Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The nature of norms and the evolution of transitional justice Winston, Carla 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


24-ubc_2016_november_winston_carla.pdf [ 3.41MB ]
JSON: 24-1.0313411.json
JSON-LD: 24-1.0313411-ld.json
RDF/XML (Pretty): 24-1.0313411-rdf.xml
RDF/JSON: 24-1.0313411-rdf.json
Turtle: 24-1.0313411-turtle.txt
N-Triples: 24-1.0313411-rdf-ntriples.txt
Original Record: 24-1.0313411-source.json
Full Text

Full Text

  THE NATURE OF NORMS AND THE EVOLUTION OF TRANSITIONAL JUSTICE  by  Carla Winston  A.B. magna cum laude, Cornell University, 2003 M.A., The George Washington University, 2006   A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  Doctor of Philosophy in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Political Science)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  August 2016   © Carla Winston, 2016 ii  Abstract There are a wide variety of contemporary international norms: some are large and diffuse, while others are concise and clear. Some norm-adopting states follow precedent, while some adapt or innovate even as they profess to follow the same norm. This dissertation attempts to explain the variation in both norms and states’ normative decisions by investigating the conceptual structure of norms. I argue that while norms are currently understood to be a single grouping of three components – a problem, a justification, and a behavior – in reality, the international community often allows deviations from those combined components as acceptable interpretations of the original norm. This dissertation makes two contributions – a concept and a decision-making model – to bridge the gap between norm structure and the outcomes of norm diffusion. The first, the “norm cluster,” expands the single norm into a group of interlinked but distinct justifications and behaviors related to a common problem. Within a norm cluster, actors have more freedom to choose the combination of components which is most appropriate for them. The second innovation, a decision-making model, helps to understand and partially predict state decisions with respect to the contents of the norm cluster, previous norm adopters, and local conditions. Because a norm cluster’s contents are determined intersubjectively, the actions of new adopters and the reactions of the interested community combine to produce normative evolution. I demonstrate the applicability of these contributions by tracing the evolution of the field of transitional justice (TJ). I expand upon existing behavioral datasets and create an additional dataset consisting of official TJ justifications. By mapping justification and behavior over space and time, I show via multiple methods that the field of transitional justice has resulted in a wide variety of normatively acceptable outcomes which reflect choices made at all points on the decision-making model. I undertake further analysis of TJ’s evolution to show the processes of innovation, discourse, acceptance and resistance which I argue shape the content and boundaries of a norm cluster, and provide an explanation for the patterns of continuity and change which have defined and changed the field.   iii  Preface I am the primary researcher, analyst, and sole author of all parts of this dissertation. The research conducted did not require ethics approval, and no part of it has yet been accepted for publication by another outlet.   iv  Table of Contents Abstract ...................................................................................................... ii Preface ....................................................................................................... iii Table of Contents ....................................................................................... iv List of Tables ............................................................................................ viii List of Figures ............................................................................................. x List of Abbreviations .................................................................................. xi Acknowledgements ................................................................................... xii Chapter 1. Introduction .............................................................................. 1 1.1. Outline of the study .......................................................................... 7 Chapter 2. Structure, diffusion, variation and choice: Norms and norm clusters ................................................................................................................ 12 2.1. The contemporary conception of the international norm..................... 13 2.1.1. Norm function and structure ......................................................... 14 2.1.2. Diffusion and the dual quality of norms ......................................... 18 2.1.3. Structures of variation, continuity and change ................................ 23 2.2. A new theoretical construct: Norm clusters ....................................... 27 2.2.1. Norm clusters in the real world: Investigating structure .................. 29 Multiple behaviors ................................................................... 30 Multiple ideations .................................................................... 31 2.2.2. Creating variation in a norm cluster: “Unfused” diffusion ................. 33 2.2.3. Expanded choice and decision-making: Costs, risks, and likelihoods . 37 Decision-making, continuity and change .................................... 39 Outcomes of the USCC process ................................................. 43 2.3. Processes of cluster emergence, evolution and limitation ................... 46 2.3.1. Norm cluster emergence and evolution .......................................... 48 Characteristics of successful innovations and change agents ....... 48 Diffusion and change ............................................................... 50 2.3.2. The limits of acceptable variation .................................................. 55 Resistance/backlash: Behaviors or ideations not accepted ........... 57 Internal contestation: Incompatibilities within the cluster ............ 59 v Intervening ideations: Reduced scope of the problem ................. 61 Freezing of boundaries: Codification/legalization ........................ 62 2.4. Conclusion ..................................................................................... 63 Chapter 3. Case and methods: Building a transitional justice dataset ...... 66 3.1. The field of Transitional Justice ........................................................ 67 3.2. Observable implications ................................................................... 72 3.3. Data collection ............................................................................... 75 3.3.1. Data availability and collection ...................................................... 77 3.3.2. Data architecture for quantitative analysis ..................................... 82 3.4. Determining the boundaries of an empirical norm cluster ................... 84 3.4.1. Problem: Past or current human rights violations ............................ 85 3.4.2. Adding behavior to problem: Transitional justice mechanisms .......... 86 3.4.3. Adding ideation to behavior and problem ....................................... 91 TJ Ideations collected and categorized ...................................... 91 Missing ideations: Alternative internalized norms ........................ 96 Missing ideations: Non-normatively justified behavior ............... 100 Alternative ideations: Non-TJ normative behavior ..................... 103 3.4.4. The norm (cluster) life cycle: The approximate tipping point.......... 107 3.5. Result: The empirical transitional justice norm cluster ...................... 111 Chapter 4. Diffusion and variation: The structure of the TJ norm cluster ............................................................................................................... 113 4.1. Cluster diffusion: The TJ norm cluster as a whole ............................ 114 4.1.1. The “problem” and the decision to act: TJ theory and evidence ..... 115 4.1.2. Diffusion leading to choice: The contagion effect .......................... 119 4.2. Examining the dataset for evidence of variation .............................. 121 4.2.1. Variation within the norm cluster ................................................. 122 4.2.2. Variation in state choice ............................................................. 125 Intra-state continuity: Combinations of ideation and behavior ... 129 Intra-state change: Multiple combinations ............................... 131 4.3. Exploring patterns of continuity and change in the TJ norm cluster ... 137 4.3.1. Continuity over time: A gold standard or bare minimum ................ 138 4.3.2. “Best precedent” and the roots of transitional justice .................... 142 vi  4.3.3. Recombination: Determinants of TJ norm component choice ......... 152 4.4. Conclusion: Many different outcomes. ............................................ 161 Chapter 5. The dynamics of cluster evolution ......................................... 165 5.1. Within-cluster dynamics: Clear continuity ........................................ 166 5.1.1. A formalized gold standard: Prosecutions for major IHL violations .. 168 5.1.2. No bare minimum: Multiple behaviors in a single state .................. 173 5.1.3. Direct, fused diffusion to similar actors ........................................ 177 5.2. Within-cluster dynamics: Some continuity, some change .................. 182 5.2.1. Unfused diffusion and new combinations ..................................... 184 5.2.2. Expanded typologies of similarity ................................................ 190 5.2.3. Changing internal hierarchies ...................................................... 194 5.2.4. Diverse choices of best precedent: Commissions of inquiry ............ 199 5.3. Evolution at the boundaries: Growing the norm cluster .................... 205 5.3.1. “New” ideations ......................................................................... 207 5.3.2. New behaviors ........................................................................... 211 5.4. Conclusion ................................................................................... 217 Chapter 6. Limits on cluster evolution .................................................... 222 6.1. Locating and distinguishing cluster contestation .............................. 224 6.2. Examples of critiques seeking to limit cluster contents ..................... 229 6.2.1. Resistance to new components: Ideations and behaviors .............. 230 6.2.2. Internal contestation .................................................................. 233 Conceptual: Amnesties for accountability ................................. 233 Practical: Rwanda and trials for reconciliation .......................... 235 6.2.3. Interference with other norms .................................................... 243 6.2.4. Scope limits ............................................................................... 249 6.2.5. Freezing via treaty ..................................................................... 254 6.3. Conclusion ................................................................................... 256 Chapter 7. Conclusion ............................................................................. 258 7.1. Conceptual contributions: The norm cluster and the USCC model ..... 259 7.1.1. Norm clusters: Theoretical and practical advantages ..................... 261 7.1.2. Decision-making and the USCC model ......................................... 265 vii  7.2. The norm cluster and the evolution of transitional justice ................. 266 7.3. Limitations and opportunities for further research ............................ 270 7.4. Conclusion ................................................................................... 276 Bibliography ............................................................................................ 278 Appendix A. Data collection and reliability ............................................. 300 A.1. Data organization: Differences from existing datasets ........................... 300 A.2. Potential remaining sources of data distortion ...................................... 305 A.3. Dataset comparisons: the TJDB and the norm life cycle ......................... 309 A.3.1. The TJDB .................................................................................... 309 A.3.2. Norms theory: The S-curve ........................................................... 312 Appendix B. Expanded tables .................................................................. 315 Appendix C. Codebook ............................................................................ 323 Appendix D. List of cases ........................................................................ 330  viii  List of Tables Table 3.1: All transitional justice mechanisms, 1970-1999. .............................................. 90 Table 3.2: Transitional Justice Ideations, 1970-1999. ..................................................... 95 Table 3.3: Transitional Justice-related mechanisms with and without ideations, 1970-1999. ..................................................................................................................... 103 Table 3.4: Amnesties by ideation, 1970-1999. ............................................................. 103 Table 3.5: Data retention rate, TJ mechanisms and TJ mechanisms with TJ primary ideations, 1970-1999. ..................................................................................... 106 Table 3.6: The Transitional Justice Norm cluster: “true TJ” behaviors and ideations, 1970-1999. ............................................................................................................ 111 Table 4.1: Means comparison on key theoretical variables, 1970-1993. .......................... 117 Table 4.2: Transitional Justice Ideation-Behavior combinations for primary ideation only, 1970-1999. .................................................................................................... 122 Table 4.3: Abbreviations for ideation-behavior combinations used in the sociograms. ...... 127 Table 4.4: TJ Ideation and behavior combinations during the emergence (1970-1993) and cascade (1994-1999) phases. .......................................................................... 140 Table 4.5: Distribution of TJ choices by context: new democracy, recent or sustained conflict, or neither. ......................................................................................... 144 Table 4.6: Ideation-behavior abbreviation codes used in the sociograms. ....................... 146 Table 4.7: Logit comparisons of selected components for key theoretical variables, 1970-1993. ............................................................................................................ 156 Table 5.1: Ideations for Commissions of Inquiry, primary only and both primary and secondary, 1994-2007 .................................................................................... 203 Table A.1: Comparison of the case universe of the TJDB and Winston (event-count) datasets, 1970-1999, by TJ mechanism type. .................................................... 310 Table A.2: Comparison of means, Transitional Justice Database Project and Winston dataset, 1970-1999. ....................................................................................... 311 Table A.3: Means comparison for key selected variables, TJ adopters (primary ideation only), behavior-only adopters, and non-adopters, 1970-1993. ............................ 315 Table A.4: means comparisons of key theoretical variables, true-TJ adopters (primary ideation only) and mech-only adopters, 1970-1993. .......................................... 317 Table A.5: Table 4.3 with additional cases of secondary ideations added. ....................... 317 ix  Table A.6: Abbreviations of ideation-behavior combinations used in the sociograms in Chapter 4. ..................................................................................................... 318 Table A.7: Combination mapping if the two "roots" are democratization and genocide/politicide history. .............................................................................. 318 Table A.8: Means comparison, GN-created groups 2 and 3: key variables related to TJ’s theoretical roots. ............................................................................................ 319 Table A.9: all “True-TJ” behavior logits, 1970-1993. ..................................................... 320 Table A.10: all primary Ideation logits, 1970-1993 ....................................................... 321 Table A.11: Logits of primary and secondary true TJ ideations, 1970-1993. .................... 322    x  List of Figures  Figure 2.1: Formal model of the tripartite structure of contemporary international norms. . 16 Figure 2.2: Visual representation of a norm cluster. ....................................................... 29 Figure 2.3: Norm cluster of the Nuclear Non-Proliferation Treaty. .................................... 31 Figure 2.4: The anti-plastic-bag norm cluster, as identified by Clapp and Swanston (2009). ............................................................................................................ 32 Figure 2.5: Modified USCC Decision tree for state choice related to the norm cluster. ........ 38 Figure 2.6: Norm cluster with some incompatibilities. ..................................................... 60 Figure 4.1: Average contagion rates of new TJ adopters and all states, 1970-1999. ........ 120 Figure 4.2: Sociogram of state choice with respect to the TJ norm cluster, 1970-1999. ... 128 Figure 4.3: Sociogram of state TJ choice by groups of similar actors, 1970-1993. ........... 147 Figure 4.4: The empirical expression of the Transitional Justice norm cluster, 1970-1999, based on the constructed dataset. ................................................................... 162 Figure 5.1: The TJ norm cluster, 1970-1982. ............................................................... 218 Figure 5.2: The TJ norm cluster, 1983-1993. ............................................................... 219 Figure 5.3: The TJ norm cluster, 1994-1999 ................................................................ 219 Figure A.1: transitional justice norm diffusion by proportion of states in the system and by count, 1970-1999. .......................................................................................... 313 xi  List of Abbreviations  ICC  International Criminal Court ICTY/R  International Criminal Tribunals for the Former Yugoslavia and Rwanda MEPV  Major Episodes of Political Violence TJ  Transitional Justice TJDB  Transitional Justice Database USCC  “Use, Select, Change, Create” UN  United Nations US  United States  Additionally, at several points in Chapter 4, specific ideation-behavior combinations are referred to via two-letter codes. They can be found here:   Accountability Human Rights Rule of Law Truth Transition/ Democracy Peace Reconciliation Amnesty AA AH AL AT AD AP AR Lustration LA LH LL LT LD LP LR Reparation RA RH RL RT RD RP RR Trial TA TH TL TT TD TP TR Commission CA CH CL CT CD CP CR  xii  Acknowledgements My time at UBC has been enriched by a countless number of people, including professors, fellow graduate students, musical colleagues, and friends. Together they have made the past seven years a rich mix of learning, teaching, thinking, performing, and relaxing, and I could not have made it through this grueling process without them. The various communities of the Department of Political Science, the Liu Institute for Global Issues, Green College, and the UBC Symphony Orchestra have provided an intellectual home within the university, and I am grateful to belong to such diverse groups of scholars, thinkers, performers, and friends. This dissertation would not have been possible without the diligent attention and support of my supervisor, Dr. Katharina Coleman. She stepped in when I was in a bind, guided me through the proposal and data collection process, advised me on the non-dissertation portions of my grad school experience, and provided many, many sets of comments on multiple chapter drafts. Her critique was invaluable in guiding me to the arguments lurking just beyond what I had actually written, and this thesis is immeasurably better for it. I am similarly grateful to Dr. Brian Job, for his continued care and attention over the past several years, including the emphasis on qualitative analysis; to Dr. Chris Kam, for needed advice on quantitative methods; and to Dr. Taylor Owen, who performed the task of last-minute addition with enthusiasm and grace. My committee has been the main driver of my intellectual work, but my friends have been my social, mental, and emotional support throughout this journey. They have provided advice, support, last-minute assistance, and no small amount of laughter, without which I would be a humorless, stressed-out, stay-at-home ball of boring. They have commiserated when times were tough or critiques seemed impossible to answer; they have helped me reformulate ideas when I thought I had been scooped; they have helped me stick to my pomodoros in order to get things done; and they have allowed me to laugh, cry, rant, and wonder as was needed. Special thanks go to Tamara Williams, Stephen Hay, Afsoun Afsahi, Katrina Chapelas, Beth Schwartz, Clare McGovern, Sule Yaylaci, Anastasia Shesterinina, Aim Sinpeng, Deborah Barros Leal Farias, Yana Ghorokovskaia, Dan Schwartz, Jennifer Allan, Adam Bower, Chris Tenove, and Shayna Plaut for that peculiar combination of academic and social interactions which have kept me both intellectually stimulated and emotionally engaged as I have written this dissertation. Of course, my biggest thanks go to my family, who have provided me with opportunities that many may never get, whose enthusiasm for my pursuit of a PhD has lent me strength when I was feeling overwhelmed, and whose own incredible achievements have pushed me to excel and to engage with the world. My aunts and uncles, grandparents and cousins have fed me, transported me, sometimes housed me, asked eagerly for updates, and always supported me. My sisters have made sure that my ego doesn’t grow too large. My grandmother Phyllis, who passed away only two weeks before this thesis was complete, boasted of me (and the rest of her family) to everyone she met. Most importantly, I have unending gratitude for my mother and father, who knew exactly what getting a doctorate was like and have encouraged and supported me anyway, in every way possible. I could not have done it without you. 1  Chapter 1. Introduction In theory there is no difference between theory and practice; in practice there is. –Anonymous In the face of unspeakable atrocities and widespread violations of human rights, by both governments and opposition forces, what does justice look like? What should justice look like, and who makes that determination? While there is a general consensus that past human rights abuses should be addressed, there is remarkable variation in both the theoretical discussion of “ideal” justice processes and in more empirically-based arguments about “possible” processes. One expert put it thus: As can be seen from the experience of emerging democracies throughout the world in dealing with the legacy of the past, there are no hard and fast rules or easy answers about how to resolve the dilemma... Official policies on this issue have been dictated not only by strict principles of justice, but also by the need to balance ethical and legal concerns with the hard realities of politics. (Benomar, 1993) The field of transitional justice emerged out of this set of tensions, and these questions remain even after more than 40 years of experience, research, and argument. Transitional Justice, as a set of norms, is both strong and fuzzy: although most actors now acknowledge that something should be done to deal with past human rights violations, exactly what should be done, and why, has varied enormously between countries and over time. Although certain principles are widely acknowledged to be important, and several potential behavioral models exist for states to choose or adapt, by and large the question of what transitional justice “looks like” in any single state, for any particular history and current political context, is relatively open. At the same time, the transitional justice community has made it clear that not just anything goes; there are limits to what “counts” as a transitional justice process, and what is an appropriate reason for pursuing it. 2  We also have a wealth of potential examples to choose from, and yet states continue to innovate within the TJ field. Trials of former government officials have been an option since the Nuremberg Tribunals after World War Two, and the International Criminal Court, founded in 1998, continues the tradition of holding leaders accountable for war crimes, crimes against humanity, and genocide. Over the past decades, new processes have been added to the pantheon of “appropriate” forms of justice: truth commissions, reparations, lustrations and vetting processes, institutional reforms, memorials and public education programs, and local traditional processes of healing, forgiveness and reintegration between victims and perpetrators. Some of these processes have achieved widespread popularity, while others remain rare. Still other options, such as amnesties, have seen their use decrease over time, or have been applied in increasingly selective ways. In addition, the goals or values with which these behaviors are identified have also varied: sometimes accountability, democracy, or rule of law is most important, while at other times, or for other states, peace, reconciliation, human rights or truth are the driving motivation behind addressing past abuses.  This variation, and indeterminacy, is not unique to transitional justice. What exactly does democracy look like? How should states fight climate change? How are gender equality issues to be addressed, and why is it important to do so? Each of these, broadly construed, is an international norm, or at the very least a set of related norms. However, there is no international treaty telling us exactly how states should design a participatory democracy, and even countries which have been democratic for decades still struggle with questions of institutional design and electoral representation.1 There are many ways in which states can                                                  1 Democracy is still an “essentially contested concept,” even as certain proponents fight for it to be declared a universal human right (Dryzek, 2016). 3  fight climate change, and the various international protocols offer states a number of options to fulfil their emissions reduction targets. Gender equality efforts, and reasons why they are important, vary so much from state to state and culture to culture that it may be nearly impossible to catalog all permutations. And yet, scholars, policymakers, and activists are all certain that these norms exist. These are all examples of variation within international norms: in definitions, in options, and in actual state practice. For the norms above, and for others, there seem to be a lot of acceptable options for “appropriate behavior.” Sometimes an innovation or a particular interpretation is met with resistance by the international community, but in a lot of cases that community seems to accept many states’ choices as different, but legitimate, attempts at implementing the very same norm. How is this variation in the content of a norm created, and why is acceptable? An additional complication arises due to the fact that these are not the only types of norms in the international system: not every norm is large and fuzzy, like democracy or transitional justice appear to be. Some norms are clear and simple, and states have no room to improvise or maneuver. Norm implementation is gauged by strict standards, and rejection or cheating leads to sanction. For example, the prohibition on starting a nuclear war is clear and absolute: there are no circumstances in which a nuclear “first strike” is acceptable behavior, and states seeking to augment that capability through nuclear weapons development, such as North Korea, are subject to severe sanctions. Sometimes states have multiple acceptable options, but sometimes they do not. We are thus confronted with yet another form of variation, this time between international norms: some norms are succinct and others fuzzy. Even the structure of norms seems to vary. Why? 4  This variation in the structure of normative behavior, both within and between states, is not merely of academic interest. It presents real questions for those who design, push for, implement, and judge the outcomes of principled action. The possibility of using multiple behaviors to pursue a common goal leads to questions of institutional choice: which policies or practices can best achieve the stated goal, especially in the diverse contexts in which different types of states will implement them? The existence of similar goals between actors can identify major influences on learning, sources of advice and potential aid, but also creates the peer group which judges whether the resulting institutional choice and design is appropriate, and whether its execution has been successful in furthering the common goal. Treaty bodies, in particular, often create this peer group by constituting tribunals or committees to judge whether treaty obligations have been fulfilled by member states. On the other hand, justification of the same behavior in several different ways suggests differences in institutional design, as certain features can be strengthened or weakened, added or omitted, to emphasize some goals over others.2 It also implies that some institutions may be inherently more malleable, or able to potentially further more goals, than others. In addition, the choice of ideation, like the choice of behavior, creates or exposes links to other states: in the form of actual or desired identity groups, similar values create bonds of like-minded actors which can share experiences, but they also lend those actors the power to judge both sincerity and success in achieving those goals with the chosen behaviors. Asserting that you are a democracy, for example, opens a state to                                                  2 Norms on campaign finance regulation, for example, include a broad variety of rules on who may fund what political efforts; the inclusion of public financing for electoral campaigns, or the exclusion of corporate donations to candidates, can be used to specifically address different types of concerns about the fairness of the use of money in politics. 5  critique from other democracies on whether derogations from certain democratic standards, such as declaring martial law, are acceptable. In non-democracies, declaring martial law would be judged by very different standards of appropriateness. Within a community of generally like-minded actors, states and non-state actors alike are continually presented with questions about how much variation in norms is acceptable: whether changes to existing precedents are “close enough” and should be accommodated, or if these changes go too far and are no longer appropriate. The construction of international treaties, for example, appears to be a double-edged sword: agreements tend to be either “narrow but deep,” with clear and easily enforceable targets but low membership, or “broad but shallow,” with a large number of signatories but much smaller, or more differentiated, required commitments (Barrett, 2002). Although each furthers normative goals, they do so differently: the first can create major changes but only for a portion of the world, leaving others without the perceived benefits, while the second is more universal but also more incremental, and potentially harder to avoid members free-riding and cheating. Advocates of normative change are thus presented with a choice: push for deep commitments (and potentially limited adoption) or universal adoption (and risk results)?3 The academic study of norms which attempts to understand these practical issues is wide-ranging. There are several different definitions for norms, and a wealth of norms research identifying specific norms empirically and determining how they emerge, diffuse,                                                  3 New innovations in multilateral cooperation have attempted to overcome this problem by incorporating both strict targets and diverse national policies for implementing it, but there is not yet solid evidence to see if it works better than the older dichotomy of treaty models. For example, see the 2015 Paris Climate Agreement.  