PENETRATING HISTORICAL DISCOURSE’S TRUTH MATRIX: A CORPUS ANALYSIS OF ORAL HISTORY TESTIMONIES

Historical Discourse’s Truth Matrix was first posited by Michel Foucault to describe the emergence of a discourse of historical events subsequent to the cessation of war and established by the most powerful arbiters of those events. This paper adapts the implements of Foucault’s toolbox to conceptualise the dimensions of subjectivity that historical events pass through from the original event to their subsequent depictions in historical writing or other media. The Corpus of Irish Historical Narratives (COIHN) is a one-million-word representative corpus of oralhistory witness testimonies taken from the Bureau of Military History Archive (BMH) which depict the events surrounding the Irish struggle for independence from the lead up to the 1916 rising to the signing of the Anglo-Irish treaty in 1921. The archive provides a rich source of both historical, social and cultural data and has been the source of many publications pertaining to events of this pivotal time in Irish history. In ceding some agency to the data itself, this study takes a hybrid corpusdriven and corpus-based approach to the analysis of epistemic modality in these statements through hedged expressions using mental process verbs and so offers a framework for the analysis of a core dimension of historical discourse’s truth matrix. The notion of ‘paradoxical authority’ is proposed, stating that the authority of a witness is boosted rather than mitigated by expressing weak commitment to the truth of a proposition.


Figure 1: Dimensions of historical discourse's truth matrix
This study focuses on the dimensions above which can be investigated through the utilisation of corpus tools to assess the extent to which truth is supported and committed to in such texts through the usage of devices which are proxies for evidence and epistemic modality (Eder, 2020). As Cubitt (2013, p. 208) states, 'The memory of an event or a historical experience begins with the event or experience itself.' This reality is the truest form of the event. Subsequent to that event exists its 'afterlife', where multiple iterations add dimensions of subjectivity, creating a 'truth vulnerability' (Rath et al., 2019) whereby reality is distanced from the narrative, these dimensions produce historical discourse's truth matrix. Inherent in each iteration is an added layer of subjectivity that is a by-product of narrative (Henwood et al., 2011), thus the preliminary interpretation of the initial event grows in import as the dominant influence on the transition from reality to history (Carr, 2016). In an archive of witness testimonies such as that investigated here, where multiple witnesses give their versions of the same events, 'competing narratives' (Golubeva, 2010) emerge that may be in conflict with each other, further blurring the possibility of a true depiction. A return to the closest proximal interpretation of the event can provide a means of reclaiming the historicity of those events (Margolis, 2016), to attempt to penetrate the dimensions of historical discourse's truth matrix. In the case of historical archives such as that investigated in this study, this entails accessing testimonies of those who witnessed and took part in the historical events.

2. The Bureau of Military History
The BMH, established in 1947 by the then Minister of Defence was an ambitious project of national importance which sought to accumulate as much information as possible about the events surrounding the Irish independence movement from witnesses and participants of those events (Doyle et al., 2002). Over the subsequent ten years, 1,773 witness statements as well as letters, photographs and press cuttings were collected by the bureau's staff, which was composed of senior army personnel as well as civilians. The staff were given training and strict instructions about how to gain an objective and accurate account from witnesses as well as a chronological account of the historical events that were to be the subjects of the statements (Ryan, 2014). Specific instructions were delivered to investigating officers pertaining to the taking of evidence, the disclosure of information, communication to the press, accuracy and literary style of witness statements. Guides were given as to how to avoid the pitfalls that are commonly cited criticisms of oral history testimonies; subjectivity and the influence of other sources such as the media, collective memory and failing memory, as is evidenced in the following instruction from the bureau's records: In listening to and recording his story you should keep an open mind. Your aim at all times must be to get from him an objective factual record of events based on his own experiences. To that end he should be tactfully questioned on every point, to ensure that what he tells is, in fact, what he knows and not something which he has imagined, read or heard from someone else. Failing memory will sometimes impart an air of unreliability to what may be a genuine story, and the utmost care must be exercised in such cases. (File S. 851, 10 May 1948;see TheEasterRising.eu [n.d.]) There is no evidence of the questions asked