Constructing Achievement in the International Criminal Tribunal for the Former Yugoslavia (ICTY): A Corpus-Based Critical Discourse Analysis

The International Criminal Tribunal for Yugoslavia (ICTY) was established by the UN Security Council in 1993 to prosecute persons responsible for war crimes committed in the former Yugoslavia during the Balkan wars. As the first international war crimes tribunal since the Nuremburg and Tokyo tribunals set up after WWII, the ICTY has attracted immense interest among legal scholars since its inception, but has failed to garner the same level of attention from researchers in other disciplines, notably linguistics. This represents a significant research gap, as the Tribunal’s public discourse (notably its case law and Annual Reports) can open up interesting avenues of analysis to researchers of law, language, and legal discourse alike. On its official website, the Tribunal claims that it has “irreversibly changed the landscape of international humanitarian law” and lists six specific achievements: “Holding leaders accountable; bringing justice to victims; giving victims a voice; establishing the facts; developing international law and strengthening the rule of the law”. While a number of legal scholars have studied and critiqued the level of ‘achievement’ actually attained by the Tribunal against these metrics and others, of interest to linguists is the ways in which this work might be conveyed discursively. In this paper, we demonstrate how methods from the linguistic field of corpus-based critical discourse analysis can be utilised to explore the discursive construction of such achievements in the language of the ICTY.


Introduction
The International Criminal Tribunal for the former Yugoslavia (ICTY) was established by The United Nations' Security Council in 1993 with the purpose of ''prosecuting persons responsible for serious violations of international humanitarian law committed in the territory of the former Yugoslavia between 1 January 1991 and a date to be determined by the Security Council upon the restoration of peace'' [50: Article 2]. It is an ad hoc tribunal that was rapidly configured in response to a specific need and with limited scope, time, and jurisdiction, and that has operated since its inception under intense pressure from the UN Security Council to work both efficiently and effectively to achieve goals set out in its mandate [13]. Since its creation, the Tribunal has indicted 161 persons; at the time of writing, proceedings are concluded on 147 of the Accused, and proceedings are ongoing for the remaining 14 cases [28].
This is not to say that the Tribunal has functioned entirely as envisioned; indeed its ongoing work is now being undertaken well beyond the initial timeframe. Under pressure to complete work, in 2003 the ICTY adopted a 3-phase completion strategy with the following milestones: investigations of war crimes should be ended by the end of 2004; all first instance trials completed by the end of 2008; and all work of the Tribunal should cease in 2010 [51,52]. The Tribunal has failed to meet these milestones, and it is anticipated at the time of writing (May 2015) that closure will take place in 2017, after which national courts in the former Yugoslavia and the UN Mechanism for Criminal Tribunals will take over and complete all resting cases [53].
In addition to time pressure-or indeed perhaps even contributing to it-is the fact that the Tribunal has been acting in the midst of highly contradictory expectations to its achievements from other outside stakeholders. The establishment of the ICTY was a decision made by the international community and not requested at the national level by the countries of the former Yugoslavia, and has therefore been critiqued by Balkan countries for being a distant and irrelevant court that is unable to bring justice and peace to the region [13]. This led to the adoption of an outreach strategy in 1999 [25], which is concerned with increasing the involvement of the local communities and emphasising the Tribunal's role in the restoration of peace and the rule of law in the former Yugoslavia. Moreover, the international community remains highly engaged in the ICTY's mission as a modern international criminal court that has inspired the establishment of other ad hoc tribunals (for Rwanda and Sierra Leone) and the permanent International Criminal Court.
As the first war crime tribunal after the Nuremberg and Tokyo tribunals following WWII, the ICTY has also been followed by the international legal community with immense scholarly interest. How would the Tribunal interpret the criminal acts listed in the Statute? What Rules of Procedure would the criminal judges-appointed from countries all over the world and belonging to different legal traditions-decide and agree on? Competing academic descriptions of the appropriate role of a war crimes tribunal (i.e. prosecuting criminals, establishing historical facts, bringing restorative justice, and ensuring international peace and security) also define the scholarly debate about the lasting contribution of the Tribunal [see 6,34,45].
Bearing this complex background in mind (and considering the contested legitimacy of the ICTY), the way that the Tribunal has reacted to the critique is an object worth studying more closely. However, to date, much of the discussion of the ICTY's achievements and limitations has been restricted to scholars of law, and despite being a transformative institution, the Tribunal has failed to attract the same attention in research traditions in other humanities and social sciences, notably linguistics.
We would like to add to the scholarship on the Tribunal by considering a linguistic perspective and re-focussing on the discourse of the ICTY in isolation. In this way, we may investigate the ways that its achievements are presented and constructed discursively in the 'voice' of the Tribunal itself.
On its official website [24], the Tribunal claims that its six main achievements are: 1. Holding leaders accountable 2. Bringing justice to victims 3. Giving victims a voice 4. Establishing the facts 5. Developing international law 6. Strengthening the rule of the law While a number of legal scholars have studied and critiqued the level of 'achievement' actually attained by the Tribunal against these metrics and others, of interest to linguists is the ways in which this work might be conveyed discursively. For instance, how does the Tribunal construct itself linguistically-in the authoritative and informative texts (judgments and annual reports) it producesin response to the expectations of the stakeholders and the UN mandate that it has to fulfil?
In this paper, we aim to demonstrate how a corpus-based critical discourse analytical approach may be illuminating in the pursuit of scientific, empirical, reproducible research on a very large amount of data made up of legal language. From a substantive perspective, we want to add a discourse analytical viewpoint to critical legal studies on international criminal courts by focussing on the way the ICTY presents itself and its work linguistically, while highlighting innovative methods that might be adopted by other scholars in the field. To this end, our broad research questions are both methodological (a) and practical (b): (a) How can methods from corpus-based critical discourse analysis contribute to examination of the language of the law?
(b) How are 'achievements' discursively constructed and manifested in two collections of texts created by the ICTY?
By exploring these questions, we will contribute to closing a gap in research within the field of legal discourse studies. The discourse of the ICTY-a court with limited, ad hoc jurisdiction, acting in the field of international criminal justice for the first time since the post-WWII trials-is a fertile area of inquiry for legal linguists with an interest in the co-evolution of law and language. Moreover, the disputed role of the Tribunal supplies the discourse analyst with valuable data for critical examination of the way language co-constructs the power and image of a legal institution.
We should make clear that our focus is on the language perspective of the lawand-language interplay. However, we hope that our results may also pave the way for asking specifically legal questions, especially concerning the way that the ICTY has contributed to the development of international criminal law. The ICTY was a prototype tribunal from which lessons have been learned, and the Rome Statute of the International Criminal Court corrects certain shortcomings of the ICTY in terms of the definition of crimes. 1 While we believe those questions to be of interest, also from a language perspective, elucidating them falls outside the scope of the present paper, though we demonstrate how these methods may also be adopted in further analysis.