6  and evolve over time and space. Some research treats norms as clear and succinct, but some attempts to understand evolution, variation, and change. Some norms identified in the real world are simple while others are complex; some are identified as single norms, while others are viewed as part of a collection. Some norms appear to develop in a linear fashion while others do not; some are created from the “top down” (via international institutions) and others from the “bottom up” (via state practice). Norms research is as varied as the phenomenon it is attempting to understand. However, I believe that the field is experiencing a disconnect between these pieces of the norms puzzle: current definitions, and the conceptual structure they imply, do not appear to allow for the incredible variety of norms, and of state practices regarding these norms, observed in the real world. Whereas definitions of the norm such as “a standard of appropriate behavior” (Finnemore & Sikkink, 1998, 891) imply a single, relatively stable unit, many diffusion theories show evolution, variation and change. While some norms identified empirically do conform to this structure, others appear inextricably bound within groups, with their exact content hard to determine. Still other norms appear one way when they first emerge, but another way later on. There are elements of both stability and continuity, and flexibility and change, within the theoretical and empirical record. Both theoretically and empirically, then, the extent of potential and actual variation between behaviors and their normative justifications, and the resulting variation in norm shapes and sizes, presents a problem. This dissertation attempts to address the theoretical and empirical issues presented above by looking more deeply at basic concepts and definitions: specifically, diving into the conceptual structure of the norm. I will argue that the singular conceptual structure implied by the prevailing definitions of a norm does not 7  match its supposed dual nature of stability and flexibility, or the evidence provided by either norm diffusion theories or the empirical record. I argue that to truly understand the variety of norms and norm-based practice in the international system, we need a conceptual structure that allows for, but does not automatically imply, variation and evolution. In other words, the conceptual structure of the contemporary international norm must be flexible but not without boundaries; it must be able to evolve but remain recognizable. It must be responsive to the dictates of the international community, but also open to new arguments. Lastly, this conceptual structure must also be able to accommodate the wealth of different types of norms, and of norm diffusion mechanisms, which have been identified through empirical research by other scholars.  That is the aim of this dissertation: to provide a conceptual structure of contemporary international norms which fully encapsulates the stunning variety of real-world norms and their evolution over time and space. In essence, the research question guiding this work is, “How does the structure of the contemporary international norm explain how states decide what ‘counts’ as ‘appropriate behavior’?” The challenge, I argue, is in how to we reconcile our understanding of what norms are with our evidence of how they diffuse. 1.1. Outline of the study  This dissertation will engage the central question posed above in three parts: first, what exactly are norms, from a conceptual and structural standpoint, such that both continuity and change  in the outcome of norm diffusion processes are possible? Second, how do states choose between continuity and change when adopting new norms? Third, 8  how do state choices and norm structure interact over time to create the variety of norm choice outcomes, and types of norms themselves, that we see in the world? Chapter 2 first engages a host of existing norms theory to argue that several of our current definitions of norms, and much of the existing work on norm diffusion, is insufficient to account for the variety we see in the real world. It lays out the prevailing concept of the international norm, which consists of three separate components: a problem to be solved, an ideation which gives the problem moral value, and a behavior which attempts to address that problem. However, I will argue that such a singular structure is inadequate to account for the variation which arises as norms diffuse throughout the international system. Instead, as Wiener (2007a) argues, norms display dual qualities: they can be both stable and flexible, implying both continuity and change when adopted by different actors. These two qualities, when combined with innovation, strategic choice, and argumentation by norm adopters, can create collections of “appropriate behavior” which include multiple potential combinations of interlinked but distinct norm components. I call this larger conceptual structure a norm cluster. It is this norm cluster which I argue most accurately encapsulates both the existing definition of a single norm and the empirical record of “acceptable variation” in norm adoption and implementation. Second, given all this acceptable variety, there must be a way for states to determine their own “best” or “most appropriate” possible option. I therefore present a decision-making process adapted from the USCC model (Use, Select, Change, Create) of institutional change (Jupille & Snidal, 2005) to account for the costs, risks, and calculations which determine whether states will simply adopt the normative choices of their predecessors or try something new. However, I do not suggest that all innovation is acceptable; the international community is responsible for deciding exactly which variations 9  “count” as norm adoption and which do not. I posit that innovations must be accompanied by argumentation which supports their inclusion into the existing norm cluster, and that contestation, critique and resistance from interested parties limits what is considered appropriate with respect to the norm cluster’s composition and application on the ground.  Third, I apply these two theoretical innovations to a single case, both to demonstrate their analytical value versus traditionally conceived norms and to show how they interact to produce normative evolution over time. Chapter 3 introduces the field of Transitional Justice (TJ) and explains my case selection (the TJ norm cluster’s structure and evolution from 1970 to 1999) as well as data collection and organization methods. The field of TJ offers an excellent example of a norm cluster: although supposedly developing from theoretically distinct roots in democratization and the law of armed conflict, I suggest that in practice these two sets of norms have, through the processes I describe in Chapter 2, formed a single norm cluster in which norm components interact relatively freely within the boundaries of what is considered “appropriate behavior” for states confronted with the problem of widespread or severe human rights violations.  Conceptualizing norms in a new way requires collecting new types of data to analyze, and changes previously accepted notions of which events are part of the “case universe” of a transitional justice dataset. Chapter 3 therefore also details the definition of the normative TJ field of practice as a collection of individual TJ events consisting of a problem, and ideation, and a behavior. There are certain methodological challenges; most importantly, not all of the required new data (ideations) may be available. I posit a number of reasons why this happens, and explain why some such omissions are actually helpful for the project I am pursuing: namely, that some ideations are missing from the record precisely because the behaviors they justify are not actually part of the TJ norm cluster. I 10  use the norm life cycle (Finnemore & Sikkink, 1998) to make this case. The end result of my data collection effort is, I argue, an accurate empirical record of the field of normative transitional justice attempts from 1970 to 1999. In Chapter 4 I begin my examination of the TJ norm cluster by looking for evidence of continuity and change in states’ choice of ideation and behavior, and exploring the variation which I would expect to result from decisions made at different steps of the USCC model. I use a variety of methods, both quantitative and qualitative, to chart these outcomes: cluster adoption, fused and unfused diffusion, and innovation. I find that the field’s history contains a stunning amount of variety, in overall ideation-behavior combinations chosen as well as in why and how individual states appear to choose TJ processes.  While some trends are apparent, including those stemming from the field’s supposed roots, the picture on the whole is one of integration of practices and diffusion based on many different determinations of “similar actors.” I also present evidence suggesting that the motivating factors behind choice of a specific behavior and a specific ideation often vary: they appear to be based on different state characteristics.  My quantitative analysis may show the shape of the TJ norm cluster, but it cannot explain how it came to be, or judge whether or not all this variation in forms of transitional justice is actually acceptable to the international community. Chapters 5 and 6 therefore explore the evolution of the TJ norm cluster over time, using mostly qualitative analysis. Chapter 5 employs the modified USCC model as a guide to chart the processes of diffusion and innovation which create new combinations of problem, ideation and behavior and, in some cases, new boundaries for the TJ norm cluster as a whole. Concentrating mainly on commissions of inquiry, and using examples from the United States, Canada, Uruguay, Uganda, Argentina, South Africa, Poland, Germany and the Philippines, as well as more 11  global overviews, it shows the evolution of specific parts of the transitional justice field of practice: via mechanisms of fused and unfused diffusion, the development (or not) within TJ of necessary standards of appropriate behavior, changing internal hierarchies of components within the cluster, the addition of new ideations and behaviors, and expanded scope of the justice problem. Chapter 6 deals with some of the limits of the TJ norm cluster: because not all innovation is acceptable, and because some formerly acceptable options may come to be seen as no longer appropriate, I explore how critique from the interested community defines the boundaries of a norm cluster and thus the limits of appropriate behavior. After distinguishing cluster-related critiques from other forms of critical discourse, I give examples of cluster limitation such as resistance to new components or combinations, interference from other norms or from other components within the cluster, scope limits on the justice problem, and the potential to reify or “freeze” certain aspects of the TJ cluster via formalization. Amnesties, trials and lustrations feature heavily in this chapter, and discourses are found coming from both domestic and international critics of the proposed TJ processes. Lastly, the conclusion situates the norm cluster concept, the modified USCC model, and the cases of transitional justice studied here within broader literatures on norms, norm diffusion, and the TJ field more generally. It summarizes the dissertation’s main theoretical and empirical contributions and some of the its limitations in terms of questions not answered, which offer avenues for potential future research.12  Chapter 2. Structure, diffusion, variation and choice: Norms and norm clusters  Norms are one of the most widely studied topics of contemporary International Relations scholarship. Norms scholars have created an extensive theoretical and empirical literature to identify, describe, analyze, and predict the emergence, diffusion and effect of norms in the international system. However, I will argue below that this scholarship is inconsistent in how it treats state choice and the outcomes of norm diffusion processes: in particular, it is divided between theories which emphasize continuity and those which allow for change in the process of norm adoption and implementation. This suggests the need for a re-examination of some of the basic assumptions underlying the study of norms.  I suggest that this problem can be remedied through a closer examination of the structure of norms themselves. I show that the prevailing understanding of contemporary international norms is that they have a tripartite structure, consisting of a problem, an ideation, and a behavior. These structural components can be combined to create norms as traditionally understood, but there may not always be a unique combination leading to a single “standard of appropriate behavior” (Finnemore & Sikkink, 1998). When multiple combinations of interlinked but conceptually distinct ideations and behaviors exist and offer multiple acceptable solutions to a common problem, I call this looser collection of norm components a “norm cluster.” Decision-making with respect to a norm cluster, rather than a traditional norm or bundle of norms, allows for more state innovation and a wider range of acceptable ideation-behavior combinations. I further argue that focussing on norm clusters, rather than on single norms as currently understood, allows for a more precise understanding of how principle-based action diffuses and evolves in the international system. 13  This chapter first investigates the structure of international norms as currently conceived and the implications of that structure for norm diffusion, arguing that norms’ dual qualities of stability and malleability create a disconnect between their singular construction and the outcomes of various diffusion processes. To remedy this problem, I then propose a new conceptual structure for understanding principled action in the international system, which I call a “norm cluster.” Section 2 explains this concept by examining both theoretical cluster structures and several real-world examples, and points to the separate diffusion of norm components as the principal enabler of norm clusters (as opposed to single norms). To understand states’ individual choices with respect to the norm cluster, I then propose a decision-making model which can help to explain and partially predict both state actions and the resulting shape and contents of the norm cluster over time. Lastly, as norm clusters are intersubjective constructs, Section 3 explores the interactive dynamics of practice and discourse which accompany choice, innovation, acceptance and limitation with respect to state actions and the contents of the norm cluster. 2.1. The contemporary conception of the international norm Norms can be thought of as “shared intersubjective understandings that make behavioral claims (Checkel, 1997),” or “standards of appropriate behavior (Finnemore & Sikkink, 1998).4” They “are used to make demands, rally support, justify action, ascribe responsibility, and assess the praiseworthy or blameworthy character of an action” (Checkel, 1997). In the international system they are also understood as “providing solutions to coordination problems, reducing transaction costs, [and] providing a language and grammar of international politics” (Cortell & Davis Jr, 2000, pp. 65-66.). With these diverse functions,                                                  4 There is no single accepted standard definition of norms, but a number of closely related and widely used definitions exist. See also (Chayes & Chayes, 1995; Jepperson, Wendt, & Katzenstein, 1996, 54). 14  norms can be understood to have both constitutive and constraining aspects (Checkel, 1997). I first argue that the interaction of these two functions creates a norm’s conceptual structure, which consists of three components: problem, ideation, and behavior. It is these three components which diffuse throughout the international system.  However, various processes of norm diffusion highlight different aspects of a norm’s “dual nature”: they are both stable and malleable. While the basic structure of a norm remains constant, the contents of that norm may or may not change due to processes of diffusion. Second, I suggest that the theories of diffusion that lead to normative outcomes each have explanatory deficiencies which reflect this duality: one type maintains object consistency, but sacrifices the possibility of variation in outcomes, while the other does the reverse. I conclude that a new conceptual structure is needed which can account for both continuity and change, and allows for variation in both processes and outcomes. 2.1.1. Norm function and structure Social constructivists such as Wiener (2008; 2014) and Wendt (1999) argue that norms have “constitutive” functions: they create categories of actors and actions, and determine those actors’ identities and interests. Norms create meaning through the construction of intersubjective (i.e. collectively held) understandings of who and what things are. This meaning includes whether or not the item in question is valued (positively or negatively).  The assignment of value, or ideational content, is subjective, but important: it is the relationship to ideations which defines problems, because material or social facts must be understood in relation to the values of a society in order to be seen as good or bad (Wendt, 1999, 111-112). This relationship accords meaning to new facts by linking them to previously established ideations (objects, practices or goals) which already have societal value. A new or pre-existing fact becomes problematized if it is interpreted in such a way 15  that the existence of this fact negatively impacts the attainment or continued practice of an ideation. In other words, no norm-conforming principled action is possible if “underlying conditions in world affairs” are not defined as problems, in relation to ideations, by actors publicly justifying their actions with respect to those ideations (Carpenter, 2007, 112).5  On the other hand, scholars also perceive norms as having a “constraint” function: they define “acceptable” justifications for behavior. If a state wishes to pursue a specific behavior to address a defined problem, acceptance by other actors of the legitimacy or appropriateness of that behavior depends on how it is publicly justified. It is rare that actors will seek to publicly justify behavior based on pure self-interest; usually actions are justified by reference to widely held ideations (Risse, 2000, p. 17). Public acceptance of that action as legitimate depends on whether external actors believe that it will actually operationalize the given (appropriate) ideation and therefore address the problem. The ideation which identifies a problem also limits the behaviors which are appropriate to solve it. Given that norms are accepted as having both constitutive and constraining functions, I echo Hurrell and MacDonald (2012, p. 61) in arguing that determining “appropriate behavior” depends on the interplay of these two processes. Actions are deemed legitimate by audiences when they are seen as “appropriate” in a given circumstance, and understanding the appropriate values to be enacted in a given circumstance guides determination of what action to take. Wiener (2008; 2014) echoes this process with what she calls “type 2” norms, or organising principles, as the construction of a legitimate link between widely held but abstract social values and much more specific rules                                                  5 I say “norm-conforming” principled action, and make reference to public justifications, in order to separate normative content as specifically discussed versus underlying motivations, which may or may not accord with publicly stated ideations. However, as I will argue in later chapters, a statement of specific ideation recognised as relevant and appropriate to addressing a given problem is a critical component of norm-conforming behaviour. 16  and procedures. Organising principles thus perform the dual function of recognizing legitimate goals and prescribing specific action to be taken for those who wish to address a problem. Therefore, I conclude that norm-building is the process of constructing a bridge between the constitutive and the constraint functions of norms such that a combined statement is reached: “Given this problem, my values dictate this behavior.”6 This statement gives a norm an intrinsic tripartite structure. First, a norm presupposes a problem, which is the issue to be addressed. Second, the norm includes an ideation, which represents a particular item of value, either current or desired. It is the enjoyment of something “good” or the avoidance of something “bad” and, as such, gives moral weight to the problem. Third, a norm enjoins a particular behavior: the action to be taken to address the given problem which allows the actor to better express or practice the ideation. In short, a problem inhibits the full enjoyment of an ideation, and necessitates a corrective behavior. Figure 1, below, formalizes these components into a logical structure. 7  Figure 2.1: Formal model of the tripartite structure of contemporary international norms. Richard Price’s account of the anti-landmines norm (1998) provides a concrete example of the interaction of these three components. The norm is the nexus between the                                                  6 Risse, Ropp and Sikkink (1999) suggest that through a process of pressure or coercion leading to behavior change, and then habituation and socialization (the “spiral model”) a behavior could also lead eventually to an ideation. 7 This conceptualization of norm structure does not address a key piece of some norm definitions: identity. The full definition of a norm according to Finnemore and Sikkink is “an appropriate standard of behavior for an actor with a given identity [italics added]” (Finnemore & Sikkink, 1998). I view identity as a conditioning agent for the whole tripartite structure: the agent’s identity affects whether or not one sees a given context as a problem, which ideation[s] which might be “activated,” and the potential range of behaviors which are possible to enact and effective at achieving the chosen ideation given the context. Given different identities, the content of a norm may change, but its underlying structure will not. If [problem], [ideation] suggests [behavior]. 17  problem (the existence of landmine technology), an ideation (the ethical need to discriminate between civilians and soldiers in warfare), and a behavior (the non-use of a certain type of weapon). Because landmines cannot distinguish between civilians and soldiers and will explode no matter who triggers them, they are a problem for societies that value civilian-soldier discrimination. Framing the norm in terms of its underlying structure results in the following directive: “If [landmines], [civilian discrimination in warfare] suggests [non-use].” This structure underlies a number of the basic definitions of norms in use in International Relations scholarship. It is clearest in March and Olsen’s “simple behavioral proposition”: Most of the time humans take reasoned action by trying to answer three elementary questions: What kind of a situation is this? What kind of a person am I? What does a person such as I do in a situation such as this? ... Fitting a rule to a situation is an exercise in establishing appropriateness (March & Olsen, 2008, 690). Similarly, Fearon argued that norms translated into ordinary language take the form “Good people do (or do not do) X in situations A, B, C…” (Fearon, 1997). “Good people” is indicative of an ideation: what makes people “good” is linked to what they think is valuable or important. “Situations A, B, C” is the problem determined by actor values. To “do (or do not do) X” is to engage in the behavior indicated by the interaction of the problem and values. Goertz (2003, 34-38) echoes this structure, but alters the order of logic, when he frames norms as a “moral syllogism”: Major Premise, Minor Premise and Conclusion. The major premise is the ideation: What is generally appropriate or valuable. The minor premise is the problem, or specific situation to which the ideation must be applied, and the conclusion is the behavior which follows.  18  Lastly, the logic of argumentation (Risse, 2000) also suggests a tripartite structure for norms and normative decision-making. The forum in which this occurs is one of “argumentative rationality,”  A communicative process in which actors collectively deliberate over their assumptions about the world, the values they share, how those assumptions and values should apply to their behavior, and whether particular behaviors actually conform with abstract standards (Payne, 2001, 783). The topics of discussion for actors engaging in this process are the constituent pieces of the norm itself: their assumptions about the world (problem), shared values (ideations), how they should act based on those assumptions and values (behavior). The last piece, whether the particular accurately reflects the abstract, acts as confirmation that the decision reached at the end of this deliberation is one that all actors agree on.  2.1.2. Diffusion and the dual quality of norms Norms may have a single conceptual structure but I contend that, as Wiener argues, (2007b; 2008; 2009),  they have a “dual nature”: they are both stable and flexible. I suggest that nowhere is this more apparent than in the process and outcomes of norm diffusion, in which states will variously interact with norms as they are (stable) or change them, either purposefully or accidentally (flexible). This duality, I argue, presents a problem for norms researchers studying the diffusion and evolution of norms throughout the international system. Wiener identifies two major, and opposing, qualities of norms: first, that norms are stable social facts, and second, that they are continuously contested and reconstituted. The logics of consequences and appropriateness, she argues, view norms as stable: once established, their validity is not contested, and their existence as social facts shapes state 19  behavior (either “norm following” or norm rejection). A third logic, that of argumentation, views norms as flexible only until their intersubjective meaning is established (“norm-setting”); after this point, they are considered stable and the first two logics structure future state behavior. Lastly, Wiener proposes a fourth logic: that of contestedness. The basic assumption of the logic of contestedness is that norms are both stable and flexible: while norms’ content and validity may be challenged at any time, recognition that a norm exists still structures behavior. In turn, that behavior may influence the content and meaning of the norm, leading to further contestation (Wiener, 2008, 37-58). As norms travel (or diffuse) throughout the world, and are adopted by states, I argue that they also display both these qualities of stability and flexibility. I suggest that the dual quality of norms offers different potential outcomes of norm diffusion, each with its own form of variation in states’ principled decisions: first, that norms diffuse exactly as they were originally structured (stability) and, second, that a norm’s content may actually change in the process of diffusion and adoption (flexibility).8 As the logic of contestedness suggests, if norms are both stable and flexible at the same time, then different states should react differently to the norms presented to them: some should adjust their behavior based on the norm’s existing meaning, while others should adapt that meaning. Diffusion outcomes over the whole group of involved actors, then, should sometimes reflect stability, and sometimes change. Diffusion has been thought of as both an outcome and a process: the spread of ideas and behaviors throughout a system, and the process by which this happens. In an                                                  8 I suggest that a norm’s structure, however, does not change: that there must be a combination of problem, ideation, and behavior remains the same even as the exact problems, ideations and behaviors themselves may change in response to contestation. 20  international system where pure coercion is relatively rare and there is not always an international organization directing action, governments make decisions in an atmosphere of “uncoordinated interdependence”: they make their own decisions, but factor in the choices of other governments (Elkins & Simmons, 2005). The outcome of a successful diffusion process is the near-universal adoption, implementation and internalization of the norm for all actors with the given problem related to that norm. International Relations scholars have charted the diffusion of a wide variety of norms, including European cooperation (Checkel, 2005), liberalization (Simmons, Dobbin, & Garrett, 2008), democracy (Huntington, 1991), women’s suffrage (Finnemore & Sikkink, 1998), election monitoring (Hyde, 2011; Kelley, 2008), anti-apartheid (Klotz, 1995), anti-chemical weapons (Price, 1998), anti-human trafficking (Lloyd, Simmons, & Stewart, 2011), and transitional justice (T. D. Olsen, Payne, & Reiter, 2010; Sikkink, 2011). I suggest that some studies of international norm diffusion implicitly assume that the tripartite structure of norms is stable, or fused together, such that norm diffusion is also assumed to include the diffusion of all three parts of a single norm: ideation, problem, and behavior. This stable construction lends itself to norm diffusion mechanisms based on the logic of appropriateness (for example, persuasion) or consequences (learning). One consistent finding among diffusion scholars, for example, is that “entities that share similar cultural attributes tend to adopt the same practices…Actors negotiating a complex set of political choices regard the actions of actors with perceived common interests as a useful guide to their own behavior” (Elkins & Simmons, 2005, 45). To rephrase the above comment using norm structure, those states with similar identities, or “cultural attributes,” will have similar goals (ideations) and common interests (problems) and will generally adopt 21  similar policies (behaviors). In essence, all three parts of a norm’s structure interact to create diffusion.  March and Olsen suggest that the process of diffusion thus works through the construction of “typologies of similarity”(2008, 693-695).9 Faced with norm choice, the precedent of the state which is most similar to the one currently making the decision is the most likely to be imitated. Antje Wiener argues that norms by their very nature have “recognizable … prescriptions for behavior (Wiener, 2008),” meaning that if norms diffuse, actors know what they have to do. Some studies of specific diffusion processes and mechanisms use the “fused” construction as well: for example, Payne theorizes persuasion by arguing that the frame of a normative debate is a key factor in an actor’s ability to persuade others. Payne uses a norm’s frame to “provide a singular interpretation of a particular situation and then indicate appropriate behavior for that context” (Payne, 2001, 39).  