payne, 2001). When a frame is launched successfully, some ideas are factored out and a part of the ideational asset is accentuated. This reduces the ideational heterogeneity among the actors and provides an ideational environment, conducive to learning. When norms are already highly value-laden, the recurrence to other values is restricted, since value conflicts arise more easily but cannot be soled under arguing as the structure of interaction (see Panke 2002). The BGA nevertheless opposed the drinking water directive and even started activities against the federal regulation (from 1990). This was considered as inappropriate by the German health ministry, which reacted in threatening the BGA with dismissals (see interview federal ministry #4). For a description of the solution between water providers and environmental groups on the one hand and the farmers associations and the chemical industry on the other hand see (Knill, 2001: 157). An example of a shaming strategy aiming at the federal level was the water quality report of an independent organization concerned with consumer interests as regards the quality of products (Stiftung Warentest). Although the value of environmental protection is relatively strongly institutionalized in Germany, re-framing of the debate in terms conducive to compliance is very difficult. The drinking water directive resembles an 'end- of-pipe approach' which does not fit to the German idea of precaution and of the polluters-pay principle (Verursacherprinzip) (see also Knill 2001: 156). Hence, new frames emphasizing the importance of the environment instead of the health, would eventually lead to additional opposition against the directive and the ECJ's judgment. This is due to the high complexity of the EIA directive, which turned the directive itself into a ‚difficult tool' for environmental protection activities (interview state ministry #2). Moreover, the EIA directive is characterized by a misfit on three dimensions at once (policy, politics and polity). This low goodness of fit to already institutionalized German environmental norms makes shaming strategies quite difficult, since the starting points for arguments on the appropriateness of the EIA directive (in the way the ECJ read it) would necessarily be very abstract. Re-framing strategies are easier to conduct than shaming strategies, because the highly technocratic character of the EIA would eventually have allowed framing the issue as a traditional cross-cutting environmental policy. Council Directive 80/777/EEC of 15 July 1980 on the approximation of the laws of the Member States relating to the exploitation and marketing of natural mineral waters. 'Member States may, in exceptional cases, exempt a specific project in whole or in part from the provisions laid down in this Directive' 1