Theoretical Background
The paper adds to research in three fields-international criminal law, Critical Discourse Analysis, and Law and Language. It does so, both in terms of the corpus linguistic methodology applied, and in terms of the critical stance taken to the language of law.
Critical discourse analysis (CDA) is a field of inquiry concerned with the interaction between language and society. Critical discourse analysts are generally interested in exposing latent ideologies and power asymmetries, and to this end, often work with institutional or other 'privileged' forms of discourse. In a list of principles governing CDA, field leaders Fairclough and Wodak posit that ''Discourse constitutes society and culture'', that ''Discourse does ideological work'' and also that ''Discourse is a form of social action'' [16: 258]. As language is not the sole preserve of linguists but rather a common thread of interest running through all of the humanities and social sciences, CDA has been successfully and often powerfully applied in fields outside of linguistics. It is our belief that each statement from Fairclough and Wodak [16: 258] remains powerfully true when we specify legal discourse as the object of inquiry: law constitutes society, does ideological work, and is a form of social action. However, from a CDA point of view, the legal field remains an under-researched area. As legal language is usually believed to be formulaic and routinized, with personal opinions of lawyers and judges neutralized and descriptions of law and fact objectified, it is often presumed that it would not be advantageous or particularly interesting to scrutinize it from a critical discourse perspective. However, the neutrality of legal language may be more a matter of belief and the ''law's desire to appear objective and authoritative'' [49: 76]. Further, the more personal language of judges in common law countries differs significantly from the more academic style and impersonal style of writing adopted in civil law countries. Nevertheless, only very few studies have contributed to the CDA agenda in law so far; the critical analysis by Kjaer and Palsbro [31] of Danish legal and media discourse on the European Court of Human Rights is among the few papers that apply an explicit CDA approach to legal discourse. This does not mean that critical views on language use in law are absent from legal research in general; within the neighbouring disciplines of socio-legal studies and legal anthropology, critical analyses of the use of language in law do exist, see e.g. the work by Mertz [39] on language ideology in American law schools, Conley and O'Barr [14] on powerful language in the legal process, and Goodridge [19] on legal discourse as a linguistics of legal power, i.e. in witness statements. While these take critical views of language of the law, they focus on oral texts (such as the language of participants in the courtroom, or professor/student interactions in the law school class-room), or treat legal language from a philosophical perspective. They do not critically analyse written texts produced by legal institutions themselves. This may be due in part to the sheer quantity of discourse produced by a court, particularly the international tribunals. To address this challenge, we adopt a mixed methodological approach: corpus-based critical discourse analysis.
Critical discourse analysis as a qualitative theory and analytical mindset has lately found a comfortable companion in quantitative corpus linguistic methods. ''Corpus linguistics sees language as a social phenomenon'' [48: 97] and offers toolkits for quantifying, visualising, and generalizing patterns of meaning in data, whereas critical discourse analysis can shape and inform analysis of these results. Computer-assisted methods in corpus linguistics allow for large-scale, systematic analyses of big data sets incorporating mathematical and statistical measures of language, and can be particularly helpful in reducing research bias and subjective slant by exposing patterns that would not necessarily be apparent to the qualitative research [7]. From the point of view of empirical legal and political studies, corpus linguistics is a novel research tool, but fits perfectly to the big-data research agenda of the field. Recent research [12,15,35,43,44] has applied computer-based network analysis of courts' citation patterns. Corpus linguistic methods have also been used to add important detail to the otherwise broad picture of judicial practice that network analysis gives [29]. Within the cross-disciplinary field of Law and Language (legal linguistics) empirical studies of legal texts are widespread, but the corpora traditionally consist of small text collections that are analysed manually: see e.g. the work on legal genres by Bhatia [8,9]. A methodological turn towards computer-based corpora and corpus linguistics is now increasingly detectable, especially within forensic linguistics [33,46], translation studies [10,11], and genre analysis [17,20,37,38]; the corpus-driven study by Kopaczyk [32] on the development of the legal language of Scottish Burghs should also be highlighted in this context. Most of that work, however, is for descriptive purposes only and lacks the critical view on language use in the legal field that is adopted in this paper.

Description of the Data
The United Nations has made a variety of resources publically available through the ICTY website. 2 These take the form (for instance) of feature-length documentaries, case summaries and individual dossiers. This is part of a concerted effort to make the Tribunal's actions known to society beyond legal professionals. Official documents created as part of the ICTY mandate-such as Annual Reports and judgements from the Trials and Appeals Chambers-are also posted online. It is on these last two text types that we base our analysis.
The main collection of texts used is drawn from the Trials and Appeals Chambers. These are available in PDF and OCR form, and though they were produced in both French and English, it is the latter that is used in this study. This is due largely to the English version's status as the authoritative version; so while ''[w]hile errors in errors of translation do occur, and smaller discrepancies between the texts abound, the parties and the public can always have recourse to the authoritative version in order to ascertain the exact meaning of a decision'' [1: 882]. In 2013, all available judgements were collected from this repository. Though they comprise a complete set (containing all judgements from a specific court), the documents are problematic for various reasons. The encoding of PDFs was not done rigorously or accurately, and special characters (for instance, Slavic accented letters) appear incorrectly both in the downloadable documents and in automatically converted text files. The texts converted through OCR are of variable quality, and noise in the headers and footers is particularly frequent. A series of programmatic cleansing scripts were devised and run to improve text quality as much as possible. 3 This resulted in a corpus exceeding 10.5 million words, in 71 texts from the Trials Chamber and an additional 50 texts from the Appeals Chamber-hereafter referred to as the Trials & Appeals corpus. This corpus represents what we might recognize as the discursive artefact of the main 'work' of the ICTY. It is important to bear in mind that though these texts represent one discourse of the ICTY, they are the work of a multitude of 'voices': three judges hear each case in the Trials Chambers; the Appeals Chamber is composed of five judges. Add to this the discourses of legal teams for the Prosecution and Defence, testimonies from victims and witnesses, and in the event of an appeal, statements from the Accused/ Appellant.
In order to triangulate an understanding of the Tribunal's 'voice' as a whole, we also incorporate a second data set: the Annual Reports corpus. The first difference is in the accessibility; though the Trials and Appeals Chambers files are public, they are difficult to find and extremely technical in nature, and real access to these is restricted to those with knowledge of international humanitarian law and technical savvy. Annual Reports-submitted to the UN Security Council and General Assembly under Article 34 of the Tribunal's Statute [27]-are more narrative in nature. Though they are clearly for a highly informed audience, the information contained is distilled to a level that is digestible by a general academic audience. More importantly, these Annual Reports offer significant insight into the changing emphases the Tribunal places upon its own work, and some information about the activities it would hope to be recognized by both the UN and the greater public. We collected reports from 1994 to 2013 and processed them into plain text as above. The resulting corpus contains 423,621 words.
The methods and tools used to analyse these two corpora are detailed below.