However, this appears to be only half the story as regards norm diffusion, as it only engages the “stable” quality of norms. Some accepted norm diffusion mechanisms can also result in normatively acceptable decision-making outcomes beyond merely acceptance or rejection of a fused norm: specifically, by emphasizing the “flexible” aspect of norms, these processes can actually change a norm’s contents as it diffuses to new adopters.10  The simplest reason behind norm variation may lie in the difficulty of determining exactly what the norm in question is. Norms are not always easy to identify: they are                                                  9 This is contrasted with multiple realizability, in which no diffusion occurs and states react in similar manners but completely independently to similar circumstances, such as natural disasters (Elkins & Simmons, 2005; Wendt, 1999, 165). 10 There may be a lag in timing (Cortell & Davis Jr, 2000), but if a state voluntarily accepts the principle (or ideation) of a norm, it will voluntarily seek to implement it (behavior). 22  “continuous, rather than dichotomous… [and] come in varying strengths (Legro, 1997, 33). In addition, as actors operate in conditions of bounded rationality (Jones, 2003) and potential psychological bias (Khong, 1992), hindrances to full information and decision-making capacity may result in imperfect identification and choice of norms by empirical actors, creating diffusion outcomes that do not exactly match previous examples.  Wiener also points out that norm diffusion implies that norms travel: they are taken out of their original (highly specific) context and applied to a new (highly specific) context. Since norms evolve through interaction, normative meaning will inevitably change slightly as context changes from actor to actor (2008, 63). She calls this the phenomenon of “meaning-in-use” (Wiener, 2009). Sandholtz (2008) echoes this argument when he posits that interpreting the existing rules to apply in new, and specific, situations constantly triggers conflicts over interpretation, and resolving these conflicts shifts the nature of the rules themselves, creating a cycle of norm change.  Differences between domestic structures and capacities can sometimes make exact emulation of an existing norm impossible. Domestic context, as Cortell and Davis (2000) argue, accounts for differences in the interpretation of the meaning of norms (ideation and perhaps problem), and differences in states’ compliance with them (behavior). If an ideation is deemed important, but the expected behavior is impossible given the norm-adopting state’s own structures and context, it might conceivably adopt the same discourse but pursue alternative behaviors.  On the other hand, a state could conceivably determine that the empirical benefits of a norm-driven behavior are desirable, but for a slightly different reason. Variation in the perceived effects of a policy is a key difference between the factors motivating states to 23  adopt norms during different phases of the norm life cycle (Finnemore & Sikkink, 1998): During phase 1, states are persuaded that the norm is “the right thing to do” because it resonates with their own identities, either pre-existing or newly redefined. However, states in phase two of the norm life cycle may also perceive an additional value in adopting the norm (for example, the duty to follow international law as the consensus decision on the standard of moral progress, if the norm is now formalized) which was not relevant for states in phase 1. Therefore, the ideational motivations of early adopters and later adopters of the same norm may be different, even if their behaviors are the same.  The same result may occur due to certain processes of norm diffusion such as grafting (Price, 1998) and localization (Acharya, 2004). Because each involves matching an ideation from a previous adopter or norm entrepreneur with an ideation already prevalent in the newly adopting state, some flexibility in interpretation is necessary: the local ideation may not mean exactly the same thing as the existing ideation, but it is conceptually close enough that the terms can be used to refer to the same expected behavior. The resulting final ideation is a mix of the imported norm and the state’s own pre-existing ideation, but is understood by all parties as comprising a part of the same norm. 2.1.3. Structures of variation, continuity and change The above types of norm diffusion each focus on one aspect of a norm’s dual quality. In this section I argue that each also has implications for how we think about the structure of that variation and of the object itself (the norm). I suggest that there are three major types of variation with respect to norm adoption as an outcome of norm diffusion. The first, fused diffusion, creates (for each specific norm), only two options: accept or reject. It limits variation, but maintains object continuity. A middle option, which engages questions of choice among a set of fused and stable norms, allows for more variation in 24  adopters’ choices while maintaining object continuity, but does not fully explain all empirical outcomes. The third, and most flexible, option, that of norm adaptation, allows for many types of variation, in both contents and state choice, but loses object continuity. I conclude that these outcomes expose a conceptual gap in the norms literature: the tripartite structure of the contemporary international norm, by itself, does not adequately capture all the possible types of continuity and change that follow from a norm’s dual nature. If norm diffusion occurs as a single, combined process for any particular norm, then states encountering a new norm in the international system have only two possible outcomes to their decision-making processes: accept the whole fused norm, or reject it. Contestation or alternative norm choice may be therefore thought of as rejection, since the existing construction of one or more of the norm’s components is rejected in favor of an alternative construction. The norm’s contents remain the same, and state choices are easily identifiable.11 This is the strictest form of both variation and continuity: variation occurs only in terms of yes and no, and the norm’s meaning itself does not change, even as actors accept or reject it. However, this is clearly not the whole story, even with respect to fused norms: much of the recent work on norms and decision-making examines the interaction of multiple related norms. These groupings are variously referred to as sets, bundles, hierarchies, frameworks, structures, regimes, and institutions.12 Finnemore and Sikkink are careful to                                                  11 Unless that state is cheating or unable to properly carry out implementation, which might make identification of acceptance or rejection difficult until a later time.  However, cheating and failure are both versions of “accept”: failure is acceptance with poor implementation due to capacity, and cheating is acceptance with poor implementation due to bad faith. 12 For example, Finnemore and Sikkink (1998) refer to both bundles and institutions, and Betts and Orchard (2014) echo their use of bundles. Crocker (1999) uses framework, while Tiemessen (2011) talks about normative structure; Acharya (2004) uses hierarchy, although this is more common among international legal scholars. Regimes tend to refer to the specific organizational structures which regulate normative behavior in the 25  note the difference between single norms and norm groups: “norm definition isolates single standards of behavior, whereas institutions emphasize the way in which behavioral rules are structured together and interrelate (a ‘collection of practices and rules’) (Finnemore & Sikkink, 1998, 892).13 Similarly, a normative framework or structure (Crocker, 1999; Tiemessen, 2011) acts as a guide for actors to determine their behavioral choices based on where their specific circumstances fall within a larger collection of norms.  Although the specific versions of the general problem wherein each fused norm might be applied are similar, and all possibilities must be considered, a single outcome from among a group of stable norms must eventually be decided based on each state’s own particular context. At the end of these decision processes, each final norm choice remains stable and identifiable: although multiple options exist, their contents remain fixed and it is relatively easy to determine which stable option a state has actually chosen. I suggest that this construction creates the appearance of variation, but it also maintains a certain amount of stability, both with respect to individual norm choices and to the structure of the norm itself: while state outcomes may vary depending on the collection of single norms within a bundle and the decision-making process used to determine which is “most appropriate,” each norm within that bundle or framework remains identifiable and constant, and can only be accepted or rejected. Scholars examining the possible choices for different states, or at different time periods, would be able to distinguish the same combinations of problem, ideation, and behavior that make up each norm in the group. In other words, larger structures (institutions, regimes) are reducible to their parts (norms),                                                  international system, such as “the human rights regime”(Donnelly, 1986). Cortell and Davis (2000) refer to two “sets” of normative structures. 13 They do note that the content of these institutions, meaning the specific norms contained in them, changes over time. 26  and each of those parts has only two options. What looks like a lot of variation, then, is actually much more uniform with respect to single objects. On the other hand, I posit that decision processes which emphasize the “flexible” quality of norms allow for change in the object as well as variation in outcomes, but lack object stability.  Theories about norm change (Acharya, 2004; Job & Shesterinina, 2014; Sandholtz, 2008; Sandholtz & Stiles, 2009) therefore explain more variation, not only of actor outcomes but of the content of the norms we actually encounter empirically. States may adopt a norm and change its contents in a number of ways, or they may reject it outright. However, change in the object itself creates an additional problem: as the norm changes it no longer has the same content as the original; either the problem, ideation or behavior are now different depending on which actor we examine. Two scholars studying the same norm at different points in time or space, without knowledge of its evolution, would not identify it as the same object of study. Flexible norms, then, do not have object continuity in the same way that stable norms do. In conclusion, I argue that the processes of norm diffusion expose a conceptual gap: norms are an object of single structure, but dual quality. If norms are regarded as stable, the tripartite construction of norms is traceable throughout the diffusion process, but variation in choice is limited. On the other hand, if norms are flexible, the end result of each diffusion process still contains a norm with its tripartite structure, but the contents of that structure will vary such that the object of the “norm” itself changes from actor to actor. Multiple studies which chronicle both types of norm diffusion exist, so each approach has evidentiary support, but these studies generally do not engage both stability and flexibility, or all three types of outcomes, within the same work. This leaves the question of their 27  prominence in the real world, and their potential interaction, unexplored. In the next section I propose a solution to this gap.  2.2. A new theoretical construct: Norm clusters The analysis above highlighted inconsistencies between the accepted structure of the contemporary international norm and the variety of accepted outcomes of norm diffusion in the real world. I argue that the solution to this dilemma is a new conceptual structure: to restructure the concept of the norm itself, and therefore norm sets, into a looser and less determinate collection of interlocking norm components, which I call a “norm cluster.” The following section outlines the conceptual roots of the norm cluster, offering an alternative way of understanding indeterminacy and choice in the international system and providing some real-world examples. Next, I explore the resulting expanded nature of diffusion and state choice with respect to a norm cluster versus the other options for decision-making, namely a single norm or bundle of single norms. Finally, I discuss the role of innovation, contestation and discourse in evolving and limiting this expanded space. I define a norm cluster as a bounded collection of interrelated ideations and behaviors linked to a general problem. The ideations and behaviors contained within it (its boundaries) delimit the expanded space for “appropriate behavior” given the general problem. Within this cluster, ideations and behaviors may be combined, depending on specific state context, into a number of distinct but acceptable iterations of problem, ideation and behavior. However, no state must necessarily directly emulate a predecessor to be seen by others as adopting and implementing the norm cluster. Lastly, although the norm cluster does have boundaries, they are somewhat malleable based on processes of discourse and learning conducted by and between relevant actors. 28  I contend that the concept of a norm cluster retains both the constitutive and constraint functions of traditional norms: while a state is constrained to do something appropriate, it is enabled by the relative freedom of choice within the cluster to determine for itself what constitutes “appropriate behavior” in its own specific context. A certain amount of different meaning-in-use is tolerated by the community as long as a legitimate option is chosen, and a new option may be legitimized if the community accepts it. Kratochwil describes this decision-making structure as one that is “indeterminate at the level of individual choice but determinate at the level of defining classes of actions” (2001, 47). This also makes the norm cluster concept consistent with a “meta-norm” (Sandholtz 2007) in which the logic of appropriateness is invoked, but specific actions remain unspecified. In essence, I offer the argument that normative decision-making may, but need not, embody a single quality of norms: a single norm may be stable for some, and flexible for others, within a range of acceptable variation as defined by the community. A certain amount of variety in outcomes, given the complexities of the real world, is not only possible but likely, and often, but not always, acceptable.  I argue that a norm cluster allows for decision outcomes which are either stable or flexible, either fused or changed, such that each state’s options may be individually different, but all can be traced back to the larger object of study. Although some actors will directly adopt a stable norm, or choose from a number of stable norms, others will innovate, leading to change in their individual normative outcome but not change in the larger object. Lastly, some forms of innovation may even alter the boundaries of the norm cluster itself. The next section will give examples of real-world norm clusters which display this structure. 29  2.2.1. Norm clusters in the real world: Investigating structure I argue that a focus on norm clusters, rather than single norms or sets of norms, helps to account for the tremendous empirical variation in how states actually adopt and implement norms and how the international community reacts to those state decisions. Some states may exactly emulate a predecessor, while others make adjustments even as they profess to adopt the same norm. Some states take only one action; others take several. Some states offer one justification for their actions, while others may attribute the same behaviors to different ideations even when applied to the same problem. Lastly, any of these norm-based decisions may change over time as new states adopt the norm cluster14 and learn from previous decisions and outcomes.  A generic visualization of a norm cluster might look like this:  Figure 2.2: Visual representation of a norm cluster.                                                   14 I define norm cluster adoption as the choice of at least one acceptable combination of ideation and behavior within the cluster. 30  Essentially, the image above represents the largest possible, and most integrated, norm cluster: every ideation and every behavior may be paired, and each combination is applicable to the problem in question. A single general problem may be identified based on its ability to impair multiple separate, but related, ideations, and there exist multiple acceptable potential solutions to that problem. Each is a combination of ideation and behavior, but these combinations need not remain fused: behaviors may link to multiple ideations, andvice versa. The boundaries of the cluster are determined by the range of ideations and behaviors which are acceptable to the relevant observers and participants. Multiple behaviors However, not every norm cluster is necessarily so unwieldy. Empirically, norm clusters consisting of a single ideation and a relatively narrow problem, but a number of different behaviors, are quite common. Indeed, this is how norms scholars often explain domestic variation in the implementation of norms which comes from variations in pre-existing domestic structures or practices. The norm embodied in the Non-Proliferation Treaty (NPT) (United Nations, 1968) is a prime example of this structure: avoiding nuclear war is the overarching goal for all adopters.15 The specific problem is nuclear technology, which could enable nuclear war. However, depending on whether states have nuclear technology, either military or civilian, the expected behaviors for this norm are quite different. The norm cluster embodied by the Nuclear Non-Proliferation Treaty looks like this:                                                   15 This problem also applies to the norm of non-use of nuclear weapons, although no such treaty exists. The norms are related, but not identical. The NPT does not forbid states to use nuclear weapons. Non-proliferation does not mean non-use, and non-use does not necessarily imply non-proliferation.  In addition, non-use includes use not in war, suggesting more ideations for non-use than for non-proliferation. 31   Figure 2.3: Norm cluster of the Nuclear Non-Proliferation Treaty. (United Nations, 1968) States which have nuclear weapons are expected not to transfer them (Article 1), and states without are expected to neither accept nor develop them (Article 2). All states are required to put in place verification, reporting and monitoring procedures to prove that they are following through on the other behaviors (Article 3). Thus, each state has different behavioral obligations based on its specific domestic context, although the goal and the general problem remain the same. Some states may have multiple duties, while others may have only one. All states are required to report, but the existing nuclear powers have different specific behavioral obligations than do non-nuclear states. States which ratify the NPT adopt the same non-proliferation “norm” but, in so doing, they agree to carry out potentially varying behaviors depending on their particular circumstances.16 Multiple ideations The opposite configuration of a norm cluster occurs as well: states may justify a single behavior with reference to multiple related, but distinct, ideations. This may be an                                                  16 An additional clause exists in the NPT which is not part of the non-proliferation norm cluster: States Parties to the NPT are allowed and encouraged to transfer peaceful nuclear technology. This clause is included in order to incorporate, rather than override, law created by a separate treaty on peaceful nuclear technology transfers. 32  attempt to please multiple audiences invested in the problem, thus gaining consensus for the adoption of the single behavior, or it may be the result of variations in framing by different norm entrepreneurs, or both. It is not uncommon to see multiple clauses in the preamble of a law stating a number of different justifications for the same piece of legislation, which may in the end be quite specific in terms of institutional choice. In addition, the choice of ideation may be instructive in the particular design of those institutions, as certain design aspects are emphasized or ignored in the pursuit of a particular goal.  Figure 2.4: The anti-plastic-bag norm cluster, as identified by Clapp and Swanston (2009). An example is the development of the anti-plastic-shopping-bag norm. As Clapp and Swanston (2009) argue, the prohibition or restriction of plastic shopping bags17 has followed a pattern similar to the phenomenon of multiple realizability, in which very different but                                                  17 Not all states implement the exact same policy to effect the reduction in bags; some ban them altogether while others impose a fee for their use. I have condensed them here for simplicity of argument, although these could technically be considered separate behaviors. On the other hand, these subtle differences in essentially the same policy might be thought of as variations in institutional design; it is possible that different ideations in this cluster correspond to either banning or fees. 33  vaguely environmental justifications were given by the early-adopting states. In addition, there was no particular attempt at persuasion or strategic framing from one state to another; environmental NGOs within each country were primarily responsible for pushing the norm and thus for generating discourse. The anti-plastic-bag behavior has evolved into an international norm with varied justifications (and thus ideations): the bags block floodwater drains (Bangladesh), pose a threat to sacred cows (India), release toxins from incineration (Taiwan), harm the tourist industry (South Africa), and finally hamper the country’s “green” image (Ireland). Each was to be avoided. 2.2.2. Creating variation in a norm cluster: “Unfused” diffusion  I argue that the key to enabling and creating a norm cluster and its resulting variety of acceptable ideation-behavior combinations, as opposed to a single norm or bundle of single norms and the continuity they imply, is the ability for norm components to diffuse separately. This implies that a new adopter will not necessarily choose the exact same combination as a predecessor. If the link between a single problem, ideation and behavior is loosened, then the process of diffusion can therefore work separately on each norm component. The possibility arises that a state may identify a problem with respect to a different ideation, or adopt the ideation of one predecessor and the behavior of another, leading to an entirely new outcome. This is how norms change, but remain conceptually related. This is in contrast to the fused form of diffusion, found with respect to single or multiple norms but also within a norm cluster, in which the content does not change. The first steps of the normative decision-making procedure are the same in both cases, in the sense that a problem is identified with respect to an ideation and a response is therefore mandated. With only a single fused norm, as discussed above, relatively little choice is 34  available: accept or reject both ideation and behavior. Within a bundle of related single norms, a greater number of “acceptable” precedent combinations of ideation and behavior exist from which to choose, but states must choose a pre-existing acceptable combination. If a new adopter merely copies the exact combination of an earlier adopter, it chooses to reproduce a particular fused norm from within the group: all three parts of the norm diffuse together. Different actors may choose different pre-existing combinations to solve the same problem, but the components of these norms do not interact or diffuse independently of one another. Object continuity is maintained (stability). On the other hand, adoption of a norm cluster offers the potential for numerous combinations of components, and much more resulting variation (or change). Although each choice leads to a single (local) norm18 with a tripartite structure, these individual choices are assembled from a range of acceptable options rather than diffused from a single absolute.  A newly adopting state has a greater wealth of legitimate examples from which to choose specific ideation and behavior. Previous adopters recognize that the new adopter is engaging in “appropriate behavior,” and the norm cluster spreads, although specific combinations within it may not diffuse directly from one state to another as a fused unit. I suggest that unfused diffusion leading to change, rather than wholesale adoption of existing “fused” precedent and the maintenance of stability, may occur for a variety of reasons. First, actors at different stages of the diffusion process may be subject to different influences. In their spiral model of human rights norm diffusion, Risse and Sikkink (1999) argue that different actors use different diffusion mechanisms over time directed at a single                                                  18 I take the term “local norm” from Acharya (2004), but with an additional meaning: the combination of acceptable norm components chosen for a specific actor at a specific time. They need not be, as Acharya suggests, pre-existing in national society rather than international. 35  state. They offer five stages at which different actors control the diffusion process19:  transnational networks at stage 1, domestic opposition at stages 2 and 4, bilateral and multilateral organizations at stages 3 and 4, and finally the state’s own apparatus at stage 5 (1999, 20). If different actors and different processes of socialization are at work throughout a state’s socialization process, but are more or less effective at different points, it would suggest that these different actors (and their different diffusion processes) will be more or less effective at diffusing different parts of the norm.  Second, the literature on implementation and compliance (Betts & Orchard, 2014; Checkel, 1997) also suggests unfused diffusion: it assumes that the processes of reaching international agreement on norms and the processes of actually implementing them are subject to different influences from different actors. If the domestic actors necessary to create law are different from those who reach international agreement, the persuasion mechanism which worked on one may not work on the other. A form of strategic bargaining may take place instead, meaning that two different diffusion mechanisms are at work on the same norm in the same state, coming from potentially different actors and working on potentially different logics.20 Third, March and Olsen argue that the logics of appropriateness and consequences may coexist, compete, and complement one another within any single institution that determines rule-based behavior. They offer two hierarchies of logics in which “the logic of appropriateness may be used subject to the constraints of extreme consequences” or, alternatively, that “rules of appropriateness are seen as one of several constraints within                                                  19 They posit that these actors all work simultaneously, and on the same norm. However, some actors are theorized to be more effective at some stages than others. 20 This strategic bargaining is similar to Putnam’s two-level game (Putnam, 1988). 36  which the logic of consequentiality operates” (2008). Neither is consistently found empirically, meaning that even within a single norm’s diffusion pattern, different states apply different logics. They also point out that the two logics depend on different abilities and resources: to determine what is “right” and to calculate the expected utility of a behavior take different inputs from different actors, following different mechanisms (i.e. persuasion vs. learning). If one actor employs moral argument and the other rational utility calculation, then a state could conceivably agree with the ideation of the moral arguer and the behavior of the rational arguer. Because both the ideation and the behavior are already included in the norm cluster (albeit combined with different components), the new combination would only have to be deemed compatible to be accepted as a legitimate expression of the norm cluster.  Separate diffusion of norm components has been demonstrated in the real world. As an example of a logic of consequences following a logic of appropriateness which leads to unfused diffusion, Lloyd, Simmons and Stewart (2011) trace the spread of anti-human trafficking legislation: global discourse on trafficking has framed the problem in either criminal terms or victimization terms. If trafficking is a criminal activity, it is a problem for societies which value the rule of law and security (ideation); if its victims are most important, then it is a problem for societies which value human rights. Either frame could diffuse via a mechanism of persuasion from any actor in the global system. However, the authors find that the most consistent predictor of the diffusion of anti-trafficking law (behavior) is the level of road linkage with a previously adopting country: they argue that the shifting of direct negative externalities onto a state will cause it to change its policies. In this case, ideational change may lead in two different directions, and come from any influential actor in the international system, but behavioral change comes due to the 37  relationship with one’s direct neighbor. Different norm components diffuse via different mechanisms, from different actors, and at different times, and each state’s local version of the anti-trafficking norm cluster would depend on whether the criminalization or human rights frame was more persuasive.  2.2.3. Expanded choice and decision-making: Costs, risks, and likelihoods The expanded space for choice and the ability of different components to diffuse via different pathways creates the variety of state choices which we see in the real world with regard to normative decision-making. Rather than merely accepting or rejecting the fused norm, or accepting one option out of several and rejecting the rest, I suggest that several more “acceptable” outcomes exist. States may accept an ideation from one state and a behavior from another; they may apply an ideation-behavior combination to a different specific poblem within the large general problem (ie a different context); they may accept an existing ideation or behavior and innovate the remaining component (arguing that the innovation should become part of the cluster); and they may repurpose a behavior from an existing acceptable precedent to fit a new ideation, or vice versa. The question for a norm-considering state, then, is how to decide between continuity and change. These choices do not all carry the same amount of cost and risk, and thus the same likelihood of being chosen by any particular state. Going from a known quantity (even if the outcome is not optimal) to a relatively unknown new choice carries the risk of failing to help to solve the problem, and the effort and political capital necessary to gain acceptance of an innovation might be better spent on something else. When faced with the option to emulate a predecessor directly or to try something new, I suggest that states follow a decision-making pathway similar to that developed by Jupille and Snidal (2005) regarding institutional choice in the international arena. They propose a “USCC (Use, Select, Change, 38  Create)” decision-making process with regard to international institutions, which I adapt to norms and norm clusters. The decision chain is modeled below, and the following sections trace states’ decision-making with respect to the original and modified USCC processes, and the outcomes they imply, with respect to individual states and the norm cluster as a whole.  Figure 2.5: Modified USCC Decision tree for state choice related to the norm cluster.  