by investigating officers or of the particular input that they had in each interview. Few of the statements were recorded using audio, oral statements were most often put on record through the collation of 'copious notes' (Ferriter, 2003). Hence, many of the statements lack the discursive features that would have existed when the witnesses were delivering their statements, but still retain elements of spoken discourse given that, 'The vast majority of witness statements were derived from some kind of interview process, and they bear all the hallmarks of oral testimony (Morrison 2009, p. 7).' Ryan comments further on the linguistic style of the statements: The great majority of the statements are narrative in style, but even within these limits, they are extraordinarily varied. Some witnesses are self-consciously reticent, for example saying 'we' when the clearly mean 'I'; others warm to their subject, becoming quite flowery in their language. The spare, laconic style of many of the female witnesses, when describing intensely emotional events, was typical of the Anglo-Irish Edwardians still surviving in the 1940s. (Ryan 2014, p. 30) In addition to those cited above, there were other issues which investigating officers had to consider, including, as Gkotzaridis (2006, p. 111) points out, 'loyalty to comrades, fear of self-contradiction, and a too happy acceptance of the myth of a happy and successful resolution.' To increase the integrity of the operation, an advisory committee composed of academics and scholars of Irish history was put in place to assist the bureau's staff if such support was needed. Witnesses were identified and attracted to the project through a publicity campaign that included the circulation of an information leaflet, a radio appeal by the Minister for Defence and notices in the national press, but the majority of participants were found through applications to the Military Service Pension scheme. Having completed the lengthy and arduous collation of statements, for reasons of protecting those involved in the project and those who participated in the events depicted in the testimonies, the archive remained largely inaccessible to the public for much of the subsequent six decades. After much debate and solicitation, the archive was finally made publicly available in 2003. The reasons (legitimate and otherwise) for the prolonged holding of these records are many and complex and set out by O'Brien (2004). Members of the public can now access the files in printed form from the military barracks in Dublin or in digital form online at www.militaryarchives.ie. This digitisation has been praised by historians, with Horne and Madigan (2013) citing it as a model of how to democratise history.
One of the key reasons why this archive is seen by historians such as McGarry (2013, p. 47) as 'the most important source to date for understanding the Irish Revolution,' is its inclusion of those who were rarely given the opportunity to contribute their voices to public record, the everyday citizens who, as McGarry states, 'rarely recorded their thoughts in letters, diaries or memoirs ' (2013, p. 47). The inclusion of this breadth of participants, as well as the indulgence provided some participants to catalogue their own life events and lifestyles extraneous to historical events, lends the archive to the curiosity of those not just interested in the historical events for which it was the remit of the bureau to document, but to a broader field of scholarship and attention as well as refocusing interest on the events which they depict, as Crowley et al. (2017, p. 869) suggest, 'the sheer volume and detail of the BMH material permittedindeed, requirednew questions to be asked'. Historians who use the archive tend to provide a caveat related to its questionable legitimacy, which is a typical concern held by historians towards witness testimonies, such as Ó Ruairc (2014, p. 113), 'As these statements were collected between twenty to thirty-five years after the events described they must be treated with caution since they may contain inaccuracies, contradictory evidence or personal bias.' Morrison (2016) likewise acknowledges this, but is also keen to highlight the accuracy which some of the statements contribute to documentation of events that was previously missing, stating that, though the previously stated detractions are present, 'this characteristic should not be exaggerated (p. 878).' Given the passage of time since both the events depicted taking place and the delivery of the testimonies compounded by the further delay in their public release, there exists little opportunity to re-establish the accuracy of the testimonies by accessing those who delivered them. What remains, in the absence of multi-modal sources, is the testimonies themselves and with such scrutiny as that proposed above by McGarry, the solitary method by which they may be verified is through reading and analysis of the language used within. Through the utilisation of corpus tools and linguistic competencies a more thorough reading of the statements can provide a rigorous and well-informed assessment of the extent to which participants were committed to the truth of such statements.