Methods and Tools
In order to illustrate the potential contributions of corpus-based critical discourse analysis to the study of legal language, we demonstrate features of two different corpus linguistic tools (SketchEngine 4 and Wmatrix 5 ) in carrying out a small range of methods: frequency, collocation, concordance, and key semantic tag analysis. SketchEngine was developed by a team led by Adam Kilgarriff, and is a powerful tool used mainly in lexicographical studies [30]. This corpus query system enables users to load their own data before automatically applying lemmatisation (grouping items by headword, e.g. run and running) and part-of-speech tagging (e.g. run_verb vs. run_noun). Frequency lists-or lists showing all words in a corpus and the number of times they occur-can be generated automatically, and may indicate interesting starting points for corpus-driven analysis of high-frequency items. Conversely, taking a corpusbased approach and searching for a specific item, users can build word sketches, or ''one-page automatic, corpus-based summaries of a word's […] behaviour'' [30: 105]. If words appear regularly in close context, and this co-occurrence is statistically significant, the items are said to be collocates. SketchEngine uses the association score logDice to calculate collocation on scalable corpus sizes; as a rough guide, a logDice score of 1 indicates that items collocate twice as often as might be expected, whereas a logDice score of 7 indicates 100 times frequent collocation.
Collocation is a central concept in corpus linguistics, and can be of special pertinence when considering legal discourse. Hunston and Francis [23: 270] argue that words have little meaning or ambiguous sense in isolation, and derive their meanings when occuring in particular phraseology. This is particularly true when considering the language of the law. For instance, responsibility may have one meaning (imbued with a multitude of folk understandings) in isolation, but in a set collocation phrase such as command responsibility, the meaning is dictated by context and jurisprudence.
Substitution for alternate collocates in these phrases dramatically shifts the meaning of the entire set, as we demonstrate in Sect. 4.3 below.
We make use of a second tool-Wmatrix-to exploit its unique interface with the UCREL Semantic Annotation System (USAS), which is not available in Sketch-Engine. USAS was developed at Lancaster University for automatic semantic tagging of input texts. The tagset comprises 21 major discourse fields and 232 further subdivisions corresponding loosely to the Longman Lexicon of Contemporary English [36]. This can be a helpful way of considering items within a corpus; as ''the semantic tags show semantic fields which group together word senses that are related by virtue of their being connected at some level of generality with the same mental concept … groups include not only synonyms and antonyms but also hypernyms and hyponyms'' [5: 1]. The existing USAS semantic lexicon is most often applied to general discourse, and we endeavoured to extend it in such a way that it would be better suited to analysis of legal language in general and to the ICTY in particular. Using a frequency list of items tagged Z99: UNMATCHED by USAS, manual tags were appended 6 to the 250 highest frequency items (occurring over 340 times in the Trials & Appeals corpus). All other expressions of particular legal interest (e.g. complicit) or appearing in standard English and not due to an OCR error (e.g. his/her) occurring at least three times in the corpus were also manually semantically tagged (semtagged). This brings the total addition to 1432 single words. An additional 150 multi-word expressions (e.g. joint criminal enterprise) were also semtagged.
Keyness is also calculated using Wmatrix. In corpus linguistics, key items in a corpus are usually calculated by comparing wordlists from one corpus (the 'target') to another corpus (the 'reference'). In Wmatrix, users may calculate key words, parts of speech, or semantic tags in a target corpus versus a reference corpus [41]. Users may either collect and load their own reference corpora, or make use one of the reference corpora pre-loaded in Wmatrix; we use the British National Corpus Written Informative Sampler, which is a large reference corpus of formal English, considered satisfactorily comparable to the ICTY. We consider positive key semantic tags, or those 'overused' in the target ICTY Trials and Appeals corpus, as opposed to negative domains are 'underused' in comparison to a reference corpus. This is measured using the log likelihood procedure [42], which demonstrates confidence of significance.
Throughout the analysis of the discursive construction of the achievements of ICTY, we hope to demonstrate how a variety of corpus linguistic tools or methods may be employed in the analysis of the language of the law. These are presented as potential components in larger studies both from linguistic or legal perspectives.

Analysis
In the sections below, we demonstrate three corpus linguistic methods and discuss feasibility of application of these methods in the analysis of legal language. In Sect. 4.1, we discuss frequency lists as a 'way in' to the corpus; in Sect. 4.2, we use collocation to trace the Tribunal's self-presentation of agency over time; in Sect. 4.3, collocation is used to explore phraseology; finally, in Sect. 4.4, we look at one key semantic tag to analyse construction of 'truth' in the ICTY.

Frequency of Human Actors
Frequency lists-or lists of all of the words, lemmas, or phrases of a certain length in a given corpus, alongside the frequency with which they occur-are well utilized as indicators for possible 'ways in' to the corpus. They are generally more useful for critical investigations when downsampled or sorted for some salient feature. For instance, we are interested in the ways that human actors are named and referred to throughout the proceedings of the ICTY. Though we have a list of the proper names and aliases of those involved, the attributes by which social actors are defined and grouped are often a telling feature of any text.
We have reviewed a frequency list of headwords and extracted the 20 most frequent ways of referring to social actors (human common nouns and attributive adjectives serving as identifiers) in the ICTY Judgements and Appeals. The most frequent of these deal with court proceedings: witness/Witness and prosecution, as well as Accused. Less prominent in frequency but more so in variety are references to position within the military hierarchy. These include: soldier, brigade, Staff (of the VP/MUP 7 etc.), police, commander, and member (of the VP/ MUP, various brigades, etc.). Lack of military affiliation is also present in nomination strategies, i.e. with civilian deaths contrasting to and resulting from military operations. More frequently, however, civilian populations are referred to using non-specific, generalised nomination strategies: man, person, people, and even population, indicating the genocidal scale of acts occurring. The importance of affiliation is also apparent in the appearance of group, which occurs both in reference to command groups undertaking destructive, violent military operations, but also to groups of villagers found to have been massacred by these forces. The ethnic nature of the conflict also comes into focus in viewing this list: Bosnian, Muslim, and Serb are all highly frequent nomination and/or attribution strategies. Ethnic/religious identity aside, only detainee and victim refer directly to those most affected by the conflict while also encoding their status into the nomination strategy (Table 1).
Considering the top item in the frequency list can also inform our analysis of the ICTY's stated achievements. The relatively high frequency of witness/Witnesshighlighting this as the most common type of social actor in the Trials and Appeals corpus-may align with ''Giving victims a voice'' [24], but only when this identity overlaps with the much less frequent victim. Contrast, for instance, line 1 (where the witness is also a victim) to line 2 (where basis knowledge of the existence of the sites is denied) excerpted from the same text, below. Only one of these (1) reflects the achievement reported by the ICTY.
1. One witness testified that she was taken out five times and raped and after each rape she was beaten. (IT-94-1-T) 2. Most witnesses for the Defence stated they had no knowledge of the existence of the camps, or if they did, they referred to them as ''collection centres''.
(IT-94-1-T) Though this paper is not a full social actor analysis, this is a rich avenue of further research. Future work may look at whether witnesses are represented directly using courtroom aliases (as in line 1), showing their direct participation in the ICTY, or whether 'witnesses' are described as being present during various events, but are not present at proceedings. This has an impact on achievement of representation of victims' voices.
We see then, how this sort of list may give some interesting 'ways in' to the data, but requires manual intervention with intent, e.g. a research question involving the quantification of certain types of nomination strategies. Another corpus method that might be more helpful is one that goes beyond sheer quantification to 'zoom in' on one social actor in particular and trace their actions (or 'agency') over time.