39 Decision-making, continuity and change This section details, step by step, the decision-making process shown above, and shows where in that process actors have opportunities for continuity or change. The first three steps are the same for all actors: first, a problem must be recognized in general, or no new principled action is necessary. A social or material fact must be problematized such that it is perceived as prohibiting the attainment or full enjoyment of a goal or value. Denial that a problem exists in the abstract or general sense is norm denial, i.e. that the status quo is acceptable and nothing should be changed. Second, a state must perceive that the general problem has specific ramifications for that state, necessitating a change in behavior. This is a question of salience (Cortell & Davis Jr, 2000): If LGBTQ discrimination is a general problem but the government says that there are no homosexual people in Iran, for example, then Iran need not implement new behaviors to address a problem it does not have (Daily Mail UK, 2007). It also engages ideational questions, although not necessarily conclusively: salience implies that the problem inhibits goals or harms national identities, but at this stage that sense may be only general and be subject to further refinement as the process of norm choice proceeds. If, however, the actor in question recognizes both a general problem and finds it sufficiently salient that it should be remedied, the question remains of what to do, leading to Step 3, “Adopt norm cluster?”. This step adds behavior to the problem and ideation: given a problem that negatively impacts enjoyment of an ideation, how is it to be remedied? This question directly mirrors the decision-making process that constructs a norm in the first place, and completes the creation of a new norm with the form “if [problem], [ideation] suggests [behavior].” If no single norm or norm cluster currently exists then the state may invent one without outside influence (i.e. multiple realizability), but if either a single norm or 40  a norm cluster exists and has been advocated or implemented by other states, then the traditional process of norm adoption or rejection follows. This stage may also be thought of as the recognition of the “metanorm” (Sandholtz, 2007): something should be done, although what exactly should be done is as yet unclear. Step 4 suggests continuity, and therefore a single norm rather than a cluster: If a state is faced with universal necessary action, in the form of either a “gold standard” or a “bare minimum,” it is highly likely that the newly adopting state will adopt, or use, that existing fused norm. I define a universal necessary action as something which is generally understood that all states must do if they adopt the norm. A “gold standard” is a single ideation-behavior combination that is applicable to all parties, regardless of context, and is the only acceptable option. No deviations or alternative choices are allowed; to act otherwise is to reject the norm cluster. This is the case with the norm of non-use of nuclear weapons: no alternatives are acceptable substitutions, and refusal to avoid using nuclear weapons brings immediate condemnation from all parties. Norm adopters shall not use nuclear weapons – period. There are no other acceptable options. As noted above, however, in this case the original conceptual structure of a norm and that of the norm cluster is analogous: because the cluster comprises only one problem, ideation, and behavior, the norm and the cluster are functionally the same object. A different form of necessary action is a “bare minimum”: something that must be done if the norm cluster is adopted, no matter what else (if anything) is also done. In treaties, this may take the form of a reporting requirement: as in the NPT example above, all parties are required to share information regardless of what other actions are specified by their own particular contexts). Whereas a gold standard suggests complete convergence between the tripartite norm structure and the norm cluster structure, however, a bare 41  minimum offers the possibility of further action, and therefore the accompanying norm cluster structure becomes relevant. States may choose additional actions beyond the required minimum, but they need not do so in order to be regarded as cluster adopters.  Rejection of either the gold standard or the bare minimum is politically costly: it signals that the norm, either in single or cluster form, has not been adopted. However, in the case of a bare minimum, I suggest that some of that cost may be negated if the state in question does take on additional normative responsibilities beyond the missing minimum standard. One example would be the United States’ response to the International Criminal Court: it is not a signatory and does not acknowledge the court’s jurisdiction over its citizens, making it a non-adopter, but it engages in some of the ICC’s practices nonetheless: it will refer non-American suspects for investigation through the UN Security Council. However, I suggest that not all clusters include a universal necessary action. A gold standard may not exist at all: in fact, this is quite common given the relative scarcity of simple, absolute norms in contemporary international relations. Additionally, although reporting clauses are quite common in modern treaties, not all norm clusters contain a bare minimum action, especially those which are not (or not yet) in treaty form. Cluster adoption in these cases is more difficult to judge, and depends on what other (and how many) actions are taken, and how they are justified. This brings us to Step 5, where the structure of a norm cluster becomes relevant with respect to decision-making and discourse becomes important in justifying cluster interpretation to the community. At this point, if a norm cluster (multiple potential ideation-behavior combinations) exists, or if a state perceives that it is possible to create one (through innovation and arguments for acceptance of that innovation), additional decision points apply. Either 42  continuity or change (or both) are possible, and the USCC model becomes a useful analytical tool for distinguishing when actors will choose continuity and when they will attempt change. Steps 5 through 7 of the USCC model show the steps actors take with respect to the available and possible choices for creating a local norm from the choices available within a norm cluster. If no gold standard or bare minimum exists, and if several acceptable precedents are available, I suggest that states move to Step 5, and select the “best fit” among existing previous examples. This best fit relies on determining similarities in ideation and possible behaviors, both based on that actor’s specific context. Each choice here reproduces norm continuity: from a set, or bundle, of precedents, states choose an option that already exists. However, variation is still created, because states with different contexts will choose different pre-existing combinations. Because these outcomes are all understood to be acceptable given the correct context, the risk of condemnation is low and the decision not particularly costly: discourse is only necessary to explain why the chosen ideation-behavior combination is the correct precedent, and the combination itself is already acceptable to the community. However, the risk of practical failure to solve the original “problem” may be higher if the similarity which determines the “best” choice, and therefore predicts the likelihood of success, is not high enough between the deciding state and the predecessor.  Steps 4 and 5 led to continuity. Steps 6 and 7, on the other hand, signal change. If a satisfactory solution does not already exist for that state’s specific version of the problem in its domestic context, the state may choose to recombine elements from within the norm cluster. This is a change in combinations of norm components. It creates a new tripartite structure (norm) within the norm cluster, and a new potential precedent for other states to emulate in Step 5. While it does not expand the boundaries of the norm cluster, it increases 43  a cluster’s internal density by creating more possible connections between ideations and behaviors. Since all components are already accepted, discourse is only necessary to explain why this new combination should also be acceptable and how it will practically address the problem. Relatively little political capital is necessary, although still more than in Step 5, but the relative risk of failure goes up because there is no clear example of this particular choice working the way it is supposed to. Lastly, if choosing another combination from within the cluster is still not sufficient, Jupille and Snidal offer the choice of “create new institution.” For a norm cluster this means innovation: add a new ideation to an existing behavior, or vice versa or both, and argue that this new combination is also an appropriate response to the problem at hand. This constitutes both change in a tripartite norm, and change in the contents of the cluster itself. In these cases the risk of both condemnation and failure is high: there is no example of previous success and no way to judge the potential impact of the innovation, and the community may be loath to allow untested innovations into the norm cluster.  There is also higher cost than at previous steps: additional political effort and/or capital must be expended in order to successfully engage in the discursive processes that result in the re-drawing of the norm cluster’s boundaries. Convincing other parties to change their minds, rather than merely to make minor adjustments to beliefs they already hold (as in Step 6), takes more work. Outcomes of the USCC process I contend that this decision-making process results in a hierarchy of likely outcomes at each step (either continuity or change, and the type of each) based on assessments of availability, similarity, risk, and cost. Earlier steps entail a high likelihood of acceptance as long as an already existing precedent is followed, and a low political cost to securing that 44  acceptance, making states more likely to end their decision-making process with continuity rather than change. In addition, the greater the similarities which can be made between the newly adopting state and those whose example it is following, the greater the likelihood of “success” of that combination. As one moves down the steps of the process, however, cost and risk increase: untested combinations may not produce the desired “solution” to the problem, and it may be politically costly to secure support for a change in combinations or an innovation. Indeed, that support may never be forthcoming: while the risk of condemnation from the community comes only in relation to non-adoption in step 4, even a good-faith recombination or innovation in steps 6 or 7 may spur rejection. This makes outcomes which produce change less likely. However, the USCC model is not meant to be predictive at the level of individual ideation-behavior combinations in Steps 5 through 7.21 If a necessary action has been taken but multiple actions are possible, or if no necessary action exists, then states must determine similarity, cost and risk according to a calculus which may not match that of other states. This can lead to very different outcomes, as variables like national identity, culture, major trading influences, regional issues, interest from the hegemon, and more combine to create highly individualized outcomes. This is certainly true of Steps 6 and 7, where no precedents exist to guide states’ choices. If states are left to themselves to make determinations of similarity rather than formally assigned to categories of action based on context (such as in the NPT), this indeterminacy also extends to Step 5. Although the USCC model illuminates decision points and available options with respect to the existing contents                                                  21 This would depend highly on the cluster in question. If little variation has taken place and most states end their decisions with steps 4 and 5, it may be possible to model state choices based on the existence of universal necessary actions or strongly defined typologies of similarity. In norm clusters with more variation, however, such modelling would be of little use. 45  of the norm cluster, and proposes the general likelihood of continuity and change in individual decision outcomes, it has no predictive capability in terms of which particular combinations will endure, emerge, or remain untested. As regards overall trends in action and the composition and evolution of the norm cluster itself, on the other hand, some predictions are possible. I suggest that the USCC model predicts the relative amount of continuity, change, and forms of variation of states acting with respect to the norm cluster. It suggests that continuity, or direct emulation of a predecessor (either a necessary outcome or an acceptable precedent) should happen most often, followed by change: recombination of existing elements within the norm cluster and by innovation as a last resort. It also suggests that innovation will happen most often in the early stages of a norm cluster’s life cycle, since no gold standard may yet exist and previous examples to copy directly will be fewer.  The model is also iterative: as states innovate, new ideations and behaviors become available for new cluster adopters or for already cluster-adopting states which would like to do more than they already have. This widens the options for step 5 and increases variation within an expanded set of recognized combinations. On the other hand, if states do not innovate or if their innovations are not accepted as legitimate, then continuity becomes by far the most likely outcome. Over time, then, the cluster (and the options available to states) may change its size and density, shrinking or expanding in response to the actions of new adopters and the existing community’s reactions. Lastly, the adapted USCC model also helps to explain the varying levels of discourse seen with respect to different states’ decisions with respect to the norm cluster: if a choice is clearly understood with reference to a universal standard or a specific predecessor, no 46  discourse is needed to explain or justify one’s choice. The need for discourse grows as states search for optimal expression of values and institutional design stage by stage, and need those views and decisions to be validated by the community. The next section details the functions of this discourse: how specific diffusion mechanisms which emphasize discourse can enable certain innovations to become accepted parts of an expanded norm cluster.  2.3. Processes of cluster emergence, evolution and limitation I have argued above that a norm cluster represents an expanded space within which multiple combinations of ideations and behaviors result in an intersubjectively acceptable means of solving the problem. I contend that this expanded space is not static, and is instead capable of evolution and change, but that evolution requires both innovation and discourse. The process of adding a new ideation or behavior to a norm cluster is akin to amending the “prevailing rules” regarding what is considered appropriate behavior and what is not. As Coleman argues, “An action that breaks previously accepted understandings of international rules can still be legitimate if the international community is persuaded to accept either a new rule or a new interpretation of existing rules” (2007, 47). So while innovation is necessary to create change, or a new option within the norm cluster, discourse is also necessary for that innovation to be accepted as a legitimate option for inclusion into the norm cluster. I argue that the boundaries of the norm cluster are dependent upon the conceptual and practical compatibility of the norm components included, which are determined intersubjectively. If an ideation (an item of social value, either currently enjoyed or desired in the future) can be reasonably argued to be important and if a behavior can be theoretically or practically demonstrated to affect the attainment or continued enjoyment of 47  this ideation, and if both are accepted by relevant actors as appropriate given the problem, then the process of argumentative rationality (Risse, 2000; Deitelhoff, 2009) is successful and the ideation and/or behavior are included in the norm cluster. This is the process of creating an “explanatory understanding,” or “locating [an act] in some set of social practices recognized as such by the relevant social collectivity – or identifying…what the act ‘counts as’ within the intersubjective frameworks held by that collectivity” (Ruggie, 1998, 860). As social constructs, norms “are contested by default” (Wiener, 2008, 63; 2014).22 The social construction of norms implies an ongoing process of constitution based on activity (Wendt, 1999; Wiener, 2008), and therefore leaves open the possibility of change in that shared understanding. When researching a norm over its life cycle, then, “it is crucial to establish whether or not the condition of ‘being shared’ holds” (Wiener, 2008). I argue that contestation about normative meaning and application, and thus cluster change and evolution, occurs both at a norm cluster’s boundaries and among its components: it involves determining how far the concept of “shared meaning” stretches. It therefore creates space for change, and for discourse about normative innovation. The following sections detail the types of contestation which take place both within and at the borders of a norm cluster. By detailing specific diffusion mechanisms which utilize discourse, I locate the processes of contestation and change within the evolving structure of the cluster and illustrate the means by which intersubjective understandings are reached concerning what “counts as” (Ruggie, 1998) belonging to a norm cluster.                                                  22 Wiener (2014, 1) defines contestation as “a social practice [which] entails objection to specific issues that matter to people. In international relations contestation by and large involves the range of social practices, which discursively express disapproval of norms.” 48  2.3.1. Norm cluster emergence and evolution I argue that actors determine what “counts as” part of a norm cluster via a combination of innovation and of diffusion mechanisms which utilize both practice, to create the change, and discourse, to determine its acceptability. Three types of potential change exist: innovation and the inclusion of new ideations or behaviors (step 7), recombination of existing elements (Step 6), or justification in new circumstances (step 5, but using different arguments about similarity and best precedent). The inclusion and legitimation of a new ideation or behavior in an existing norm cluster, or the recombination of existing elements, is dependent upon the development of intersubjective understandings which redefine the boundaries and contents of that cluster. Characteristics of successful innovations and change agents I propose that the likelihood of acceptance of these innovations is, for the most part, dependent on the nature of the innovation and the innovator. I suggest that the characteristics which make a state’s choices likely to be accepted and emulated are the same as those which drive norm diffusion in the existing literature. This may be the actor’s moral stature or particularly relevant relationship to the problem: just as certain norm adopters are more “critical” in getting a single norm to reach a tipping point (Finnemore & Sikkink, 1998, 901), so may they be more critical in changing a norm cluster’s boundaries. While not every state making an argument for change or inclusion will have all of these characteristics, I suggest that those states which are successful in changing the content and structure of a norm cluster will have one or more of these traits. Which specific trait, however, may vary widely.  Some are related to voice and power: for example, states which are more prominent in the international system – be they large economies, powerful militaries or former 49  colonizers – tend to have more influence in international negotiations (Ikenberry & Kupchan, 1990; Sandholtz & Stiles, 2009, 16). States whose arguments have an additional form of domestic legitimacy behind them, such as democratic political will or legal rulings, also carry additional deliberative weight (Bower, 2014; Hurd, 1999; Waters, 2004). Norm entrepreneurs with the argumentative skill to persuade others (such as through framing or grafting) are better at gaining acceptance than those who are not (Payne, 2001; Price, 1998; Risse, 2000). Actors with particular moral or symbolic prominence due to the severity of their experiences, their position as stakeholders, or the global recognition of their struggle – such as Nobel Peace Prize winners or victims themselves – are given a normative microphone which other actors are not (Alford, 2008; Brysk, 1995; Keck & Sikkink, 1998, 22-23). Lastly, states which clearly take a “costly moral action” (Kaufmann & Pape, 1999) are more likely to be seen by others as acting in good faith, and thus their innovation will be seen as more acceptable than one proposed by a state which loses nothing. Material and practical reasons which enhance the likelihood of acceptance also exist. If the innovation itself is backed by scientific evidence or an epistemic community (Haas, 1992), it is more likely to be accepted because the practical outcomes are better understood and therefore considered less risky. For landmines, an addition to the norm cluster might be the inclusion of funding for a particular type of landmines detection and removal technology which has been proven safe and efficient; for climate change, a new form of carbon-capture or a manufacturing process which limits emissions. Additionally, states which can demonstrate “success” or “competency” in accomplishing the prescribed policies (or in creating political advantage for those who enact them) offer positive examples for politicians in other states seeking both practical and political gain (Acharya, 2004; Gilardi, 2010).  50 Diffusion and change I contend that change in a norm cluster’s contents or boundaries can take multiple forms: justification in new circumstances which expands the context for the “best” precedents (Step 5), recombination of different pre-existing components within a norm cluster (Step 6), or innovation of completely new ideations and behaviors (Step 7). Each of these change processes includes the use of discourse and argumentation as well as practice: the actor must convince the audience that what it is proposing “counts” as appropriate. This section briefly touches on the nature of these discourses and where in a norm cluster’s life cycle they may be found. Single norms are first contested, with respect to other single norms, during their emergence phase. Finnemore and Sikkink (1998) argue that new norms emerge into an already norm-rich environment and must fight for relevance and primacy with regard to norms already in existence. Therefore, norm entrepreneurs must convince others that their option is the best among many, thereby adding to the repertoire of accepted combinations and changing the internal hierarchy of a cluster such that their option becomes the “gold standard. Additionally, studies of deliberation also note internal contestation over normative meaning from the norm-setters themselves: those who invent a norm for others to follow first engage in a deliberative process to determine the shared meaning that will be diffused. This is the logic of argumentation: norms are flexible until their meaning is agreed upon (Wiener, 2008, 44). Both of these processes may create a single “gold standard” (step 4) by determining a single, most appropriate, response to a problem. Change may happen at Step 5 not by creating new combinations of ideation and behavior, but by broadening the types of similarities which determine the most appropriate combination among existing precedents. In effect, this is a change in the scope of the 51  “problem” to which similar actors are responding. In Step 5 of the USCC model, the choice of “best” precedent may hinge on the specificity of the problem in which each precedent applies. A narrowly drawn problem limits the number of times a particular option will be appropriate; a wider problem suggests more widely applicable standards, and thus more instances in which the precedent should be followed.  The development of new chemical weapons over time offers an example of a widening problem, and thus a change in the applicability of a “best” precedent, based on new technologies. Mustard gas and chlorine gas23 have been banned weapons since just after World War I, when agents like Sarin or VX gas did not yet exist. These are now banned as well; they have been added to the “problem” of chemical weapons based on their similarity to existing weapons, and thus the relevant appropriate response to their manufacture or use. Current debates about what qualifies as a chemical weapon – i.e. the problem to which our ideations about indiscriminate and overly destructive methods of warfare apply – include the use of white phosphorus as an incendiary weapon or obscurant (Human Rights Watch, 2009; U.S. Department of Defense, 1991). The problem of “chemical weapons” to which the norm cluster applies will expand, or not, based on whether these are accepted as similar enough to existing chemical weapons such that the norm cluster applies.24 As the norm cluster diffuses, the possibility of augmenting a norm cluster through recombination (step 6) arises: norm entrepreneurs may contribute to the increased                                                  23 Chlorine gas when used as a weapon falls under the category of chemical weapons, but chlorine also has many civilian uses so its manufacture is not prohibited. In other words, when chlorine does not negatively impact suffering in the conduct of war (ideation), it is not a problem. 24 To the extent that the new prohibition is accepted as part of the same interrelated set of ideations, contexts, and behaviors, the norm cluster expands. If the new prohibition is accepted as similar but not interrelated, Price’s (1998) theory of norm grafting and the more general advancement of prohibition of different forms of weapons applies. 52  flexibility of their own proposed (single) norm through their use of strategic framing to persuade new adopters. Frames are used to create or explain social meaning, or the relevance of a particular act to a broader societal feature or goal (Payne, 2001). However, societies and cultures differ, and a frame that is accepted in one society may meet with resistance in another. No single frame appears to be universally persuasive,25 and outcomes (such as norm adoption) are uncertain even when applied to seemingly analogous situations. Therefore norm entrepreneurs who wish to induce a behavioral change in many states will employ multiple frames,26 sometimes simultaneously, in their efforts to convince others to adopt a new behavior. Thus, “a single desired outcome can be explained by multiple frames and any given frame can conceivably justify more than one possible outcome” (Payne, 2001). Although adopters agree that the same behavior is an appropriate way of dealing with a general problem in a number of similar contexts, they may not agree on exactly why that behavior is appropriate. Within a wider but not infinite range of relative similarity, multiple ideations may be recognized as applicable by both the norm entrepreneur and the norm adopter.  Differing state structures and capacities, as discussed above, may also contribute to cluster evolution via innovation. In cases where it is not possible to exactly emulate a predecessor, a newly adopting state may believe in – or at least assert - the same ideation, but choose to adapt the norm to include a new behavior which it is actually able to carry out. However, inclusion of this new combination in the norm cluster remains dependent on                                                  25 Although some frames may be more persuasive than others: Keck and Sikkink suggest that issues which can be framed as causing bodily harm to vulnerable individuals or threats to equality of opportunity are particularly compelling (1998, 27). 26 Frames and ideations mean nearly the same thing, but they are not identical. Frames problematize a social fact, as ideations do, but they need not problematize it in moral terms. Ideations, on the other hand, involve the attainment or impeding of moral outcomes. In other words, non-cluster-consistent behavior may be framed, but contain no ideations. 53  public justification: the state must be able to argue that the new behavior is also capable of operationalizing the already-accepted ideations. For example, when the Rwandan government decided to use gacaca to pursue accountability for crimes related to the genocide, it had to justify to other actors that this version of a court system was capable of delivering accountability in a manner acceptable to other states. Considerations involved the ability of the gacaca to deliver fair and speedy trials as well as the relative level of harm to someone unjustly punished. Rwanda argued that, because the alternatives were norm failure (the country’s overwhelmed legal system could not deliver fair and speedy trials for so many people) or norm rejection (accountability cannot be achieved via trials so we will not attempt to pursue accountability), the relative harm of adapting its behavior was less than the harm of the other two options. While the international community was certainly split on whether Rwanda’s justification was acceptable and, therefore, whether the gacaca system was an acceptable adaptation to the “transitional justice” norm cluster, I argue that Rwanda nevertheless had to at least make the argument, open to public scrutiny, before it went ahead with its plans.27  Lastly, I suggest that norm diffusion may broaden the range of ideations included in a norm cluster by allowing cluster adoption by actors whose identity may be different from original adopters. The empirical effects of a norm-driven behavior may be desirable to actors whose values and goals are different, although generally related, to the goals of the initial adopters. For example, Finnemore and Sikkink describe the motivations of actors in Stage 1 of the norm life cycle as “altruism, empathy, ideational commitment” while they                                                  27 Although the gacaca remains controversial with respect to its ability to deliver justice without infringing on due process and other legal rights, I suggest in the conclusion that the transitional justice normative cluster has evolved to include a set of behaviors which are “traditional” or “localized” to the countries engaging in the TJ process. 54  characterize the motivations of adopters in Stage 2 as “legitimacy, reputation, [and] esteem” (Finnemore & Sikkink, 1998, 898). Thus, while a state adopting an anti-landmines norm in its emergence phase might refer to civilian non-discrimination because it genuinely believes that such a thing is important, a state adopting later during the norm’s life cycle might instead explain its decision with reference to participation in a system of international humanitarian law. In this case, while civilian non-discrimination might be important, the driving force behind cluster adoption would be inclusion in a group of states which agree to play by the same moral rules.28 In turn, previous adopters recognize that international law is an acceptable reason to adopt the anti-landmines cluster, even though they may not think it is the most important reason. These ideations, while related because the anti-landmines norm may result in the creation of international law banning certain weapons which do not discriminate against civilians in war, are not exactly the same thing. The structure of the “norm” evolves over time and according to the identity of the adopter.  