3. The Corpus of Irish Historical Narratives (COIHN)
In compiling a corpus from a large archive such as the BMH, certain criteria must be met to ensure that the corpus meets the needs of the study undertaken. As it was not deemed necessary to use the whole archive in order to get an impression of the linguistic characteristics of the archive, a sample of the archive was selected to provide an optimal representation of its entirety. To create a one-million-word corpus from the statements available in the BMH, which comprises over eleven-million words, many considerations were regarded to ensure that this sample was representative of the collection. Without such consideration, potential conclusions drawn during analysis may not be deemed typical of the collection as a whole. This required a three-staged corpus-design process. As regards representativeness, Leech (2007, p. 135) argues that, 'Without representativeness, whatever is found to be true of a corpus, is simply true of that corpusand cannot be extended to anything else.' As this study does not analyse the entire archive, the sample that is taken from it must be designed to represent the entirety of the archive in order to make inferences about the full corpus. Thus, models of representativeness were adhered to when undergoing corpus design. One aspect of determining the adequacy of a sample to represent the archive is the size that the corpus needs to be in order to suit the purposes of this study, i.e to provide a sample large enough to investigate epistemic modality in the data. Vaughan and Clancy (2013) discuss the merits of using small, domain-specific corpora for pragmatic research and contend that large corpora are unsuitable for some research. This necessitates the return to the purpose of the study to ascertain the optimum amount of words to generate desired results. To this end, a certain amount of trial and error is necessaryconducted via the following stages. First, a corpus was constructed that provided a random sample of the archive, then, after a preliminary analysis and observation of meta-data, this was honed and reformulated into a corpus that was suitable for the purposes of the study. After determining parameters related to corpus size, choices pertaining to text selection were made. As the initial stage of corpus construction did not involve any text selection criteria (the texts were selected randomly by choosing the first one-hundred statements listed alphabetically by surname), this was used to gauge how this approach did not represent the archive as a whole in terms of sociolinguistic factors; gender, location and age. This was compared with the dispersion of these factors in the archive as a whole and used to influence the reselection of texts based on these criteria. For example, 11% of the BMH archive is composed of testimonies delivered by female participants, and this must be reflected in COIHN.
Further considerations when selecting texts included those outlined by Biber (1993), suggesting thought should be given to the kinds of texts included, the number of texts, the selection of particular texts and the length of text samples. In constructing COIHN, text length was deemed an important factor; as some statements originally included were in excess of 70,000 words, it was judged that too large a proportion of the corpus would be represented by some individuals, rather than the many potential voices; as Clancy (2010, p.85) states, 'a relatively small corpus can be skewed by a relatively large text.' Following the removal of a number of statements for these reasons, the size of the corpus was reduced significantly from 699,015 words to 275,010 words. Though the removal of statements for reasons related to length caused the most significant change in the corpus at this stage, there were other reasons for the exclusion of statements. By selecting statements by surname, there was a prevalence of participants from certain geographical regions. For example, the inclusion of all participants with the surname Barry resulted in six participants from Cork being included, where this surname is particularly common.
In rebuilding the corpus, statements were chosen more selectively, paying attention to the geographical location of each participant. The resultant range of geographical locations is broad, at thirty-two, with stronger weight given to regions of higher population and dominance of participation in the events depicted in the statements. Finally, the corpus was reconstructed to represent the archive with consideration given to the factors mentioned here. A summary of the three stages of the construction of COIHN is given in Table 1  Having constructed the corpus, COIHN was investigated to determine the density and type of items used which connote epistemic modality. A hybrid of corpus-based and corpus-driven approaches (Tognini-Bonelli, 2001) was taken as the items chosen for investigation were those which both emerged as salient in COIHN (through analysis of keyword lists) and those which were deemed pertinent to the aims of this study (such as those presented in the subsequent section). While this approach ceded some agency to the corpus itself in directing the research, much of the analytical direction remained in pursuit of predetermined aims and hypotheses regarding linguistic items that contribute to commitment to truth. The three items analysed in Section 5 below are presented to reflect the heteroglossic nature, diversity and complexity of expressions which establish epistemic modality.

4. Epistemic modality
Though hedging is not synonymous with epistemic modality, Coates (1982, p. 49) asserts the interrelationship between epistemic modality and hedging, stating that, 'epistemic modality is always a hedge.' Epistemic modality refers to a speaker's degree of certainty to the truth of a proposition or, as Coates (1987, p. 112) describes it, 'the speaker's confidence or lack of confidence in the truth of the proposition expressed'. Epistemic modality, like hedging, is connected to the act of saving face (Fraser, 1990), in that the degree to which the truth of a proposition is committed to determines the level of risk incurred if the proposition is deemed false or unreliable (Vullioud et al., 2017). The risk, in this situation is to the perception of the speaker as reputable or reliable (Mazzarella et al., 2018 Various devices are employed in the extract above to connote epistemic modality; approximations, vagueness to details highlighted with question marks, modal verbs and adverbs. All of these are worthy of analysis for their expression of commitment to truth (and are the subject of a forthcoming monograph), however, this study focuses on the aforementioned mental process verbs highlighted in bold.