Key (Contrastive) Collocations Indicating Diachronic Shift
Many researchers are interested not just in the social actors, but also types of processes commonly occurring within a text; this may give rise to further analysis demonstrating, for example, agency of or impact upon groups of actors previously listed. We are interested not just in how often the Tribunal represents the voices and experiences of others, but also how they represent themselves as an institution. To do this, we turn briefly from the Trials and Appeals corpus to the Annual Reports corpus, which represents the unadulterated voice of the Tribunal, and which allows analysis of self-representation to the wider public. The Annual Reports corpus has the additional benefit of taking place across a diachrony, allowing us to examine any potential changes in the discourse of the ICTY from its inception toward its closure. To test for changes in discourse over time, we have split the Annual Reports into two subcorpora (1994-2003 vs. 2004-2013), which may then be compared to one another using forms of keyness analysis. SketchEngine allows for the generation of contrastive collocation tables, which indicate the strength of collocation relative to subcorpus. As one of the main functions of the Annual Reports-if not the primary goal-is to represent the ICTY's actions and activities in the past year, we have taken the Tribunal itself as our search node in this test analysis.
In Table 2, we reproduce the 'Subject of' section of the WordSketch of Tribunal. The items listed in Table 2 are actions undertaken in grammatical constructions where the Tribunal is the subject, occurring with greater-than-expected frequency as measured by the LogDice statistic. Items in italics indicate a preference for one subcorpus or the other, with strongest preference appearing at the top (for 1994-2003) and the bottom (for 2004-2013) of the table. Items which are not in italics share a more equal preference between the two subcorpora, and appear in the middle of the table, with two LogDice scores indicated.
We shall discuss in turn, those items associating with the Tribunal more strongly in the first half of the Annual Reports (1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003), then those with a stronger association with the Tribunal in the second half of the Annual Reports (2004-2013), before briefly touching upon those shared more equally between the two subcorpora. Illustrative examples (or concordance lines) are accompanied by the year of their corresponding annual report.

Agency in 1994-2003
Many of the items showing preference for the Tribunal in the first half of the Annual Reports can be clearly linked to the ICTY's early status, for instance: begin, become, and establish. The structure of the Tribunal itself is described repeatedly in concordance lines containing comprise, indicating lack of widespread familiarity with its makeup: 3. As is well known, the Tribunal comprises three organs: its judiciary, consisting of 11 Judges assigned to two Trial Chambers and one Appeals Chamber, the Office of the Prosecutor and the Registry. (1996) Even from the outset, the Tribunal constructs itself prominently and positively, establishing bureaucratic assemblies (''an Inter-sessional Working Group'', ''a working group'', ''a witness protection programme'') and judicial processes (''a code of international criminal procedure''), but also social connections (''valuable contacts'') and its own existence and reputation (''itself as a fully operational international criminal court'', ''its credibility''). The starts of processes are likewise positively evaluated, as in line 4 below, where the Tribunal claims that it dispenses justice and achieves tangible results. Towards the end of the first half of the Annual Reports, the Tribunal reflects instead on what it has become (e.g. in line 5), though evaluative lexis remain positive, reinforcing selfconstruction as important and efficient: the ICTY is ''fully functioning'', providing fair trials while maintaining protection for vulnerable parties. 4. Through the development and application of concrete procedures, the Tribunal is beginning to dispense that justice, achieving tangible results for both victims and accused. (1998) 5. During the reporting period the Tribunal has become a fully functioning international criminal court, providing fair trials to the accused, while maintaining a high degree of protection for victims and witnesses. (1999) Despite this positive self-presentation as playing an active and important role in international humanitarian jurisprudence, the verbs appearing in and preferring the first half of the Annual Reports do not reflect such unbridled success in the actual work of the court. Firstly, only two (indict and bring, as in bring justice) relate to the central, judicial work of the Tribunal. In viewing the concordance lines, these are also presented with reservations, as the early Tribunal encountered problems in meeting these goals: 6. That mandate has not yet been properly fulfilled because the vast majority of persons indicted by the Tribunal are still at liberty, ignoring their indictments with seeming impunity. (1997) Indeed, in the 1994-2003 subcorpus, it is difficult to determine exactly which processes the Tribunal is (successfully) undertaking in regards to jurisprudence. In six out of the seven key instances of do, this item is followed by not. In the single instance not followed by negation, the clause is completed with the foil ''undone'' (see line 7 below), again underscoring its own lack of agency or efficacy. Here This strikes us as unusual given that the texts are designed to summarise the (judicial) activities of the Tribunal in the previous year; it seems that it would behove the ICTY to stress material/legal processes. Instead, we observe a rather unexpected number of mental/behavioural processes. In concordance lines where the Tribunal is the agency of expecting, these expectations are of States (80 %) and NGOs (20 %) cooperating fully with judicial principle and process by participating without influence, and assisting in the fulfilment of shortcomings, e.g. of facilities and services (see line 9 below). Beyond expectation, the Tribunal also states plainly that it relies 'heavily' on this State Support, particularly as ICTY workload increases (line 10). It is made clear in the Annual Reports that success is contingent upon cooperation and effort on the parts of various other parties.
Constructing Achievement in the International Criminal… 537 9. Lacking an incarceration facility, the Tribunal expects States to provide facilities to imprison persons whom the Tribunal convicts. (1998) 10. The Tribunal relies heavily not just on the cooperation of States of the former Yugoslavia but on all States for its daily operations and it proceeds under the assumption that States will provide their full and unreserved support. (1997) The Tribunal enjoys many forms of cooperation and support (e.g. ''immunity'', ''privileges'', ''the usual exemptions'', ''the support of the United Nations'', ''a high degree of administrative support'') and welcomes further assistance in matters both legal (e.g. ''surrenders'', ''statements'', ''the establishment of a truth and reconciliation commission in Bosnia and Herzegovina'') and administrative (''two additional judges''). Towards the latter portion of the first half of the Annual Reports, the Tribunal constructs itself as reaching landmarks of efficacy and recognition: 11. After 10 years, the Tribunal has reached a point of some institutional maturity, as the events of the past year demonstrate. (2003) However, this trajectory does not continue in agency expressed between 2004 and 2013. We discuss the radically differing self-construction of the Tribunal below.