The process of norm localization suggests the same effects. Acharya explains norm localization as  A complex process and outcome by which norm-takers build congruence between transnational norms … and local beliefs and practices." In this process, foreign norms, which may not initially cohere with the latter, are incorporated into local norms (Acharya, 2004).29  Actors presented with a transnational norm, including problem, ideation, and behavior, may see the problem applicable to their own state, and the behavior both                                                  28 It might even be argued that the first adopter acts via a logic of appropriateness and the second via a logic of consequences. 29 Acharya continues, “The success of norm diffusion strategies and processes depends on the extent to which they provide opportunities for localization.” He suggests that the most widely diffused norms are those which provide the opportunity for local interpretation; namely, that the existence of multiple ideations within a norm’s shared meaning is an essential facet of its ability to diffuse successfully.  55  normatively appealing and politically desirable. However, they may be unable to convince other domestic political actors to adopt the ideational component of the norm because it does not fully resonate with pre-existing local identities and goals. In other words, international and local ideations are not congruent. Acharya notes, however, that by a discursive process of reconstructing and adjusting both the local and international norms, congruence can be built and the new practice incorporated into domestic usage. Consequently, while the state will ultimately adopt the new behavior, its ideation is no longer either the state’s own original ideation or the one offered to it from abroad. The multiplication of ideations attached to a single behavior changes the contents of a norm cluster over time. 2.3.2. The limits of acceptable variation The previous examples of norm cluster change showed how actors can create an expanded space for alternative ideations and behaviors in response to a common problem. However, I contend that choice with regard to a cluster, even within an expanded range of appropriate behavior, is not infinite. Not just any potential answer is acceptable: complete relativism does not automatically follow from the relaxation of rules on uniqueness. Boundaries of a norm cluster do exist, and they are created and enforced not merely by logical argumentation or practical compatibility but also by evidence of success and the acceptance of those arguments by a community of interested parties.  I posit that determinations of inclusion, based on plausibility and reasonable argument, reside not with the actor making the argument, but with the relevant portion of the international community seized of the issue at hand. This may mean preceding adopters, actors of particular moral or political standing, practitioners, judicial bodies, or 56  even norms scholars. Because norms are intersubjective, inclusion or exclusion of an innovation in a norm cluster must by definition involve more than the innovating actor. I argue that resistance to a cluster’s evolution is also displayed through discourse and practice; in this case, however, it is the audience (rather than the innovator) which is most important. When a particular innovation which seeks to add a new component, or combination of components, is determined by the relevant parties to be an unacceptable variation, the interested community responds negatively to the proposing party. This may take legal, legislative, or other political form, but a norm’s limits are discoverable through other parties’ reactions to state actions.30 Additionally, I argue that this resistance takes place at various points within and at the boundaries of the norm cluster’s component parts: it may occur with respect to innovation at the boundaries, incompatible combinations within a cluster, gradual narrowing of the scope of the problem, or freezing of boundaries via treaty creation. Several examples are discussed below. It is important to note that activists or states will not necessarily have the same conceptual boundaries in mind at the same time, even when identifying the same cluster. Because the determination of plausibility, competency, legitimacy and/or appropriateness carries a strong element of subjectivity, it stands to reason that each actor or observer may interpret the relationships between specific problems, ideations and behaviors, as well as the actors choosing them, slightly differently. Although a cluster’s contents and borders are constructed intersubjectively, that intersubjective understanding may not be universal at a particular point in time.31 While most actors may agree on a particular interpretation, some                                                  30 Deitelhoff (2009) suggests that legal discourse may be especially powerful, because it “devalues arguments based on power and interests and requires normative justifications, based on legal precedent and analogy” (p. 35). 31 For example, opinion is divided as to whether Kosovo is an independent state, and therefore whether the set of practices and responsibilities which accompany a recognition of state sovereignty apply. If Kosovo is not 57  may not, and would protest when others choose the interpretation that they find inappropriate.32 Variation in intersubjective understandings may therefore occur both in state or activist interpretations of action, leaving norm cluster boundaries contested until (and if) such an intersubjective agreement can be reached. This is in fact an essential element for the evolution of norm clusters, as contestation “at the boundaries” is what leads to the addition or subtraction of norm components and combinations from the roster of accepted interpretations of the norm cluster. Lastly, I do not argue that such variety and breadth in normative outcomes is always the case. The community may simply reject any and all proposed additions or changes to the original norm, leaving fused norm diffusion, and continuity, as the only acceptable option. As noted above, the norm on the non-use of nuclear weapons has only one possible behavior: do not use nuclear weapons. All deviations are unacceptable. Indeed, the smallest possible configuration of a norm cluster consists of only a single ideation and behavior for a given problem.  The sections below detail some of the specific ways in which the community may reject proposed changes, or may narrow existing choices, for norm component combinations within a cluster. Resistance/backlash: Behaviors or ideations not accepted  States and other international actors may limit the contents of a norm cluster by refusing to accept certain ideations and behaviors (either newly proposed or already                                                  independent, a different set of rules of engagement is necessary. This is analogous to determining whether a given “problem” is similar enough that the norm cluster applies to it, obligating the state to engage with the meta-norm. 32 This corresponds to the “critical mass” at the tipping point, and the remaining non-adopting states, which signify that a norm (or norm cluster) has reached general acceptance by the community. In both cases, the behavior of particularly critical actors in the community serves as a signal to others. 58  accepted) as legitimate, or by limiting the scope of the problem to one which generates a smaller range of ideations and behaviors in the first place. This form of contestation is called “norm resistance” (Elgström, 2000). A commonly found version of norm resistance comes from those actually trying to diffuse the norm in the first place. Norm entrepreneurs trying to diffuse a specific, fused version of a new norm may argue not only that their norm supersedes old norms, but that it is the only acceptable interpretation (or “gold standard”). This is an attempt to limit the creation of a norm cluster: by limiting acceptable interpretations to a single option, a universal “gold standard” is created and the possibility of choosing different outcomes depending on context is eliminated. This ends the USCC process at Step 4. Another example of resistance comes from new adopters, who may also refuse to acknowledge the legitimacy of certain options within the cluster. Rather than acknowledging all possible options and choosing the “best,” this action reduces the number of acceptable precedents (Step 5) for states to choose in the future. For example, Capie writes that when norm entrepreneurs such as Canada and Japan proposed a broad range of measures related to the prevalence and use of small arms, for a number of related reasons, ASEAN countries and China strongly objected to all attempts to frame the small arms issue as anything but terrorism- and security-related. When ASEAN addressed a UN conference on the issue, it stressed that, for Southeast Asia, “the problem of the illicit trade in small arms and light weapons…relates principally to trans-national crimes” (Capie, 2008, 644). In this case, the small arms norm cluster diffused only partially because the new adopters accepted as appropriate only one of a larger number of possible ideations, and with it only a small range of behaviors. Other ideations within the cluster, and any behaviors unrelated to transnational crime, were actively contested and their status as “acceptable behavior” 59  questioned. For Asian adopters, the norm cluster has different boundaries than for Western adopters: only some pieces of the norm cluster remain universal. Any future-adopting states would have a different characterization of the norm cluster’s contents based on this change; a smaller range of actions at Step 4 and a larger range of possible, but not necessary, options at Step 5. A situation of active backlash to the attempted inclusion of new components to a cluster occurred when much of the world criticized the George W. Bush Administration of narrowing the definition of torture. In narrowing the definition of torture, I argue that the Bush Administration attempted to add new behaviors, such as waterboarding, to the set of practices falling under the category of “acceptable forms of interrogation.” Using the ideation of anti-terrorism,33 key figures within the administration argued that these practices should be allowed (McKeown, 2009). However, this argument was hotly contested both within and outside the United States, use of the argument and the practices decreased over the next several years, and after the Bush Administration left office, the new Obama Administration quickly reversed many of the criticized Bush-era policy directives (Sikkink, 2013). The behavioral change argued, namely that certain interrogation practices should be seen as appropriate because the goal fighting terrorism makes them so, was not accepted, and the norm cluster’s original boundaries remain intact. Internal contestation: Incompatibilities within the cluster As explored in section 2.1.3, it is possible that not every combination of ideation, behavior and problem within a norm cluster will be conceptually compatible or complementary. A particular behavior may just not be justifiable with reference to a                                                  33 Sikkink refers to this as a “counter-norm” (2013, p. 146) but, as she says it was used to “justify” torture, I contend that it is an ideation used to justify a behavior. 60  particular ideation: the conceptual relationship makes no sense or no plausible process exists to link the behavior to the attainment of a specific goal. Alternately, a process of learning over time may lead to the discovery of practical incompatibility even when a theoretical argument can be made. I suggest that the likelihood of this incompatibility appearing grows with the complexity of the norm cluster. As a cluster spreads and more states adopt it, the space for interpretation of similarity and appropriateness widens based on the experiences of previous adopters and the ability to discern empirical effects of the behavior. Specific combinations are adopted because particular policy makers believe that these behaviors will “solve the problem” and further the given ideational goal. If this turns out not to be the case, these combinations may be dropped, creating a “feedback loop” which alters the internal density of the cluster. In both of the above cases, the visual map of a norm cluster alters as well:   Figure 2.6: Norm cluster with some incompatibilities.  61  In the figure above, Ideation E is conceptually or practically incompatible with all behaviors except Behavior E. Attempts to justify other behaviors with Ideation E would either be viewed as bad faith, nonsensical or unrelated to the larger cluster, even though all the component parts fit comfortably within the larger cluster. I suggest that this outcome is quite common, either through resistance to certain combinations during the decision-making process itself or as the product of learning from previous experience. For example, the norm cluster on climate change includes the ideations of “reduce oil production” and “reduce excess carbon dioxide in the atmosphere.” It also includes the behaviors of plastics recycling and reforestation. While states could argue that they want to reduce oil production by recycling plastic and reduce carbon dioxide in the atmosphere by reforesting, they cannot reasonably (i.e. to other environmentally friendly states) justify a recycling program with the goal of reducing carbon dioxide, because recycling does not directly prevent carbon dioxide emissions. Intervening ideations: Reduced scope of the problem Another form of limitation is narrowing of a norm cluster’s boundaries through the intervention of ideations, either from within or outside the cluster, which are deemed more important than others. By making certain ideations more important than others, combinations which advance the secondary ideation but harm the primary one become inappropriate, thereby making certain options inappropriate in solving the general problem.  As an example, the use of the death penalty in the United States is a behavior meant to address the problem of serious crime. Relevant ideations might include punishment and deterrence. However, the United States Supreme Court holds the constitutional guarantees against cruel and unusual punishment to be normatively more important than punishment and deterrence. As the definitions of what constitutes cruel and 62  unusual punishment have changed over time, the Supreme Court has, over the last 40 years, increasingly restricted the use of the death penalty in contexts where its application is considered cruel and unusual.34 This narrows the scope of the original problem, changing the definition from “serious crime” in all cases to “serious crime AND when cruel and unusual punishment can be avoided.” In fact, the United States may be moving in the direction of most of the world’s countries in deciding that it is not possible to use the death penalty and avoid cruel and unusual punishment; this would mean eliminating the death penalty as an acceptable means of dealing with serious crime, thus removing it from the cluster completely. Freezing of boundaries: Codification/legalization  Lastly, I suggest that the process of treaty creation or other forms of legalization (Abbott, Keohane, Moravcsik, Slaughter, & Snidal, 2000) is, in many ways, an attempt to reify the limits of a norm cluster and introduce an element of stability to norm adoption and implementation. By formally specifying the scope and nature of the problem, legitimate aims of States Parties, and the behaviors they legally bind themselves to carry out, an international treaty defines what is considered “appropriate behavior” for all its signatories. Gold standards or bare minimums may be conclusively identified, categories of actors are created which identify the relevant “best” precedent combinations, and failure and/or cheating are made easier to determine through text regulating how behaviors are to be carried out and the designation of an entity empowered to make that determination. Activity                                                  34 These specific contexts include uneven application, mandatory sentencing, its use on those where minors when the crime was committed, its use on the mentally unstable or challenged, where the participant did not kill or intend to kill, and when the victim survives the crime committed against them (Woodson V. North Carolina, 428 U.S. 280, 1976; U.S. Supreme Court, 1972; U.S. Supreme Court, 1977; U.S. Supreme Court, 1982; U.S. Supreme Court, 1986; U.S. Supreme Court, 1988; U.S. Supreme Court, 2002; U.S. Supreme Court, 2005; U.S. Supreme Court, 2008) 63  outside of the treaty’s boundaries is not considered norm-compliant and may be subject to sanction.  I do not assume that legalization necessarily stops contestation over cluster evolution altogether. Insofar as the process for changing a cluster’s boundaries is made more formal and specific, it may be more difficult for actors to affect cluster evolution than it was before the treaty’s existence. In addition, the use of legal language as the dominant mode of normative communication may also hamper certain forms of argumentation or make certain ideations and behaviors less likely to be successfully added to the cluster (Bower, 2014). However, many treaties specify processes for amendment (which would change the cluster) and offer states the opportunity to “opt out” of certain provisions via formal reservations (changing “universal necessary actions”). These legal processes are the specific ways in which the norm cluster which serves as the treaty’s base content can change, and their inclusion in many treaties suggests that legalization is not an impenetrable obstacle to that change. 2.4. Conclusion I have argued in this chapter that contemporary international norms, although they have a singular conceptual structure, have a dual nature: they are both stable and flexible. While some scholars assume that norms consist of stable one-to-one-to-one linkages between norm components that diffuse as a single unit, others view them as flexible interpretations of appropriate behavior which change from actor to actor and over time. Each implies different forms of variation in outcomes, for both actor and object. In order to accommodate the possibility of both stability and change in the processes and outcomes of norm diffusion, I introduced the concept of a norm cluster as a group of interlinked but conceptually distinct ideations, behaviors and problems. The norm cluster offers actors 64  more breadth in their decision-making, the opportunity to justify or reject innovations, and a wider variety of precedent and influence from which to learn, without sacrificing continuity of the general object itself. The examples used in this chapter are ideal representations of the links between problem, ideation, and behavior which arise through a process of argumentation and contestation and define the space of “appropriate behavior” for states in the international system. It is this norm cluster, rather than the single norm, that I contend should form the base unit of analysis in studies of norm emergence, diffusion, and adoption.  In addition, I have provided a decision-making model to help explain both actors’ individual decision outcomes with respect to the norm cluster and, when applied iteratively, a means of understanding cluster evolution over time and the patterns of continuity and change created by its adopters. To further explore the dynamics of cluster evolution, the last section provides insights into the processes of innovation and discourse, and the qualities of actors and arguments, which can expand or limit a norm cluster’s contents based on both adopters’ innovations and the acceptance or rejection of those innovations by the international community.  The next chapter of this dissertation will define and justify the analysis of a specific norm cluster, that of “Transitional Justice” (TJ). By mapping and investigating the field of TJ, as a set of normative decisions made by states to address past human rights violations, I will locate the applicability of the norm cluster concept as regards principled decision-making and explain the data-gathering and coding processes which have led to my quantitative TJ dataset. Chapters 4 will use that TJ dataset for quantitative analysis, following the steps of the USCC model to search for patterns of continuity and change in the choice of ideation and behavior by TJ-adopting states. Chapters 5 and 6 will draw on specific qualitative examples from the TJ norm cluster to trace states’ decision-making 65  processes as they determine whether to adopt or adapt existing precedents within the cluster, and investigate the processes of diffusion, learning, innovation and contestation which have led to the cluster’s evolution.  66  Chapter 3. Case and methods: Building a transitional justice dataset To illustrate the utility of the norm cluster concept and examine the processes of innovation, discourse and choice which lead to normative continuity and change, this dissertation will explore the field of scholarship and practice of “Transitional Justice” (TJ). I have chosen transitional justice due to the relative likelihood of it being a norm cluster (making it an “easy case”), but more importantly because of the existence of particular evolutionary dynamics which provide evidence in favor of the cluster concept’s analytical utility.35 Given the richness of scholarly debate on the subject, and the duelling imperatives of international obligations and domestic pressures, I suggest that the TJ field is likely to contain individual cases (state TJ choices) which demonstrate most, if not all, of the possible outcomes of the USCC model. Additionally, TJ makes an excellent case study because existing quantitative data collection in the field of TJ is robust but until now has been, I argue, unable to clearly distinguish principle-based action from other types of similar behaviors. I contend that my method of data collection and delineation, in which ideations serve as an additional necessary condition for inclusion in the “norm cluster” dataset, provides a way to separate relevant normative action from other types of behavior. Lastly, I suggest that the particular construction of the field, namely its dual history, offers a clear way of quantitatively testing for the continuity and change created by Steps 5 and 6 of the USCC model, and thereby determining the norm cluster concept’s analytical utility relative to previous concepts (single norms and sets of norms). The field’s construction is a conjunction of two very different, and (it is thought by some) differently diffusing, sets of                                                  35 Other potential norm clusters considered included human trafficking, climate change, and various weapons treaties. No other case had, at the time of decision, an existing robust dataset, large scholarly literature, and contention between the cluster’s continuity in interpretation (hypothetical roots) and variation in outcomes (local applications), which is necessary to test the concept’s analytical utility. 67  pre-existing norms: international humanitarian law and democratization/domestic human rights. These roots offer a way to test some of the USCC model’s implications for norm choice and evolution. This chapter will chronicle the process of defining, constructing, and delimiting the boundaries of the TJ norm cluster for quantitative and qualitative analysis, as distinct from traditional quantitative behavioral analysis. Section 1 will define the field and explore its utility as an explanatory case for the concept of a norm cluster and the modified USCC model. Section 2 discusses the observable traits of the norm cluster and USCC model: what do norm clusters look like empirically? What signals continuity, and what change, and where is the relevant discourse found, and from whom? It therefore determines where evidence to support or oppose my theoretical contributions may be found. Section 3 details data sources and their collection and organization for quantitative analysis. Lastly, Section 4 takes the reader through a process of theoretically informed boundary-setting to delimit the contents of the TJ norm cluster, as opposed to existing datasets that measure only behaviors. The result is a new dataset, including for the first time transitional justice ideations, which provides the quantitative evidence used in this dissertation, and a point of departure for the qualitative evidence explored in later chapters. 3.1. The field of Transitional Justice  For the purposes of this dissertation, I define transitional justice (TJ) as the “field of activity and inquiry focused on how societies address legacies of past human rights abuses, mass atrocity, or other forms of severe social trauma, including genocide or civil war, in order to build a more democratic, just, or peaceful future” (Bickford, 2005). This definition includes a set of ideations (democracy, justice and peace) and a broad problem (severe social trauma comprising human rights abuses and mass atrocity). While not specifying 68  behaviors, there is no sense that a single behavior is a standard or only approach. The most commonly accepted transitional justice behaviors include trials (international and domestic), amnesties, reparations, lustrations and institutional reforms, and Commissions of Inquiry (often called truth commissions or truth and reconciliation commissions). Transitional justice, as a field of practice and academic inquiry, is relatively new. Although accounts differ as to when exactly it began, generally its roots are traced either to post-World War II justice processes or the actions of newly democratic Third Wave countries in the 1970s and 1980s (Arthur, 2009). Since then, it has grown with surprising speed and is now regarded as having entered the “cascade” phase of a norm life cycle (Finnemore & Sikkink, 1998; Sikkink, 2011). Sikkink and Walling (2007) found that transitional justice mechanisms, both independently and as a group of behaviors, grew from the 1970s to the late 1990s and leveled off in the early-mid 2000s. A number of other authors, using their own databases of transitional justice behaviors and qualitative research, have corroborated this finding (H. Kim, 2008; T. D. Olsen et al., 2010; Sikkink, 2011; Wiebelhaus-Brahm, 2010). Transitional justice appears to be firmly established in practice. Academic literature on the phenomenon began to appear in the late 1980s and early 1990s (Henkin, 1989; Kritz, 1995; Orentlicher, 1991), and by the 2000s was firmly established as a separate field of inquiry combining theory and practice from both its roots. Today there are numerous scholarly programs, non-governmental organizations, and internationally recognized principles of the field, and a host of academic publications on the subject, as well as two journals, the International Journal of Transitional Justice and the Transitional Justice Review. 69  Importantly for this study, transitional justice has never been thought of as a single norm. Most transitional justice scholars agree that the field has emerged from two distinct historical roots: the development of international criminal law and the spread of democracy, specifically the Third Wave.36 The first root of the TJ norm cluster is international humanitarian law (IHL), specifically the law of armed conflict. The development of IHL from the late 19th century is an example of “vertical diffusion” (Sikkink, 2011, 249): it has been an international, rather than transnational, process, and has generally followed the norm life cycle identified by Finnemore and Sikkink which stresses international actors and processes. This root most often concerns trials for grave breaches of international laws of war, but may also cover amnesties and reparations, particularly state-to-state reparations. It has resulted in formalized international agreements from which the norms’ impact on domestic discourse and action should be able to be traced. Specific international legal documents include the Charters of the Nuremberg and Tokyo Tribunals (1945), the Convention for the Prevention and Punishment of the Crime of Genocide (1948), the Geneva Conventions (1949) and Additional Protocols (1977), the 1993 and 1994 UN Security Council resolutions creating the International Criminal Tribunals for the Former Yugoslavia and Rwanda, and the 1998 creation of the International Criminal Court.  The second set of roots of the TJ norm cluster comes from the Third Wave of democratization (Huntington, 1991) in which states emerging from authoritarian rule were confronted with the dual political challenges of dealing with remnants of old regimes in order to stabilize the new political systems, and of responding to newly empowered                                                  36 This is similar, but not identical, to the pathways which Kathryn Sikkink argues have created the diffusion and legalization of the “Justice Cascade” (Sikkink, 2011, 96-125). Like Sikkink, I believe that if the norm of trials for accountability for human rights violations is a single norm, it had diffused and cascaded by the time the Statute of the International Criminal Court came into force in 2002. 70  domestic constituencies, including those pushing for human rights and restitution for atrocities carried out by previous governments.37 These would include many types of serious or widespread human rights violations, such as torture, extrajudicial executions and disappearances, arbitrary arrest, unlawful expropriation of property, restriction of civil rights, and more. They may be related to the conduct of armed conflict within the country (in which case domestic armed opposition groups could also be culpable), but they need not be.38 Potential domestically-led behaviors to address these atrocities have included domestic trials, de jure and de facto amnesties, commissions of inquiry, reparations, and lustrations and institutional reform, focused on specific government actors, opposition and/or paramilitary actors, as well as on wider systemic structures which may enable injustices.  The Third Wave is an example of “horizontal diffusion” (Sikkink, 2011): there is no international treaty on democracy or democratization, and the process has been theorized to be mostly one of domestic politics, with a strong geographic effect on diffusion based on common language and identity, spillover effects, information and learning (Gleditsch & Ward, 2006; Huntington, 1991). Although much of the world is nominally democratic now and there has been extensive discussion on democracy as a norm or set of norms, there is no specific document or agreement by which to identify a tipping point or to regulate the use of any of the above practices.39 In addition, although human rights is the subject of the                                                  37 Many of these rights are the subject of two previously existing international legal documents, namely the International Covenants for Civil and Political Rights (ICCPR, 1966) and Economic, Social and Cultural Rights (ICESCR, 1976), but had not been respected by previous regimes. While the two Covenants have the force of law, they cannot be used in domestic courts to force compliance or redress, and the Committees to which States submit reports have no sanctioning power associated with a state’s failure to uphold human rights. There are a small number of cases, principally in Latin America, in which individuals have taken governments to court for non-compliance with regional human rights treaties, but they are not included in my case set unless a state government fulfills its international obligations by passing a domestic law granting reparations to the plaintiffs. 