5. Analysis
Holmes (1986, p. 13) proposes three categories to explain degrees of certainty and doubt; certain, probable and possible. Similarly, in her analysis of modal certainty, Simon-Vandenbergen (1997, p.344) suggests degrees of commitment assigned to three values: 1) neutral, expressing lack of commitment (median value) 2) tentative, expressing weak commitment (low value) 3) assertive, expressing strong commitment (high value)

Concordance 1: Examples of I think in COIHN in combination with other hedged expressions
Different levels of commitment in COIHN are exemplified in the concordance lines with I think as the node in the above in descending order of certainty. Halliday (1994) highlights the use of I think in proximity with modals to create a cumulative expression of uncertainty; while it must have been adds certainty in (1) I remember mitigates it in (2), further mitigation is achieved with I'm not sure in (3) and this is compounded with both as far as I remember and I'm not quite sure with I think in (4).
Other studies use differing terminology to scale these values, but the use of three values is consistent, such as low, intermediate and high (Cornillie, 2007); strong, moderate and weak (Nuyts, 2006). A blend of these terms is employed below to describe the use of mental process verbs in the BMH statements, starting with I think. Simon-Vandenbergen (2000) observes the difficulty in categorising the functionality of I think, indeed, the numerous categorisations of the functions of I think tend to suggest conflicting uses. Both Fetzer (2014) and O'Grady (2017) analyse the use of I think in political contexts; O'Grady's (2017, p.297) focus on political debates found four functions of I think in initial position: 1) projecting a speaker's commitment to a proposition by allowing him to seem confident and assured 2) uncertainty or tentativeness

I think
3) evaluation of a proposition (when used in proximity of a modal verb) 4) as a hesitation device (when used with pauses or hesitation markers).
The first two functions here highlight the conflicting and somewhat contradictory nature of the use of I think and the difficulty that this poses in categorising its function with certainty as it expresses both assuredness and uncertainty. Elsewhere, I think has been described as expressing both certainty by increasing authority and objectivity as well as uncertainty by connoting tentativeness (Preisler 1986, Holmes 1990, Aijmer 1997. Brown and Levinson (1987, p. 171) state that, by using I think a speaker is, 'not taking full responsibility for the truth of his utterance.' While Holmes (1990) suggests that I think boosts assertion, Kaltenböck (2007, p. 239) states that it is deliberative 'only in initial positions if it has a nuclear tone or prosodic booster or if it is followed by a thatcomplementiser.' In Kaltenböck's (2010) investigation of the hedging properties of I think, he asserts that the effect of such a term on the listener is most often that the listener assumes this connotes that the statement is an approximation of the truth and that the user is attempting to produce a statement that is as close to the truth as they can attempt to be despite limitations of memory or knowledge. The data here also suggests expressions with remember have a similar effect (see Section 5.2). Zhang's (2011Zhang's ( , 2014 'elasticity framework' proposes a scale of functions of I think that are stretchable and can overlap. The five functions as presented in Zhang (2014, p. 228) are shown below and incorporate the functionality that is articulated in much of the previously stated literature:

elasticity framework
These malleable categories, according to Zhang, work in order of descending strength and I think is chosen by a speaker due to the fluidity of these categories; it is precisely this malleability which makes I think a preferred expression due to its multiple applications and the awareness that a reader/listener may attribute any of these to its use. Concordance 1 (above) shows how these categories manifest in the data when I think is used in combination with other hedged expressions. I think occurs 579 times in COIHN and predominantly functions as an epistemic modal expression rather than as a marker of opinion, which accounts for the majority of its 1,975 occurrences in the spoken, Limerick Corpus of Irish English, (also a 1-million word and used as the key comparative corpus for this study [see Farr, Murphy and O'Keeffe, 2004]). The distinction between opinion I think and epistemic modal I think is made by determining whether it is used to express a positive/negative judgement for the former or degree of certainty to determine it as the latter. The below samples show its usage in expressing tentativeness towards detail in initial position: (1) I think that would be the beginning of May 1921 but it was before O'Malley came to visit us.