Agency in 2004-2013
In contrast to the early subcorpus (of which the Tribunal was construed as the agent of a multitude of processes), collocates in 2004-2013 are few, including only host, approach, and conclude. Once more, these have surprisingly little relation to the judicial processes expected from the texts; host, for instance, refers to the ICTY hosting bureaucratic/social events, such as ''visits'', ''conferences'', and ''ceremonies''. As the Tribunal approaches the completion of its work, the Annual Reports contain frequent references to the number of cases concluded (as in line 12 below). This is the most transparent representation of the Tribunal as an active agent of legal change and due process found in this subcorpus. We move now to consider processes of which the Tribunal is an agent with relative equity across both subcorpora. Many of these are common auxiliary verbs (i.e. be, take, and have). However, further analysis into open class verbs allows us a view of similarity in the corpora, after distinguishing differences. A greater degree of agency is discovered in verbs sharing near preference between the early subcorpus and the late subcorpus of Annual Reports. Of particular interest are those related to the law: issue (''orders'', ''directives'', and ''warrants'') and adopt (''rules of procedure and evidence'', ''a largely adversarial approach to its procedures'', ''an amendment to rules 72 and 73 of the Rules of Procedure and Evidence'').
A continuation of the court's 1994-2003 self-representation as an entity reliant upon outside support is echoed here. The Tribunal describes itself as receiving administrative and financial assistance (''support'', ''donations''), as well as staff and facilities (''seconded personnel'', ''two servers and 50 network computers'').
These findings can be demonstrated in microcosm when viewing concordance lines for continue. In all of its ongoing work, the Tribunal describes itself as continuing to benefit from the support of others (e.g. line 13) and to achieve important, tangible results (see line 14). However, the process has not been without significant difficulties, which have hindered the court from operating at the level of efficacy it had hoped or expected to achieve (line 15).
13. Throughout the reporting period, the Tribunal has continued to benefit from the services of gratis personnel, that is, personnel provided at no cost to the United Nations by donor Governments or non-governmental organizations.
(1997) 14. Notwithstanding periodic setbacks resulting from obstructionism by some States, the Tribunal continued to achieve tangible results. (1999) 15. The Tribunal continued to encounter certain difficulties with respect to the application and interpretation of its headquarters agreement, specifically in relation to the privileges and immunities that judges and staff members receive in comparison to those working for other international organizations. (2003) Therefore, we find that both the early and late subcorpora, as well as the processes in between, depict self-representation strategies focussing more on stating the importance of the ICTY, rather than demonstrating it by detailing action. This was visible in the early subcorpus (where the process of establishing the court was more in focus than the processes taking place inside it) as well as the late subcorpus (which contained little agency to speak of). The achievements of ''Developing international law'' and ''Strengthening the rule of the law'' are difficult to uncover in the very public discourse of the ICTY, and this is a problem that is exacerbated as time moves along.
To take another view of how the ICTY might construct their contribution to ''Developing international law'', we move back to the Trials and Appeals corpus to analyse a different achievement: ''Holding leaders accountable'' [24].

Variations in Phraseology: The Case of Responsibility
One powerful way to use corpus linguistic methods (which might be of particular interest to those investigating the language of the law) is to investigate variations in phraseology. Perceived lack of variation-or firm preference for formulaic phraseology-is a reason that many scholars do not consider legal texts to be rich fodder for deep linguistic analysis. However, in the example below, we will demonstrate that variation does exist, and that this can be (critically) meaningful.
The development of international humanitarian law is seen (at least by the Tribunal itself) as one of the ICTY's main achievements [26]. Specifically, the delineation and distribution of accountability-encompassing all levels of the social-military scale from citizen-soldiers up to heads of state, and allowing for both individual and joint responsibility-was to be a defining feature of the Tribunal. At the outset of the ICTY, this was pursued through the use of the mechanism of 'command responsibility' or 'superior responsibility', through which those in superior (military) positions are held criminally responsible for failing to adequately punish or prevent crimes committed by persons under their command or authority. Descriptions of individual responsibility and command responsibility can be found in Article 7(1) and Article 7(3) of the ICTY Statute, respectively [27]: Article 7: Individual criminal responsibility 7(1) A person who planned, instigated, ordered, committed or otherwise aided and abetted in the planning, preparation or execution of a crime…shall be individually responsible for the crime… 7(3) The fact that any of the acts…was committed by a subordinate does not relieve his superior of criminal responsibility if he knew or had reason to know that the subordinate was about to commit such acts or had done so and the superior failed to take the necessary and reasonable measures to prevent such acts or to punish the perpetrators thereof.
In the later years of the Tribunal, the concept of 'joint criminal enterprise' has overtaken command/superior responsibility in ICTY jurisprudence (see Ambos [2] for a legal analysis of the relationship between these forms of responsibility and additional explanation for change in usage over time). However, in the Tribunal document detailing its accomplishments, the ICTY states that application of Article 7(3) specifically 'has removed uncertainty' about distinctions of criminal responsibility in a war-time environment [26: 5]: The Tribunal has applied the modern doctrine of criminal responsibility of superiors, so-called command responsibility. It has clarified that a formal superior-subordinate relationship is not necessarily required for criminal responsibility. In the same vein, the Tribunal has removed uncertainty about the level of knowledge to be expected from a superior whose subordinates were about to commit crimes or actually committed them […] However, we will demonstrate here that a level of uncertainty about the boundary between individual and command responsibility perpetuates in the ICTY documentation, in part due to mixed terminology employed throughout the cases. This has potential consequences for the claimed achievements of ''Holding leaders accountable'' and ''Developing international law''.
In Table 3, we have replicated a section of the WordSketch for responsibility in the Trials & Appeals corpus. These 'modifiers' give an insight into the compound nouns (or n-grams) comprising responsibility, occurring with unusual frequency. The highest-ranked of these feature expected results, those endorsed and contained within the ICTY Statute: individual, superior, command, and (joint) criminal (enterprise). Yet beyond these results, we see great variation in near-synonyms: direct, personal, ultimate, overall, full, primary, and immediate. The meanings of these will be explored in greater depth below to establish some understanding of the (legal) ramifications of their use in context.

Items Indicating Singular Responsibility
The official ICTY term under Article 7(1) for direct, singular involvement in wartime crime is individual responsibility. This is the most frequent iteration (occurring 842 times) and the most statistically strong association (with a score of 11.82). Two alternatives also appear on the WordSketch: direct and personal.
Concordance lines featuring the collocation between direct and responsibility show a clear synonymy with individual responsibility. In the majority of these 49 cases, Article 7(1)-which refers to individual responsibility-is clearly referenced. Arising from this alternative is an additional complementary phrase: indirect responsibility. This corresponds to Article 7(3). An example of these items in use can be found in Line 16 below.