38 As long as those violations in armed conflict were not severe enough to rise to the level of “grave violations of international humanitarian law,” in which case they would be related to the IHL root. 39 Some authors would also add the development of international human rights norms as an additional root. While it would be correct to suggest that the harms for which TJ is considered an appropriate approach in the context of democratization are those from the two International Covenants, these documents and norms apply 71  Universal Declaration (1948) and at least two major international Covenants, both are non-self-executing (meaning that governments must pass domestic legislation to implement the laws contained in them), and contain an incredible breadth of defined rights with no sense of hierarchy or relative “seriousness” of violations. This means that, on paper at least, there was (at the time the field emerged) no international consensus on the “severity” of violations which might fit under the requirements of the definition of transitional justice. The one major human rights convention with that specificity, namely the Convention Against Torture (1984), was created well after the Third Wave began. There is, however, much disagreement within the field on the extent to which these two roots are exclusive and therefore lead to different practices based on different contexts (the presence of IHL violations and democratization). Given the high degree of overlap, or cases in which states democratize after a major conflict or the form of government oppression includes selective crimes against humanity, it is unclear whether states are, now that the field is well established, truly choosing behaviors based on these roots, or whether the field today displays much greater integration and variation in practice. This is one of the principal reasons why I have chosen TJ as my case study, for the roots provide an opportunity to search for continuity in state choice (if the roots are determinants of the typologies of similarity which influence the choice of precedents at Step 5 of the USCC model) or variation (if they are not). It will be a goal of this dissertation, particularly in Chapter 4, to try to tease out whether transitional justice as a field of practice is, or indeed ever was, two separate groups or one large norm cluster. I will suggest that it displays                                                  to all states regardless of political system and the ICCPR in particular has almost universal membership, meaning that there is little variation in the potential source of these norms.  72  characteristics of both: certain patterns suggest two roots, but in general, the field appears to be a well-integrated single unit (cluster). 3.2. Observable implications Because I argue that norm clusters include outcomes where norms proved both stable and flexible, the examination of any particular norm cluster should be able to reveal evidence of both continuity and change, in different states and corresponding to different outcomes along the USCC decision tree. I will use the development of the field of transitional justice to search for this evidence, both quantitatively and qualitatively. In particular, I suggest that TJ, as an inherently normative field of practice with a high degree of localization of decision-making and a low degree of international coordination in the early stages of its development, is likely to exhibit a much higher incidence of change (Steps 6 and 7) than of continuity (Steps 4 and 5). This makes TJ an excellent choice to demonstrate the relative utility of the norm cluster concept, as opposed to the single norm or bundle concepts, as the similarities and differences between the concepts will be clearly identifiable. One observable set of outcomes of the USCC choice process, which does not differ from decision processes applied to single norms or bundles of norms, is a set of similarities leading to continuity: because the general problem which motivates cluster adopters, and the conditions which enable them, are generally similar, the related characteristics of TJ cluster adopters in my dataset (such as past conflict intensity or recent democratization) should mirror the characteristics of TJ adopters in existing qualitative and quantitative accounts of the field of transitional justice. My dataset should therefore mirror the findings of previous TJ scholars with respect to the characteristics of TJ-adopting states, providing a test of my reliability in data collection and coding. 73  I will also examine if the TJ norm cluster has developed a necessary action, either in the form of a gold standard or a bare minimum, in which all actors make the same choice regardless of circumstances. This corresponds to Step 4 of the USCC, and should either of these exist it would show evidence of the continuity of single norms, rather than the malleability and choice which accompanies sets of norms or norm clusters. However, as I have theorized that additional cluster-based activity is possible beyond the bare minimum, finding one would not necessarily detract from the relative utility of a norm cluster as regards possible choices made beyond Step 4.  For transitional justice as a field, the two theorized “roots” of IHL and democratization offers a clear test for Step 5 of the USCC (choice among “best” precedents): an analysis of the field which examines the constituent parts of the TJ norm cluster should be able to identify these roots and the continuous choices that states make with respect to those roots, if indeed they began or have remained as separate sets of traditional norms. In this case states would be selecting similar ideation-behavior combinations depending on their root context, as would be expected if TJ is two separate sets of fused norms which rarely overlap. However, if a substantial portion of the decisions made at Step 5 are not direct emulations of other states based on root context, then the norm cluster (and the dynamics which have evolved the TJ cluster specifically) becomes analytically useful for understanding those choices. I argue that TJ therefore provides a clear test of the relative utility of the norm cluster concept as opposed to sets of traditional norms. If my mapping produces clear patterns of choice outcomes based on democratization and IHL violations, then the norm cluster does not add much analytical value. If there are few or no discernable patterns related to these roots, then the 74  malleability of choice implied by the norm cluster concept (and the latter steps of the USCC model) becomes relevant for explaining decision outcomes. The second set of observable implications regards the sources of difference and change, and lies at the heart of the utility of the norm cluster concept: An analysis of the specific combinations of ideation and behavior of TJ adopters, and the state characteristics which might have motivated or allowed them to do so, may point to either diversity in adopters (if unfused diffusion occurs) or similarity (if it does not). If the majority of decisions are due to unfused diffusion, the two root contexts will not always result in different strategies of transitional justice. In essence, there will be very few distinct patterns of adoption based on the characteristics of adopters: what makes a “best” precedent will vary, and even states with the same “root” may make very different choices.40  I will also examine the motivations for choice of ideation and behavior separately; unfused diffusion also suggests that different influences may be at work on different component parts. In other words, the variables which influence behavioral adoption may be different than those which lead to ideational choice. This would be evidence of recombination (Step 6), as actors with multiple varied influences could make choices for one component based on one influence, and the other component based on something else. Such evidence would favor the relative utility of the norm cluster, and not fused norms, as                                                  40 This may significantly complicate further quantitative analysis, especially in terms of modelling outcomes. If state motivations for choice do not follow similar patterns, then any attempts at modelling specific combinations (or even single component choice) will be relatively insignificant or will not explain much of the variation. In addition, depending on the size of the dataset, analysis of specific combinations or components may be hampered by their status as “rare events.” The utility of more sophisticated quantitative techniques may therefore, I suggest, be at least partially determined by much simpler tests such as counts, correlations, and bivariate analyses. As will be shown in Chapter 4, TJ is this type of norm cluster, and I heavily qualify my conclusions resulting from logistic regression. On the other hand, wide variation in outcomes and influences offers tantalizing clues for process tracing, and a place for network analysis in identifying specific sources of influence. 75  states making similar choices based on fused norms would presumably demonstrate similar characteristics if those components are analyzed separately.  As regards Step 7, or innovation, the nature of data collection for my quantitative dataset does not allow investigation of this possibility because many of these choices are likely to be undertaken only once or a limited number of times, meaning they cannot be evaluated quantitatively due to a small numbers problem. Therefore, evidence in favor of innovation, as well as examples of the decision-making processes which lead to continuity and change and the community’s acceptance or rejection of that change, will be presented in qualitative form in Chapters 5 and 6. I will look specifically at the impact of certain actors and arguments on component and combination choice, and whether those choices are accepted by the community based on argumentation and execution. I will also explore features of institutional design and their role in convincing the community of the appropriateness of certain behaviors or ideation-behavior combinations. 3.3. Data collection Because I have argued that norms and norm clusters are constructed of three types of components – ideations, behaviors and problems – the collection of evidence in support of my theory must span all three of these elements. To date, no quantitative TJ study has engaged with the question of whether individual ideations, or specific meanings, are attached to specific behaviors.41 In addition, the datasets which do exist all treat problem and context slightly differently. I have therefore both sifted through existing datasets compiled by other scholars, and built on them to compile a new dataset which includes transitional justice ideations. I argue, and will demonstrate through the process of                                                  41 “Political realists generally conflate the question of why a given state action is taken with that of what response is possible.” (Teitel 2000, 3) 76  boundary-setting later in this chapter, that the inclusion of ideational data allows the researcher to separate out intrinsically normatively defined behaviors, of the type being studied, from behaviors which are purely political or which belong to normative processes which are not part of the cluster being studied. I do not purport to capture underlying motivations, which may be different from stated intentions, and would influence when and how states adopt the norm cluster. While underlying motivations may explain individual choice to some degree for those states which are not “true believers,” the choice of any individual combination within a norm cluster includes the choice of ideation, by either true principled intention or rational calculation. If states choose not to engage all parts of the norm cluster, they do not engage the norm cluster as a whole. In other words, actors choose to normatively define their action or not, and to engage a particular norm cluster rather than a different one. Regardless of whether the underlying motivation is normative or not, I argue that those choices of stated ideation then influence the content and evolution of the norm cluster and the potential choices of successive actors. In particular, they show the boundaries of socially acceptable justification for action, create potential typologies of similarity based on shared values, and provide clues to the sources of influence (and thus potential leverage) on decision-making. In addition, although the concept of a norm cluster does not necessitate a state-centred focus, my particular study attempts to document state-to-state norm diffusion. Data availability for non-state and sub-national actors is both scarcer and less comparable across multiple possible independent variables than state-based data, making the inclusion of non-state practices in a comprehensive quantitative analysis nearly impossible. Further qualitative analysis, on the other hand, might benefit from the inclusion of non-state actors 77  by more accurately tracing diffusion of specific norm components within the cluster, and such information is included in Chapters 5 and 6 where it is relevant. 3.3.1. Data availability and collection In the last several years a number of datasets and collections have appeared for TJ-related behaviors. I follow in the tradition, and draw on the work of, a number of excellent transitional justice scholars. The datasets by these scholars provided the starting point of my investigations by supplying behavioral data with similar definitions of the “problem” of transitional justice. They include the Transitional Justice Database (TJDB) (T. D. Olsen, Payne, & Reiter, 2014), the Amnesty Law Database (Mallinder, 2010), the Truth Commissions Digital Collection (United States Institute of Peace, 2011), the Handbook of Reparations (De Greiff, 2006), and the Justice Cascade trial database (Sikkink, 2011). I search for potential “transitional justice events”: behaviors previously identified by scholars, addressing past human rights violations, which may be publicly justified in a way which conforms to the types of “acceptable” justifications of TJ identified by the scholarly community. I begin my dataset in 1970 for two reasons: first, the largest previously existing dataset, namely the TJBD, also begins in 1970, making straight comparisons easier to interpret.42 Second, two events related to the conceptual roots of the TJ field occur in the mid-1970s: the beginning of the Third Wave (Greece, 1974) and the creation of the Additional Protocols to the Geneva Conventions relating to non-international armed conflict (1976). If each of these events (and the processes which led both to and from them) has contributed a “root” which has changed states’ definitions of problems, relationship with                                                  42 The appendix includes a comparison of my dataset to the results compiled by the Transitional Justice Database Project as a reliability check. 78  ideations, and possible behavioral choices, then those changes would begin to be reflected in a dataset within a few years before and after those events. In addition, those scholars adopting either a democratization or conflict lens (i.e. adding an additional contextual restriction to adoption of the norm cluster) would also begin their datasets in the 1970s based on these events, and although I do not adopt either of these specific contextual restrictions, I do adopt the same time frame for comparability reasons.  I end my data collection in 1999 (except for commissions of inquiry, used only in Chapter 7) for two reasons as well: one theoretical, and one practical. Most important is the theoretical reason: I have stopped at 1999 because I am certain that this date captures the entirety of the norm emergence phase and the tipping point into the cascade phase of the norm cluster’s life cycle. I argued in the previous chapter that innovation (and thus the latter stages of the USCC model) are most likely to occur in the early stages of a norm cluster’s development, so the emergence phase will be of the greatest utility for discovering examples of change, as opposed to continuity, because precedents will be scarce and no gold standard may yet have emerged. In addition, studying the emergence phase offers potential explanations for why some norm clusters are large and diffuse while others remain tightly controlled; it is in the emergence stage of a norm cluster’s development at which we might expect to find the kind of debates about the borders of the cluster which determine its overall specificity. Those debates might be largely resolved before a tipping point or formalized treaty, so studying the emergence phase offers the greatest chance to explore 79  the implications of the norm cluster concept on variation between norm clusters as well as variation within norm clusters.43 Second are the time and resource constraints inherent in any major research project. Finding official documentation and matching it to existing datasets for these 30 years of practice took nearly a year of research and two three-week trips to the Library of Congress; to add years beyond 1999 would be resource-draining for relatively little additional analytical value. In addition, going backwards any further in time lessens the likelihood of obtaining good primary source documents, while going forward in time lessens the availability of secondary research to provide additional case details even as it may dramatically increase the number of cases (if the cluster grows as the norm life cycle theory would expect).   I corroborate information from these existing datasets, and add ideational data to the behavioral and contextual data identified by the previous datasets, to determine “actual TJ events,” or those which contain all three norm components (problem, ideation and behavior) and thus definitely engage the TJ norm cluster (versus other norm clusters or non-normatively identified behaviors). I collect ideational data, or publicly stated justifications for the behaviors, by searching for primary sources themselves: founding legislative documents and presidential decrees. Since transitional justice is relatively new, and developed relatively quickly, all primary sources with the exception of the Nuremberg Tribunal Charter and the Geneva and Genocide Conventions have appeared within the last 45 years. This means that many primary sources and high-quality secondary sources, such                                                  43 Future work, adding to the dataset from 2000 onwards, may illuminate more evolution in the norm cluster or different dynamics with respect to post-ICC choices, but I am confident that most of the potential choice outcomes of the USCC have been captured within the 1970-1999 time frame. 80  as news reports, can be found online and in library-accessible databases such as LexisNexis and ProQuest. Documentation was sometimes available through previously existing published material, including reports of several Commissions of Inquiry themselves as well as the 1995 edited volume Transitional Justice: How Emerging Democracies Reckon with Former Regimes (Kritz, 1995), which along with the Handbook of Reparations included reprints and translations of some primary source documents.44 If primary sources were not available online or through previously published materials, I turned to the U.S. Library of Congress as a repository of official government gazettes.  When no primary source documentation was available, information was corroborated through secondary sources, especially news outlets. The LexisNexis Academic news database and the New York Times Historical Archive (accessed through ProQuest Historical Newspapers) were the sources of this secondary information. Very rarely, reputable third-party websites offered reprints of news or government documents when no other source could be found. I was unable to corroborate the existence of approximately 20 amnesties, so they have been dropped entirely from my dataset. Where two databases disagreed on the year of an institution’s creation I used the year of the founding document rather than on the date activities began or verdicts were rendered. From each primary or secondary source, I collected information on the law’s personal and temporal jurisdiction (who it applied to and over what time period), the specific behavior it mandated, the justifications given for this behavior, and any accompanying notes on limitations and relationships to other mechanisms. Where more than one ideation was given, the first stated ideation was recorded as the “primary” ideation                                                  44 For a discussion of coding issues related to language and translation, please see Appendix A. 81  unless another ideation occurred more than once in the same preamble or text. In these cases, secondary ideations were recorded, and are included in some parts of this dissertation as additional evidence in favor of continuity, change, similarity or innovation. However, the primary ideation is the one used for most quantitative analysis.45  Despite these efforts, I was unable to find ideational data for 149 cases out of 418 (36% of all corroborated cases). I view this as partially a data collection issue, but also, and more importantly, as a confirmation of the theoretical importance of studying ideations in norms research. The hypothesized reasons for this “missing” data, the relationship of those reasons to my overall concept, and their effect upon the resulting norm cluster dataset, which includes only those cases with ideations, will be explored thoroughly in the next section.  This collection of ideations represents the first effort of its kind with regard to the quantitative study of transitional justice. While such statements are commonly collected in qualitative work on a small number of cases, or in large datasets on unrelated topics such as political party platforms (Benoit, Laver, & Mikhaylov, 2009), the link between specific behavioral and ideational data has not yet been made on this scale in the field of international relations. The resulting dataset is offered for the further study of transitional justice specifically, with the hope that similar ideational data may be collected for other norm clusters in the future. It may be found in Appendix D.                                                  45 This could potentially have the effect of excluding some theoretical ideations from the empirical dataset, or of biasing the results in favor of more variation: for example, if peace and reconciliation are consistently paired, but which comes first varies randomly, amnesties look divided between two groups when they are really part of one large group. However, I show in the next chapter that this is relatively rare, and robustness checks with secondary ideations included can be found in Appendix B. 82  In addition, datasets from unrelated studies of conflict, peace, governance, rights, violence, and structural and historical conditions have been used to analyze the relationship of my collected dataset (transitional justice choices) to potential independent and intervening variables which might impact those choices. A codebook containing detailed information on all external datasets and variables is included in Appendix C. Where specific cases are discussed in greater detail in Chapters 7 and 8, I draw upon additional primary and secondary sources related to the country or practice in question. As discussed in section 3.2, quantitative data cannot illuminate the decision-making processes which lead to specific outcomes, particularly with respect to innovations or to combinations which appear only rarely; in these cases, and to illustrate even more common outcomes, qualitative analysis becomes a valuable research tool. 3.3.2. Data architecture for quantitative analysis The collected data are generally presented in an event-count dataset in which potential TJ events (behaviors with or without ideations) comprise the rows of data. The event-count dataset allows the separation of TJ mechanisms when more than one occurs in a given year, especially if not all behaviors pursued in a given year have an ideation. Separating the data in this way means that the stated motivations of individual mechanisms may be matched together, which is necessary for the boundary definition process in this chapter as well as the analysis carried out in Chapter 4. This also means that those states which employ multiple mechanisms, especially in the same year, have more analytical weight in the resulting dataset; for example, Argentina appears 13 times in the event-count dataset while Canada appears only once. If the conditions which make states more likely to employ TJ also affect their decisions on how many mechanisms to pursue, this data architecture can help to illuminate that finding.  83  Additionally, I perform some of my quantitative tests using only data from the emergence phase, which I will identify in the last section of this chapter, or by splitting my analysis into separate tests for the emergence and cascade phases.46  As I discuss in the next section, states may, over time, internalize the reasons for adopting a normative behavior. Therefore, the ideation associated with it may disappear from legislation or cease to be publicly justified in the same way as it was in earlier decisions. This may happen for early adopters during later stages of the life cycle, if these states are still engaged in normative activities, but I suggest that internalization is unlikely to happen during the emergence stage. However, as the TJ norm cluster’s life cycle advances, I am increasingly likely to mis-categorize these behaviors as “non-TJ” when they are in fact fully “internalized” TJ. This means that while information for new adopters would be correctly counted, including any new variations in the combination of problem, ideation, and behavior, information on additional (non-innovative) events for pre-existing TJ adopters would not.47 Additionally, the countries included as “adopting TJ mechanisms” would change, meaning that analyses of their characteristics would also be different. Because it could skew my data in these cases, I restrict or separate certain analyses by their norm cluster life cycle phases. In one table only, a comparison of the basic characteristics of non-adopters to mechanism-only and True-TJ adopters which appears in Appendix B, I use country-year data architecture. This is because non-adopters, with no “events” to record, cannot be captured in an event-count dataset. In addition, I convert the data to two-mode matrix                                                  46 I use the terms “emergence” and “cascade” to refer to the field as a whole, rather than to individual events or states; states themselves are not “emergent” or “cascading.” States which adopt TJ norms in the emergence phase of the life cycle are referred to as “early adopters.” In addition, I show in Appendix A that a purely mathematical understanding of the tipping point (as an inflection point at which the rate of adoption relative to the number of previous adopters becomes positive) does not necessarily exist in this case. 47 If an existing TJ adopter tried a combination that was new to that country, on the other hand, it would have to be publicly justified and would therefore show up in my dataset. 84  format for two network analyses in Chapter 4; nodes are countries and ideation-behavior combinations. Descriptive statistics and logistic regression (both country-year and event-count) are carried out with Stata version 12 (StataCorp, 2011). The network analysis uses Ucinet/Netdraw (Borgatti, Everett, & Freeman, 2002/2011). 3.4. Determining the boundaries of an empirical norm cluster This section details the process of defining and coding the contents of the empirical expression of the transitional justice norm cluster, as collected via the methods discussed above. In defining the transitional justice norm cluster empirically I ask the following question: what is the problem, and what behaviors and ideations which respond to that problem are both acceptable in the theoretical literature on transitional justice and found in the real world? I have argued that the structure of a norm cluster is based on the association and interaction of these three norm components; therefore, the empirical study of any particular norm cluster, such as transitional justice, must collect observed instances of these combinations to create a dataset. In states which appear to have experienced the relevant problem, what specific form has it taken, what are the behavior or behaviors undertaken to address it, and how are they publicly justified? In sections 3.4.1 through 3.4.3 I detail how the theoretical content and empirical cases of the TJ norm cluster are collected, coded, and organized. In collecting and delineating this dataset, I will also explore the implications of the tripartite structure of norms themselves, and of the norm life cycle, on data retention and case selection. I also use this process to distinguish truly TJ norm-consistent action, i.e. events which engage the contents of the norm cluster to be studied, from similar non-normatively and other-normatively-identified behavior which appears in behavior-only datasets. Specifically, I suggest that a dataset comprised of only those cases in which the problem, ideation and 85  behavior all correspond to the theoretical version of a transitional justice norm cluster reduces the possibility that other norm clusters (emerging, cascading and internalized) and non-normative decision-making interfere with the empirical determination of what “counts” as the transitional justice field of practice.  This section engages in a process of theory-based delimitation of the information gathered, in order to create and organize the data collected into two subsets: transitional justice behaviors accompanied by a TJ-related ideation, and those without. By necessitating the inclusion of all three components, I offer a method for constructing a quantitative dataset which measures both individual normative TJ actions and, when combined, the TJ norm cluster as a whole.  3.4.1. Problem: Past or current human rights violations For the purposes of this dissertation, I define the “problem” addressed by transitional justice to be severe or widespread human rights violations in a country’s past. This is quite broad; I take a maximalist approach to defining the field.48 I assume that at some point during the last 100 years most states have experienced such human rights violations49; this means that any state could be a candidate for transitional justice                                                  48 As compared to a more limited approach which adds the additional scope condition of democratization (T. Olsen, Payne, & Reiter, 2010), which I do not define as a problem but rather as a context which creates problem salience, or of one which adds the necessary condition that such violations have occurred during a civil war or authoritarian regime. 49 The 100-year time limit is a somewhat arbitrary, but also conceptually linked, scope condition. Although I do not argue that abuses must be recent or immediate, I do acknowledge that the likelihood of pursuing transitional justice measures may be related to the passage of time based on the availability of individuals and the relative stability of state practice.  Although state decision-making is the subject of investigation in this dissertation, I acknowledge that state decision-making is driven by individuals agitating either for or against change. As such, within a 100-year time limit, both victims and perpetrators are potentially still alive as to advocate for or against, and to be affected by, transitional justice measures which apply to individuals, making such activities more likely. For those types of transitional justice which affect state structures and practices, the 100-year limit offers an upper bound within which the state structures in question remain continuous or identifiable. Pursuing TJ measures in order to alter state practice does not make sense if a state practice has already altered so far that the original oppressive structures and practices no longer exist. Lastly, within the data collected there has been only one empirical case where TJ measures other than apologies have been undertaken 86  measures. The set of states to which this problem applies is therefore “all states,” meaning that the transitional justice “problem” is potentially universal, and that any state could conceivably have reason to address it by adopting the TJ norm cluster. 