(2) I think it was on the morning (afternoon?) of Wednesday that Tom Weafer got a bullet in his (3) I think I am very vague about this that I had a short interview with him; but before I took up (4) I think Dermot O'Hegarty was there. Leo Henderson was there, also Liam Carroll. I distinctly The last of these, (4), suggests a distinguishing between levels of commitment; while doubt is expressed in the first sentence, this is contrasted with the second that suggests a higher degree of certainty when devoid of I think. As with I suppose (see Section 5.3 below), I think also has properties which are used for framing evaluations of the thoughts or feeling of others, as in example (5): (5) I think she was taken aback by my attitude and by the fact that I had not realised it earlier As with the other mental process verbs outlined here, I think, when used in proximity with other items that connote epistemic modality such as modals (6, 7), which suggest a degree of propositional evaluation aligning with O'Grady's framework above, or with approximators (8,9) or with a combination of these (10), create a cumulative effect of tentativeness: (6) I think it must have been Holy Saturday afternoon that my brother-in-law, Joe Connolly, went (7) I think I must have been given a cast-off. There was an electric shot firer to be got at the mines (8) I think that was about one week later after our trial we were brought back to Kilmainham Jail (9) I think that approximately two to four men from each Company were selected to form it. I was (10) I think it must have been about May 1921, we went down to Fidown in the 8th Battalion area Álvarez-Gil and Alonso (2019) describe the relationship between modality and evidentiality expressed by I think. They define evidentiality as 'the set of linguistic mechanisms involved in expressing the source of information from which the speaker has obtained the information to formulate the specific propositional content he or she expresses, as well as the way in which the information has been attained (p. 48)'. Therefore, while evidentiality is an expression related to the source of knowledge, epistemic modality is related to commitment to the reliability of that source (de Haan, 2005).
Evidentiality can inform expressions of epistemic modality as it is concerned with providing the source from which a proposition may be judged as true and so is often dealt with as a sub-category of epistemic modality. Álvarez-Gil and Alonso describe these dual characteristics of I think and found that evidential I think is used as a politeness strategy reducing the risk of face-threatening results, while epistemic I think is used to preserve the positive face of the writer.
It is somewhat difficult to categorise with confidence the occurrences of I think that suggest evidence or epistemic modality. As has been previously stated, the literature on each show that these categories are not always mutually exclusive. However, Diewald et al. (2009, p. 190) propose that evidentiality should be treated externally to epistemic modality, 'Evidentiality is concerned with indicating the information source the speaker is relying on to make a claim. This places this category next to epistemic modality without, however, merging them into one.' This intersective stance is one which corresponds with the perspective of this study in relation to the treatment of I think as, unlike the other two mental process verbs analysed here, I think is used to express a tentativeness towards the truth of the proposition while also attributing the source of that knowledge to the self. Influencing the judgement of the credibility of that source is the memory and subjective experience of the participant. I think is employed in the data, just as those devices explicitly expressing fallibility of memory (see Section 5.2 below) to express the modality of evidence. This infers a merging of evidential and epistemic modality in I think to establish a concurrent multi-functionality. In keeping with that expressed by Nuyts (2001), I think is used as a subjective expression of evidence i.e. the knowledge available to the self is the evidence that informs the narrative. The awareness of the participant as to the degree to which this is reliable helps them decide the degree of epistemic modality to employ to maximally save their face. Thus, three functions of I think emerge, which may be employed independently or concurrently: 1) designation of the self as the source of knowledgesubjective evidential 2) acknowledgement of the fallibility of that knowledgeepistemic modal 3) reduction of face-threatening riskpoliteness In relation to the second of these, Lyons (1977) distinguishes between objective epistemic modality, which expresses an objectively measurable chance that a proposition is true or not and subjective epistemic modality, which involves a purely subjective guess regarding its truth. Others such as Coates (1983) and Palmer (2001) imply an inherent subjectivity in epistemic modality, while Nuyts (2001) proposes a framework of intersubjectivity in epistemic modality as it involves the perspectives of both the speaker and the hearer. In addition to the above explication, it should be noted that I think, as an almost universally applicable truth predicate, reflects that anything can be perceived as true or untrue because this indicates both the source of knowledge as thought and a degree of tentativeness towards that knowledge as being the truth. This holds that I think can be employed as a device from which the 'equal validity paradox' emerges in discourses of not only taste and opinion as suggested by Coliva and Moruzzi (2014), but also in discourses of personal experience where the speaker is not only the producer of an utterance, but also the source of evidence informing it. Even though two statements may provide contrary content about the same event, when the participants use I think they enter into a domain of 'faultless disagreement' where both statements may be deemed equally valid (Kölbel, 1997). If the Law of Non-Contradiction (LNC) states that both sides of contradiction cannot be true (Grim, 2004), then to accept the nature of I think as a self-justifying truth predicate that cannot be disputed, the pursuit of a dialethic perspective that takes into account paraconsistent logic where statements can be deemed both true and untrue is necessary. Though the extracts which exemplify uses of I think in this section are not what Coliva and Moruzzi (p. 38) describe as, 'disputes about what is tasty, beautiful and morally right', these are nonetheless similarly faultless as they are declarations of the self as the sole source of evidence to substantiate the truth of the events described and so can not be contradicted by an audience who does not have access to this intangible source. 'The passing of forty years has dimmed the memory of the events of those historymaking days of 1915 and reminiscences of the true facts are somewhat vague.' The above extract highlights the keen sense of the influence of memory felt by some of the participants of the BMH and the impact of this on their ability to commit fully to the truth of their statements. Though not always stated as explicitly as this, memory is designated throughout the statements as a source of tentativeness to commitment through various expressions. Figure 3 below presents the allocation of the 1,113 occurrences of I remember into strong, weak or moderate memory categories, in accordance with the framework proposed by Nuyts (2005) above, through observation of its usage in context. These include occurrences where the pronoun and verb are separated, for example, I well remember and I can not remember.