In general, Article 7(1) concerns the accused's direct responsibility while Article 7(3) deals with his indirect responsibility. (IT-05-87-T)
While direct responsibility does illustrate variation in use of legal terminology, the more interesting collocate of singular responsibility is personal. This is due to differing (and distinctive) uses in the voices of the accused, appellants, and of the Tribunal itself. The appearance of personal responsibility as a report of the appellant's statements is an almost flawless predictor of text type. In all six cases where this appears in the 'voice' of the Accused alone (i.e. without the secondary indirect influence of the Tribunal), it occurs in a plea bargain. Further, in each of these six cases, the acknowledgement of personal responsibility is accompanied by moral/ emotional discourse indicating remorse (see line 18) and self-confessed wrongdoing (see line 19). 18. Furthermore, where the plea, and the circumstances in which it came to be made, involves a profound acknowledgement of personal responsibility, it may demonstrate that an accused is genuinely remorseful. (IT-95-17-S) 19. Bralo has not, however, alleged any form of duress emanating from his superiors such that he was compelled to commit the crimes of which he has been convicted…He has taken full personal responsibility for those crimes and has acknowledged that he knew them to be wrong. (IT-95-17-S) With nearly perfect precision, the use of personal responsibility by the Accused will indicate that the text type is a plea; indeed, in only one instance is this phrase used in a plea in the voice of the Tribunal. However, the concept of remorse appears immediately afterwards in this case: 20. In the Todorović case, it was stated that: ''In order to accept remorse as a mitigating circumstance, the Trial Chamber must be satisfied that the expressed remorse is sincere.'' In this regard, the Chamber takes account of Duško Sikirica's statement during the Sentencing Hearing, in which he said: ''I deeply regret everything that happened in Keraterm while I was there. I feel only regret for all the lives that have been lost and the lives that were damaged in Prijedor, in Keraterm, and unfortunately, I contributed to the destruction of these lives.'' (IT-95-8-S) When used by the Accused, the declaration of personal responsibility holds the legal ramifications of individual responsibility under Article 7(1), but also carries an emotional element that seems to have been incorporated to improve the possibility of reduced sentencing with demonstrated remorse during the plea. This emotive component is unique of individual (specifically personal) responsibility, and does not seem to appear in concordance lines arising from collocates indicating 'group' responsibility, discussed in Sect. 4.3.2 below.

Items Indicating Collective Responsibility
Based upon Article 7(3) of the ICTY Statute [27], expected (and endorsed) terminology to indicate criminal responsibility of the known actions of a group of subordinates would include superior and command responsibility. Once more, these are the most common, occurring 496 and 450 times, respectively. As an aside, it is interesting to note the near parity in these frequencies. The inclusion of two termswhose distinction may have been meaningful in other courts (i.e. post-WWII)used simultaneously and/or interchangeably is a contributing factor to imprecision and ambiguity in the Statute and the proceedings. This is further aggravated by the repeated use of two additional terms: ultimate responsibility and overall responsibility.
Both forms are relatively infrequent, likely owing to the existence of highly frequent, endorsed alternatives: superior, command, and joint criminal enterprise. Ultimate responsibility collocates 13 times, and overall responsibility collocates eight times. These instances cannot be linked as neatly to the existing Statutes as above. Rather, inspection of the concordance lines must be undertaken to determine that ultimate responsibility and overall responsibility are being taken for the actions of (military) subordinates or as acting (military) superior. 21. The Prosecution argues that by virtue of this position, ultimate responsibility for the conduct of ABiH soldiers rested with Rasim Delić and that he had more power than any other person in the ABiH to ensure that his subordinates were punished for their misdeeds and prevented from perpetrating other criminal conducts. (IT-04-83) As evidenced in line 22 above, these uses are cause for ambiguity in the court system themselves. Use of the phrase overall responsibility in the Trial Chamber is revisited in the Appeals Chamber when the Appellant's role as deputy commander is questioned. As this does not directly connect to the Statute through shared terminology or direct reference to the Article 7(3), there is some obscurity in the wording of the Tribunal. In the next section, we discuss three further collocates with ambiguous or contradictory meanings in context.

Ambiguity in Responsibility
Given the preponderance of terminology around responsibility, it is nearly inevitable that cases of irreconcilable ambiguity should arise.
When primary and responsibility collocate, the resulting meaning could be synonymous with command responsibility, but seems to relate more closely with the folk (or common) meanings associated with principal functions. In concordance lines 23 and 24 below, we see two such examples. In line 23, it is Mucić (as a commander) who is found by the Trial Chamber to have had primary responsibility for civilian detention; this is defined by terms similar to those laid out in Article 7(3), but is not linked to criminal enterprise. In line 24, it is the Trial Chamber itself who declares its own primary responsibility (for evaluating evidence), quite outside of the established Statutes of international humanitarian law. This instance is quite similar in form to the final collocate in the ambiguous set: immediate. Each of the 11 instances of the collocation between immediate and responsibility occur in the Trial and Appeal of Veselin Š ljivančanin. This term is actively problematized by Š ljivančanin, who reduces the accusation of immediate responsibility to a literal reading of personal accountability for acts perpetrated upon prisoners of war. This misinterpretation stems from the failure of the Trials Chamber to utilize the form command responsibility, with its legal distinction inclusive of his position over other soldiers.
28. The circumstances of his conduct which led to his conviction have been identified. In particular, they reveal a failure to act to protect from severe criminal abuse the prisoners of war who were his immediate responsibility.
(IT-95-13/l-ES) 29. The Prosecution responds that Š ljivančanin fails to show how the Trial Chamber's use of the words ''immediate responsibility'' is ''in any way discordant with its factual findings'' regarding his responsibility for the tortures of the prisoners of war at Ovčara and that his allegation that other JNA soldiers had some responsibility over the prisoners is irrelevant. (IT-95-13/1-A) Though it is clear that the ICTY is working under extreme pressure, the erosion of terminology established in jurisprudence is demonstrated to have led to ambiguity in what, exactly, the Accused are acknowledging or admitting to, as well as inhibiting understanding of case basis of the Prosecution. This tarnishes the achievements of ''Holding leaders accountable'' by obscuring accountability and fails in ''Developing international law'' by contradicting terminology laid out in previous jurisprudence.