3.4.2. Adding behavior to problem: Transitional justice mechanisms I use the academic TJ literature as a guide to begin my behavioral data collection, as a general scholarly consensus has emerged that certain activities definitely “count” as transitional justice behavior. Scholars have pointed to international and domestic trials, Commissions of Inquiry (commonly called Truth Commissions), blanket and selective amnesties and pardons, official apologies, reparations (both collective and individual), memorials, revised history texts, lustrations/purges and vetting laws, interpersonal encounters, traditional ceremonies, reintegration processes, and more, as potential behaviors which fall within the “transitional justice” range of practice. As a means of conceptually bounding the transitional justice norm cluster, this study will focus on the five most commonly acknowledged behaviors, all of which have been used by states and captured in the Transitional Justice Database Project (TJBD): amnesties, trials, lustrations, reparations, and commissions.50  Trials (or more accurately here, domestic trials) are the pursuit of retributive justice for human rights violations through the national judicial system. Such processes may be started by citizens themselves or on the directive of the executive branch of government. There are two different definitions of “human rights trials” in the transitional justice literature: one includes all trials whose charges involve human rights violations, such as the                                                  for abuses more than approximately 100 years old: New Zealand’s 1995 reparations for unlawful seizure of Maori territory in the 1860s (New Zealand, 3 November 1995). 50 Many of the other methods mentioned above have been used only sparingly, and are not well-documented in a cross-national dataset. The effort required to comprehensively add these to the study would far exceed their analytical value, which would be to further show the field’s variation in outcomes. 87  Greek “torturers’ trials” of the mid-1970s (Amnesty International, 1977), and the other includes trials whose procedures are carefully tailored to uphold the human rights of those on trial as well as those bringing suit, regardless of the charges. Although the latter has been used in democratization and human rights research as evidence of commitment to human rights norms in contrast to the excesses of a prior regime (Sikkink, 2011), I use the former definition. Because capacity and other problems may hinder the execution of a fair trial, intention and outcome may not be the same. Additionally, if the most general intent of transitional justice mechanisms is to deal with past abuses (rather than specific ideations such as avoiding future abuses or engendering the rule of law), then addressing those past violations should be the critical benchmark by which trials are judged to be part of transitional justice. Amnesties are a controversial phenomenon for transitional justice studies. They are the explicit legal commitment not to pursue trials or other forms of personal accountability for illegal activities. Additionally, pardons remove criminal sanctions already imposed. Amnesties have become both more accepted and more conditional in contemporary TJ practice (Mallinder, 2008; McEvoy & Mallinder, 2011), making their evolution within the theoretical TJ norm cluster quite interesting. Many early transitional justice activists and democratization scholars, mostly with legal backgrounds, saw amnesties as a political tool to ensure stability or democratic transition rather than to pursue real justice (Henkin, 1989; Huntington, 1991). However, later scholars have shown amnesties’ applicability to justice particularly in civil conflict situations and in the restoration of civil interpersonal relations where societal violence has been severe (Mallinder, 2009b; Reiter, 2014).51                                                   51 Their application to human rights in all three situations was detailed in a 1985 report by the United Nations Special Rapporteur on the Administration of Justice and the Human Rights of Detainees (Joinet, 1985). 88  In addition, the use of amnesties for transitional justice has become more refined. Whereas before the 1990s blanket amnesties (i.e. amnesties for a large variety of crimes applicable to all persons) were common, they have increasingly been limited to crimes not amounting to grave breaches of international law. Statements from the United Nations Secretary-General and the creation of the International Criminal Court, among other things, are evidence that amnesties are no longer considered appropriate for those accused of the gravest of violations such as war crimes, crimes against humanity, or genocide (Office of the United Nations High Commissioner for Human Rights, 2009).52 While all of these amnesties count as transitional justice efforts in my dataset, some of this evolution will be explored in Chapter 6. Lustration is the practice of removing from public office, both elected and appointed, those persons either involved in or complicit in human rights violations. Lustrations are closely linked to vetting policies, which apply the same criteria (past involvement in human rights violations) to exclude potential new entrants to public service. This may mean the forced retirement or removal of senior military officials, the banning of particular persons and/or political parties from competing in elections, or the removal from the government bureaucracy of persons in particularly sensitive human rights-related positions. Lustrations and vetting policies were quite common in Eastern Europe after the fall of Communist regimes, as various newly democratic governments attempted to remove Communist actors, sympathizers and collaborators from positions of power, but have been used less frequently in other regions.                                                  52 If an amnesty is revoked or overturned in order to make way for trials, those trials are recorded in the dataset, but the amnesty remains in the dataset as well. 89  Reparations are a government’s attempt to “repair” or “restore” things lost in the course of a conflict or oppressive regime (De Greiff, 2006). State-based reparations can take many forms: they can be symbolic, monetary, or in-kind, and they can be transferred to individuals or to groups. They can also include a combination of these elements. Specific types of reparations include monetary payments for unlawful internment, repayment for lost property and income, access to healthcare and social services when the need for them was created by the violations themselves, the creation of public memorials and museums, the issuance of official apologies, the rehabilitation of lost civil and political rights, and the dedication of additional state funds and government efforts towards communities disproportionately impacted by the human rights violations. Lastly, Commissions of Inquiry, more commonly called truth commissions or truth and reconciliation commissions, are “official bodies set up to investigate a past period of human rights abuses or violations of international humanitarian law” (Hayner, 1994). They may be created by the legislative or executive branches of government. They typically collect testimony and uncover official documents and other material sources of evidence, and their findings are collected in a report. That report is presented to the government, but it may also be released to the public. I refer to these bodies as commissions of inquiry for two reasons: first, the enabling legislation for many of these commissions comes out of a pre-existing and more general law on commissions of inquiry which is not always used to investigate human rights violations. Second, adding “truth” or “truth and reconciliation” to a commission’s title automatically assumes that the commission’s goals are truth and/or reconciliation, when the legislation itself may not reflect those ideations. If enabling legislation for a particular commission specifically refers to the institution as a truth commission, it is reflected in my database. However, as a category, I use the phrase 90  “commissions of inquiry” to guard against this assumption. Finally, standing commissions which are created to deal with future violations only, or which are not specifically mandated to address human rights violations but may acknowledge them in their report, are not included.53 The commission’s mandate must include addressing past human rights violations. Based on the restrictions and counting methods discussed in the data collection section above, and using the data sources described in the methodology section, I recorded 418 instances of potential transitional justice events, or TJ mechanisms, from 1970 to 1999: amnesties, lustrations, reparations, trials, and commissions.54  Mechanism Freq. Percent Amnesty 251 60.05 Lustration 37 8.85 Reparation 36 8.61 Trial 60 14.35 Commission 34 8.13 Total 418 100 Table 3.1: All transitional justice mechanisms, 1970-1999. I argue in the next section that further adjustments are necessary to obtain a more precise picture of the transitional justice norm cluster, which is comprised of problem, behaviors, and ideations. In particular, I suggest that a problem-behavior dataset without further adjustments based on the presence and nature of the stated ideation accompanying                                                  53 For example, the 1997 Liberian Commission on Human Rights (Liberia, 27 October 1997) has no mandate to investigate human rights violations which occurred before it was founded. This restriction also excludes the 1972 Pakistani Commission of Inquiry, commonly known as the Hamoodur Rahman Commission. Convened to determine the reasons behind Pakistan’s loss of its most recent war with India and the resulting independence of East Pakistan (Bangladesh), the report specifically discussed “moral crimes” in the Pakistani military and government’s activities with regard to East Pakistan. However, the Commission’s mandate did not expressly dictate that human rights violations should be investigated (Commission of Inquiry (Pakistan), 1974). 54 As a reliability check, a comparison of this distribution of behaviors to that of the TJDB may be found in Appendix A. I also remind the reader that each data-point in the event-count version of this dataset is a potential “transitional justice event”: a behavior which addresses widespread or severe human rights violations. States may be represented multiple times if they used multiple mechanisms. In a country-year dataset, states are represented even if no TJ behavior takes place.  91  these behaviors suffers from two potential overcounting problems: normative overlap and non-normatively identified behavior. Insisting on the presence of a relevant ideation, therefore, serves to more accurately identify the principle-based action which is the subject of both the norm cluster concept and the USCC decision process. 3.4.3. Adding ideation to behavior and problem Transitional justice theorists and activists have identified a large range of possible goals for TJ processes. While some scholars assume that these goals are implicit, I argue that the presence of a commonly accepted, TJ-related ideation cannot be assumed for any particular behavior included in the potential TJ cluster.  This section will detail the range of ideations, or normatively driven goals, which are most commonly invoked and recognised as appropriate rationales for transitional justice processes. Once added to behavioral data, I will demonstrate how having all three components of a norm, including an openly stated ideation, enables the researcher to weed out cases which do not correspond to the norm cluster under review. TJ Ideations collected and categorized The laws, edicts, pronouncements, speeches and other discursive material which accompany the behaviors in my dataset include a number of ethical justifications for action. I have collected the claims made publicly about why behaviors are undertaken (given of course the limitations of my data collection efforts), condensed them where necessary if they are very similar, and removed those justifications which occur only once,55 in concert                                                  55 I cannot infer via quantitative means that an ideation appearing only once has been generally accepted as a legitimate part of the TJ norm cluster. However, it is likely that some of these cases were in fact accepted as legitimate, and therefore some examples will be explored in later chapters to show the process of cluster evolution with respect to the inclusion or rejection of new ideations. This is somewhat problematic as regards the boundaries of the quantitative dataset, as it may exclude actions which should in fact be included as part of transitional justice (thus undercounting). However, qualitative work which shows the acceptance of new ideations into the norm cluster (even if they occur only once) would enable a second round of evaluation, this 92  with only one behavior,56 or have most definitely not been accepted as a legitimate goal of transitional justice (such as in honor of the anniversary of the state’s founding). The remaining seven ideations form the core of my ideational dataset. They are accountability, rule of law, human rights, truth, democratic transition or democracy, peace, and reconciliation.57 The addition of these ideations to my dataset allows me to identify what I am calling “True-TJ” event: TJ behaviors, justified with an appropriate ideation, which address severe or widespread human rights violations. Accountability, particularly individual accountability, is one of the oldest professed goals of transitional justice, and has its roots in the argument that the state has a “duty to prosecute” human rights violations based on customary international law and obligations created by accession to the major international human rights treaties (Orentlicher, 1991). It is most often associated with legal proceedings, criminal penalties and trials, although this is not absolute. If a law is broken, the person who broke it must pay a price, often but not always involving the loss of freedom or money. Penalties are paid to the harmed party by the transgressor; usually in the justice field that harmed party is the state (whose laws were broken) rather than the victim (whose rights were violated). It does not mean revenge, for revenge is based on victims’ desire to see reciprocal suffering; the state takes no part in the delineation of transgression or the exacting of payment. The term “punishment” walks a fine line between the two, as the penalties exacted by the state are an imposition on the                                                  potentially correcting the undercounting. Given the amount of research required for each of the 39 “other” entries and the relative analytical value of pursuing this second round, however, this further work was not pursued.  56 “Repatriation” occurs only with respect to amnesties. This will be explored further below. 57 I have not included “justice” in this group, although it appears several times in discourse. I see “justice” as an expression of the larger norm cluster, which is called “transitional justice.” Just as “transition” need not mean a democratic transition, I view “justice” as the larger concept to which each of these smaller ideations belongs. 93  transgressor, but the term also carries emotional weight related to the desire to exact suffering. Also associated with legal proceedings, although again not exclusively, is the improvement or institution of the rule of law.  Generally, it refers to “regularity, stability, and adherence to settled law (Teitel, 2000, 11),” but in transitional situations the term also may refer to a set of adjustments to the legal and political system after a period in which the laws did not serve key aims such as the protection of human rights, in order to regularize the protection and promotion of those rights. States pursuing a goal of rule of law are those which seek to create regularised processes of governance, rather than rule-by-decree, rubber-stamp parliaments, or general lawlessness. Rule-of-law-related non-TJ behaviors might include institutional reform or new constitutions. Similar terms which have been coded in my dataset under rule of law include “constitutionality” (Uruguay) and “good governance” (Uganda). The promotion and protection of human rights would seem an obvious reason for justifying behavior meant to deal with human rights violations. It would also seem indicative of a “human rights norm cluster” rather than a TJ cluster, especially if no transition has taken place. However, some of the specific discussions of human rights in legislation refer not merely to the past commission of abuses, but to international human rights treaties to which the state is a party and thus has current obligations (Argentina), to reaffirming a commitment to human rights which had been ignored (United States), and to the prevention of recurrence of human rights abuses in the future (Czechoslovakia). This preventative goal, in particular, is one of the foundations of the transitional justice field (Zalaquett, 1990). 94  The Transition/Democracy category includes democratic transitions, recommitments to democracy even if the implementing state is technically already democratic, and transitions from previous regimes no matter what the new regime is. Empirically, it is very difficult to separate out “democratic transition” and “democracy” from many documents, especially since many states which espouse “democracy” as a goal are in fact going through a democratic transition at the time. In addition, not all states aiming for a democratic transition are successful: some laws are passed with the stated ideation of democratization but the state itself fails to democratize. Lastly, states which actively seek a “clean break” (Nigeria) are included in this category as evidence of a desired “transition,” whether they are democratizing or not. Peace and Reconciliation are perhaps the goals which are most identified with conflict management and resolution rather than with democratic transition, showing evidence of the dual roots of the transitional justice field. However, even democratizing states have professed both goals, particularly when an authoritarian regime and civil war have coexisted. The terms ‘peace’ and ‘reconciliation’ may be also used somewhat interchangeably by particular actors, and are often mentioned together in discourse. However, they have separate definitions. I therefore have chosen to keep them separate in my dataset. Peace refers to the cessation of hostilities, the demobilization and reintegration of troops and militias (Democratic Republic of Congo), stability (Philippines), and/or the restoration of normal relations between opposing forces (Croatia). This definition is usually restricted to situations of armed conflict. Reconciliation is a deeper process, including the restoration of social trust and the ability of different actors to work together within a single political system. It may, but need not, be invoked with regard to civil war. “Coexistence” (Colombia) and “National Unity” (Germany, Somalia, Zimbabwe) are also commonly used, 95  and have been categorized as a variant of reconciliation. These two categories are by far the most often cited in primary source documents, comprising more than half of the total justifications within the TJ spectrum, and are the largest single group of primary-secondary ideation combinations (18 out of 41 total behaviors justified with two ideations). The table below recounts the ideations, both primary and secondary, which form part of the empirical transitional justice norm cluster.58 A Pearson Chi2 test of the distributions of primary and secondary ideations scored a 16.1809 with 6 degrees of freedom, and a p-value of p=.0103; the distribution of secondary ideations is significantly different from the distribution of primary ideations. However, there are relatively few secondary ideations, and a comparison of the distributions of primary ideations only and combined primary/secondary ideations reveals a Chi2 score of 1.4577 (6 degrees of freedom) and a p-value of .962, suggesting that the inclusion of secondary ideations does not significantly change the distribution of ideations overall.  Primary Ideations Secondary Ideations All Ideations Ideation Type Frequency Percent Frequency Percent Frequency Percent Accountability 16 7.58 1 2.44 17 6.69 Human Rights 17 8.06 6 14.63 23 9.06 Rule of Law 31 14.69 2 4.88 33 13.78 Truth 11 5.21 1 2.44 12 4.72 Trans/Democ 29 13.74 5 12.20 34 13.39 Peace 53 25.12 5 12.20 58 22.83 Reconciliation 54 25.59 21 51.22 75 29.53 Total 211 100 41 100 252 100 Table 3.2: Transitional Justice Ideations, 1970-1999.                                                  58 Two additional categories of ideation, namely “repatriation” and various forms of “other,” are not part of transitional justice. Their numbers and their exclusion from the final dataset will be discussed in the following sections. I also do not assume that these are the only possible acceptable TJ ideations; more could be added to the empirical TJ cluster in the future if a state chose them and their arguments were accepted by the relevant actors. See the above reference to ideations occurring only once. Again, each data point is a potential “TJ event;” the inclusion of a TJ-acceptable ideation would make it a “true” TJ event. 96 Missing ideations: Alternative internalized norms Once these ideations were amassed, categorized, and connected to their accompanying behaviors, more steps remained before a final version of the dataset mapping the empirical norm cluster of TJ could be used in analysis. Specifically, I argue that it is necessary to ensure that the behavioral and ideational data overlap: that each case in the final dataset includes information on all three norm components. Although some scholars believe that “talk is cheap,” I argue that it is critical for delineating newly chosen cluster-consistent behavior from that which is not identified as principled behavior at all, or from principled behavior stemming from existing and internalized norms.59 Therefore, failing to ensure that relevant ideations are present may mean that events are added to a dataset which do not actually engage the norm cluster in question, thus lowering validity and hampering accurate analysis. I will use the history of amnesties and their related ideations, and their place in transitional justice processes versus other processes, to explain both the problem and the solution. Amnesties were by far the most frequently observed behavior in the first pass at construction of my dataset, comprising just over 60 percent of the total.60 They were more than 4 times more frequent than the next most popular mechanism, trials. While some of this difference is likely due to amnesties’ relatively low cost of implementation versus other                                                  59 Understanding the supposed reasons behind subtle variations in institutional design, even within the same behavior, may provide both a way to judge “success” and hypocrisy. If a commission of inquiry is justified with reference to accountability, then, it would be hard to argue that the effort is sincere if individual perpetrators and the leadership most responsible are not named publicly. While this dissertation does not examine questions of success or hypocrisy, I suggest that the addition of ideational information could provide valuable insight for future studies. 60 Amnesties were included in the problem-behavior dataset for this dissertation if human rights violations occurred in the conflict for which the amnesty is used, if the amnestied persons had been accused of human rights violations, or if the imprisonment of those amnestied has been called a human rights violation itself. In other words, their inclusion is directionally neutral: it covers amnesties both of and by human rights violators. All three types correspond to the uses set out in the Joinet report (Joinet, 1985). This is in partial contrast to the TJBD, which includes amnesties for the state and the opposition but codes and analyzes them separately. (T. D. Olsen et al., 2014) 97  mechanisms, I additionally call attention to the fact that amnesties are potentially suitable for addressing many different normative goals, not all of them TJ-related.  Indeed, amnesties as a behavior are potentially part of at least two additional (and much older, more internalized) norm clusters: management of armed conflict and use of executive power (Joinet, 1985; Mallinder, 2009b; Reiter, 2014). They are often also used to secure political ends which are not publicly identified as such because they are unlikely to be regarded by the public as legitimate. Without accompanying data on ideations, therefore, it is impossible to tell which amnesties belong to the TJ cluster in particular and which to other clusters, or none at all.61 This section will use several examples to provide greater detail to the argument. The fact that some of these additional norm clusters are older and more internalized than the TJ norm cluster points to another way in which the addition of ideational data to behavioral datasets allows for a more accurate final collection. Recall that the norm life cycle (Finnemore & Sikkink, 1998) and the spiral model (Sikkink & Risse, 1999) both theorize that once a norm has been internalized, states no longer question the reasoning or the justification behind their behaviors; it is simply “the thing to do” to address the problem. Therefore, overt reference to an ideation may be missing. However, for emerging or not yet internalized norms, I argue that a behavior is more likely to be accompanied by an explicit                                                  61 A potential complication exists with regard to amnesties with the stated goals of peace or reconciliation, as these ideations are part of both the conflict management and the transitional justice norm cluster. In cases of problem overlap (where a conflict includes human rights violations), if the conflict management cluster is not yet internalized by the state in question, it will not be possible to separate amnesties for peace and reconciliation into their separate clusters. This is a relatively large number of cases, but not of states: of 91 amnesties for peace or reconciliation, 25 occurred in one of 9 states with at least 5 major episodes of civil violence in in the same year and over 20 episodes in the 5previous years. 22 of those amnesties, in 8 countries, occurred before 1994. (Center for Systemic Peace, 2013). This could point either to the gradual internalization of amnesties in the conflict management cluster or to the declining use of amnesties within the transitional justice cluster. The issue will be explored further in a later chapter. 98  statement of ideation, in order for that behavior to be understood as comprising part of a “new” norm instead of a previously existing one. Because amnesties (and other mechanisms) potentially form part of the accepted group of behaviors for other pre-existing and relatively settled norm clusters, I argue that separating those mechanisms with stated ideations from those without offers a way to use discourse to determine with greater accuracy which behaviors are part of the norm cluster and which are not. The effect is to separate out events for which “no explanation is needed” because the norm they embody has already been internalized by the state in question, versus those for which a justification is still provided, whether or not it has been internalized. Although states internalize norms at different rates, I assume that because transitional justice is the newest norm cluster, emerging only in the 1970s or 1980s, it is the cluster least likely to have been internalized by any particular state and therefore the most likely to have ideations stated rather than implicit. Therefore, I exclude amnesties without stated ideations (87 out of 251) from my TJ norm cluster dataset.62 They include cases such as India (1975), Portugal (1976), Thailand (1981) and Chad (1992). Amnesties are not the only mechanism which may form part of an older, internalized norm that overlaps in some way with transitional justice. I suggest that some of the trials included in The Justice Cascade (Sikkink, 2011) dataset also demonstrate this unintentional conflation of old and new norms. Even under Sikkink's carefully crafted rules for including observations - human rights trials in transitions to democracy - not all trials have expressed justifications, suggesting that they were expressions of an already internalized norm such as                                                  62 Norm internalization is not the only reason why ideations may be missing; see the next section for an additional explanation for missing data. I assume that some of the 87 amnesties excluded are because they are internalized alternative norms and that some are actually not identified as normative behavior, but I have no way of quantitatively determining which is which. 99  “trials after a regime change to mark a difference between regimes,” regardless of human rights-related problems or transition to democracy. My problem-behaviour dataset includes at least 34 of her trial countries, but 11 of these sets of trials have no stated reason at all. Without ideational data about the goals of the trial or the values expressed by the use of a trial rather than another mechanism (or nothing), it is impossible for the quantitative scholar to know precisely whether such a trial is part of a transitional justice process or the expression of an older, more internalized alternative norm.63  Sikkink herself provides qualitative evidence that some of the early trial cases (namely Greece and Portugal) may not actually have been adopting the new norm. Neither case has an expressed justification for trials in my dataset, although other practices occurring in the same time frame have associated ideations (the 1976 Portuguese amnesty was for national reconciliation).64 In investigating the reasons for these trials, Sikkink tests several potential ideations: the desire for punishment or revenge coming from the populace, the new president wanting to make a clean break from his past, and a desire for national reconciliation.65 However, the conclusion Sikkink reaches is that, for each of these cases, trials were simply “the thing to do,” and that the governments “had no choice” (2011, p. 50). There may have been some problem-based innovation, namely pursuing trials for torture rather than only state-related crimes like treason, but the countries did not see                                                  63 For the countries in Sikkink’s dataset the potential TJ norm may be expressed as “If [human rights violations + democratization], [______] suggests [trials].” The focus of Sikkink’s book is on individual criminal accountability for gross human rights violations, so she might argue that “accountability” is the value expressed by this norm. The scope conditions also have an additional necessary component, democratization, on top of the problem of violations. However, without a direct expression of the “accountability” ideation by the actors themselves, it is not certain that Sikkink’s version of the norm is the one being enacted through these trials. 64 Portugal, Decreto-Lei n. 758/76 (22 October 1976): “Os Governos e, em geral, os Órgãos de Soberania posteriores a 25 de Abril de 1974, sem embargo da sua legitimidade revoluconária, não conseguiram reuneficar a familia portuguesa desgarrada pelo anterior regime em torno de um claro projecto e de um vivo sentimento de unidade nacional” (Italics added). (Portugal, 1976a) 65 Reconciliation, however, is speculated to have been related to the commutation of death sentences after trials were finished (Sikkink, 2011, p. 47). 100  themselves as innovators, or making a new normative choice. Their practical example may be the basis for the conduct of human rights trials in many other countries, but these countries did not regard themselves as setting a normative precedent.  