I remember
As can be seen from Figure 3, while occurrences of expressions of strong memory are frequent, participants are more likely to express hyperthymesia, which tallies with Achugar's (2008) analysis of the construction of memory in military discourse. In a departure from the stylistic norm of the statements, I remember often collocates with now to perform a degree of metalepsis. Often used in literary criticism, metalepsis is what Genette (1988, p. 234) describes in a somewhat literary fashion as, 'a boundary that is precisely the narrating (or performance) itself: a shifting but sacred frontier between two worlds, the world which one tells, the world of which one tells.' Metalepsis is, for the most part, absent throughout the statements (as evidenced by a lack of present simple constructions), yet in expressions of memory, participants are inclined to use now. This divergence from the relative conformity of the dominant tense (past simple) to frame these expressions into 'the world which one tells' shows a particular proclivity on the part of the participants to express their degree of memory despite the departure from the stylistic norms of the statements inherent in doing so. This necessitates further observation then as to the motives of the participants' willingness to part from the rhetorical norms of the statement to express the degree to which their recollection is reliable. These expressions perform more than just the standard functions of hedges but point towards a manifestation of a discursive construction of truth. Not all statements contain an abundance of these expressions, some statements are far more direct in their recounting of events with a comparatively lower density of hedged expressions. The cognitive acts of thinking and remembering are referred to as spontaneously triggered sources during the act of delivering the testimony: (11) assembled at Walterstown. As far as I can remember now, there were about fourteen or (12) ere of the times pointed that way. I cannot remember now who made the remarks that by (13) enses. Amongst the members whom I can remember now were Christopher Lynam, Jas The cognitive process of remembering while engaged in delivering the testimony is exemplified in the above and further illustrated with broader context in the extract below with both thought and memory emerging as influential processes: (14) Thinking about this now, it may have been before O'Malley's arrest that Treacy was arrested, because I remember that Treacy was the officer commanding the Brigade during the attack on Hugginstown R.I.C. Barracks which took place in March 1920, and I think that was the last action he took part in before his arrest.