Key Semantic Domain Analysis
Wmatrix offers an easy-to-use interface allowing researchers to calculate key semantic tags in their data compared to a reference corpus. The very high preference for the G2.1 tag: LAW AND ORDER-appearing in the first position, indicating the lowest p value-is reassuring; this is precisely the sort of finding to be expected when comparing a corpus of specialized legal language to one of general written informative  (Table 4). By way of example we will analyse the language of one semantic domain, A5.2?: EVALUATION: TRUE, listed in the seventh position on the key semantic tag list. It is not surprising to find it among the most dominant semantic domains in the discourse of a criminal court. 'Finding the truth' is what criminal courts are supposed to do, and ''Establishing the facts'' is one of the Tribunal's own advertised achievements. Indeed, ''legal discourse is paradigmatically concerned with truth, both in terms of evidence or verification, and also, more generally, in terms of the definition or delimitation of power and powers in the discourse of the rights, duties, capacities and procedural forms generally of both public and private law'' [19: 192]. However, methods of discursively constructing this 'truth' are of interest to both linguists and lawyers.
The automated calculation of key concepts in the Trials and Appeals corpusparticularly when presented in semantic categories-reveals interesting patterns in the language of the court in this domain. Thus, when one looks more closely into the words and phrases chosen by the court to express its evaluation of the 'truth', the limited lexical variability is striking. Only few terms are applied, and the exact same phrases are reproduced repeatedly. Even more surprising is the fact that truth is not among the most frequently used words in the semantic field, even though ascertaining the truth of what happened-and who did it-is the main task of a criminal court.
With 37,372 occurrences (3533.4 per million words), evidence tops the list in terms of frequency, followed by fact (freq. 11,457 or 1083.2/million), prove (freq. 3015 or 285.1/million), factual (freq. 1916 or 181.2/million), and proof (freq. 1434 or 135.6/million). In comparison, the word truth amounts to only 358 instances, including its use in non-legal phrases such as 'in truth'. This seems to indicate reluctance on the part of the tribunal to describe what it is doing in terms of discovering the 'truth', let alone declaring universal truth. However, the relatively few instances of truth in the language of the judgements give a clear picture of how the Tribunal constructs its role as a legal institution that must meet the evidentiary standards of a criminal court in the specific context of the war crimes committed in the former Yugoslavia. We can 'drill down' further by investigating the constituent word forms semantically tagged as A5.2?: TRUE. The concordance lines below were discovered by using this method. In line 28, three A5.2? semtagged items (evidence, truth, facts) appear. We can see the standard narrative of the role that evidence plays in a criminal trial, which is to assist in ascertaining ''the truth of the facts'': 28. Every criminal trial involves two issues: first, that the crimes charged have been committed and, second, that an accused is responsible for those crimes.
The object of evidence is to ascertain the truth of the facts with respect to these two issues… (IT-99-36-T) However, in the specific context of the ICTY, other goals must also be achieved: to put an end to war crimes, to take effective measures to bring to justice the persons who are responsible for them, and to contribute to the restoration and maintenance of peace [27]. Like all criminal courts, expediency if a requirement of fair trial rights [21,22]. The task of ascertaining the truth is therefore limited by the need to handle the cases efficiently: 29. The Appeals Chamber recalls that the right to cross-examine witnesses is a fundamental right […] Relevant to the exercise of this right is the trial chamber's duty to exercise control over the mode and order of witness examination so that it facilitates the ''ascertainment of truth'' and avoids ''needless consumption of time.'' (IT-05-87-A, with reference to Rule 90(F) of the Rules of Procedure) A manual analysis of the concordances in which truth is embedded confirms the impression given by its underuse that the word does not belong to the standard vocabulary of the tribunal. It turns out that truth is unevenly spread across the case law of the tribunal and occurs only in a limited number of specific contexts. Indeed, the tribunal tends to use the word truth only when arguing for the mitigating effect of guilty pleas; see line 30 for an extract from the tribunal's judgement in an early case. The comparatively low frequency of truth indicates that there is a tension between the legal language of a criminal court and the language of a war crimes tribunal whose task is not restricted to establishing whether or not the Accused is guilty of the crimes committed or not. As it highlights itself, the Tribunal's mandate is to contribute to establishing the truth about the conflict. However, as a court of justice it can do so only in the context of fact-finding that may prove or fail to prove the guilt of the Accused. Other textual features identified by the automated corpus analysis supports this finding.
The widespread use of particular phrases repeated as formula reveal a considerable degree of standardization of the juridical language applied in this domain. While some of the phrases are general criminal legal language formula (e.g. ''The Prosecution has proved/failed to prove''), others are specific to the language of evidence created and reproduced by the ICTY. As such, they express the Tribunal's construction of the procedural steps to be taken in order to establish the truth. The concept of 'standard of unreasonableness' is a case in point. The following are quotations from the case law of the Appeals Chamber in which the standard was developed and established. Again, the specific texts are identified through the concordance lines in which the high frequency words belonging to the semantic domain A5. Investigating the concordance lines of fact, factual and evidence uncovers the frequent co-textual use of the phrase reasonable trier of fact. A word search shows that it has been used 1127 times in the full Trials and Appeals corpus, making it almost as standardized and repetitive in the language of the Tribunal as (prove) beyond reasonable doubt (used app. 3000 times) with which it often co-occurs (linguistically) and which it modifies (conceptually).
Use of methods from corpus linguistics can also expose interesting deviations from the norm that may not have been accessible to the reader. In establishing the normal (high frequency) constructions of fact in the key A5.2? semantic tag and examining the surrounding context, we discover an interesting 'slip of the tongue'. The use of a problematic alternative phrase-''prove as best as it can''-also confirms that truth is not the part of the vocabulary of evidence. It is a non-technical expression which is used by the court in a Judgement only once, even if it is probably more in accordance with the actual difficulties of fact-finding in the context of a war crime trial indicated by the legal standard ''prove beyond reasonable doubt'' (used in 575 instances out of a total of 3015 occurrences of prove).

Concerning Martić's allegation that Judge Moloto's comment ''that the
Prosecution must proceed and prove its case as best it can'' revealed the Trial Chamber's (mis)understanding of the ''beyond reasonable doubt'' standard, the Appeals Chamber notes the following. As the parties had not reached an agreement on certain facts, the Presiding Judge simply remarked that the only way to proceed was to let the Prosecution prove its case ''as best as it can''. In doing so, he did not articulate the standard of proof to be employed by the Trial Chamber, but simply recalled that the burden of proof rests on the Prosecution. (IT-95-11-A) The formulaic language used in this particular semantic domain is revealing of the way the ICTY constructs evidence and truth. The legal effect following directly from the text of the judgements makes judicial language in the field of criminal law an obvious object of routinization and ritualization. Is the Accused guilty or not guilty of the crimes with which he or she is charged? The Tribunal's answer to that question has immense legal and human consequences. This is why the court cannot conclude that the Prosecution has ''proved as best as it could'' that the Accused is guilty. The standard formula prove beyond reasonable doubt as well as the ICTYcreated standard of no reasonable trier of fact could have reached a different conclusion make it possible for the Tribunal to speak authoritatively about the 'facts' of the case, without referring directly to 'truth' and without casting doubt on the possibility of ''Establishing the facts'' with straightforward modifications like prove as best as it can.
''Establishing the facts'' is one of the self-promoted achievements of the ICTY. The key semantic domain analysis confirms that fact-finding is indeed one of the core topics of the judgments. Of course, this is not a surprising result, but variation in semantic construction of 'fact' and preference shown therein can be telling. The complex historical and political context in which the Tribunal has acted is mirrored in the language adopted in the judgments. Of particular interest is the fact that the word truth is almost exclusively used by the court in arguments concerning guilty pleas, while the Tribunal is otherwise reluctant to address its fact-finding mission in terms of 'truth'. What emerges is an interesting built-in conflict between the public and political expectations of the ICTY and of the way it actually fulfills its roles, constrained as it is and was by the difficulties of ascertaining evidence of events, the time pressure under which the Tribunal has had to work, and the complex relationship between the desire to bring justice to the victims and the need to meet the standards of the rule of law including giving the accused a fair trial; cf. the legal debate on the different roles of Courts of Law and Truth Commissions [18,45].

Concluding Remarks
The purposes of this paper have been twofold: to demonstrate how corpus linguistics methods can contribute to analysis of the language of the law; and to explore how the ICTY's own stated 'achievements' are discursively manifested and constructed in two collections of texts created by the Tribunal: the authoritative texts of its trial and appeal judgements as well as the informative annual reports. We believe that some interesting results and some limitations have come to light in approaching both of these goals, and we discuss them in turn below.