I conclude that, based on Sikkink’s account, these countries were simply following an internalized norm of trials, human rights-related or not, after a regime change. The ideation (although unstated due to it being already internalized) might be to mark a difference between regimes, discrediting the old and enhancing the legitimacy of the new, but not really in pursuit of “transitional justice” as defined here. They are therefore excluded from my TJ norm cluster dataset. I follow a similar set of assumptions regarding the exclusion of reparations, lustrations and Commissions of Inquiry from the final TJ dataset. Missing ideations: Non-normatively justified behavior  In addition to alternative settled norms I argue that there is another way in which the potential TJ events included in a problem-behavior dataset may not actually be part of the empirical norm cluster of transitional justice, which provides an additional reason for exclusion of cases without stated ideations. I contend that non-normatively justified behavior may also lead to an overcount of seemingly TJ-related activities. Not all behavior is driven by publicly justifiable goals: some actions are coerced and others are the result of a political arrangement to forestall an outcome unrelated to the normative goals of TJ. For example, granting an amnesty to avoid a coup has a goal of remaining in power or forestalling unconstitutional regime change, but the public is unlikely to accept these justifications as legitimate goals of a transitional justice process (or as part of other norm clusters). Moreover, both TJ-related and non-normatively identified activities may occur at the same time, especially during a democratic transition when decisions must perhaps be 101  negotiated with other parties. While a strictly problem-behavioral dataset might miss this distinction, I argue that the addition of ideational data allows me to separate out most cases of non-normatively identified behavior, or behavior which at least publicly justifies itself as normatively consistent, from the final dataset. Specifically, in cases of some cases it is less likely that an ideation will be given at all. Consequently, this dataset eliminates cases where no ideation is given for this reason as well as that of internalized other norms.  Examination of a specific country’s experience, namely Argentina’s democratic transition, provides further evidence in for the value of this exclusion. Jose Zalaquett’s analysis of the series of mechanisms which were enacted in Argentina from 1983-1987 describes an overall goal of “national pacification,” and attaches certain specific ideations to specific mechanisms. Truth, in particular, is easily identified: the establishment in December 1983 of the National Commission on Disappearance of Persons and the creation in 1984, after the Commission’s report was delivered, of a bureau within the Ministry of the Interior to continue investigations into disappearances and related crimes. His stated justifications for the combination of high-level trials and lower-level amnesties is divided, however: trials are associated with punishment, with the larger goal of “overcoming the past,” while amnesties were born of the realization that “a policy of unrestricted prosecutions and punishments was neither conducive to national pacification nor possible to carry out without military unrest turning into open rebellion” (Zalaquett, 1989, pp. 24-26).  Zalaquett’s descriptions of behavior carry elements of both TJ-consistent and non-TJ consistent ideational considerations: overcoming the past and national pacification are considered appropriate aims of transitional justice processes, but punishment (as opposed to accountability) and prevention of a military rebellion are pragmatic and political aims that are unrelated to the TJ norm cluster. Punishment is associated not with justice but with 102  revenge, and seeking to avoid military rebellion (coup?) is a tactical decision related to the stability of government. Examination of official documents is necessary to determine whether the TJ-consistent normative goal was publicly explained as most important for a particular behaviour or not. In this case, trials were explicitly instituted for the goal of “rule of law,” while the amnesties have no stated ideation (Government of Argentina, 1983; 1986; 1987). They appear to be legal or administrative procedures which are not publicly justified by a higher normative aim. In short, according to the Argentinian government, amnesty was not part of its pursuit of transitional justice.66 Therefore, I exclude the amnesties – but not the trials - from my dataset.  Following the logic of this and the previous section, I contend that a high proportion of the cases in which there is no stated ideation are likely to be either expressions of settled alternative norms or non-normatively justified behavior, i.e. an ideation is not stated because it is either internalized or because no publicly acceptable ideation is involved. Therefore I exclude from the final dataset all those behaviors without official justifications (150 potential TJ events). This leaves a total of 268 events out of 418, for a 64% overall retention rate.67 Except for commissions, the retention rate is between 1/2 and 2/3 of all cases. The Pearson Chi2 value of 1.6520 (6 degrees), with a p-value of .799, indicates that the distribution of behaviors with ideations is not significantly different than the distribution of behaviors without.                                                  66 Interestingly, the Commission on Disappearances also has no stated ideation in its implementing legislation, suggesting that the use of a commission of inquiry was not an expression of a new norm, and that “truth” was merely a procedural necessity before trials or other TJ mechanisms could take place. 67 This significantly restricts the amount of cases in the final dataset, but I show in Appendix B that the resulting separate datasets (i.e. the TJ norm cluster and the remaining non-ideational behaviors) comprise distinctly different sets of actors, making decisions based on very different sets of circumstances and influences. These differences act as a robustness check against my arguments for the exclusion of these cases. Undercounting in this way may carry the additional risk of missing the complete spread of TJ norms, or of changing the year of the approximate tipping point, but it cannot change analysis of the structure of the norm cluster or of the variety of combinations, as the data simply does not exist. 103   All mechanisms Mechanisms with stated ideations Retention rate (Percent)  Freq. Percent Freq. Percent Amnesty 251 60.05 164 61.42 65.34 Lustration 37 8.85 21 7.87 56.76 Reparation 36 8.61 21 7.87 58.33 Trial 60 14.35 34 12.36 55.00 Commission 34 8.13 28 10.49 82.35 Total 418 100 268 100 63.68 Table 3.3: Transitional Justice-related mechanisms with and without ideations, 1970-1999. Alternative ideations: Non-TJ normative behavior Lastly, I suggest that certain events which meet the “problem” and “behavior” criteria for inclusion in a TJ dataset, and which have stated ideations, may actually be expressions of other norms and thus should also be excluded. I argue that an analysis of ideational data, beyond merely whether or not such data exists, helps to separate out those cases in which a specific but unrelated norm is accidentally swept up with transitional justice practices. Again, the practice of granting amnesty helps to illustrate my point. The table below illustrates the range of ideations, both primary and secondary,68 associated with amnesties in my dataset:  Primary Ideations only Primary and Secondary Ideations Ideation Freq.  Percent Freq.  Percent Accountability 0 0 1 0.51 Human Rights 4 2.44 7 3.57 Rule of law 10 6.10 10 5.10 Trans/Democ 14 8.54 18 9.18 Peace 47 28.66 51 26.02 Reconciliation 44 26.83 58 29.59 Repatriation 18 10.98 19 9.70 Other 27 16.46 32 16.33 Total 164 100 196 100 Table 3.4: Amnesties by ideation, 1970-1999.                                                   68 Test of the similarity of the distributions: Pearson chi2 (7) =2.0253, p=.958. 104  Based on this table, I suggest that adding the third component of a norm (ideation) helps to determine which instances of amnesty count as transitional justice and which do not. All of these amnesties happened in countries with a history of human rights violations, and those amnestied were either the subjects or perpetrators of some of those violations. However, not all the amnesties listed above actually attempt to “solve” the human rights problem, even though they interact with it. The first six categories have all been espoused by transitional justice theorists as legitimate goals of a transitional justice process. However, they account for just over 50% of the total number of amnesties recorded. I argue that the 19 cases of amnesty which have an ideation of “repatriation,” and the 32 marked as “other,” belong to separate normative endeavors. Amnesty for the purposes of repatriation has barely been mentioned in the theoretical and empirical literature on transitional justice (Harris Rimmer, 2009; Roniger, 2011; Wolfson, 2005), which suggests that it has not yet been accepted as part of the norm cluster. Therefore, I posit “amnesties for repatriation” can be excluded from the final version of the TJ norm cluster empirical dataset. There are 19 such cases in the collected data (18 primary ideations and one secondary).69 There are also no commissions, trials or lustrations for repatriation, or even reparations, which would conceptually seem to be a possible match since additional funds and services could help refugees resettle in lost homes. On the other hand, if more scholars make and accept this argument, and if states which explicitly link repatriation to the wider practice of transitional justice (including by                                                  69 18 primary ideations and one secondary ideation (Croatia 1995; the primary ideation was Rule of Law). Those with repatriation as the primary ideation have been removed from the “True TJ” dataset, while Croatia is not, as it is listed instead as an amnesty for rule of law. 105  using it to justify activities other than amnesty) are positively received, then repatriation will enter the norm cluster as an acceptable ideation. There is also an “other” category of justifications. While this category is not exclusively limited to amnesties, “other” reasons are attached for amnesties more than any other behavior in the TJ norm cluster, both absolutely and relatively. These other reasons include but are not limited to repentance (Turkey), family reunion (Bangladesh), rescuing youth (El Salvador), self-determination (France), forgiveness (Iran), political progress (Jordan), confidence-building (Israel), compassion (Philippines), and the inauguration of a new president (Madagascar). Some of these may conceptually fit the larger concept of transitional justice, but they appear infrequently in both the scholarly literature and in empirical practice. If future actors cite these ideations as their motivations behind the adoption of transitional justice behaviors, and are accepted as legitimate TJ actors by scholars and the wider political community, then they would be included as TJ ideations in the future once evidence of that acceptance had been established. The process of removing behaviors without accompanying ideations, or with ideations unrelated to the theoretical TJ norm cluster, is also applied to the other behaviors in the dataset. Remaining totals and retention rates for specific behaviors are detailed in the table below. The remaining TJ events total 211 cases over the 30-year period of investigation.    106   All Mechanisms  “Mech only” count “True TJ” Count “True TJ” Percent “True TJ” Retention rate Mechanism Freq. Freq. Freq. Percent Percent Amnesty 251 164 119 56.4 47.41 Lustration 37 21 18 8.53 48.65 Reparation 36 21 18 8.53 50 Trial 60 34 30 14.22 50 Commission 34 28 26 12.32 76.47 Total 418 268 211 100 50.478 Table 3.5: Data retention rate, TJ mechanisms and TJ mechanisms with TJ primary ideations, 1970-1999.  While amnesties remain predominant, they are a slightly smaller percentage of the whole (56.4%) than they were without ideational knowledge to exclude non-TJ mechanisms (60.05% as noted in table 1), and occur only four times as frequently as the next most popular mechanism (trials).70 Reparations and lustrations occur least often, with trials and commissions in the middle. A Chi-squared comparison of the two distributions of mechanisms, however, does not reveal statistically significant differences (Chi2 (6) =2.9155, p=.572): the distribution of behaviors in the “true-TJ” events of my final dataset is very similar to the original group of “potential TJ events” collected from behavioral data only. In total, these exclusions represent a large proportion of the initial dataset. I have provided theoretically-based explanations for these restrictions throughout the previous sections, but it must be acknowledged that they might in some way bias my final dataset, and thus my findings, in ways which differ significantly from those of other datasets. Appendix A details a series of tests comparing the distributions of behaviors, and the retention rate of behaviors based on “True-TJ” ideations, in my dataset to that of the                                                  70 This dataset may still overestimate the occurrence of amnesties within the transitional justice norm cluster by including amnesties, particularly for peace or reconciliation, which occur in the context of armed conflict (with human rights violations). As such, these practices are on the boundary between norm clusters, which has been discussed above. According to my theory, since conflict management is a relatively settled norm it should not need justifying and therefore be excluded in the dataset, but if the conflict management norm has not already been internalized, amnesties may still be justified in this manner. This means that the boundaries of the two norm clusters, and which amnesties fit in which cluster, cannot be completely determined without further restrictions on the dataset or additional evidence regarding the intent of the governments instituting these amnesties.  107  Transitional Justice Database Project (TJDB). I show that my data are not biased by the addition of “true-tj” ideations as a necessary criterion for inclusion.71 It is also possible that I may be undercounting the true number of TJ events due simply to missing data, which would impact estimations of the tipping point as well as hiding the true spread of the norm cluster, but as will be seen in the next section, my estimate of the tipping point accords well with other determinations based on non-quantitative measures. These exclusions may also expose differences in the characteristics of the actors employing these TJ events versus those which employ behaviors without ideations, and I suggest that they should: actors with ostensibly similar normative goals should be similar to one another in key ways, and different from those acting for other reasons, be they non-normatively or other-normatively identified. I therefore include in Appendix B some comparisons of the base characteristics of non-adopters, mechanism-only adopters, and “true-TJ” adopters on key variables which transitional justice scholars have identified as relevant to the pursuit of TJ-normative behavior. While not meant to imply causality (which will be addressed in Chapter 4.1), they show that in many ways, TJ norm adopters are different from other actors, providing further evidence that cluster-consistent actions and non-cluster-consistent actions may be generally separated in behavioral datasets by the addition of ideational data. 3.4.4. The norm (cluster) life cycle: The approximate tipping point Lastly, in order to perform certain analyses, I distinguish between the emergence and cascade phases of the norm cluster. The development of a norm over time, as a                                                  71 Appendix A details general issues with data collection and coding, explains differences between myself and other sources, and presents two sets of comparisons for reliability purposes: one to the TJDB, and one to general theory on the norm life cycle. 108  mathematical function of the number or proportion of adopters, follows a theoretical pattern. In the emergence stage of a norm’s development, adoption is supposedly slow and the total number of adopters is small. At the “tipping point”, a critical mass is reached and the rate of adoption (and number of adopters) grows dramatically; this is called the cascade stage. As the total number of adopters reaches the total number of actors in the system and only a few holdouts remain, adoption rates slow down and states begin the internalization stage (Finnemore & Sikkink, 1998; Simmons et al., 2008). As I contend that many of the same diffusion dynamics present for single norms are applicable to norm clusters, I extend the concept to the TJ norm cluster in my own work. There is a risk that some early-adopting states may internalize their TJ norms before the beginning of phase three of the norm life cycle, thus removing ideation from legal or other texts earlier than other states (Finnemore & Sikkink, 1998). As argued above, as states internalize a norm it ceases to need overt justification, and pursuing a certain behavior to deal with a certain problem is just “what is done.” The ideation is no longer overtly included in the text of relevant laws. Such a case would drop out of my dataset, but be retained in a behavior-only dataset. However, I suggest that limiting certain analyses to the emergence phase only should minimize this risk. Only the earliest of adopters would be at risk of this happening, rather than early and later adopters if both phases are analyzed together.72                                                  72 The decision to adopt a particular combination remains in the dataset no matter how many additional actions are taken. If investigating the variations inherent in a norm cluster, as I am attempting to do, multiple iterations of the same combination by the same state would add weight to the relative importance of a particular combination within the cluster, but it would not change when that combination appeared within the dataset, nor which states had used it. Much of the analysis conducted in this dissertation is therefore not affected by the absence of this data. 109  Therefore, I must find an approximate “tipping point” at which to divide my dataset for certain tests. Theorists are not united in their determination of exactly what constitutes a “critical mass,” or how many and which states must adopt a norm before it “tips” into a cascade. Some scholars (Gilligan & Nesbitt, 2009; Locher, 2002; Weldon & Htun, 2013) choose a simple numerical cut-off: Finnemore and Sikkink suggest that a tipping point rarely occurs before 1/3 of the states in the system have adopted the norm (1998, 901).73 I therefore use both a 1/3 mathematical cutoff and a “critical states” criterion for approximating the tipping point. Determining the tipping point in this manner requires counting TJ-adopting states rather than TJ behaviors adopted.74 As I argued above, most states have had grave or widespread human rights violations in the past 100 years, making all states potentially able to adopt the TJ norm cluster. Therefore, I count state adopters as a proportion of total states in the international system. Following the precedent of the TJDB, I begin counting “states in the system” beginning in the first year of my dataset. This means that adoption rates for 1970 are calculated based on the number of states which had adopted the norm cluster in 1970. That is 1 (Peru) out of 130 states in the international system, or 00.77%.                                                  73 It might theoretically be possible to determine a tipping point for a norm cluster statistically via a Markov process. However, I do not use that method here for two principal reasons: first, as I do not count (and do not expect, during these earlier stages) norm exit, the requirement of a Markov process that actors be able to move freely between states does not appear to hold for this dataset. Second, while a model could conceivably be created which accounts for states’ choice to pursue TJ behaviors, there exists no data which helps to account for the ideological motivation which separates “true adopters” from those undertaking TJ-behaviors for other reasons. An accurate model of norm cluster adoption, as separate from behavioral adoption, is not achievable, and therefore I would not regard any tipping point reached via a Markov process using this model as trustworthy. 74 Because a state may choose multiple combinations of (specific) problem, behavior and ideation, some states may appear several times in an event count database while others appear only once. A simple count of events would therefore show an increase in transitional justice adoptions, but these would be diffusions of specific combinations within the TJ norm cluster rather than the whole norm cluster itself. 110  Using these criteria, the approximate tipping point occurs in 1993, with 190 current and former states in the system and 64 states which had adopted TJ at any point since 1970.75  A secondary validation method is checking for the inclusion of “critical” states. For example, Finnemore and Sikkink posit that a critical state for the anti-landmines norm would be one that produced or used landmines (ibid). In terms of potentially “critical” states adopting TJ through 1993, two processes from each of TJ’s roots suggest that a tipping point occurs at least near 1993, if not in that exact year: first, the wave of adoptions from the newly democratic states of the former Eastern Bloc and Soviet Union (mostly 1990-1991) seems a good indicator of widespread acceptance of a new normative environment as relates to emerging democracies, if not human rights violations more generally. Second, the United Nations Security Council’s adoption of Resolution 827 of May 25, 1993 created the Statute of the International Criminal Tribunal for the former Yugoslavia (ICTY), which for the first time asserted that gross violations of the laws of armed conflict should be addressed by individual criminal trials (United Nations Security Council, 25 May 1993).76 Based on these two secondary “critical” adoption points, I conclude that 1993 is a fair approximation of the TJ cluster’s tipping point. This also means that the rest of the cases in my dataset are part of, but not the entirety of, the cascade phase of the cluster’s life cycle. This portion of the data includes the 63 cases of “true TJ” events which occur from 1994-1999, leaving 148 “true-TJ” events in                                                  75 A “tipping point” also implies a mathematical relationship between phases: a graph of the cumulative percentage of adopters out of the whole community of possible adopters looks like an S-curve, with the changes between phases as inflection points in the graph. For an examination and discussion of my dataset as it relates to the S-curve and the norm life cycle more generally, see Appendix A.  76 Although they are not domestic trials, as the “IHL” root of the field is thought of as an instance of vertical norm diffusion, I believe it is appropriate to use international criteria to determine a critical change in the practice. 111  the emergence phase. The table below details their breakdown in terms of both mechanisms and ideations. Emergence (1970-1993) Cascade (1994-1999) Mechanism Freq. Percent Mechanism Freq. Percent Amnesty 87 58.78 Amnesty 32 50.79 Lustration 12 8.11 Lustration 6 9.52 Reparation 13 8.78 Reparation 5 7.94 Trial 20 13.51 Trial 10 15.87 Commission 16 10.81 Commission 10 15.87 Total 148  Total 63        Ideation Freq. Percent Ideation Freq. Percent Accountability 10 6.76 Accountability 6 9.52 Human Rights 11 7.43 Human Rights 6 9.52 Rule of Law 23 15.54 Rule of Law 8 12.7 Truth 7 4.73 Truth 4 6.35 Trans/Democ 25 16.89 Trans/Democ 4 6.35 Peace 37 25 Peace 16 25.4 Reconciliation 35 23.65 Reconciliation 19 30.16 Total 148  Total 63  Table 3.6: The Transitional Justice Norm cluster: “true TJ” behaviors and ideations, 1970-1999.  3.5. Result: The empirical transitional justice norm cluster After the series of data restrictions imposed above (and further detailed in Appendix A), the total number of cases left in the norm cluster out of the initial set presented in the problem-behavior database is significantly reduced. Recall that between 1970 and 1999, the initial population of cases with a matching problem and behavior was 418. By eliminating cases due to a lack of or a “non-TJ” ideation, that number was reduced to 211. This group of TJ events comprises my most accurate record of the empirical expression of the transitional justice norm cluster. I argue that the resulting dataset, created by more accurately measuring TJ events, is a valid representation of the empirical expression of the transitional justice norm cluster. Specifically, the removal of behaviors without 112  accompanying ideations, or with ideations which are not part of the accepted canon of “transitional justice”-related goals and values, offers a more precise look at how states have attempted to engage with the norm cluster as a whole.77  The next chapter will analyze the collected dataset in order to show that variation in ideation-behavior combinations occurs which is both widespread and not attributable to the two potential bundles created by the supposedly separate problems of TJ’s roots: IHL violations and authoritarian legacies. It will do so by exploring quantitatively the characteristics of those states which have adopted the transitional justice norm cluster and their resulting ideation-behavior choices.                                                  77 Appendix A contains a series of reliability tests which compare my dataset and the Olsen, Payne and Reiter dataset as well as the “true-TJ” portion of my dataset with complete non-adopters and behavior-only TJ events. It finds, in addition to table 4.1., that the addition of ideational content creates a set of adopters which are more likely than both the OPR dataset and my behavior-only dataset to conform to the theoretical arguments about which states undertake TJ activities, and why. This provides clearer evidence than the Olsen, Payne and Reiter dataset that TJ adopters are a generally coherent group of actors. 113  Chapter 4. Diffusion and variation: The structure of the TJ norm cluster This chapter analyses my large dataset in order to understand the contents and evolution of the transitional justice norm cluster, and to search for evidence of continuity and change as states work through the steps of the USCC decision process. By examining the cluster as a whole, cluster adopters, specific norm components, and how they combine with one another over time, in groups, and with respect to specific motivating factors, I find that the field of transitional justice carries elements of both continuity and change, making the norm cluster concept a helpful conceptual innovation and the USCC a useful tool for determining how each occurs. Lastly, the cluster’s mapping helps to identify specific cases for further qualitative analysis, to trace the processes of decision-making and discourse which have shaped the TJ norm cluster over time. The chapter proceeds according to the steps of the USCC itself, with each step examined with respect to its own observable outcomes and using its own test or exploration of evidence. First I examine steps 2 and 3 of the model, which relate to the recognition of a problem as salient and the diffusion of the norm cluster as a whole. I include analyses of both cluster adopters and non-adopters in order to provide evidence as to why certain states are “true-TJ” adopters and some are not. I then examine the overall structure and contents of the TJ norm cluster, and its adopters, for evidence of variation and continuity. This section is meant to familiarize readers with the “true-TJ” dataset before proceeding to analyze the reasons behind certain specific outcomes based on the later steps of the model. I show variation in outcomes of combinations of ideation and behavior, of the number and type of combinations chosen by individual states, and variation of these combinations even within states which engage the norm cluster more than once. Lastly, I proceed through steps 4-6 of the USCC model to look for decisions which reflect gold standards or bare 114  minimums, are based upon particular typologies of similarity, or are constructed according to the potentially separate diffusion of norm components. I observe several trends which suggest continuity and direct diffusion. However, I also find that a great deal of variation exists in the choice of ideation-behavior combinations which is not attributable to TJ’s theorized roots, suggesting unfused diffusion and change. I also find that predictors of specific behavioral choice do not predict ideational choice and vice versa; this also implies unfused diffusion. This all suggests that states sometimes imitate each other’s’ practices exactly, and sometimes do not: the forces leading to continuity and change in norm diffusion outcomes exist for each actor, and states may choose both. I conclude that transitional justice is indeed well-characterized as a norm cluster, and that it is a rich example of variety in norm diffusion outcomes. 4.1. Cluster diffusion: The TJ norm cluster as a whole This section examines transitional justice norm adopters as a particular group of actors as compared to non-adopting, or behavior-only, states. In doing so I lend support to my argument from Chapter 2.2.4. on the nature of principled decision-making. I suggested that the first three stages of the modified USCC decision-making process (“General Problem Identified” through “Adopt Norm cluster?”) are the same for norm diffusion processes which ultimately create either continuity or change in a state’s ultimate choice of ideation-behavior combination. I find that the TJ norm cluster, as a whole entity, conforms to existing theory on transitional justice and diffusion: states which adopt it are more likely to have significant experience with the general “problem” of human rights violations and may potentially find it more salient due to that problem’s severity or because of a democratic transition. In addition, states which have a greater opportunity to learn about the norm cluster from their 115  neighbors are also more likely to adopt the TJ norm cluster than those who do not, meaning that the TJ norm cluster as a whole shows evidence of diffusion across states.  4.1.1. The “problem” and the decision to act: TJ theory and evidence This section examines whether the data I have collected suggests that TJ-adopting states are motivated to solve common problems. This corresponds to the second and third sections of my decision-making model: that states recognize that a given problem applies to them, and that they decide to do something about it. Transitional justice mechanisms are theorized as the most “appropriate” way to deal with the “problem” of severe human rights violations, often engendered by repressive government or civil war. Therefore, those states with more recent or more devastating experiences of violations should find the problem more salient, and therefore be more likely to create TJ mechanisms to deal with it. In addition, the literature suggests that salience may be increased due to a democratic transition and the desire to cement certain political values which were not present before. I therefore hypothesize that states for whom either the “problem” of gross human rights violations or the opportunity to address it (or both) is more salient, due to recent severe conflict or democratization, are more likely to adopt the TJ norm cluster than those which adopt only behaviors without associated TJ ideations.  To test this hypothesis, I examine the characteristics of states across a number of variables operationalizing specific aspects of transitional justice theory. Using an event-count dataset for