The act or performance of the testimony brings the participant to the point of accessing knowledge that had been dormant and participants, as seen in the above extract, are willing to express this flow of emerging knowledge. This also provides evidence of the episodic memory that is prevalent in the articulation of the accounts as opposed to merely event memory or semantic memory (Keven, 2016). Mirativity essentially describes the expectation of knowledge in discourse (Hyslop, 2017). Though Aikhenvald (2012) mainly deals with mirative strategies as prevalent features of languages other than English, she highlights an aspect of mirativity as being the 'unprepared mind' which appears to influence the expressions of spontaneous recollection exemplified in the above samples. Moreton (2016, p.330) highlights the emotional significance of I suppose in Irish famine-era emigrant letters, showing that it both adds to the interactive nature of the letters and influences the construction of an imagined world from shared knowledge of family and friends. This emotional weight of I suppose is less present in the BMH documents due to their monologic nature directed towards an 'imagined audience' (Goffman, 1974), rather than the close personal relationship between the author and recipient of the letters under analysis by Moreton and others (see Hickey and Amador-Moreno 2020). Of the 38 occurrences of I suppose in COIHN, 30 of these collocate with approximators or it is used as a discourse connecting hedge (Fetzer, 2010 Goffman (1967) terms 'connectives'. Connectives, according to Goffman, are used to frame reported speech in phrases such as he said that and according to. Such connectives are seen to hold particular weight in terms of attribution of social responsibility (Goodwin, 1990) in both written and spoken discourse (Rae and Kerby, 2007). The connectives in the examples shown here are referring to thought, experience or knowledge of others and so tally with these phenomena; he knew, he thought, he felt, he was afraid. In accordance with Goffman's participation framework, the person whose thoughts are being accounted for aligns with a principal, though they are not shown to be responsible for the talk as the principal role suggests, they are shown to be responsible for the thought, showing an articulation of the presumed stance of the principal rather than the author. There is an element of external evaluation (Labov, 1972) that is more commonly associated with quotations in these examples, where the main narrative thread is somewhat distanced in order to evaluate the thoughts of others. In framing these propositions with I suppose the participants are showing an awareness of what Schiffrin (1993, p. 233) describes as the, 'speak for yourself rule'. This rule implies that, to speak for another, there are certain devices which need to be employed to, 'show displacement in person, space and time'. As with I think, I suppose reflects a consciousness of the fallibility of the self as the evidence of the thoughts or feelings of another and a conscientious willingness to express this. An added weight to this in the current data may be the effect of knowing that these expressions are attributing knowledge or feelings to those who were killed in the events depicted or not given the opportunity to contribute their own version of events to public record and so commitment to this attribution as fact is mitigated through the usage of I suppose.

6. Conclusion
In summation, it is worth discussing the collective effect of the three mental process verbs presented above. It has been outlined that I think expresses epistemic modality with evidential properties, I remember aligns with the notion of mirativity, while I suppose is used with connectives to framing reported thought (Britt 2018), feeling or experience. A liberal usage of such verbs, in accordance with the literature on hedging might initially assume the effect of mitigating the authority of the witness as reliable presenters of historical events. Paradoxically, the effect on the overall impression of the witness who is willing to express a conscientious appraisal of their own knowledge results in a heightened integrity and an increased impression of reliability. This paradoxical authority is achieved through a presentation of self that is both self-aware and aware of the broader context to which their testimony contributes, boosting, rather than reducing their authoritativeness of their accounts. This study has presented the multi-dimensional matrix of historical discourse and shown how three mental process verbs connoting epistemic modality in testimonies of witnesses to historical events contribute to the linguistic construction of truth in sources of historical significance. Other devices such as fillers, modal verbs, reported speech and passive voice have epistemic modality properties and are worthy of future analysis. In addition, the assessment of context is central to the interpretation of truth, '[t]he discourse of truth is determined by the historical context in which it is also interpreted. This context seems to play a major role in what can be named as the verification of truth' (Deligiaouri 2018, p. 305). An awareness then of the environment in which a text is not only created, but in which it is presented influences its factual integrity.
In terms of the current study, this involves multiple contexts with a seemingly inexorable extension; the context of the event, the context of the depiction of the event, the context of the construction of this depiction, the context of the presentation of this constructed depiction, the context of the representation of this in historical adaptations, the context of the representations in which these adaptations are referencedad infinitum. This study has shown that the analysis of mental process verbs, in conjunction with an awareness of both the context they emerge from and are subsequently utilised can provide a clearer lens through which historical sources may be judged. Rather than looking at the words of history or historical writing (Mazzi, 2014), this study analysed the words that compose the sources of history which subsequently inform historical writing. This has shown that a corpus-assisted (Partington, 2006) framework, when used in combination with established historical methods, may add depth to historical research that involves oral history texts, establishing it as a 'curative science' rather than a 'handmaiden of philosophy' as suggested by Foucault (in Bouchard and Simon 1977, p. 156). While historians engage with texts at a level that is conducive to their discipline's understanding of what is necessary to extract what is needed as sources to inform historical narratives, this engagement likewise pertains to the knowledge of a data-set required to fulfil a corpus-assisted approach (Partington et al., 2013) to textual analysis that is counter to the distanced observation of data-sets in solely corpus analytical frameworks (Sinclair, 2004). Though the complex layers of the multi-dimensional truth matrix remain impenetrable in totality, this study has offered evidence that linguistic tools and competencies can redress the lack of reciprocity between linguists and historians alluded to by Schiffrin (2003), and offer substance to the assessment of the integrity of sources prior to their subsequent dissemination.