Considerations and Limitations
Law is performed in language, but to analyse a Tribunal through the texts it generates is to treat discourse as an artefact. In this study, we have demonstrated several methods from corpus linguistics and triangulated the various ways in which the ICTY functions (e.g. as a finder of facts) by looking at the textual evidence it leaves behind. Because legal language must be so precise, this sort of study is better suited than most to treat discourse as an artefact. But this still cannot fully bring out the intricacies of the actual court in action. For instance, in order to establish the 'fact' of conviction, two out of three judges must reach an agreement; this means that one judge may not agree that the 'truth' has been established, or that the document created necessarily reflects its entirety. We have also considered discourse as a process, where the Tribunal constructs itself and shapes the law over a diachrony. In order to do this, we must have a good awareness of the powers and tensions that occur outside the texts but are reflected inside of it.
There are a number of additional considerations when taking a corpus linguistic approach to critical discourse analysis; scholars in the field have acknowledged a number of limitations. The most pertinent for this work is that sections of the corpus are largely decontextualized: while law scholars may find it critical to know, for instance, where on the page (e.g. which section or paragraph) a certain argument appears or case is cited, most corpus processing programs treat constituent texts (e.g. judgements) as a whole. In the case of the ICTY, we must also be aware of the language issues. The Tribunal has two official languages; the English version of the texts that we have analysed is, in some cases, translated from a French original text. Moreover, the many participants in the Tribunal (especially witnesses and accused) speak a number of languages that are all translated into English and French both in the court and for the official documentation. 8 Though we take this at face value, another interesting avenue of research may be in taking a critical view of these translations as a potential source for flattening or increasing variation in phraseology, as well as identifying mistranslations and aspects of evidence which are lost in translation.
These limitations might be addressed in future work, particularly if scholars are inclined to take a corpus linguistic approach. Corpora can be annotated with metadata such as time period, the makeup of the court, dissenting judges, paragraphs and sections. Citations and social actors may be cross-referenced across various texts. This is time-consuming but can be rewarding if research questions deal with issues such as time series analysis, argumentation, disagreement, citations, and document structure.
In our study, cleansing of the texts and removal of the footnotes was considered sufficient. However, the process of finding, downloading, cleaning, and loading texts into programs could have been significantly aided by the ICTY and the UN itself, if data legacy were added to the considerations of a court. Though posting PDFs online is a step in the right direction, we believe that a real contribution to the people of the region and to scholars in the field would be the development of a queryable portal that would make the work of the court more accessible and transparent. This would aid further work as well as replicability.

Discussion
The Tribunal has had to navigate in the midst of competing expectations of its role and functioning, while especially living up to the mandate given to it by the UN Security Council. A close-up view of the way it uses language gives the researcher a unique chance to see how the Tribunal handles this mandate.
By utilising a variety of methods from corpus linguistics and critically analysing the results, we have discovered discursive remnants of the Tribunal's stated 'achievements', but also uncovered problematic language use that raises doubts about the court's operation and its contribution to the international legal community-or at least documents the difficulties that the ICTY has faced. In legal terms, it was the first war crimes tribunal after the Nuremburg and Tokyo trials, and as such had to develop modern international criminal law almost anew, or in other words to make ''creative use'' of the sources of international law, which is seen by some as unjustifiable in criminal law; see in this regard Swart [47]. In political terms, it has had to cope with the dual pressures exerted at both the international and national levels. The UN mandate in itself has not been a sufficient basis for it to claim its authority vis-à-vis the population in the former Yugoslavia. At the same time, the tribunal has had to deliver fast and tangible results in terms of convictions of responsible leaders. The instable terminology that we have documented for the different categories of the concept of responsibility reflects the immaturity of the court's case law in this regard.
Future corpus linguistic research may elucidate the further development of the concept of responsibility, including the contested ''joint criminal enterprise''. This may be done by including the case law of other criminal courts, both the other ad hoc tribunals (the ICTR and the Special Tribunal for Sierra Leone) and the permanent criminal court ICC. Thus, corpus linguistic can add a discourse perspective to legal studies of responsibility in international criminal law (for an overview of both history and development of international criminal law, see e.g. Ambos [3,4]).
Use of the most basic tool of corpus linguistics-frequency analysis-does indicate that the Tribunal discursively encodes one of its 'achievements', that of ''Giving victims a voice''. Witnesses who are also victims appear regularly in the texts (see Sect. 4.1), and they are directly quoted in evidence, which is an empowering role. However, the Tribunal has not been as successful in demonstrating its own agency. Using collocation analysis, we found that the Tribunal is markedly less active over time in presenting itself in Annual Reports. As an agent, the ICTY is occupied by the work of self-creation in the first half of its mandate, and appears side-lined by administrative tasks in its later years (Sect. 4.2). The Tribunal's importance is stated rather than demonstrated, and in this way, the achievements of ''Developing international law'' and ''Strengthening the rule of the law'' are dubiously portrayed in reports circulated to the UN and to the greater public.
In another view of ''Developing international law'', we analysed one way in which the ICTY strode towards ''Holding leaders accountable'' (Sect. 4.3). While terminology was asserted in jurisprudence, the court was seen to deviate from phrases of accountability, resulting in confusion, appeal, and general erosion of the contribution to international law. A similar pattern was exposed in key semantic domain analysis of A5.2?: EVALUATION: TRUE. The word truth is dispreferred by the ICTY, as it does not belong to the standard vocabulary of legal procedure. However, even in ''Establishing the facts'', 'slips of the tongue' can derail the court and may cause scepticism about effective fact-finding and therefore, appropriate administration of justice (Sect. 4.4). In addition to this, despite the importance of the truthtelling function of the court, ''[i]nformation about the trials in general has been poorly disseminated: To the extent that peoples in the former Yugoslavia are denied access to the proceedings of the ICTY, the truth exposed through the judicial process may have no appreciable impact on interethnic reconciliation'' [40: 87].
In general, formulae in legal language are used for reasons of language economy and efficiency, or for reasons of institutionalization and bureaucratization, where personal emotions or even doubt are hidden behind an objectified, neutralized and standardized language that does not leave room for self-reflection or individual concerns. In the context of war crime tribunals, this laconic, formulaic legal language seems inadequate in relation to the emotions and circumstances that made people commit crimes against humanity in the war-torn communities and the immense quest for redress among the victims of the mass atrocities.
Nonetheless, the achievements of the ICTY are encoded and performed in language, and some forms more clearly serve the mandate than others. Once formulaic language is established in the jurisprudence, deviations from this language damage the integrity of the court and erode its legacy. The ways in which social actors (including witnesses and even the Tribunal itself) are discursively constructed offer interesting insights into the ways that these roles are conceived and enacted. The ICTY's failure to portray itself as active and focussed in the Annual Reports and Trials and Appeals Chambers is iterated in its relative inactivity (overrunning) and lack of focus (straying from and being drawn back towards its mandate) outside of the texts. This has proven an interesting exercise in triangulating representation using multiple methods and data sets, and indicates a rich area for further exploration. like to thank Ioannis Panagis, Sigrun Valderhaug Larsen, Kerstin Carlsson, and Matt Fisher for their contributions to this work.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.